EL CURSO TEMPORAL DE LA NEGACIÓN
Transcripción
EL CURSO TEMPORAL DE LA NEGACIÓN
EL CURSO TEMPORAL DE LA NEGACIÓN: PROCESAMIENTO Y REPRESENTACIÓN (The Temporal Course of Negation: Processing and Representation) ISABEL ORENES CASANOVA Director: Carlos Santamaría Moreno Universidad de La Laguna. Facultad de Psicología. Departamento de Psicología Cognitiva, Social, y Organizacional. Área de Psicología Cognitiva. 2 de Diciembre de 2013 2 D. Carlos Santamaría Moreno, Profesor Titular de la Universidad de La Laguna. CERTIFICA que la presente Tesis Doctoral, realizada por Isabel Orenes Casanova con el título “El curso temporal de la Negación: Procesamiento y Representación”, ha sido elaborada bajo su dirección y supervisión en el Departamento de Psicología Cognitiva, Social y Organizacional de la Universidad de La Laguna, y que dicha Tesis reúne, a su juicio, los requisitos académicos y científicos necesarios para proceder a su presentación y defensa pública ante el tribunal que se designe al efecto. La Laguna, 29 de octubre de 2013 Fdo. Carlos Santamaría Moreno 3 4 Índice ‘*’ indicates that the chapter is written in English Agradecimientos ............................................................................................................... 7 Prefacio: el puente entre el conocimiento y la realidad .................................................. 11 *General Abstract........................................................................................................... 19 1. Negación: Concepto y Uso ........................................................................................ 27 2. Procesamiento de los enunciados negativos ............................................................... 31 2.1. Curso temporal de la negación: doble procesamiento ............................... 32 2.2. El tamaño del conjunto complementario ................................................... 36 2.3. La tarea de verificación: Integración de enunciados e imágenes .............. 39 3. Representación del operador negativo ........................................................................ 42 4. Estudiando la negación a través de los movimientos oculares ................................... 46 5. Objetivos..................................................................................................................... 54 6*. How negation is understood: Evidence from the visual world paradigm ................ 57 6.1. Abstract ..................................................................................................... 57 6.2. Introduction ............................................................................................... 58 6.3. Experiment 1 ............................................................................................. 65 6.3.1. Method .......................................................................................... 66 6.3.2. Results ........................................................................................... 69 6.3.3. Discussion ..................................................................................... 75 6.4. Experiment 2 ............................................................................................. 76 6.4.1. Method .......................................................................................... 77 6.4.2. Results and Discussion ................................................................. 77 6.5. Experiment 3 ............................................................................................. 80 6.5.1. Method .......................................................................................... 82 5 6.5.2. Results and Discussion ................................................................. 83 6.6. General discussion ..................................................................................... 85 6.7. Conclusion ................................................................................................. 89 7*. Visual concepts impede negation ............................................................................. 91 7. 1. Abstract .................................................................................................... 91 7. 2. Introduction .............................................................................................. 91 7. 3. Experiment 1 ............................................................................................ 94 7.3.1. Method .......................................................................................... 94 7.3.2. Results and discussion .................................................................. 98 7. 4. Experiment 2 ............................................................................................ 99 7.4.1. Method ........................................................................................ 100 7.4.2. Results and discussion ................................................................ 101 7.5. General discussion ................................................................................... 103 8*. Inconsistencies make negation plausible: Evidence from the VWP ...................... 105 8.1. Abstract ................................................................................................... 105 8.2. Introduction ............................................................................................. 105 8.3. Method..................................................................................................... 109 8.4. Results ..................................................................................................... 112 8.5. General discussion ................................................................................... 120 9*. General Conclusion ................................................................................................ 123 References .................................................................................................................... 135 6 Agradecimientos Me gustaría dedicar esta tesis a mis padres por confiar siempre en mí. También quisiera agradecer al resto de mi familia, amigos y compañeros de trabajo por compartir sonrisas, buenas conversaciones, sabrosas comidas y mensajes de ánimo para seguir adelante en este proyecto. Gracias también al Ministerio de Ciencia y Educación de España por otorgarme una beca de formación del profesorado universitario (FPU) y a la fundación canaria Manuel Morales, por becarme para seguir mi formación como investigadora durante el año 2012. A la Universidad de La Laguna y sus profesionales por facilitar la ejecución de este trabajo. A Alberto Avilés, Jon Andoni Duñabeitia, Manuel Gutiérrez, y Christoph Scheepers por ayudarme con la técnica de movimientos oculares, técnica principal de la tesis. A Phil Johnson-Laird, Sunny Khemlani, y Linda Moxey, con quiénes tanto he aprendido sobre psicología cognitiva. A la Universidad de Glasgow, donde he trabajado muchos meses. Y a los estudiantes de la facultad de psicología, que muy amablemente participaron en los experimentos. En especial, mi agradecimiento a Carlos Santamaría por su inteligencia, conocimiento, y aportación en esta tesis así como su espíritu crítico que permite analizar la realidad de forma alternativa. Y a David Beltrán, GRACIAS por TODO. 7 8 A mis padres 9 10 Prefacio Prefacio: el puente entre el conocimiento y la realidad Me gustaría comenzar la tesis guiando al lector sobre lo que se va a encontrar en ella. En este mismo primer apartado reflexionaré muy breve y libremente sobre un tema que desde siempre me ha fascinado y que me ha llevado a adentrarme en el estudio del pensamiento humano: la capacidad del ser humano para construir una sociedad como la actual, de convertir piedras en instrumentos, sonidos en símbolos, y en definitiva, de comprender y trasformar el mundo que le rodea, y esto sólo ha sido posible, a mi entender, a través del conocimiento. Seguidamente, el lector entrará ya de lleno en lo que constituye el tema de la presente tesis, la negación, encontrándose en primer lugar con la introducción, en la que se plantea un marco teórico en relación a su procesamiento y representación. A dicha introducción le seguirán los tres bloques de estudios experimentales que suponen el núcleo principal del presente trabajo, los tres escritos en inglés y en formato adecuado para su publicación. Y finalmente, la conclusión general, donde se expondrá los principales resultados obtenidos en los diferentes estudios experimentales así como sus principales aportaciones teóricas, también escritas en inglés (igual que el Resumen de la tesis), por exigencias de la modalidad de mención de doctorado internacional a la que se adscribe la presente tesis doctoral. 11 Prefacio Quisiera señalar que el concepto de mente es el punto oscuro que llegó a obsesionar a los intelectuales de Occidente cuando al fin renunciaron a ese otro punto oscuro, el concepto tecnológico de Dios. El carácter inefable de lo mental cumple la misma función cultural que el carácter inefable de lo Divino: sugiere vagamente que la ciencia no tiene la última palabra (Rorty, 1982). Recuerdo que la primera vez que leí las palabras de Rorty me quedé atónita. Tanto escepticismo te invita a reflexionar sobre el concepto de mente, que no es algo banal ya que es el objeto de estudio de la psicología. Desde mi punto de vista siguen sin estar claras preguntas básicas como: ¿existe realmente la mente?; si existe, ¿cómo definirla?; y finalmente, la pregunta más importante para un investigador en el campo de la psicología, ¿cómo estudiarla? Quizás fue Descartes el autor que distinguió más claramente entre la mente y el cuerpo. Una idea mucho menos racionalista y más romántica fue la que algunos defendieron al situar la mente en el corazón. Probablemente hoy en día se concibe, fundamentalmente dentro de la Neurociencia Cognitiva, y siguiendo la frenología y la concepción más empirista de Locke, que la mente se reduce al cerebro. Las neuronas son las responsables de procesar información en el cerebro. Son unas células bastante peculiares ya que son las únicas que no se reproducen y tienen la misma vida que el propio individuo. Mientras que la mayoría de las células se regeneran, por otras que son capaces de realizar la misma función a través de la información genética y la síntesis de aminoácidos, esto no ocurre para las neuronas. Lo que diferencia a estas células del resto son las dendritas y los axones a través de los cuáles se comunican usando señales electroquímicas. Y digo bien, señales 12 Prefacio electroquímicas, porque toda la información que reciben nuestros sentidos, bien sea a través de energía luminosa o química, es traducida al código electroquímico de las neuronas. Ésta es la información que parece manejar el cerebro, por lo que uno se sigue preguntando cómo estudiar la información, entendiendo ésta, ya no como código eléctrico-químico, sino como conocimiento. Cuando estudiamos los procesos cognitivos a través del cerebro, éste nos informa sobre las partes que se activan, la actividad cerebral que produce o el neurotransmisor implicado; sin embargo, en mi opinión, y espero que en la de otros muchos también, esta información no explica la función cognitiva en sí, entendiendo por explicación un principio que causaría dicha función cognitiva. Por ejemplo, la caída de un cuerpo responde al principio que propuso Newton y está relacionado con la masa, distancia y la constante de gravedad. El área de Broca parece que se activa con la producción del lenguaje, pero su activación, por ejemplo con la técnica de estimulación magnética (TMS), no podría producir directamente que una persona diga de manera específica la palabra ‘hola’. Hoy por hoy sigue siendo más sencillo saludar a la persona para obtener tal respuesta. Una cuestión menos complicada que la producción del lenguaje, podría ser las causas de la ansiedad. Recuerdo estar en una clase de psicofarmacología donde la profesora explicaba los mecanismos fisiológicos de la ansiedad, pero la realidad es que es imposible descifrar si tal estado está provocado por la boda o el despido de alguien, a menos que el propio sujeto informe de ello. Es evidente que estos mecanismos neurofisiológicos solamente nos cuentan una parte de la historia, para obtenerla completa se necesita algo más. Recientemente, en un artículo científico publicado en Proceedings of the Royal Society B: Biological Sciences, Susan Edwards y Stephen Pratt (2009) describían el 13 Prefacio paralelismo que podría existir entre la conexión de neuronas y una colonia de hormigas. Ambos conjuntos carecen de inteligencia de forma aislada, sin embargo, en conjunto dan lugar a conductas inteligentes y conscientes. Desde este punto de vista, resulta complicado localizar la mente en algún lugar del cerebro, más bien la mente surge o es un potencial de la interacción de varias variables. Esta visión me recuerda a la concepción que existe en física cuántica sobre los electrones, los cuales pueden estar en varios sitios a la vez, pero ahí no acaba todo, parece que el lugar donde se encuentran depende del observador. Esto sólo es posible si los electrones en vez de concebirlos como materia se conciben como información. Si entendemos la mente como información en este mismo sentido, yo me pregunto entonces si la mente puede estar en varios sitios a la vez. Quizás sea algo parecido a la mente social que proponía Jung, y sobre todo si la información depende del observador, y éste es nuestro objeto de estudio, ¿podría haber varias mentes posibles? Concebir la mente como información es una influencia de la teoría de la información y la computación. Algunos restringen la información a patrones de activación como sería la teoría conexionista, pero, personalmente mi interés reside en conocer cómo el ser humano es capaz de transformar la naturaleza en una sociedad, y desde este punto de vista es necesario concebir la información como conocimiento. Definición acertada o no, es una definición y probablemente el primer paso para preguntarse cómo estudiarla: ¿cómo se adquiere el conocimiento? ¿Cómo se organiza y representa? ¿Cómo se procesa y memoriza? ¿Qué información podría dar el cerebro sobre el conocimiento? Etc. En la presente tesis nos proponemos estudiar fundamentalmente los formatos de representación del conocimiento, específicamente nos interesamos por la posibilidad de 14 Prefacio que los individuos puedan representar la negación. Los niños a una edad muy temprana (en torno al año) usan la negación cuando apenas tienen diez palabras en su vocabulario. El primer vocabulario que producen suele estar relacionado con palabras esenciales con un claro referente. Por ejemplo, es frecuente escucharles decir mamá, papá, tata, teta, leche, agua. Es interesante descubrir o más bien darse cuenta que entre estas primeras palabras se encuentra la negación, una palabra sin un claro referente, sin embargo su función es fundamental porque se aplica a cualquier enunciado (proposición) y modifica su significado (cambia su valor de verdad de falso a verdadero y viceversa). Además la negación está presente en cualquier lenguaje, natural o artificial, lo que nos hace pensar que su función es fundamental para nuestro desarrollo como humanos. Descubrir o esclarecer cómo las personas comprendemos la negación es el objeto de estudio de esta tesis, concretamente nosotros vamos a plantear fundamentalmente cómo nuestro sistema cognitivo la procesa y representa. 15 16 El Curso Temporal de la Negación: Procesamiento y Representación (The Temporal Course of Negation: Processing and Representation) 17 18 General Abstract General Abstract Representation has been a central issue in cognitive science since the beginning (see Bruner, Goodknow, & Austin, 1956). Nowadays, it is commonly accepted that iconic (or analogical) representation is essential to language comprehension (Barsalou, 1999; Glenberg, Meyer, & Lindem, 1987; Johnson-Laird, 1983; Zwaan & Radvansky, 1998). This means that individuals understand language by representing an analogical structure in their mind that corresponds to the structure of the world being described. However, there are some components of natural language, such as negation, that do not have any analogical correspondence in the real world. Following Wittgenstein’s classical example (1953), one could well employ an image of a red figure with a superimposed cross to represent the sentence ‘The figure is not red’, but one would have still to know that the cross symbolizes negation, and nothing in the image could tell us that. Furthermore, one would have to know what negation itself means. Negation is a syntactic operator that takes an argument (e.g., Spain won the Confederations Cup final in 2013) and reverses its truth value. This means that if an argument is false, then its negation is true (e.g., Spain did not win the 2013 Confederations Cup) and vice versa. This meaning can only be captured symbolically. However, embodiment theory, one of the frameworks that has risen to dominance in cognitive science today, holds that negation is only understood iconically. The main contribution to knowledge of the present thesis is the finding that individuals are able to understand negation symbolically. Although previous studies have shown this using reaction time (Clack & Chase, 1972), we present evidence of the unfolding processing of this kind of representation using the tracking of eye movements. The assumption is 19 General Abstract that there is a direct relation between what the eyes look at and what the mind processes. Other contributions of the present thesis relate to how the comprehension of negation is modulated by the representation of the negated argument and the context in which it occurs. The meaning of negation is difficult for most individuals to understand, leading to longer processing times and higher error rates compared to the corresponding affirmation. A variety of factors have been proposed to account for this fact, ranging from purely grammatical mechanisms to the influence of affective connotations. But accounts based on semantic and pragmatic factors are surely the most supported by empirical evidence (Miller, 1962; Wason, 1965). Semantic accounts attempt to answer the question of how the meaning of a negative sentence is understood by a representational mind. As a result, they explain the differences between negation and affirmation on the basis of mental processes and representations. A considerable part of this dissertation is dedicated to delving further into these types of cognitive explanations of how negation is understood. First, we make use of a new methodology, the visual world paradigm, to reconcile two well-established cognitive theories: embodiment and mental model theory. Second, we offer a new perspective of the semantic function of negation and examine its impact on the comprehension of concepts that vary in their degree of concreteness. Pragmatic accounts confront the issue of negation complexity from a different perspective. They focus on providing an answer to the questions of why and how people use negation in a communicative context, and consider that comprehension is definitively influenced by how much appropriate (felicitous) information is being used in a given context (Givon, 1978; Wilson & Sperber, 2004). The main pragmatic account 20 General Abstract (the denial hypothesis) claims that negation is used to deny what a hearer might be mistakenly keeping in mind. In the third part of this dissertation, the role of denial in comprehension is evaluated, again by means of the visual world paradigm methodology. The following is a summary of the different experiments run to study the three main predictions comprising the experimental part of this dissertation. 1. How negation is understood: Evidence from the visual world paradigm. As explained above, the first goal of the present dissertation was to show that individuals are able to understand negation symbolically. To test this, we studied the process of understanding negation and the resulting mental representations. The twostep theory (Kaup, Lüdtke, & Zwaan, 2006), based on the embodiment theory, holds that the understanding of a negative sentence, like the figure was not red, leads to the construction of two iconic mental representations: first, the negated situation (e.g., a red figure) and next, the actual situation (e.g., a green figure). Remarkably, this theory claims that the two representational steps are compulsory, in that they must occur whenever the comprehension of a negative sentence is at stake. In this way, the inclusion of symbolic operations is avoided. Alternatively, the mental model theory, while accepting that individuals will represent the actual situation to understand negation whenever possible, also holds that they can represent the negated situation and apply a symbol for negation. Hence, this theory recognizes two ways of dealing with negation, as well as the possibility of including symbolic operations in its comprehension (Khemlani, Orenes, & JohnsonLaird, 2012). Our hypothesis was that the comprehension of negative sentences is modulated by the availability of the actual situation. In our experiment, this actual situation was 21 General Abstract made available in what we called the binary context, in which a disjunctive sentence like the figure could be red or green appeared before a negative sentence like the figure was not red. But it was unavailable, in the sense of underspecified (or indefinite), in what we called the multary context, wherein the same negative sentence followed a sentence like the figure could be red, or green, or blue, or yellow. Also, we made use of the visual world paradigm methodology, which allowed us to track semi-continuously participants’ visual attention over a set of four figures shown in the screen, while they were hearing the passages containing both the context (binary or multary) and the target sentence (affirmative or negative). For the binary context, the same outcome was predicted by both the two-step and the mental model theories: that there would be an initial focus of attention on the figure representing the negated situation (a red figure) followed by a focus on the figure representing the actual situation (a green figure). In contrast, different predictions were anticipated for the multary context: under the mental model theory, the prediction was that there would be persistent attention on the figure representing the negated situation, while the two-step theory predicts a progressive decrease of attention on that figure. In Experiment 1, we confirmed predictions of the mental model theory and found that people shifted their visual attention toward the figure representing the actual situation in the binary context, while focusing exclusively on the negated figure in the multary context. This latter finding, the focus of attention on the negated figure, was confirmed using different tasks in two follow-up experiments (Experiment 2: recognition; Experiment 3: verification) in which negative sentences were presented without any preceding context. The finding of a persistent focus on the negated figure suggests that the comprehension process was relying on the representation of the 22 General Abstract negated situation, and hence that the actual situation is of no use for understanding negation outside of a binary context. Thus, these findings pose a problem for the two-step theory. They show that the two simulation steps are not necessary to understand negation. Instead, they tend to support the mental model theory, which claims that negation can be understood by keeping the negated situation plus a symbolic tag indicating that this situation is negated. 2. Visual concepts impede negation. The second goal of the present dissertation was to demonstrate that an argument that is symbolic could make negation easier to understand. We predicted that the negation of arguments that involve visual words (related to the representation of an image) would be slower to process than negation involving non-visual words (although visual words are ordinarily easier to process). To test this hypothesis, we carried out two experiments. In Experiment 1, we presented a context that alternated between negation of the argument involving the visual (iconic) word: the boy was brave and he was not asleep and negation of the non-visual (or symbolic) argument: the boy was awake and he was not cowardly. The next sentence (the target) consisted of a similar conjunctive phrase with the same meaning, but with the previously negated word replaced by its contrary visual argument (e.g., the boy was brave and he was awake). The target sentence was similar for both conditions, and we measured the time that people took to read this target. As predicted, results showed that participants took longer to read the target sentence after the negation of the visual argument than after negation of the non-visual argument, although visual concepts had been found to be easier to comprehend than non-visual concepts in a previous lexical decision task. 23 General Abstract In Experiment 2, we corroborated the previous result using a similar design. In this case, we presented a context that alternated between negation of the argument involving visual word (e.g., the boy was brave and he was not asleep) and negation of the non-visual word. The next sentence (the target) consisted of a similar conjunctive phrase, but without negation (e.g., the boy was brave and he was asleep). This was done to test that the effect found in Experiment 1 was due to the type of representation of the negated word (e.g., asleep), rather than its cancellation. Together, these two experiments lead to the conclusion that iconic representation could impede some symbolic operations such as negation. 3. Inconsistencies make negation plausible: Evidence from the visual world paradigm. The third goal of the present dissertation was to prove whether the context in which the argument is presented could make negation predictable and easier to understand. A special case of this prediction occurs when the negation functions to deny a misconception. Outside of this context, negation is more difficult to understand than affirmation (Givon, 1978). To test this hypothesis, we presented three types of context followed by affirmative or negative sentences (e.g., her dad was (not) poor) while presenting images of the affirmed (e.g., a poor man) or denied entities (e.g., a rich man). In a neutral context (e.g., her dad lived on the other side of town) the participants looked at the affirmed entities before the denied entities. This effect was magnified when the context corresponded to a contextually true belief, that is, when the context and the target sentence were consistent (e.g., She supposed that her dad had enough savings followed by her dad was not poor). However, people looked at the object that corresponded to the negation within the same time frame as they looked at the object that corresponded 24 General Abstract to the affirmation when the sentence referred to a contextually false belief, that is, when the context and the target sentence were inconsistent (e.g., She supposed that her dad had little savings followed by her dad was not poor). Those results corroborated our prediction. Negation was as easy to comprehend as affirmation when it was used to deny a false belief (or misconception). Outside of this context, negation was more difficult to understand than affirmation. The findings of this section confirmed denial hypothesis predictions, leading to the conclusion that there is an asymmetry in the usage of the main semantic function of negation (truth-value reversal) that modulates comprehension. In particular, they indicate a preference of using negation to reverse the value of mistakenly held ideas. It is likely that this asymmetry in use is related to the important maxim in communication of telling the truth. Conclusions. The empirical analysis addressed in this dissertation has yielded three main findings with important consequences not only for the topic of negation, but also for the more general topic of language comprehension. As a whole, these findings lead us to conclude that the comprehension of negative sentence is modulated by the core semantic meaning of negation (truth-value reversal). This meaning can be only represented symbolically, from whence follows that any representational theory of human language has to consider including symbolic representation for connective concepts like negation. Our results fit better with the mental model theory than with the two-step theory. Also, we have revealed for the first time that iconic (visual) representations slow the comprehension of negative sentences. This finding could be generalized to reasoning and would indicate that symbolic representation could be essential in certain complex tasks. This type of representation 25 General Abstract allows for ease of operation, because it reduces the processing load on working memory. And finally, we have shown that although negation is more difficult to understand than affirmation, the meaning of negation is easily understood when it is used in a denial context. 26 Procesamiento y Representación de la Negación 1. Negación: Concepto y Uso Desde Aristóteles, la negación -expresada generalmente en castellano con la palabra nose define como un operador sintáctico que se aplica sobre un argumento (palabra o frase) y revierte su valor de verdad, de falso a verdadero y viceversa (ver Aristotle, 1984). En términos lógicos un enunciado y su negativo se contradicen el uno al otro, es decir, si uno es verdadero, el otro es necesariamente falso. Por ejemplo, si la frase afirmativa ‘Jefferson inventó el pararrayos’ es falsa entonces su correspondiente negativa ‘Jefferson no inventó el pararrayos’ debe ser necesariamente verdadera. Fíjese el lector que la negación se usa para señalar que un argumento es falso: ‘Es falso que Jefferson inventó el pararrayos’. La función de la negación quizás sea falsar enunciados, pero la negación y la falsedad son dos términos diferentes. Primero, porque un enunciado negativo (por ej., No estoy en España) puede ser tanto verdadero como falso y, paralelamente, un enunciado falso (por ej., Es falso que estoy en España) podría expresarse tanto en modo afirmativo como negativo (por ej., Es falso que no estoy en España). Y segundo, porque la negación es parte del lenguaje, un elemento sintáctico del mismo con consecuencias semánticas, mientras que la falsedad es un aspecto del metalenguaje, es decir, no es expresado en el lenguaje en sí sino más bien dice algo sobre éste (ver Khemlani, Orenes, & Johnson-Laird, 2012). Con respecto al uso o función de la negación, una idea bien establecida es que no tiene ningún sentido negar un enunciado a menos que exista alguna razón para creer que lo afirmado por dicho enunciado es falso (véase por ejemplo, Horn, 1989). Según esta idea, la única razón para que alguien dijese que ‘Jefferson no inventó el pararrayos’ 27 Procesamiento y Representación de la Negación es que previamente se hubiese afirmado, o bien se aceptase más o menos implícitamente como verdadero que ‘Jefferson inventó el pararrayos’. De no ser así, seguramente el hablante habría utilizado directamente la frase afirmativa verdadera: ‘Franklin inventó el pararrayos’ – por supuesto, siempre y cuando estuviese en posesión de dicho conocimiento. Por tanto, el uso de la negación tiene como función principal falsar un mensaje, o en un sentido más amplio corregir la idea de un hablante (Givon, 1978; Strawson, 1952; Wason, 1965). La función correctora de la negación sugiere que el hablante hace uso de ella para corregir la opinión de un interlocutor, posiblemente con la esperanza de que éste acabe rechazándola y abandonándola. No obstante, dicho distanciamiento quizás no se produzca, y no tanto por culpa del apego que el interlocutor le tenga a su propia opinión, como por la importancia que ésta adquiere al caer dentro del foco de la negación. Negar una idea significa dotarla de cierto protagonismo en el ejercicio de la comprensión. Si o no ello conduce al rechazo y consecuente olvido de dicha idea depende de factores que en ocasiones pueden pasar inadvertidos para el hablante. Así, si se dice a alguien que ha sufrido un accidente ‘No pienses en el accidente’, lo más probable es que su pensamiento acabe precisamente centrado en el accidente, por más que nuestro objetivo no fuese precisamente ese. Por tanto, parece poco eficaz el uso de la negación cuando el objetivo es promover el abandono por parte del interlocutor de una idea o conducta. Aunque, como bien indicó Freud (1925), el hecho de que alguien niegue una idea de manera repentina e inesperada quizás deba tomarse precisamente como indicio de aquello entorno a lo que realmente giran sus preocupaciones. En lugar de sustituirla por otra más apropiada, Freud recomienda al psicoanalista centrar el análisis sobre lo que 28 Procesamiento y Representación de la Negación está siendo negado. De ahí, ya emergerá por asociación lo que bulle en su subconsciente. El estudio del desarrollo del lenguaje deja bien claro que existen varios usos diferentes de la negación, no solamente el que lo vincula a una función correctora (Horn, 1989). Por ejemplo, cuando los niños empiezan a utilizar la palabra no la usan para indicar que un objeto ha desaparecido: ‘No está’, y también se utiliza tempranamente para oponerse o rechazar algo. Como cuando la madre le dice ‘Vamos a dormir’ o ‘Ponte la camiseta roja’ y el niño le responde que ‘No’. En este sentido, el niño trata de comunicar que no está de acuerdo con lo que dice la madre o no está motivado para hacerlo. Además los niños están familiarizados con la negación porque los adultos la utilizan para prohibir, ‘No metas los dedos en el enchufe’ o ‘No cruces la carretera solo’ (Morris & Hasson, 2010). Por tanto, la negación puede ser usada con otros objetivos como la comunicación de no-existencia, rechazo o prohibición. De hecho, la función de falsar, su principal uso, es la última que se adquiere en el desarrollo cognitivo (para una revisión, Horn, 1989). Hasta ahora hemos visto ejemplos en los que la negación se aplica sobre todo el enunciado. Actualmente sigue abierto sin embargo el debate sobre si la negación podría aplicarse a un argumento menor que la proposición en su conjunto (Khemlani y cols., 2012). La realidad es que la negación se puede aplicar a diferentes partes del enunciado. Por ejemplo, en el verbo o en el adjetivo. En relación a este último ejemplo, es relevante el caso de los adjetivos marcados. Un adjetivo marcado es aquel que contiene una negación implícita (Clark, 1969; 1974). Por ejemplo, el adjetivo ‘ausente’ es marcado porque contiene la negación ‘no presente’. El uso de los adjetivos marcados es de interés porque su uso no es necesario, es decir, éstos pueden ser expresados a través de 29 Procesamiento y Representación de la Negación la negación (por ejemplo, no encendido-apagado; no grande-pequeño; no alto-bajo; no rico-pobre), sin embargo las personas prefieren utilizar otra palabra, y no incluir directamente la negación. Hay algunos conceptos contrarios que se han formado sustituyendo el operador negado no por un prefijo: im-, des-, in-, di-, etc. Lo relevante es que las palabras y sus contrarios suelen estar muy asociadas, es decir, unas evocan a otras y viceversa (Fernández y cols., 2004). Este hecho es importante en el procesamiento de la negación porque permite sustituir la negación por su contrario, y como veremos más adelante, un enunciado afirmativo se procesa y memoriza mejor que un enunciado negativo. Ahora bien, no siempre una negación puede ser sustituida por su contrario. Por ejemplo, cuando se escucha en los medios de comunicación que ‘los funcionarios no son ricos’, rara vez se entiende que ‘ellos son pobres’, más bien se entiende que ellos no son ni pobres ni ricos sino que pertenecen a la clase media. Este ejemplo nos indicaría que la negación tiene la función de modificar o mitigar, no siempre contradecir, el argumento sobre el que se aplica. Este sería el caso de adjetivos como caliente y frío. Si uno pide un café en el bar y dice ‘el café no está caliente’ no significa que está frío, más bien que no está todo lo caliente que le gustaría o que está templado. La negación en el lenguaje natural, además de contradecir, modifica el argumento sobre el que se aplica en un sentido más amplio y variado, el cuál dependerá en muchos casos de la pragmática de la comunicación (Giora, 2006; Schul, 2011). 30 Procesamiento y Representación de la Negación 2. Procesamiento de los enunciados negativos La negación ha sido tradicionalmente estudiada en lingüística y filosofía (ver Horn, 1989). Sin embargo, a partir de los años 60 y 70 los psicólogos empezaron a interesarse por cuál podría ser el efecto del operador negativo sobre el procesamiento cognitivo. Los resultados han mostrado de forma consistente que los enunciados negativos son más difíciles, producen mayor tiempo de reacción y mayor número de errores que los enunciados afirmativos (Wason, 1959, 1961). En un primer momento se interpretó que el mayor tiempo en la lectura de un enunciado negativo podría deberse a la presencia de una sílaba extra, pero el tiempo que se estima para leer una sílaba es de 25-90 ms mientras que el efecto de la negación es de más de 100 ms (Clark & Chase, 1972). También se ha destacado que la dificultad de la negación podría deberse a que ha sido asociada con connotaciones negativas (por ej., prohibiciones; Wason & Jones, 1963). En este sentido, Eifermann (1961) realizó un experimento en hebreo, en cuya lengua existe dos palabras para el no: lo que se usa para prohibir y eyno que se usa como negación. Los resultados mostraron tiempos de lectura más largos y más errores asociados con lo que con eyno. Este experimento iría además en contra de aquellos que consideran que la dificultad de la negación es debida a su poca frecuencia ya que lo es más frecuente en el lenguaje hebreo que eyno. Otra posible explicación ha sido que la negación requiere de un mayor número de trasformaciones gramaticales comparado con los enunciados afirmativos (Klima, 1964). Aunque Miller (1962) mostró que factores semánticos, y no sintácticos, son los responsables de la complejidad de la negación, al encontrar que las diferencias entre los enunciados afirmativos y negativos se reducían cuando eran oraciones pasivas frente a activas. 31 Procesamiento y Representación de la Negación A pesar de las numerosas interpretaciones que han tratado de explicar por qué los enunciados negativos resultan más difíciles que sus correspondientes enunciados afirmativos, probablemente la que más peso ha tenido en la literatura, y la que se detallará en el siguiente apartado, ha sido que para comprender enunciados negativos se requiere de un doble procesamiento que no es necesario para la afirmación. Esta idea viene del lógico Bertrand Russell quien escribió que ‘cuando digo “esto no es azul” de alguna manera considero que es azul primeramente y después lo rechazo, considerando un color diferente al azul’ (Russell, 1948). 2.1. Curso temporal de la negación: doble procesamiento Siguiendo las propias palabras de Russell se podría considerar que para la comprensión de enunciados negativos, tales como ‘la puerta no está abierta’, primero se procesa el argumento negado, abierta, y después se aplica el operador negativo no sobre dicho argumento para inferir su conjunto complementario, cerrada. Esta inferencia requiere de tiempo adicional comparado con su correspondiente enunciado afirmativo, ‘la puerta está abierta’, cuya comprensión no requiere de ninguna inferencia, y de ahí su menor dificultad. Recientemente, Kaup, Lüdtke y Zwaan (2006) corroboraron esta idea a través de una tarea de naming. Los autores presentaron enunciados negativos con predicados contrarios (por ej., La puerta no está abierta) y después de 750 o 1500 ms presentaron una imagen que correspondía (por ej., una puerta cerrada) o no con el enunciado (por ej., una puerta abierta). La tarea de los sujetos era nombrar el objeto de la imagen, en ambos casos ‘puerta’. Encontraron que a los 750 ms, los sujetos nombraban más rápido ‘puerta’ cuando veían una puerta abierta que una cerrada, mientras que los resultados 32 Procesamiento y Representación de la Negación eran opuestos cuando el intervalo entre enunciado y figura fue de 1500 ms. En este caso nombraban más rápido ‘puerta’ cuando ésta era una puerta cerrada que cuando era abierta. Estos resultados parecían confirmar que los individuos representan primero el argumento negado (por ej., abierta) y después el conjunto complementario (por ej., una puerta cerrada). Además, demuestran también que tal inferencia podría darse entre los 750 y 1500 ms, con lo cual la comprensión de la negación podría ser tardía. Moxey y Sanford (1987) también mostraron que para la comprensión de ciertos cuantificadores era requerido el doble procesamiento. Por ejemplo, los autores encontraron que los individuos después de leer ‘Unos pocos de los fans fueron al partido’ inferían ‘Ellos gritaron en el campo’, mientras que cuando leían ‘Pocos fans fueron al partido’ inferían ‘Ellos se quedaron en casa’. Esto es, el cuantificador ‘unos pocos’ hace referencia a los que van a ver el partido mientras que ‘pocos’ es un cuantificador negativo que hace referencia al conjunto complementario, es decir, a los que no van a ver el partido. En este sentido, el uso de estos cuantificadores sería el mismo que el de la negación. Tal y como se ha planteado antes, sólo si alguien piensa que ‘Jefferson inventó el pararrayos’, se diría que ‘Jefferson no inventó el pararrayos’. De la misma manera sólo en el caso de que alguien pensara que se iba a llenar el campo entonces se podría decir que ‘Pocos fans fueron al partido’. Igualmente, se ha encontrado que en la comprensión de condicionales contrafactuales (por ej., Si los deportistas hubieran ganado el partido entonces ellos hubieran tenido unas vacaciones) y también de aquellos enunciados que expresan deseos (por ej., A los deportistas les gustaría ganar el partido) se procesan tanto la situación que se describe como su complementaria (ver Byrne, 2005). Esto es, se produce un doble procesamiento similar al antes visto para la negación y los 33 Procesamiento y Representación de la Negación cuantificadores. Aunque, recientemente Nieuwland y Martin (2011) han obtenido resultados contrarios que parecen indicar que no es necesario el doble procesamiento para su comprensión. La necesidad de un doble procesamiento también ha sido puesta en duda para la negación recientemente. Por ejemplo, Nieuwland y Kuperberg (2008) presentaron una tarea de lectura usando la técnica de potenciales evocados, con la peculiaridad de que los enunciados afirmativos y negativos eran plausibles (por ejemplo, ‘Viajar a Bagdad no es seguro por la guerra’ podría ser similar a ‘Viajar a Bagdad es peligroso por la guerra’ según indicaron los participantes en un estudio normativo, es decir, ambos enunciados eran considerados igualmente plausibles). Los autores encontraron efectos principales de verdad en un componente cerebral relacionado con la integración semántica (N400). Es decir, encontraron que los enunciados falsos presentaban mayor N400 que los enunciados verdaderos con independencia de la polaridad, esto es, de si eran afirmativos o negativos. Este efecto principal de verdad solamente es posible encontrarlo si la interpretación de la negación ya está disponible en el momento que se lee la palabra clave, lo que sugiere que la negación está siendo integrada inmediatamente en el proceso de comprensión. Tan rápida integración de la negación no es congruente, sin embargo, con el doble procesamiento, el cual según han indicado algunos estudios requeriría de entre 750 a 1500 ms (Kaup y cols., 2006). Tian, Breheny, y Ferguson (2010) encontraron también efectos tempranos de negación utilizando una tarea diferente. En este caso, se presentaron dos tipos de enunciados negativos: hendidos (por ej., Fue Jane quien no cocinó espagueti) y no hendidos (por ej., Jane no cocinó espagueti) y 250 ms después, aparecía una imagen que correspondía (por ej., Un plato de espagueti sin hacer) o no (por ej., Un plato de espagueti hecho) con el enunciado. La 34 Procesamiento y Representación de la Negación tarea de los sujetos era indicar si el objeto de la imagen había sido nombrado. Los resultados mostraron interacción entre el tipo de enunciado y la correspondencia (o no) entre enunciado e imagen. Es decir, en el caso de los enunciados no hendidos, las respuestas eran más rápidas cuando la imagen no correspondía que cuando correspondía, lo que confirma el doble procesamiento de la negación. En cambio, para los enunciados hendidos los resultados fueron opuestos, lo que indicaría que a los 250 ms, la negación había sido comprendida. Una vez más, estos resultados indicarían que la comprensión de la negación es temprana bajo ciertas situaciones, y que posiblemente el doble procesamiento o bien se realiza rápidamente o bien no es necesario. Algunos autores han señalado que estos efectos tempranos de la negación se deben a la influencia de la pragmática (por ej., Dale & Duran, 2011; Nieuwland & Kuperberg, 2008; Tian y cols., 2010). Nosotros estamos de acuerdo con esta postura, es decir, el papel de la pragmática es facilitar la comprensión de la negación al igual que facilita el procesamiento en general. Ahora bien, ésta no puede explicar por qué los enunciados negativos son más difíciles que los afirmativos cuando se presentan sin contexto, en todo caso la pragmática podría explicar la facilitación de la negación comparado con la afirmación en contextos favorables. Nuestra postura es sin embargo que tal dificultad, la que es indicada por las diferencias entre afirmativas y negativas, se debe a factores semánticos (Miller, 1962). No hay ningún caso en el cual un enunciado afirmativo (por ej., Es rojo) y su negativo (por ej., No es rojo) sean semánticamente idénticos, ya que el operador negativo aumenta el significado del argumento sobre el que se aplica. Ahora bien, la frecuencia de uso o un contexto que anticipe el concepto negado podría facilitarla, como veremos en el siguiente apartado, y 35 Procesamiento y Representación de la Negación hacerla más parecida a su homóloga afirmativa en lo que a coste de procesamiento se refiere. 2.2. El tamaño del conjunto complementario Es frecuente encontrar expresiones en el lenguaje que requieren de la realización de inferencias para su comprensión (Johnson-Laird, 1983). Por ejemplo, en el caso de las anáforas, tal como ‘María se va a ir de viaje y ella se está preparando la maleta’ se tendría que inferir que ella se refiere a María para su comprensión. Dicha inferencia puede complicarse si nos dicen que ‘María se va a ir de viaje con Amalia y ella se está preparando la maleta’ En este caso sería difícil saber si ella se refiere a María o a Amalia. Esta ambigüedad es frecuente en los enunciados negativos cuando el conjunto complementario se refiere a varias posibilidades (Condición Múltiple). Por ejemplo, para comprender el enunciado negativo ‘La puerta no es azul’ los individuos podrían inferir que el conjunto complementario de azul podría hacer referencia a marrón, verde, roja, amarilla, etc. Listar este conjunto de posibilidades llevaría mucho tiempo para su procesamiento, además de suponer una sobrecarga para la memoria. Debido a que la memoria de trabajo es limitada, resulta predecible que los individuos tiendan a representar el menor número posible de posibilidades, para lo que en determinadas situaciones pueden verse ayudados por el contexto, el cuál puede ayudar a desambiguar y por tanto conducir a una reducción en el número de posibilidades (Khemlani y cols., 2012). No obstante, otra forma, quizás más eficiente, de reducir el número de posibilidades sería que los individuos no realicen tal inferencia y mantengan el operador 36 Procesamiento y Representación de la Negación negativo en memoria, es decir, en estos casos los individuos comprenderían, por ejemplo, ‘La puerta no es azul’ manteniendo el argumento negado, y de esta manera reducirían tiempo y memoria para su procesamiento (Giora, 2006). Es más, en este caso seguiría resultando más difícil la negación que su correspondiente afirmativa porque aunque no requiera de una inferencia (o doble procesamiento) es, como ya hemos dicho, semánticamente más compleja. De la misma manera, las personas podrían comprender el siguiente enunciado: ‘Ella se está preparando la maleta’ sin necesidad de saber quién es ella. Estos ejemplos nos llevan a concluir que hay diferentes grados de comprensión de un enunciado negativo, más o menos específico, dependiendo de la accesibilidad de su complementario. Así, en el caso de la negación es posible comprender ‘La puerta no es azul’ como la puerta no es azul o la puerta es amarilla (u otro color), dependiendo de la información disponible. Esta propiedad o característica no está presente para las afirmativas, las cuales suelen tener un significado más específico. Glenberg y colaboradores (1999) encontraron que los enunciados afirmativos y negativos (por ej., El abrigo (no) era negro) podrían ser procesados igual cuando previamente presentaban una situación plausible, tal como, ‘Ella no estaba segura si el mejor abrigo podría ser claro u oscuro’. Mientras que, por el contrario, los enunciados negativos eran más difíciles que los afirmativos después de leer un contexto no plausible ‘Ella no estaba segura de qué tipo de material sería el abrigo’. Los autores consideraron que un contexto plausible era aquel que presentaba la misma temática que la frase crítica, en este caso el color. No obstante, dicho contexto fue presentado en este experimento de forma binaria (claro u oscuro), y aunque los autores no atienden a este hecho explícitamente, la realidad es que este contexto binario reduce el número de 37 Procesamiento y Representación de la Negación posibilidades, y convierte a la negación en específica, al igual que la afirmación. Por tanto, podríamos concluir que la comprensión de la negación se ve beneficiada cuando los individuos pueden convertir la negación en una afirmación. Probablemente fue Wason (1961) el primer psicólogo en descubrir que los individuos convertían los enunciados negativos binarios, esto es, aquellos que tienen una alternativa accesible, por ejemplo ‘El 9 no es par’ era sustituido por ‘El 9 es impar’ antes de decidir si un enunciado era verdadero o falso. Watson (1979) presentó imágenes con alguna diferencia entre ellas a niños de entre 7 y 9 años, y encontró que los infantes utilizaban la negación (66%) para expresar ausencia o pérdida de una imagen con respecto a otra, por ejemplo, ‘El caballo no tiene cola’ o ‘El caballo no tiene las riendas’; mientras que ellos preferían utilizar la alternativa afirmativa cuando había una o varias posibilidades, por ejemplo, el tamaño (por ej., ‘Es grande’ frente a la otra imagen en la que aparecía un objeto pequeño). Este mismo resultado fue encontrado en Beltrán, Orenes, y Santamaría (2008), donde también se presenta una tarea de producción pero en este caso con adultos. Además se encontró que los participantes preferían utilizar más la negación para la condición múltiple, la cual no tenía un conjunto complementario específico, que para la binaria. Estos estudios parecen demostrar que los individuos a veces, sobre todo cuando el conjunto complementario es específico (condición binaria), tienden a inferirlo, lo que confirmaría el doble procesamiento para la comprensión de la negación; sin embargo, cuando el conjunto complementario no es específico (condición múltiple), es posible que los individuos no hagan ninguna inferencia y mantengan la negación a menos que el contexto indique de alguna u otra manera cuál es la situación verdadera específica. 38 Procesamiento y Representación de la Negación 2.3. La tarea de verificación: Integración de enunciados e imágenes Una de las tareas más frecuentemente usadas para el estudio de la negación ha sido la verificación frase-imagen. En ella, los participantes tienen que indicar si un enunciado (afirmativo o negativo) es verdadero o falso con respecto a una imagen. Clark y Chase (1972) entendían que para realizar con éxito esta tarea, los participantes deberían comparar el enunciado con la imagen en el mismo formato de representación. En el caso de que ambas representaciones coincidiesen, la respuesta se consideraba verdadera; mientras que cuando ambas representaciones no son similares, entonces la respuesta sería falsa. Uno de los resultados más consistentes que se ha obtenido con la tarea de verificación ha sido la interacción entre el valor de verdad y la polaridad (ver Carpenter & Just, 1975; Clark & Chase, 1972; Trabasso, Rollins, & Shaughnessy, 1971; Wason & Johnson-Laird, 1972; Wason & Jones, 1963). Esto es, los enunciados afirmativos verdaderos son más fáciles que los enunciados afirmativos falsos ya que mientras que hay una correspondencia o matching entre los enunciados afirmativos verdaderos (por ej., A está encima de B) y la imagen (por ej., A/B); no hay tal correspondencia (o se produce un mismatching) entre los enunciados afirmativos falsos (por ej., B está encima de A) y la imagen (por ej., A/B). El caso opuesto ocurre para los enunciados negativos, es decir, los enunciados negativos falsos son más fáciles que los negativos verdaderos por la correspondencia frente a la no correspondencia o mismatching entre el enunciado y la imagen. Lüdtke y colaboradores (2008) presentaron una tarea de verificación usando potenciales evocados. Los autores presentaron enunciados afirmativos y negativos (por ej., En frente de la casa (no) hay un gato) y después aparecía una imagen que 39 Procesamiento y Representación de la Negación correspondía o no con el enunciado. Ellos manipularon el tiempo de presentación entre los enunciados y la imagen. En la mitad de los ensayos la imagen aparecía a los 250 ms y en la otra mitad a los 1500 ms. Los resultados mostraron la clásica interacción entre el valor de verdad y la polaridad cuando la imagen aparecía 250 ms después del enunciado. Estos datos indicarían correspondencia o matching entre el enunciado y la imagen, que sería más fácil que la no correspondencia. Sin embargo, cuando analizaron la imagen presentada 1500 ms después del enunciado entonces aparecían efectos de negación, por lo que los autores concluyeron que la negación podría ser comprendida en periodos tardíos. Una pregunta que surge es si los efectos encontrados en este estudio están relacionados con la propia comprensión o más bien con la verificación. Es decir, una vez que los individuos han comprendido el enunciado, al ver la imagen, tienen que contrastar o comparar el enunciado con la imagen y decidir si el enunciado es verdadero o falso con respecto a la imagen, y este proceso requiere de tiempo, especialmente para la negación ya que no hay ninguna imagen que corresponda directamente con un enunciado negativo. El caso que podría resultar más sencillo es cuando el enunciado negativo es falso, ya que si después de leer ‘Enfrente de la casa no hay un gato’ aparece un dibujo con un gato enfrente de una casa es relativamente sencillo identificar que es falso. Ahora bien, cuando leían el enunciado negativo ‘Enfrente de la casa no hay un gato’ y aparecía un gato arriba de la casa, en principio, tendrían que indicar que el enunciado es verdadero, pero es igual de verdadero que si aparece abajo, detrás o dentro de la casa. Ante tantas posibilidades verdaderas, los participantes en la tarea debieron aprender que el gato podría aparecer enfrente o arriba de la casa (condición binaria), por tanto una estrategia que podría facilitar la tarea sería convertir el modelo negativo (por ej., Enfrente de la casa no hay un gato) en un modelo alternativo afirmativo (por ej., 40 Procesamiento y Representación de la Negación Arriba de la casa hay un gato) que pudiera comparar con la imagen directamente, y esta inferencia podría requerir de tiempo, según los autores hasta 1500 ms. La pregunta teórica clave es si dicha inferencia que parece útil para la tarea de verificación es necesaria para la comprensión del enunciado. Para responder a esta pregunta se va a presentar primeramente una tarea neutra como puede ser el reconocimiento más que una de verificación, la cual podría facilitar, como hemos indicado previamente, la conversión de un modelo negativo a uno afirmativo. Fueron Young y Chase (1971) los que anticiparon que ciertas tareas o instrucciones podrían facilitar la conversión de los enunciados negativos a afirmativos y además mostraron que la conversión requiere de tiempo, lo que podría confirmar que tal proceso se realice entre los 750 y 1500 ms. En nuestra opinión, y atendiendo a los estudios analizados, todas aquellas tareas en las que se requiere integrar enunciados negativos con una imagen van a facilitar su conversión para mejorar la comparación. Esto es lo que ocurriría en la tarea de verificación frase-imagen; pero también en tareas que sin requerirlo en algún nivel se integran ambas, y esto podría ocurrir en la tarea de naming de Kaup y colaboradores (2006) cuya tarea solamente era nombrar el objeto de la imagen, pero tal tarea se hacía más rápida cuando el objeto de la imagen correspondía con el enunciado que se había presentado previamente. En conclusión, la hipótesis de que la comprensión de la negación requiere de doble procesamiento y efectos tardíos (750-1500 ms) parece tambalearse; primero porque se han encontrado efectos tempranos (250-400 ms) y segundo porque los efectos tardíos encontrados podrían deberse a la inferencia o conversión desde el modelo negativo al modelo alternativo afirmativo, inferencia que pudiera no ser necesaria para la comprensión de la negación. Nuestra hipótesis y la que vamos a tratar de defender en 41 Procesamiento y Representación de la Negación la primera parte de la tesis es que el doble procesamiento depende de la información que hay disponible, concretamente del tamaño del conjunto complementario o número de alternativas, o incluso una tarea como la verificación podría favorecer la disponibilidad. En líneas generales, podríamos suponer que nuestro sistema cognitivo sustituye el modelo negativo por otro modelo alternativo afirmativo siempre que sea posible como ocurre en aquellos casos donde hay una alternativa o contrario claro, accesible (condición binaria). Por tanto para estos casos, es predecible que se facilite el doble procesamiento. Sin embargo, cuando esto no es posible, el sistema cognitivo podría mantener el modelo negativo más que sustituirlo por varios modelos alternativos afirmativos (condición múltiple), los cuales podrían suponer una sobrecarga para la memoria (Johnson-Laird, 2001). 3. Representación del operador negativo Uno de los axiomas más importantes en Ciencia Cognitiva es que las personas son capaces de representar el conocimiento que adquieren del mundo que les rodea. Una representación, en términos de Peirce (ver Peirce 1931-1958), y probablemente la concepción más aceptada por la mayoría de los psicólogos cognitivos, es una triada formada por un signo, un intérprete y un objeto, es decir, para que una representación funcione se necesita un sujeto (intérprete) que interpreta una realidad (objeto) a través de un signo. La relación entre el objeto y el signo que lo representa puede ser a través de un icono o un símbolo. Un icono es una representación en la que el signo mantiene las mismas propiedades y relaciones que el objeto que representa. El ejemplo más sencillo sería una imagen (p.e. un círculo rojo). El símbolo es una representación abstracta de 42 Procesamiento y Representación de la Negación cualquier nombre, adjetivo o verbo, y se caracteriza por no mantener las mismas propiedades y relaciones que el objeto que representa. Por ejemplo, el lenguaje en sí (ver Von Eckardt, 1993 para una completa revisión del concepto). Hay varias teorías generales que han planteado cómo podrían ser las representaciones mentales. La teoría corpórea defiende que son icónicas, es decir, son analógicas al objeto que representa (Barsalou, 1999; 2005; Zwaan & Radvansky, 1998). Una propuesta completamente opuesta sería la teoría proposicional, la cual supone que las representaciones son simbólicas, es decir, no tienen las mismas propiedades que el objeto que representan, sino que se componen de proposiciones (Carpenter & Just, 1975; Clark & Chase, 1972). La teoría proposicional, sin embargo, no niega la existencia de las imágenes, sino el hecho de que éstas sirvan para explicar el funcionamiento de nuestro sistema cognitivo, es decir, de alguna manera las considera epifenoménicas. De la misma manera, la teoría corpórea defiende que el lenguaje en sí, o cualquier abstracción derivada de éste, no es la base del significado lingüístico. Un modelo híbrido sería la teoría de modelos mentales. Su principal promotor, Johnson-Laird (1983), entiende que la información puede ser representada en diferentes formatos: el modelo mental, la proposición, y la imagen, dejando fuera las reglas lógicas. Una imagen representa un objeto, mientras que una proposición sería la descripción del objeto o diferentes descripciones de éste; la imagen guarda una relación analógica con el objeto que representa mientras que este no es el caso para la proposición; una imagen puede ser rotada y transformada haciéndola corresponder con el objeto mientras que una proposición puede ser verdadera o falsa con respecto al objeto. La imagen sería un tipo específico de modelo mental. Un modelo mental es una representación icónica o analógica al objeto que representa, y según esta teoría sería el 43 Procesamiento y Representación de la Negación modo con el cual las personas comprenden y razonan sobre el mundo mientras que las proposiciones sólo permitirían una comprensión superficial (vocabulario y gramática). Además, desde este enfoque se admite que los modelos mentales pueden combinarse con representaciones simbólicas. Este sería por ejemplo el caso de la negación como estudiaremos más adelante (para una revisión, Khemlani y cols., 2012). En resumen, podemos considerar la teoría de modelos mentales un enfoque flexible sobre la comprensión. La diferencia entre la teoría de modelos mentales y la proposicional es que mientras que la primera defiende que las personas para comprender construyen un modelo mental, la segunda defiende que las personas representan proposiciones. Para la teoría de modelos mentales, las proposiciones serían descripciones de los modelos mentales. Tanto la teoría de modelos mentales como la teoría corpórea defienden la existencia de los modelos mentales (o de situación, Zwaan & Radvansky, 1998); la diferencia es que mientras que la teoría corpórea entiende que la comprensión se basa exclusivamente en representaciones icónicas, la teoría de modelos mentales acepta la existencia de otros formatos de representación, y su combinación. Según la teoría proposicional la negación se representaría de forma lingüística y encapsularía el argumento sobre el que se aplica. Por ejemplo, ‘El círculo no está a la derecha del triángulo’ se representaría algo como: No(a-la-derecha-de (círculo-triángulo)). Según la teoría corpórea el operador negativo no puede ser representado explícitamente, por lo que las personas construirían un modelo alternativo afirmativo, por ejemplo, podría representar la frase anterior a través de un modelo con un círculo a la izquierda del triángulo. La teoría de modelos mentales defiende que la negación 44 Procesamiento y Representación de la Negación puede ser representada de forma icónica, pero esto sería una de las formas en que se podría representar. Es decir, el individuo podría representar el círculo a la izquierda o arriba o abajo o detrás o delante, etc., pero cada una de las representaciones no podría captar el significado de la negación en sí, por lo que también defiende una representación simbólica. Es decir, las personas podrían representar un modelo mental de un círculo a la derecha de un triángulo precedido por un símbolo que representara la negación. ¬ Durante un tiempo se consideró que los seres humanos éramos seres racionales que procesábamos y representábamos proposiciones. Fue a partir de los años 70 cuando los psicólogos experimentales empezaron a mostrar que la comprensión implicaba la construcción de modelos mentales, y no solamente la formación de representaciones proposicionales (Johnson-Laird, 1983). Los primeros indicios que hicieron cambiar de perspectiva fueron estudios sobre memoria (Barclay, 1973; Bransford, Barclay, & Franks, 1972; De Soto, London, & Handel, 1965). En estos estudios se presentaban historias para recodar y encontraron que más que retener la información superficial lingüística, se retenía información semántica profunda a través de modelos mentales, y después describían esos modelos, a veces, con palabras diferentes a las que se habían utilizado en las historias. Esta evidencia sugiere que las personas construyen modelos mentales tal como la teoría de modelos mentales y la teoría corpórea defiende. La diferencia entre ambas teorías sería que mientras que la teoría de modelos mentales defiende otros formatos de representación, este no es el caso para la teoría corpórea. Consecuentemente, uno de los objetivos que nos proponemos en la presente tesis es testar ambas teorías. Este debate puede resultar novedoso ya que la mayoría de la 45 Procesamiento y Representación de la Negación literatura sobre ambas teorías ha tendido a contrastarlas de manera casi exclusiva con la teoría proposicional (simbólica). 4. Estudiando la negación a través de los movimientos oculares En los estudios que presentaremos más adelante el procesamiento y la representación de la negación es estudiado a través de medidas conductuales como los errores que se cometen en la realización de una tarea y el tiempo que se tarda en desempeñarla. Estas dos medidas están relacionadas con la complejidad del proceso cognitivo que se está midiendo, tal que a mayor dificultad en el procesamiento de una tarea determinada le suelen seguir un mayor número de errores y tiempo de respuesta más largos. Al ser estas dos medidas ampliamente estudiadas en ciencia cognitiva (Donders, 1969), en este apartado describiremos otra medida conductual que es menos conocida, la fijación de la mirada al objeto de interés. Esta medida además de informar sobre la complejidad del proceso cognitivo en general, registra información espacial y temporal del objeto que está siendo atendiendo visualmente en cada momento y esta información está directamente relacionada con el procesamiento cognitivo. La relevancia de la fijación de la mirada para el estudio de la cognición fue promovida por Just y Carpenter (1976) bajo lo que se conoce como la hipótesis fuerte de la relación ojo-mente. Según ésta, existe una relación directa e inmediata entre fijación y procesamiento cognitivo, tal que lo que está siendo fijado, por ejemplo una palabra, es instantáneamente procesada, y lo es además durante tanto tiempo como dure la propia fijación. Los trabajos que aquí presentamos comparten lo esencial de esta hipótesis, aunque somos también plenamente conscientes de que la información que no es fijada visualmente también puede ser 46 Procesamiento y Representación de la Negación procesada, como bien indica la amplia literatura sobre atención encubierta (por ej., Posner, 1980). En general, los ojos son los órganos a través de los cuales se conoce el mundo que nos rodea. Para llevar a cabo esta función, los ojos realizan principalmente dos tipos de movimientos: unos, los movimientos de la fijación, se dirigen a compensar el movimiento de la cabeza y los objetos del campo visual, con el objetivo de mantener fija una imagen visual en la fóvea. Los otros, los movimientos sacádicos, que permiten el barrido del campo visual y, por tanto, la captación de los objetos de interés sobre los que la mirada queda finalmente fijada. Los movimientos sacádicos son movimientos rápidos cuya función principal es detectar el objeto de interés y mantenerlo en la fóvea. Una vez que el objeto es detectado, el ojo se mantiene relativamente estable para atender y procesar dicho objeto. A este periodo estable del ojo se denomina fijación y se relaciona con varios procesos cognitivos como la atención, el lenguaje, y la memoria entre otros (Allopenna, Magnuson, & Tanenhaus, 1998; Altmann & Kamide, 2004; Duñabeitia y cols., 2009; Hayhoe & Ballard, 2005; Lawrence, Myerson, & Abrams, 2004; Liversedge & Findlay, 2000; Rayner, 1978; 1998; Tanenhaus y cols., 1995). En nuestro caso, las fijaciones de los ojos se van a medir a través de un aparato llamado SR Research EyeLink II, que consta de un ordenador donde se presenta el experimento, un ordenador que registra el movimiento de los ojos, y un casco que consta de tres cámaras, una para cada ojo más otra que registra la pantalla (Figura 1). Antes de empezar cualquier experimento hay que calibrar y validar la medida. Para ello aparece en la pantalla nueve puntos uno detrás de otro y el participante tiene que mirar al punto. Lo que se registra entonces es la desviación o no entre donde aparece el punto y donde miran los ojos, su coincidencia indica que el registro es válido. Además de esta 47 Procesamiento y Representación de la Negación medida obtenida en la calibración, la velocidad y la aceleración media es la que se tiene en cuenta para saber lo que pasa durante el experimento. Por ejemplo, una velocidad y una aceleración del ojo que sea superior a la calibración es considerada un movimiento sacádico, mientras que el no registro del movimiento de los ojos se considera un parpadeo. Cuando no es un movimiento sacádico ni un parpadeo se considera una fijación. Figura 1. Laboratorio de movimientos oculares En los experimentos que se presentan más adelante utilizamos el paradigma de ‘mundo visual’ (del inglés ‘Visual World Paradigm’) para el estudio del impacto del procesamiento de la negación sobre la exploración visual del medio. Este paradigma consiste en la presentación oral de una palabra o enunciado mientras que los sujetos 48 Procesamiento y Representación de la Negación están viendo en la pantalla elementos visuales que pueden estar o no relacionados con la información presentada de forma oral (para una revisión: Huettig, Rommers, & Meyer, 2011). La suposición básica, en línea con la hipótesis ojo-mente, es que las personas exploran la pantalla en consonancia con la información que les ha sido presentada verbalmente, y que este proceso de relación entre interpretación y acción (fijación de la mirada) ocurre de manera automática (Cooper, 1974; Tanenhaus y col., 1995). El uso del paradigma de ‘mundo visual’ ha producido varios hallazgos interesantes en relación a los mecanismos lingüísticos. Por ejemplo, Altmann y Kamide (1999), en un estudio dirigido al análisis de los mecanismos anticipatorios del lenguaje, mostraron que sus participantes, cuando escuchaban la frase ‘el niño se comerá un pastel’, fijaban la mirada hacia el pastel que aparecía en una escena visual ya en el momento de escuchar el verbo comer, antes incluso de escuchar la palabra ‘pastel’. A nivel léxico, se ha encontrado que las fijaciones pueden dirigirse hacia objetos visuales relacionados con las palabras oídas. Así, por ejemplo, Duñabeitia y colaboradores (2009) encontraron que las personas miraban a una nariz cuando escuchaban la palabra ‘oler’. En relación con el objetivo de la presente tesis, estos hallazgos sugieren que este paradigma puede ser útil en el estudio de la negación, puesto que permiten predecir que los enunciados afirmativos y negativos quizás conducen a fijaciones sobre elementos diferentes de una escena visual, o también a patrones de fijación distintos sobre un mismo elemento. Por ejemplo, ante la presentación oral de la frase ‘La figura es roja’, es de esperar que las fijaciones se dirijan hacia una figura roja, manteniéndose en ella tanto tiempo como dure el proceso de comprensión. Mientras que para su versión negada (‘La figura no es roja’) se puede predecir que primero miren a la figura roja y después cambien a otros elementos o figuras de diferente color al rojo. En este sentido, uno de 49 Procesamiento y Representación de la Negación los objetivos de la tesis es explorar si para la negación el cambio de atención, tal y como éste es indicado por la fijación de la mirada, automática y necesariamente va desde el elemento visual que representa el argumento u objeto negado hacia un elemento diferente o alternativo. La importancia de profundizar en el procesamiento de la negación estriba en su conexión con las teorías de la representación, como se ha comentado previamente. Antes de terminar este apartado, es necesario abordar, dada su controversia, el tema del análisis de las fijaciones en el contexto del paradigma de ‘mundo visual’. La mayoría de los estudios utilizan pruebas paramétricas como el análisis de varianza (ANOVA), sin embargo, algunos autores defienden que éste no es un método adecuado ya que los movimientos oculares, y particularmente las fijaciones, no cumplen con los supuestos de estas pruebas estadísticas (ver: Barr, 2008; Jaeger, 2008). Una hipótesis frecuente en psicología es aquella que utiliza una variable independiente categórica (por ejemplo, la polaridad de un enunciado: afirmativo o negativo) y una variable dependiente continua, tal como el tiempo de reacción o el número de errores. Sin embargo, cuando se utiliza la técnica de movimientos oculares estamos tratando con una variable independiente continua (el tiempo) y una variable dependiente categórica (la región de interés sobre la que se produce la fijación). Lo que generalmente se suele hacer para usar el análisis de varianza es ‘discretizar’ la variable continua tiempo en intervalos y así tenemos intervalo 1, intervalo 2, etc. Esta solución puede originar conclusiones erróneas relacionadas con efectos anticipatorios o de ratio (ver Barr, 2008). Para evitar este problema se ha recomendado, entre otras alternativas, el uso de modelos de regresión para analizar el cambio de la variable dependiente en función del tiempo. Adoptando esta posible solución queda pendiente todavía el 50 Procesamiento y Representación de la Negación problema de que la variable dependiente es categórica y como tal sigue una distribución multinominal más que normal. Este problema es compartido por el análisis de varianza (ANOVA), para cuyo uso es necesario que la variable dependiente cumpla unos supuestos, que son: 1. La variable dependiente debe medirse al menos a nivel de intervalo 2. Independencia de observaciones 3. La distribución de los residuales debe ser normal 4. Homogeneidad de varianza. Para solucionar el problema de que la variable dependiente es categórica (formada por cada una de las regiones en la se divide el espacio de la pantalla) se utiliza la probabilidad de que el sujeto mire a dicha región. Pero utilizar la probabilidad plantea otro problema ya que los valores que puede tomar están limitados, entre 0 y 1, y por tanto el error de la varianza es proporcional a la media. En cambio, el ANOVA supone que el rango de valores de la variable dependiente es ilimitado y que los errores se distribuyen de forma normal e independiente de la media. Para solucionar este nuevo problema se transforma las proporciones en odds y se obtiene su logaritmo (y=ln(p/1-p), donde ln es el logaritmo neperiano, p es la proporción de fijaciones hacia el objeto de interés e y es el valor obtenido de la transformación de p, Jaeger, 2008). Llegados a este punto, seguimos teniendo otro problema ya que las observaciones son dependientes. Es decir, la probabilidad de que se mire a la posición de interés es dependiente de que no se mire al resto de posiciones, y todas ellas suman 1. Para solucionar este problema, los modelos de efectos mixtos (o multiniveles) son adecuados ya que modelan la no dependencia. Por tanto, Barr (2008) propone que el mejor modelo para analizar los movimientos oculares es el modelo de regresión 51 Procesamiento y Representación de la Negación logística de efectos mixtos (o multinivel), ya que los modelos de regresión son adecuados para acomodar la variable continua del tiempo, los modelos logísticos son necesarios para acomodar una variable dependiente categórica, y los modelos de efectos mixtos atienden a la no dependencia de observaciones. Otros análisis también han sido propuestos. Por ejemplo, Scheepers (2003) utilizó modelos lineales logísticos para analizar los movimientos oculares. Estos modelos son óptimos para el análisis de variables categóricas puesto que no asumen normalidad u homogeneidad de varianza. Scheepers, Keller y Lapata (2008) utilizaron el modelo logístico ‘logistic power peak’ (LPP) en otro estudio. Heller, Grodner, y Tanenhaus (2008) calcularon la proporción de fijaciones en el target con respecto a la suma de la proporción de fijaciones en el target y los distractores. Para evitar problemas relacionados con la proporción de los datos, las medias de los participantes y los ítems fueron transformadas utilizando modelos quasi-logit antes del análisis de varianza (ANOVA). Ferguson, Scheepers, y Sanford (2010) obtuvieron la probabilidad de fijación en el target usando su logaritmo y luego analizaron los datos utilizando análisis de clúster y finalmente ANOVA. Mirman, Dixon, y Magnuson (2008) también propusieron utilizar un análisis de curvas de crecimientos para no sólo analizar el tiempo correctamente sino también tener en cuenta las diferencias individuales de los sujetos. Esta diversidad de modelos estadísticos utilizados para el análisis de los movimientos oculares podría originar la imposibilidad de comparar unos estudios con otros, por lo que es necesario que los investigadores se pongan de acuerdo sobre cuál es el mejor modelo para analizar el curso temporal de las fijaciones. Aunque, también es cierto que el uso de un modelo u otro debe depender de la hipótesis del investigador y 52 Procesamiento y Representación de la Negación los datos. Por ejemplo, uno de los objetivos de la tesis es explorar si la conversión de un modelo negativo a su alternativo afirmativo es necesaria, es decir, tenemos dos objetos de interés: el que corresponde al modelo negativo y a su alternativo afirmativo, en algunos casos incluso todos los elementos son de interés porque había varios alternativos posibles. Este tipo de exploración no puede hacerse con el mejor estadístico para el análisis de los movimientos oculares propuesto, el modelo de regresión logístico de efectos mixtos, porque éste no explora el momento en el que cambia una distribución de un objeto a otro, más bien indica el incremento o disminución de un objeto de interés o las diferencias de ese objeto entre las condiciones. Por otro lado, el análisis de funciones no lineales tampoco se puede utilizar para explorar nuestra hipótesis porque, como veremos, algunas de las manipulaciones que hemos introducido conducen a distribuciones de fijaciones diferentes para las condiciones que se pretenden comparar, y por tanto requeriría el uso de funciones diferentes para cada una de ellas. Además estos análisis indican si hay diferencias en la proporción de fijaciones, pero no indican el momento temporal, y nosotros estamos muy interesados en el curso temporal de la negación. Por tanto, aunque parece que hay un modelo estadístico adecuado para analizar los movimientos oculares, es importante tener en cuenta que los datos y la hipótesis son los que realmente deben determinar qué método estadístico es el adecuado y eso hemos tratado de hacer en los trabajos con movimientos oculares que presentaremos a continuación. 53 Procesamiento y Representación de la Negación 5. Objetivos El principal objetivo de la presente tesis se dirige a deshojar una de las palabras esenciales del lenguaje, la negación, para estudiar su complejidad en la comprensión. Son muchos los estudios que han analizado diferentes factores para dar respuesta a esta cuestión, sin embargo, al día de hoy sigue siendo una cuestión irresoluble. Para avanzar es su conocimiento, nosotros vamos a atender al concepto propio de la negación para inferir las primeras conclusiones que se derivan de él y que son las que vamos a testar en la presente tesis. Desde Aristóteles, como hemos dicho previamente, la negación se define como un elemento sintáctico que coge un argumento (por ej., rojo) y revierte su valor de verdad (por ej., no rojo) de falso a verdadero, o viceversa. Este concepto, revertir el valor de verdad, es simbólico. Es decir, no hay nada en el mundo real (externo) que corresponda con dicha función. El primer objetivo que nos planteamos entonces es demostrar que las personas pueden comprender la negación (por ej., La figura no es roja) de forma simbólica, es decir, representando exclusivamente el modelo negado (por ej., no figura roja). Ahora bien, cuando la alternativa sea accesible, los individuos tenderán a convertir la negación en su alternativa afirmativa (por ej., una figura verde) puesto que como hemos comentado previamente, es más fácil recordar información en afirmativo que negativo. Estas dos formas de comprender la negación estarían relacionadas con la representación de la negación, en el primer caso, la representación sería simbólica ya que la negación está representada explícitamente y la única forma de representarla es simbólica; y en el segundo, icónica, puesto que la negación no se representaría explícitamente. Como consecuencia este objetivo nos 54 Procesamiento y Representación de la Negación permitiría discriminar entre teorías representacionales: la teoría de los modelos mentales y la teoría corpórea. De la definición de negación también se desprende que la negación coge un argumento verdadero o falso y lo convierte en falso o verdadero. Esto significa que la negación se aplica a una proposición, la cual puede ser verdadera o falsa, sin embargo este no es el caso de la imagen. Tal y como hemos indicado previamente, una imagen puede ser rotada y transformada haciéndola corresponder con el objeto mientras que solo una proposición puede ser verdadera o falsa con respecto al objeto. Con lo cual, la propia definición de la negación nos lleva a pensar que la negación no se aplica directamente sobre una imagen (o representación visual), principal ventaja de las palabras concretas frente a las abstractas, y esto nos lleva a hipotetizar que argumentos simbólicos (o no visuales) pudieran facilitar la comprensión de la negación frente a argumentos icónicos (o visuales). De corroborar esta hipótesis, la idea de que la negación sea simbólica se vería fortalecía. Con estos dos trabajos habríamos descartado que la dificultad de la negación se deba a su doble procesamiento o su representación simbólica como algunos estudios han propuesto ya que habríamos encontrado que la negación se puede comprender en un solo paso representando la situación negada y además habríamos probado que la representación simbólica aunque suele ser más difícil que la icónica, no sería el caso para la negación, que podría incluso facilitarla. El tercer y último objetivo de la tesis es demostrar que la dificultad de la negación con respecto a la afirmación es semántica (Miller, 1962), es decir, la negación añade información semántica con respecto a su correspondiente afirmativa. La negación coge un argumento afirmativo y revierte su valor de verdad de falso a verdadero o viceversa. Para demostrar esta hipótesis 55 Procesamiento y Representación de la Negación utilizaremos la pragmática presentando ambos tipos de enunciados en varios contextos. Si nuestra hipótesis es cierta, cuando la afirmación se presente en un contexto para falsar la información previa, principal uso de la negación, entonces ambos tipos de enunciados serán procesados en el mismo momento temporal y resultarán iguales de fáciles o difíciles. Fuera de este contexto, la negación se mostrará más difícil que la afirmación porque su principal uso es falsar o contradecir la información previa, y ésta función es más compleja que la afirmación, que generalmente presenta información consistente con la previa conversación. 56 How negation is understood 6. How negation is understood: Evidence from the visual world paradigm 6.1. Abstract This chapter explores for the first time how negation is understood using the visual world paradigm. Two-step theory assumes that the comprehension of negative sentences, such as the figure was not red, necessarily involves the simulation of the actual situation (e.g., a green figure), while mental model theory claims that individuals might retain a symbolic representation of the negated situation (e.g., a non-red figure) when an analogical representation is awkward or unfeasible. Our hypothesis is that people will switch to the actual situation whenever possible, but will be able to maintain a symbolic representation of the negated situation when needed. To test this, we studied the role of context and task in the availability of the actual situation in three visual world paradigm experiments. Experiment 1 presented either a specific context (binary: the figure could be red or green) or an unspecified context (multary: the figure could be red or green or yellow or blue). Then, affirmative and negative sentences (e.g., the figure is (not) red) were heard while four figures were shown on the screen and eye movements were monitored. We found that people shifted their visual attention toward the actual situation in the binary context, but focused on the negated situation in the multary context. This latter result is corroborated in Experiments 2 and 3, in which negative sentences were presented without any context using two different tasks: recognition and verification. Our findings suggest that individuals are able to represent negation symbolically, but readily switch to the actual situation when the context permits. 57 How negation is understood 6.2. Introduction There is growing evidence in the literature that language comprehension makes use of analogical representations in the form of models (Barsalou, 2005; Glenberg, Meyer, & Lindem, 1987; Johnson-Laird, 1983; Zwaan & Radvansky, 1998). However, there are at least two different views on how these representations are used. On the one hand, embodiment theory holds that all representations, including those relating to abstract concepts, must be iconic to the world (Barsalou, 1999; 2005; 2012; Glenberg et al., 1987; Lakoff & Johnson, 1980; Zwaan & Radvansky, 1998). According to this theory, symbolic representations are foreign to the human mind. On the other hand, mental model theory, while holding that most representations are iconic, also allows for the use of combinations with symbolic representations (Johnson-Laird, 1983; 2006). An example of this combination can be seen in the processing of negation (Khemlani, Orenes, & Johnson-Laird, 2012). Both views posit the existence of iconic situation models, but differ in that the mental model theory also claims that symbolic representations exist, while the embodiment theory holds that representations are exclusively iconic. In the present study, we will contrast these two views in three experiments using negation. Our main hypothesis is that people will switch to the actual situation whenever possible, but will be able to maintain a symbolic representation of the negated situation when needed. This represents a new approach to the question: in the past, the two theories in question have been compared with propositional theories using exclusively symbolic representation systems (Clark & Chase, 1972; MacDonald & Just, 1989), but this is the first time they will be compared with each other. 58 How negation is understood Negation is an operator that takes a single argument and reverses its truth value. In other words, if A is an argument that is false (e.g., Jefferson was the president of the United States), then Not-A is necessarily true (e.g., Jefferson was not the president of the United States), an analysis that goes back to Aristotle (1984). As the meaning of negation has no direct equivalence in experience, it represents a challenge for embodiment theory. Nevertheless, a theory of negation has recently been proposed (the two-step simulation theory or two-step theory, for short) that, in the spirit of the embodiment view, excludes symbolic and linguistic-like representations of any kind (Kaup, Lüdtke, & Zwaan, 2006). According to the two-step theory, negation cannot be represented explicitly, and so the comprehension of a sentence such as the door is not open must be accomplished through two representational steps (Kaup et al., 2006). The first step is represented via a perceptual simulation of the negated situation that corresponds to the affirmative equivalent (e.g., an open door); in a second step occurring 750-1500 ms later, this perceptual simulation is changed to the actual situation implied by the negative (e.g., a closed door). This theory also holds that these two steps are mandatory for comprehension. This requirement that there must be a shift from one simulation to the other would appear to be justified for sentences such as the one used in the above example (the door is not open). For this sentence, a simulation of the actual situation (a closed door) is readily inferred from its binary (or complementary) predicate: in the case of ‘not open’, there are only one alternative, ‘closed’. Negative sentences are not always this informative about the actual situation, however. For instance, the multary (or contrastive) predicate in a sentence such as the door is not blue has many alternatives (‘red’, ‘green’, ‘yellow’, etc.), and so there is no clear actual situation that can be 59 How negation is understood simulated. In fact, it seems impossible to know what the door actually looks like, unless additional information is provided. The two-step theory holds, however, that even in cases of uncertainty such as the one given above, a simulation of the actual situation might take place, although this simulation could be empty (see Barsalou, 1999). For example, the door is not blue could well be simulated by representing a door of an unspecified color (Kaup, Zwaan, & Lüdtke, 2007). However, this case has not yet been shown to occur. Mental model theory agrees with the possibility that people might simulate the actual situation for multary predicates, but holds that they are able to retain the negated situation while applying a symbol that represents negation (Khemlani et al., 2012). Consider the following assertion: the circle is not to the right of the triangle. Mental model theory argues that individuals, in order to understand this situation, will construct an iconic mental model of the corresponding affirmative assertion and then apply the negation symbol. The resulting model might look like this: ¬ In this sense, Giora and colleagues have also proposed the suppression/retention hypothesis, which argues that suppression following negation (i.e. the change toward the actual situation) is not mandatory, but sensitive to discourse goals and requirements (Giora, 2006; Giora et al., 2007; Giora et al., 2009; see also for a similar idea Schul, 2011). Wason (1961) was the first psychologist to discover that individuals replace the negated situation with the actual situation when it is available, such as in the case of binary predicates. Watson (1979) also observed that children did the same for multary predicates, although they retained the negated situation in the case of adjectives that 60 How negation is understood express loss or absence, such as the horse has no tail. Beltrán, Orenes and Santamaría (2008) corroborated those results in adults, and also found that individuals produced more negations for multary predicates than for binary predicates. These studies support the idea that people are able to represent negation through either its actual situation or its negated situation, as mental model theory claims. What they represent in each case will depend on the availability of the actual situation. In the case of binary predicates (e.g., open-closed), the actual situation is readily available, whereas in the case of multary predicates (e.g., blue-red-green-yellow), the correspondence with the actual situation, unless it is obvious from the context, is not straightforward. As those studies used a production task, it is necessary to test whether those results are similar for comprehension. Kaup and collaborators (2006) showed that people changed to the actual situation when people understand negation using binary predicates, the main goal of this study then is to know whether people maintain the negated situation when they understand negation using multary predicates. To test this, we studied the role of context and task in the availability of the actual situation in three visual world paradigm experiments. In a typical visual world task, verbal and visual inputs are presented simultaneously while eye movements are recorded. Experiments have shown that when linguistic input matches visual input, the eyes will begin to move automatically towards the corresponding visual input (Cooper, 1974; Tanenhaus et al., 1995). These results suggest that it is the conceptual overlap between the word and the object that mediates between language and eye movements. These data on semantic relatedness rule out an account based solely on phonological overlap between the unfolding word and the names associated with the objects in the scene (Altmann, 2011). Eye movements are thus presumed to reflect interactions 61 How negation is understood between linguistic and visual representations that occur at the level of conceptual representation (Salverda, Brown, & Tanenhaus, 2011). In keeping with the above assumptions, it follows that when people are exposed to negative assertions in a visual world paradigm, they will look more frequently at what they are using to process and represent the assertion. The advantage of the visual world paradigm for studying negation is that all the objects (whether or not they are related to the comprehension of the sentence) are displayed on the screen, and so people are free to turn their attention to whatever object is required for processing and representation. With respect to the timing of the display of the input, most of the paradigms applied thus far in research on negation have opted for using a discrete delay between the sentences and the pictures. For example, Lüdtke and collaborators (2008) employed a sentence-picture verification paradigm using event-related potentials (ERP) in which participants were presented with affirmative or negative sentences (e.g., in front of the house there was a/no cat) followed by a matching or mismatching picture 250 or 1500ms later. The results showed that when the delay in the picture presentation was short (250 ms), verification latencies as well as the ERPs evoked by the picture showed an interaction between negation and truth value. In contrast, when the delay was long (1500 ms), main effects of truth value and negation were observed, in addition to the interaction effect in the N400 time window. These results suggest that negation is fully integrated into sentence meaning only at a later stage of the comprehension process. As we have seen, the authors interpreted this to mean that the comprehension of negated sentences happens around 1500 ms. However, sentence comprehension should not take that long, and so it might be worthwhile to use another methodology to record what is happening in that lengthy period of time. 62 How negation is understood Using the visual world paradigm, we presented affirmative and negative sentences (e.g., the figure was (not) red) via a loudspeaker while four colored figures (e.g., yellow, blue, green, and red) were shown on the screen. One of the colors represented the negated situation (e.g., the red figure) while the others stood for the actual situation (e.g., the yellow, blue, and green figures). With this design, the participants (and not the researchers) were able to manage the pace of their comprehension, thus allowing us to determine the temporal course of negation. Also, our use of the visual world paradigm allowed us to determine what people were looking at when comprehending negative sentences: did they focus on the negated situation or the actual situation? This approach has clear implications for helping to discriminate between the two theories under comparison. We carried out three experiments in all. The goal of Experiment 1 was to determine how people process negative sentences (e.g., the figure was not red) in binary (e.g., the figure could be red or green) and multary contexts (e.g., the figure could be red, or green, or blue, or yellow). According to mental model theory, the predicted outcome would be that the participants would shift their visual attention from the negated situation (e.g., a red figure) toward the actual situation (e.g., a green figure) in binary contexts (as Kaup et al. (2006) found), while they would keep their visual attention on the negated situation in multary contexts. In Experiment 2, we presented affirmative and negative sentences without any verbal context. This condition corresponds with the multary context in Experiment 1, and so we expected that the findings would be the same. The purpose of this second experiment was to control for any possible influences of context. Many studies have shown that the processing of negation is facilitated when assertions are presented in a suitable context (Dale & 63 How negation is understood Duran, 2011; Givon, 1978; Glenberg et al., 1999; Johnson-Laird & Tridgell, 1972; Nieuwland & Kuperberg, 2008; Strawson, 1952; Tian, Breheny, & Ferguson, 2010; Wason, 1965). Also, Salverda and Altmann (2011; Soto & Humphreys, 2007) showed that context activates the concepts that guide processing; it is therefore appropriate to test the role of context in the processing of negation. Experiments 1 and 2 employed a recognition task (e.g., were the figures circles?) which was the same as the one used by Kaup et al. (2006). In both studies, participants were asked about the identity of the picture that was presented on the screen; the question was not related to the meaning of the sentences themselves (see also Zwaan, Stanfield, & Yaxley, 2002). This task should be considered more ecological than a task requiring comprehension (such as a verification task), which could modulate the processing of negation; it is well known that such tasks affect processing (Clark & Chase, 1972; Jones, 1966; 1968), and moreover, language comprehension does not usually require any specific task to be carried out. Some may question our decision to use a recognition task, arguing that this type of task does not imply full comprehension of the sentence’s meaning, with the resulting danger that participants might not necessarily be interpreting the affirmative and negative assertions, but merely making an associative response to them. In support of this argument, it should be pointed out that Evans (1972) found that individuals showed a tendency to pay attention only to the arguments that have been presented in the premises, regardless of whether they are affirmative or negative (this effect was called the ‘matching bias’). To control for this unlikely effect, we decided to replicate Experiment 2, but with a verification task, in which people needed to understand the sentences in order to give the right response. This is what we did in Experiment 3. We 64 How negation is understood predicted that the type of task would not influence the results and that the findings for Experiments 2 and 3 would be the same. 6.3. Experiment 1 The goal of Experiment 1 was to test whether, on hearing negative sentences in the binary (e.g., the figure could be red or green) and multary (e.g., the figure could be red, or green, or blue, or yellow) contexts, participants would shift their visual attention from figures representing the negated (mentioned) situation towards figures representing the actual (alternative) situation. To that end, we designed a visual world experiment in which participants heard affirmative and negative sentences while exploring a display with four colored figures, and examined the patterns of participants’ fixations on the figures when these sentences were preceded by one of two different spoken contexts: 1) a binary context that mentioned two of the four figures in the display, and 2) a multary context that mentioned all the figures in the display. The predictions supported by current theories differ for binary and multary contexts. Mental model theory makes the following predictions about performance in the task: if an affirmative assertion, such as the figure is red, occurs in either context, there should be an increase in fixations on the red figure referred to in the assertion. If a negative assertion, such as the figure is not red, occurs in a binary context (e.g., the figure could be red or green), there should be an increase in fixations on the actual (alternative) figure, i.e. the green one. This result would support the idea that people switch to the (analogical) representation of the actual situation whenever possible. However, if the negative assertion occurs in the multary context (e.g., the figure could 65 How negation is understood be red, or green, or blue, or yellow), there should be an increase in fixations on the red figure referred to in the assertion. This latter prediction is crucial. Its corroboration supports the possibility of a symbolic interpretation of negation as posited by mental model theory, rather than the hypothesis that negation always calls for the simulation of the actual situation, as held by the two-step theory. Indeed, two-step theory would predict the same results as mental model theory as explained above, except in the multary context, where it predicts that the results will be similar to the binary context for negative sentences, i.e. that there will be an increase in fixations on the alternatives. Kaup et al. (2007) also claimed that it would be possible for the actual situation to remain empty or unspecified (Barsalou, 1999). In that case, the theory predicts that the proportion of fixations on the mentioned color (e.g., the red figure) and the alternatives would be similar, because none of the figures is being processed. 6.3.1. Method Participants. Thirty-one native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. All of them had uncorrected vision or wore soft contact lenses or glasses. Materials. Sixty-four experimental trials were presented to participants distributed in four conditions. For each trial, a different display with four colored figures was shown on the screen (see Figure 1). These figures occupied distinct quadrants of the display and were always of different colors (yellow, blue, green and red). The same colors were used in all the experimental displays, but with the quadrant position counterbalanced. In 66 How negation is understood contrast, the figures’ shape (diamonds, triangles, circles, or squares) varied across displays, but were always the same for the four figures in each display. In other words, all figures were of different colors but an identical shape within trials, and of the same color but different shapes across trials. The experimental conditions were generated from the combination of two spoken sentences, both presented via loudspeakers. In each trial, the first sentence established the context, which was either binary (e.g., the figure could be red or green) or multary (e.g., the figure could be red, or green, or blue, or yellow), and the second sentence was the target, which was referred to in either an affirmative (e.g., the figure is red) or negative polarity form (e.g., the figure is not red). The resulting conditions were hence: binary affirmative, binary negative, multary affirmative and multary negative. In total, there were 16 experimental trials per condition. For all trials, the target word was the color word mentioned in the second sentence. This target word established the onset from which the fixation time-course was analyzed. In addition to the figure display, a written question was also shown on the screen at the end of each trial. This question involved the detection of the shape of the figures in the display (e.g., were the figures circles?) and was unrelated to the spoken sentences, which always referred to the color. In this way, the impact of speech comprehension on fixations could be detached from the impact of the task goal. Apparatus and Procedure. Participants’ eye movements were recorded at a rate of 500 Hz using an SR Research EyeLink II head-mounted eye-tracker connected to a 21-inch color CRT for visual stimulus presentation. Procedures were implemented in the SR Research Experiment Builder. Calibration and validation processes were carried out at 67 How negation is understood the beginning of the experiment and repeated several times per session. Trials started with the display of a central fixation dot for drift correction (see Figure 1) while participants listened to the context sentence that took 3500 ms. After that, a display with the four colored figures appeared for 5500 milliseconds. The target sentence was delivered after one second of display preview, with the target word starting 1500 ms later for affirmatives, and 1600 ms later for negatives. The trial concluded with the appearance of the written question (e.g., were the figures circles?), to which participants had to answer either ‘yes’ or ‘no’, with each response being correct for half of the trials. There was a practice block before the experimental block. The trials of this block (eight in total) followed the same structure but used a different set of colors (gray, violet, brown and orange). The entire experiment lasted approximately 30 minutes. Figure 1. Procedure of Experiments 1: Recognition task (the sentences have been translating from the Spanish). 68 How negation is understood 6.3.2. Results Behavioral data. Response accuracy and latency were both analyzed separately using two-way ANOVAs with context and polarity as factors. The aim of these analyses was to test whether there was any influence of the spoken sentences on the recognition task. Accuracy analysis failed to show any significant sign of interference. In contrast, latency analysis did reveal that the response times were modulated by the spoken materials. Specifically, it showed that response latencies were slower after negative sentences (1288 ms; sd=41) than after affirmatives (1242 ms; sd=38; F(1,30)=6.987, p=.013, η2=.189), and were also slower after binary contexts (1282 ms; sd=41) than after multary contexts (1248 ms; sd=38; F(1,30)=6.425, p=.017, η2=.173). These main effects of polarity and context provide evidence of participant involvement in the comprehension of the spoken materials. Eye movement data. First, we will describe the procedure for analyzing the eyemovement data generated by the EyeLink system. Bitmap templates were created for identifying each of the regions of interest (color circles) in each display. The output of the eye-tracker included the x and y-coordinates of participant fixations, which were converted into region codes using templates. The time period analyzed was from 500 ms before the onset of the critical word (the color word mentioned in the affirmative or negative assertion) to 1500 ms after the word. This time window was chosen to guarantee that there would be enough time for participants to comprehend negations (e.g., Kaup et al., 2006; Lüdtke et al., 2008). This period was divided into 20-ms time slots. For each time slot, the number of fixations on each color was counted and 69 How negation is understood converted into fixation probabilities obtained from the four figures and the background (Figure 2). Fixation probabilities were obtained for display quadrants containing either 1) a figure of the color that was mentioned in the target sentences (‘Mentioned color’; e.g., a red figure for sentences like the figure was (not) red), 2) a figure of one of the colors mentioned in the context sentence but not in the target sentence (‘Alternative color’; e.g. a green figure for a context sentence like the figure could be red or green or the figure could be red, or green, or blue, or yellow, followed by target sentences like the figure was (not) red), or 3) a figure of a color different from any of those mentioned in the context and target sentences (‘Other colors’; e.g., a blue figure for context sentences like the figure could be either red or green followed by a target sentence like the figure is (not) red). Figure 2 shows the impact of the spoken sentences on the time course of fixation probabilities for the Mentioned, Alternative and Other color figures. In this figure, the graphs for affirmatives (2.a and 2.b) show that there was an increase in fixation probabilities on the Mentioned color accompanied by a decrease in fixation probabilities on the Alternative color in both binary and multary contexts. In contrast, the graphs for negatives (2.c and 2.d) illustrate patterns of fixation that differ as a function of context. Specifically, these graphs describe a time course for negative sentences in the multary context that is similar to that of affirmatives, with an increase in fixation probabilities on the Mentioned color and a decrease in fixations on the Alternative color. At the same time, they reveal a reversed time course for negatives heard in the binary context, with a decrease in fixations on the Mentioned color and an increase in fixations on the Alternative color. This latter pattern is preceded by a short period of increased fixations 70 How negation is understood on Mentioned color figures, which quickly fades into the stabilization of both the increase in fixations on the Alternative color figures and the decrease in fixations on the Mentioned color figures. Thus, a visual inspection of the time course of fixations suggests that contextual information only affected the fixation pattern elicited by negative sentences. To explore the statistical significance of the observed pattern, first we calculated the mean fixation probabilities for 100-ms time regions from 100 to 1500 ms after critical word onset. Given that 180-200 ms are needed to account for an eye programming latency (Martin, Shao, & Boff, 1993), the mean of the first time region (100-200 ms) was considered to be the baseline and was used to conduct statistical comparisons against the means of all other 100-ms time regions (for a similar method, see Huettig & Altmann, 2010). This correction to baseline allowed us to control for any bias in the pattern of fixations on figures caused by the context. Tables 1 and 2 detail mean fixation probabilities of each time region with respect to the chosen baseline for both Mentioned and Alternative color figures and also the significances of the corresponding t-tests against the baseline. The statistical comparisons gave results that corresponded nicely with the visual pattern described above. 71 How negation is understood Figure 2. Temporal course of affirmative (top) and negative sentences (bottom) for Binary context (left) and Multary context (right). The onset of the critical word (the color word mentioned in the affirmative and negative sentence) is represented by 0 on the horizontal axis. Error bars represent 95% confidence intervals by subject, such that no overlap between conditions indicates a significant difference. A pre-inspection of fixation probabilities on the Other colors for binary and Alternative color for multary context indicated no differences over time (ps>.2), so the average of these is shown as a single line. 72 How negation is understood Table 1. Summary of the difference of probabilities of fixation between each region and the baseline for the Mentioned color (MC) and Alternative (A) for affirmative and negative sentences for the binary context. Asterisks indicate significant differences (p<.05) with respect to the baseline and the minus sign ‘-‘ indicates a decrease of the proportion of fixation with respect to the baseline. 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 1000-1100 1100-1200 1200-1300 1300-1400 1400-1500 Affirmative_MC Negative_MC Affirmative_A Negative_A .0010 -.0026 .0006 -.0032 .0490* .0145 -.0261 -.0232 .1594* .0416 -.1235* -.0494* .2742* .0758* -.2126* -.0555 .3410* .0777* -.2571* -.0258 .3671* .0506 -.2655* .0223 .3839* .0155 -.2723* .0742 .3940* -.0129 -.2736* .1068* .3814* -.0302 -.2646* .1255* .3782* -.0398 -.2701* .1293* .3601* -.0613 -.2638* .1416* .3498* -.0875 -.2707* .1603* .3394* -.1070* -.2762* .1546* Paired t-tests showed that negative sentences in the binary context generated an initial increment of fixation probabilities on the Mentioned color figure (first significant value at 500-600 ms time period, t(30)=2.481, p=.019). Yet, from 900 ms onward, this tendency changed and fixation probabilities started to increase for the Alternative color figure (t(30)=2.512, p=.018) and decrease for the Mentioned color figure, with this change being significant starting at 1400 ms (t(30)=-2.101, p=.044). Thus, on hearing negative sentences in the binary context, participants shifted their visual attention from the Mentioned color figure toward the Alternative color figure. 73 How negation is understood Table 2. Summary of the difference of probabilities of fixation between each region and the baseline for the Mentioned color (MC) and Alternative (A) for affirmative and negative sentences for the multary context. Asterisks indicate significant differences (p<.05) with respect to the baseline and the minus sign ‘-‘ indicates a decrease of the proportion of fixation with respect to the baseline. 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 1000-1100 1100-1200 1200-1300 1300-1400 1400-1500 Affirmative_MC Negative_MC Affirmative_A Negative_A .0065 .0156 -.0018 -.0049 .0292 .0122 -.0044 -.0073 .1106* .0497* -.0252* -.0049* .2287* .1141* -.0547* -.0343* .3487* .1676* -.0853* -.0553* .4184* .2277* -.1030* -.0647* .4767* .2267* -.1152* -.0592* .5088* .2208* -.1224* -.0535* .5245* .2071* -.1279* -.0521* .5296* .1861* -.1313* -.0484* .5240* .1531* -.1339* -.0423* .5193* .1430* -.1323* -.0407* .5067* .1371* -.1279* -.0402* In contrast, statistical comparisons revealed that there was no shift of attention from the Mentioned to the Alternative color figures in the case of negative sentences heard in the multary context. For these sentences, fixation probabilities revealed instead a prolonged increase in visual attention on the Mentioned color figure (t(30)=2.480, p=.019), which was accompanied throughout by a decrease in fixations on the Alternative color figures (t(30)=-2.554, p=.016) starting at 400 ms. Thus, on hearing a negative sentence in a multary context, participants looked at the figure corresponding to the Mentioned color, at the expense of a reduction of attention toward the figures of an Alternative color. Finally, an identical pattern of fixations was obtained for affirmatives in both binary and multary contexts (see Tables 1 and 2). Just as for negatives in the multary context, affirmative sentences induced a lengthy increase in fixation probabilities on the 74 How negation is understood Mentioned color as well as a decrease in the fixations toward the Alternative color figures. Thus, on hearing an affirmative assertion, participants’ visual attention shifted toward the figure of the Mentioned color, irrespective of whether a binary or multary context had been provided. 6.3.3. Discussion Experiment 1 showed that hearing negatives in a binary context causes participants’ visual attention to shift from a figure of the negated (mentioned) color towards a figure of the actual (alternative) color. This finding confirms the two stages predicted by the two-step theory, and suggests that during the processing of a negative assertion, there must be a shift in the respective saliency of the negated and the actual situations. Moreover, the specific time courses of fixations indicate that the shift was completed around 900 ms after critical color word onset, which is in agreement with some previous reports that the instantiation of the actual situation takes around 1 second (e.g., Kaup et al., 2006). Importantly, Experiment 1 also revealed that hearing negatives in a multary context induces participants’ visual attention to remain trained on the figure of the negated (mentioned) color throughout the period examined. This finding, unlike the previous one, is indeed at odds with a mandatory two-step process and confirms, in opposition to this view, that the shift from the negated to the actual situation is not a mandatory dynamic for negation processing. As predicted by mental model theory, when the actual situation is underspecified or indeterminate, it is the negated situation that remains as the salient content in the course of processing. 75 How negation is understood Supporting this conclusion are the qualitatively similar patterns of fixation exhibited by negatives and affirmatives in multary contexts. Irrespective of context, affirmative sentences generated a deployment of visual attention characterized by a progressive increase in attention to the figure of the Mentioned color accompanied by a parallel decrease in attention to figures of the Alternative colors. This suggests that, as in the case of negatives in multary contexts, hearing an affirmative assertion leads to the progressive instantiation in the mind of the content that has been mentioned explicitly in the sentence. 6.4. Experiment 2 Experiment 1 yielded evidence against a mandatory two-step dynamic for the processing of negatives. There, we gave participants an initial verbal context that varied in terms of the availability of the actual situation to which the subsequent affirmative or negative sentence could be referring (binary and multary context). The goal of Experiment 2 was to test whether the deployment of visual attention for isolated negative sentences (without contextual constraints) would reproduce the pattern observed for negatives in the multary context of the previous experiment. To that end, we presented affirmative (e.g., the figure is red) and negative sentences (e.g., the figure is not red) without any previous verbal context while a picture with four colored figures (e.g., red, blue, green, and yellow) was shown on the screen. The predicted outcome was that there would be an increase in the proportion of fixations on figures of the Mentioned color in both affirmative and negative sentences, accompanied in both cases by a decrease in fixations on the Alternative colors. 76 How negation is understood 6.4.1. Method Participants. Thirty-one native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. All of them had uncorrected vision or wore soft contact lenses or glasses. Materials and Procedure. The instructions and procedure were the same as in Experiment 1. The only difference here was that we presented affirmative and negative sentences (16 trials of each) that were not preceded by any verbal context. Hence, the time given to explore the display before hearing the sentences was not conditioned by any previous verbal information. 6.4.2. Results and Discussion Behavioral data. One participant was eliminated from the analysis because she did not answer any questions. We only found significant differences for reaction times (F(1,29)=5.483, p =.026, η2=.159). Participants were slower to respond to the task questions after negative sentences (95%; 1345 ms; sd=50) than after affirmatives (93%; 1292 ms; sd=48). Again, this result can be taken as indicative of participants’ involvement in the comprehension of the verbal material, even though this was not required by the task. 77 How negation is understood Eye movement data. The procedure for analyzing the eye movement data was similar to that used in Experiment 1. We eliminated one additional subject due to problems in the recording of her eye movements. Figure 3. Temporal course of the Mentioned color and Other colors for affirmative (left) and negative sentences (right) without verbal context (recognition task). The onset of the critical word (the color word mentioned in the affirmative and negative sentence) is represented by 0 on the horizontal axis. Error bars represent 95% confidence intervals by subjects such that no overlap between conditions indicates a significant difference. A pre-inspection of fixation probabilities on the Other colors indicated no differences over time (ps>.2), so the average of these is shown as a single line. Figure 3 shows the effect of the spoken sentences on the time course of eye fixations on Mentioned color and Other color figures. The graph for affirmatives (Fig 3, left) shows that there was an increase in fixations on the Mentioned color figure and a decrease in fixations on the Other colors. We observed the same pattern for negative sentences (Fig 3, right), with an increase in fixation on the Mentioned color and a decrease in fixation on the Other colors. Table 3 shows mean fixation probabilities of each time region with respect to the baseline for both Mentioned color and Other color figures. For negative sentences, it is observed that participants increased fixations on the Mentioned color (t(28)=3,068, 78 How negation is understood p=,0059) and decreased fixations on the Other colors (t(28)= 3,649, p=,001) starting at 400 ms. The same pattern was observed for affirmatives, but the increase with respect to the baseline for the Mentioned color (t(28)= -2,266, p=.031) and the decrease for the Other colors (t(28)= 2,863, p=,008) started at 300 ms. Table 3. Summary of the difference of probabilities of fixation between each region and the baseline for the Mentioned color (MC) and Others (O) for affirmative and negative sentences without verbal context. Asterisks indicate significant differences (p<.05) with respect to the baseline and the minus sign ‘-‘ indicates a decrease of the proportion of fixation with respect to the baseline. 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 1000-1100 1100-1200 1200-1300 1300-1400 1400-1500 Affirmative_MC Negative_MC Affirmative_O Negative_O -.0033 -.0175 -.0067 .0062 .0358* -.0015 -.0211* - .0047 .1343* .0644* -.0501* -.0294* .2625* .1621* -.0897* -.0597* .3780* .2349* -.1222* -.0791* .4484* .2707* -.1378* -.0879* .4734* .2833* -.1398* -.0908* .4524* .2846* -.1391* -.0944* .4388* .2759* -.1417* -.0915* .4011* .2147* -.1372* -.0869* .3849* .2147* -.1323* -.0764* .3551* .1988* -.1260* -.0747* .3139* .1894* -.1189* -.0691* To sum up, Experiment 2 showed the same pattern as the one observed in Experiment 1 for the multary context. Therefore, it must have been the availability of the actual situation and not the verbal context that modulated the shift toward the actual situation (or alternatives). It was found that people kept their visual attention on the Mentioned color when the actual situation was not available, regardless of whether a verbal context was presented or not. This finding is important, because many studies have shown that negation is facilitated by context or pragmatics (Dale & Duran, 2011; 79 How negation is understood Nieuwland & Kuperberg, 2008; Tian et al., 2010). However, the prior presentation of the dimension (i.e. the color) or of the negated and the actual situation in unspecified way (i.e. the figure could be red or green or blue or yellow) does not seem to facilitate negation as some authors have claimed (see Glenberg et al., 1999; Lüdtke & Kaup, 2006), unless this information specifies the negated situation or the actual situation, such as is the case for the binary context. We infer from our results that although negation is processed in a qualitatively similar way to affirmation, a quantitative difference emerges. For example, when we compared affirmative and negative sentences, we observed an increase in fixations on the Mentioned color that was significantly higher for affirmative sentences (13%) than for negative sentences (7%) starting at 400 ms (t(28)=3.116; p=.004). This also occurred for the multary condition in Experiment 1: the increase in attention on the Mentioned color was 11% for affirmatives and 5% for negatives starting at 400 ms (t(30)=2.160, p=.039). This result could indicate that even when affirmative and negative sentences are processed in a similar way in those conditions (in both cases people focused on the Mentioned color), the proportion of attention to the Mentioned color is higher for affirmative sentences than for negative sentences. The higher fixation on the Mentioned color for affirmative sentences (16%) as compared to negative sentences (4%) was also observed for the binary condition in Experiment 1 starting at 400 ms (t(30)=4.058, p<.001). 80 How negation is understood 6.5. Experiment 3 The previous experiments support the idea that people focus on the negated situation both when understanding negation in a multary context (Experiment 1) and when no context is provided (Experiment 2). However, as we used a recognition task (with no deep comprehension demands) in these experiments, it might be argued that the attention to the mentioned color might be due to a lack of full comprehension of the sentences. Of course, this hypothesis would hardly explain the effects obtained in our binary conditions. However, the goal of Experiment 3 was to test whether the deployment of visual attention for isolated negative sentences using a verification task would reproduce the pattern observed in the previous experiment for isolated negative sentences using a recognition task. If so, this would disprove the aforementioned alternative explanation, while lending additional support to our hypothesis. To that end, we presented affirmative (e.g., the figure is red) and negative sentences (e.g., the figure is not red) without any previous verbal context while a picture with four colored figures (e.g., red, blue, green, and yellow) was shown on the screen. We predicted an increase in the proportion of fixations on figures of the Mentioned color in both affirmative and negative sentences, accompanied in both cases by a decrease in fixations on the Alternative colors. 81 How negation is understood 6.5.1. Method Participants. Thirty-one native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. All of them had uncorrected vision or wore soft contact lenses or glasses. Materials and procedure. The instructions and procedure were the same as in Experiment 2. The only difference here was that participants were asked to complete a verification task (e.g., could the figure be red?) instead of a recognition task (see Figure 4). Figure 4. Procedure of Experiment 3: Verification task (the sentences have been translating from the Spanish). 82 How negation is understood 6.5.2. Results and Discussion Behavioral data. Five participants were eliminated from the analysis as their data were more than 2.5 standard deviations from the mean. The results showed that the percentage of accurate responses was lower for negative sentences (80%) than for affirmatives (92%; F(1,25)=37.306, p<.001, η2=.599). Also, the time that participants took to read the comprehension question and answer it was longer for negative sentences (1562(354) ms) than for affirmatives (1456(287) ms; F(1,25)=10.067, p=.004, η2=.287). Figure 5. Temporal course of the Mentioned color and Other colors for affirmative (left) and negative sentences (right) without verbal context (verification task). The onset of the critical word (the color word mentioned in the affirmative and negative sentence) is represented by 0 on the horizontal axis. Error bars represent 95% confidence intervals by subjects such that no overlap between conditions indicates a significant difference. Eye movement data. The procedure for analyzing the eye movement data was similar to that used in Experiment 1. Figure 5 shows the increase in fixation probabilities on the Mentioned color that was observed for affirmative and negative sentences; however, the 83 How negation is understood proportion of fixation is lower for negatives than for affirmatives. The lower level of attention on the Mentioned colors for negatives as compared to affirmatives is a stable pattern that is observed across all contexts. Table 4 shows the mean fixation probabilities of each time region with respect to the baseline for both Mentioned color and Other color figures. For negative sentences, it is observed that participant fixations increased for the figures of the Mentioned color (t(25)=3.462, p=.002) and decreased for the Other colors (t(25)= 2.720, p=.012) starting at 400 ms. The same pattern was observed for affirmatives, but the increase with respect the baseline for the Mentioned color (t(25)=3.277, p=.003) and the decrease for the Other colors (t(25)= -2.375, p=.026) started at 300 ms. Table 4. Summary of the difference of probabilities of fixation between each region and the baseline for the Mentioned color (MC) and Others (O) for affirmative and negative sentences without verbal context. Asterisks indicate significant differences (p<.05) with respect to the baseline and the minus sign ‘-‘ indicates a decrease of the proportion of fixation with respect to the baseline. Affirmative_MW Negative_MW Affirmative_O Negative_O 200-300 300-400 400-500 500-600 600-700 700-800 800-900 900-1000 1000-1100 1100-1200 1200-1300 1300-1400 1400-1500 .0058 .0426* .1508* .2906* .4226* .5339* .5492* .5450* .5328* .5050* .5041* .4802* .4689* -.0016 .0115 .0905* .2006* .2934* .3248* .3046* .2741* .2435* .2142* .1712* .1321* .1013* 84 -.0053 -.0167* -.0395* -.0776* -.0993* -.1094* -.1086* -.1047* -.1066* -.1026* -.0997* -.0981* -.0988* -.0073 -.0143 -.0323* -.0576* -.0671* -.0576* -.0445 -.0362 -.0337 -.0273 -.0154 -.0072 .0012 How negation is understood 6.6. General discussion Two-step theory has claimed that negation cannot be represented explicitly, and that the only way to comprehend negated sentences (e.g., the door is not open) is by means of a shift of attention from the simulation of the negated situation (e.g., an open door, or in our case the mentioned color) toward the simulation of the actual situation (e.g., a closed door, or in our case the alternative; Kaup et al., 2006). This form of processing negation fits well with the notion of a perceptual simulation (Barsalou, 1999), and our results confirm that the participants tend to do that whenever possible. Mental model theory also claims that most representations are perceptual-motor simulations or iconic. However, it also allows for the occasional use of symbolic representations and, by extension, the idea that negation could be represented explicitly. This means that individuals should be able to retain the negated situation in their minds in order to understand negation (Khemlani et al., 2012). Both two-step and mental model theory share the same main assumptions; the essential difference between the two is that while the former claims that people always change to the actual situation (or alternative), the latter also supports the possibility that people might retain the negated situation (or mentioned word). The present study investigated, over three experiments, the time course of attention shifting to the alternative color in the comprehension of negative sentences. We found that participants do shift their attention to the alternative whenever possible, but they also are able to retain the negated situation, as predicted by mental model theory. In Experiment 1, we presented affirmative and negative sentences in binary and multary contexts while participants viewed four different colors on the screen. The 85 How negation is understood results showed that people shifted their attention to the Alternative color starting 900 ms after hearing negative sentences in a binary context, while they looked at the Mentioned color the whole time in a multary context. In Experiment 2, we presented affirmative and negative sentences in isolation (i.e. without a context), and the results replicated those observed for the multary context in Experiment 1. As our participants had demonstrated their understanding of the negative sentences, they must have retained the Mentioned color, applying a symbolic negative operator to it (Giora, 2006; Khemlani et al., 2012; Schul, 2011). This idea was confirmed in Experiment 3, where we used a verification task to ensure participants’ full understanding of the sentences and obtained the same results as those observed in Experiment 2 and in the multary context in Experiment 1. This would therefore appear to indicate that the negation was also understood in Experiments 1 and 2, even though only a recognition task was used. In addition, it is important to consider the stable pattern observed in the understanding of negation; individuals increased their attention to the negated situation starting at 400 ms in all experiments. The N400 component is related to semantic features in event-related potentials (Kutas & Federmeier, 2011), and future research should examine this possible relation using co-registration. The act of retaining the negated situation to understand negation, as evidenced by the present findings, is opposed to the mandatory view of two-step theory, which predicts that individuals will always understand negation through the actual situation (e.g., representing the door is not open by a closed door). However, the results of all three experiments presented here corroborate the claim that although people will prefer to change the negated model for the affirmative model (or alternative) in a binary context and a sensorimotor representation of negation is therefore possible, there are 86 How negation is understood cases (e.g., multary contexts) in which the negative assertion is not informative about the actual situation and many alternatives are therefore available. It would appear that in such cases, people might retain the negated situation and represent the negated operator symbolically (Clark & Chase, 1972; Giora, 2006; Khemlani et al., 2012; Mayo, Schul, & Burnstein, 2004; Schul, 2011). It seems then that while the cognitive system tends to cancel a negated model by an affirmative model (or alternative) in binary contexts, it would prefer to retain the negated model instead of many different affirmative models (or alternatives) which could overload the memory in multary contexts. This would happen as consequence of the limitations of working memory: multiple models can overload its processing capacity and lead to errors in which reasoners fail to consider some models of the premises. As drawing Johnson-Laird (2001), the fewer the number of models needed for an inference, the easier the inference should be by the effect of limit memory. Our results are also consistent with those of certain studies that, while not opposing the embodiment theory, have identified some exceptions (Anderson et al., 2010). For example, Tian and collaborators (2010) presented affirmative and negative sentences that were either cleft (‘it was Jane who didn’t cook the spaghetti’) or non-cleft (‘Jane didn’t cook the spaghetti’), with a matching or mismatching picture appearing 250 ms later. The participants were to indicate whether the object in the picture had been mentioned in the preceding sentence or not. The results showed an interaction between the type of sentences (cleft or non-cleft) and the matching condition. That is, after a non-cleft negated sentence, responses were faster when the image mismatched than when it matched, a result that supports the two-step simulation. However, participants were faster in response to matching images after hearing cleft negated sentences. This result poses problems for two-step processing (Kaup et al., 2006, 2007). 87 How negation is understood One result that has been consistent in the three experiments was the Mentioned word is less attended visually for negative sentences than affirmatives. A question that arisen is whether this fact would be related to the longer time that was required to answer the recognition or verification task after listening to negative sentences compare to affirmative sentences, result that is consistent in literature (see Carpenter & Just, 1975; Carpenter et al., 1999; Clark & Chase, 1972; Trabasso, Rollins, Shaughnessy, 1971; Wason & Johnson-Laird, 1972; Wason & Jones, 1963). Also, the fixation on the Mentioned color could be related to the literature on recall. For instance, the embodiment theory claims that negative concepts are less accessible (through recall) than affirmative concepts, because the negated situation is replaced by the actual situation (Kaup, 2001). On the other hand, propositional theory holds that negation is an explicitly represented operator that takes a whole proposition in its scope. As the negation operator encapsulates the negated concept, it is less accessible (Fillenbaum, 1966; Hasson & Glucksberg 2006; Lea & Mulligan, 2002; MacDonald & Just, 1989). This latter point of view fits better with our results, because the negated situations (or the Mentioned color) received less attention than the affirmatives in both the binary and multary contexts, while an inhibition of the negated situation due to the change toward the actual situation only happened in the binary context. In sum, the lower amount of visual attention paid to the Mentioned word for negative sentences could be related to the fact that the negative concepts are harder to process and recall. Finally, it is important to consider whether the results of this study (which uses exclusively perceptual features, i.e. colors) could be generalized to include more abstract concepts. In our view, a shift of attention toward the actual situation (or alternative) such as that tested in our study depends more on the availability of the 88 How negation is understood actual situation than on the format used to represent the concepts. As the reader can corroborate in any association corpus, such as the Spanish free-association norms (Fernández et al., 2004), the two words of binary predicates are highly associated in the case of both perceptual and abstract concepts; this association should facilitate the shift of attention to the alternative, which will be modulated by the accesibility of the actual situation regardless of the format used to represent the negated argument (e.g., the figure is not red or the door is not open). Also, literature has shown that the shift of attention to the alternative can be modulated by context. In our experiments, the context was logically dichotomized by adding disjunctions, and the binary context showed similar results to those obtained by Kaup et al. (2006), while the multary context showed results that we would expect to see when the actual situation is not available. We found that the multary condition gave results that were similar in the three experiments, whether the verbal context was explicit or not, therefore the availability of the actual situation would appear to be more important than the verbal context for shifting attention toward the actual situation (or alternative). In sum, the representation of negation (the negated situation or the actual situation) depends on the availability of the actual situation, and when this is unavailable, the mental representation of the sentence calls for symbolic tags. 6.7. Conclusion In conclusion, this study has shown that individuals are able to retain the negated concept and consequently, they should be able to represent the negation operator. However, when the alternative concept is available, they prefer to shift attention to the 89 How negation is understood analogically representable concept. This result supports a theory that assumes that a symbolic representation of negation is possible (e.g., the mental model theory; JohnsonLaird, 1983). 90 Visual concepts impede negation 7. Visual concepts impede negation 7. 1. Abstract Many studies have showed the advantage of processing visual concepts over nonvisuals. The present paper reports the case of negation in which, presumably, imagery could slow down processing. Negation reverses the truth value of an argument from false to true or vice versa. Propositions and not images are true or false. Consequently, an image cannot be negated directly. In contrast, negation can be readily attached to a linguistic representation as a syntactic operator. This leads to a paradoxical hypothesis: despite the advantage of visual words for general processing, the negation of arguments involving them (related to the representation of an image) would be harder than negation involving non-visual words. Two experiments support this hypothesis by showing that sentences with a previously negated visual argument took longer to be read than sentences with non-visual arguments. 7. 2. Introduction There is growing evidence supporting the idea that iconic (analogical) representations are crucial in understanding language. This means, for instance, that individuals tend to construct an image of a red figure in order to understand the sentence: the figure is red (Barsalou, 1999; 2005; 2008; Glenberg, Meyer, & Lindem, 1987; Johnson-Laird, 1983; Zwaan & Radvansky, 1998). This evidence connects with the classical concreteness effect: concrete concepts are processed more quickly and accurately than abstract 91 Visual concepts impede negation concepts in a variety of tasks, such as word recognition, lexical decision, recall, problem-solving, and reasoning (for a review, see Denis, 1989). Indeed, this effect has been related to an iconic representation (an image code in the words of Paivio, 1991, 2013) that could be implicated in the processing of concrete concepts but not of abstract concepts. The advantage of concrete concepts over abstract concepts could be related to the greater processing demands needed to construct images (cf. Johnson-Laird & Bethell-Fox, 1978). This type of representation is more structured and elaborated than verbal representation, and consequently leads to a richer and deeper semantic processing (cf. Craik & Tulving, 1975). Furthermore, this dual coding theory (Paivio, 1991) has found support in recent neuroimaging studies showing that the neural representation of abstract concepts relies more heavily on the verbal system, while concrete concept representation involves more mental imagery and relies more heavily on the perceptual system (see Wang et al., 2010 for a meta-analysis). In sum, the role of imagery is currently considered crucial in language comprehension. However, by their very nature, some components of natural language cannot be represented iconically. One significant example of this is negation. Negation is a syntactic operator that takes a proposition and reverses its truth value. In other words, if the proposition A is false (e.g., Steve Jobs is alive), then Not A is necessarily true (e.g. Steve Jobs is not alive) and vice versa; this is an analysis that goes back to Aristotle (1984). Propositions are the smallest units of knowledge that can be true or false. Negation reverses the truth value of arguments from false to true or vice versa such as propositions or sets of propositions, but cannot apply directly to icons or images, as they are not true or false arguments. However, the mental model theory considers that negations can be stored as ‘mental footnotes’ (see Johnson-Laird, 2001). 92 Visual concepts impede negation For example, ¬RED denotes a model of the possibility in which a figure is not red. But people soon lose track of these footnotes because they are hard to remember. There is no such thing as an image of falsity. From the meaning of negation and image, it could be inferred a paradoxical situation: while concrete concepts will generally be easier to process, the negation of arguments involving abstract concepts, which are related to the verbal system, would be easier to process than the negation of arguments involving concrete concepts, which are related to the perceptual system or imagery. There are no studies that have focused on this novel prediction, at least to our knowledge, but if it is true, the negation of arguments involving visual words should be slower than negation involving non-visual words. To test this, we carried out two experiments using a subtle methodology. The goal was to present the same target for both conditions to avoid possible lexical and sub-lexical effects, among others. To achieve this, the experimental texts in Experiment 1 consisted of a sentence formed by an initial conjunctive phrase connecting a visual and a non-visual word. We alternated between negation of the argument involving the visual word: the boy was brave and he was not asleep and negation of the non-visual argument: the boy was awake and he was not cowardly. The next sentence (the target) consisted of a similar conjunctive phrase but with a replacement of the previously negated word with its contrary. The resulting sentence contained the same words for the visual condition: the boy was brave and he was awake and the non-visual condition: the boy was awake and he was brave. The dependent variable was the time that participants took to read this target sentence. The predicted result was that the participants would take longer to read the target sentences after the arguments where the visual words were negated. 93 Visual concepts impede negation In Experiment 1, the participants had to change from a negated situation (e.g., the boy was not asleep) to a situation with the same meaning but using the contrary term (e.g., the boy was awake). This processing of negation would be related to cancellation of the negated concept (asleep). To rule out the possibility that our results were a consequence of the cancellation of concepts rather than the use of negation of each type of argument, we designed Experiment 2, in which the negative concepts were preserved in the target sentences. Here, we presented a sentence formed by an initial conjunctive phrase connecting a visual and non-visual word, one of which was presented in negated form (e.g., John said that the boy was brave and he was not asleep). The next sentence established whether John was right or wrong (50% each), and finally the target sentence was presented (e.g., the boy was brave and he was asleep). As seen, the last sentence repeated the same words from the first one (brave and asleep) and was the same across conditions. The predicted result was the same as for Experiment 1. 7. 3. Experiment 1 The aim of Experiment 1 was to test whether the negation of arguments involving visual words would be slower than the negation of arguments with non-visual words. 7.3.1. Method Participants. Eighty-two native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. 94 Visual concepts impede negation Material and Procedure. Two normative studies were carried out before the experimental study. The goal of the first normative study was to select the material, which consisted of visual and non-visual words with a clear contrary. We presented 80 sentences (20 pairs of visual and non-visual words and their contraries (20*4=80)). Seventy-six students of the University of La Laguna (64 females; mean age: 22 years) had to write the contrary of each word and estimate the degree of visualization in the choice using a Likert scale from 1 to 5. For example, the boy was asleep. The boy was… The underlined words were obtained from the Spanish free-association norms (Fernández et al., 2004). The selection criterion for visual and non-visual words was that they must have a clear contrary, i.e., the percentage of agreement about the contrary must be above 90%. Following this criterion, half of the items were eliminated, leaving 40 words for use in the experimental study. Visual (20) and non-visual (20) words were equated for word frequency, number of syllables, and letters (see Table 1). We used BPAL to determine the words’ characteristics (Davis & Perea, 2005). There was also no difference between visual and non-visual words with respect to the percentage of participants’ agreement on the contrary. The only difference between them was the degree of visualization (t(75)=14.157, p<.001; see Table 1). Table 1. Mean values for characteristics of stimuli. Experiment 1 Visual Non-visual Frequency 1.46 1.65 Number of syllables 3 3 Number of letters 7.10 6.60 Agreement on the contrary 98% 97% Degree of visualization 4.12 2.63 95 Experiment 2 Visual Non-visual 1.63 1.44 3 3 6.22 6.67 98% 94% 4.29 2.73 Visual concepts impede negation The goal of the second normative study was to reproduce with our material the typical concreteness effect observed in lexical decision tasks: faster decision times for visual than non-visual words. Fifty-three native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment study in exchange for course credits. We presented the 80 words (40 visual and 40 non-visual words) used in Experiments 1 and 2 and 80 pseudowords that were created from these words, which thus kept the same properties. Each trial started with a fixation crosshair for 400 ms, followed by the string that remained on the screen for 2000 ms or until a response was given. Finally, a black screen appeared with the indication ‘please, press the SPACE bar to continue’. The experiment started with six practice trials and all stimuli were presented in a random order. To indicate whether or not a string was a Spanish word, participants were asked to position their left and right index fingers on two response buttons marked yes/no. The experiment lasted approximately 5-10 minutes. The participants completed the experiment in an individual cubicle and the instructions and trials were presented via the program E-PRIME 2.0 in a PC with CRT screen. An ANOVA comparing lexical decision latencies across words (663 ms; sd=83) and pseudowords (788 ms; sd= 112) indicated a significant lexical effect (F(1,52)=179.359, p<.001): words were identified faster than pseudowords. In addition, lexical decisions were faster for visual words (656 ms; sd=87) than for non-visual words (670; sd=83; (F(1,52)=6.468, p=.014). This result showed the classical concreteness effect and the well-known advantage of visual words in lexical decision tasks. Based on the results of the first normative study, we presented 40 experimental trials preceded by four practice trials. Also, 32 filler sentences were presented. Each trial had two sentences. The first presented the description of either a person or an 96 Visual concepts impede negation object through two words, one visual and one non-visual, one of which was negated. For example: the boy was awake and he was not cowardly. Participants had to read the sentence and press the button as soon as possible after reading it. Then a second sentence with similar meaning appeared, but in this case the negated word was replaced by its contrary (e.g., the boy was awake and he was brave). Half of the sentences negated the arguments involving the visual word (e.g., the boy was brave and he was not asleep/the boy was brave and he was awake) and half negated the non-visual argument. The second, target sentence was similar for both visual and non-visual conditions (only the order was altered; see Table 2). The predicted result was that the response to the target sentence after the negation of arguments with visual words would be slower than with non-visual words. The only task given the subjects was to read the sentences and press the button as soon as possible. Some comprehension questions also appeared to check that people were actually reading the sentences. The participants completed the experiment in an individual cubicle and the instructions and trials were presented to them via the program Presentation 12.1 in a PC with CRT screen. 97 Visual concepts impede negation Table 2. Examples of experimental sentences of Experiments 1 and 2 Visual Condition Non-visual Condition Exp 1 The boy was brave and he was not The boy was awake and he was not asleep. cowardly. The boy was brave and he was The boy was awake and he was brave. awake. Exp2 John said that the boy was brave and he was not asleep. John was right. The boy was brave and he was awake. John said that the boy was awake and he was not cowardly. John was right. The boy was awake and he was brave. John said that the boy was brave and he was not awake. John was wrong. The boy was brave and he was awake. John said that the boy was awake and he was not brave. John was wrong. The boy was awake and he was brave. 7.3.2. Results and discussion We removed two subjects whose reaction times were more than 2.5 standard deviations from the mean, thus we analyzed 80 participants. 88% of the comprehension questions were responded to correctly, thus indicating that participants were reading and comprehending the sentences. We found that the visual target sentences (1734 ms; sd=365) took longer to read than the non-visual target sentences (1693 ms; sd=341), and these differences were significant (t(79)=2.295, p=.024). Fifty-two participants showed this pattern. 98 Visual concepts impede negation 7. 4. Experiment 2 Experiment 1 showed that the negation of arguments involving visual words took longer to read than those involving non-visual words. This result is likely due to the fact that negation cannot easily apply to images and since visual words have this kind of representation, this could slow down the processing of negation. However, there is another possibility. It has been shown that in negation with contraries – the material that we used in Experiment 1 – the negated concept is cancelled by an alternative; that is to say, individuals understand the boy was not asleep to mean he was awake (Kaup, Lüdtke, & Zwaan, 2006). Also, it is well documented that visual words have two codes: image and verbal, whereas non-visual words have only one code: verbal (Paivio, 1983; 1986; 1991; 2007; 2013). The advantage enjoyed by visual words in many tasks is related to these two codes. This means that the negation of arguments involving visual words would be more difficult than negation involving non-visual words by the effect of cancellation of two codes instead of only one. To examine this alternative hypothesis, we designed Experiment 2, similar to Experiment 1, but where we presented the same words in the first and second sentences so that individuals would retain the negated concept instead changing to the alternative. If the results of Experiment 2 were similar to Experiment 1, this would corroborate the effect of negation on visual words and not to the cancellation. 99 Visual concepts impede negation 7.4.1. Method Participants. Ninety-two native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. Material and Procedure. We took 80 words from the first normative study, including 40 visual and 40 non-visual words. We added more words in this experiment for two reasons: 1) to strengthen our data and 2) because the percentage of agreement about the contrary was not important in this experiment, given people had to retain the negated word and not change to the alternative. In the case of Experiment 2, the percentage of agreement on the contrary was higher for visual than for non-visual words (98%/ 94%; t(78)=3.055, p=.003) and there were differences in the degree of visualization (4.29/2.73; t(78)=17.529, p<.001). The words were equated for frequency, number of syllables, and letters (see Table 1). Eighty trials were presented to participants. Each trial started with the description of a person, in which either a visual argument (e.g., John said that the boy was brave and he was not asleep) or a non-visual argument (e.g., John said that the boy was awake and he was not cowardly) was negated. The next sentence established whether the description was right (e.g., John was right) or wrong (e.g., John was wrong). The trial finished with the target sentence, which was similar for all conditions but with the order altered. In sum, four conditions were created (see Table 2). In the first two conditions, people had to retain the negated concept: 1) when negation involves a visual word and the information is wrong (e.g., John said that the boy was brave and he was not awake. John was wrong. The boy was brave and he was awake); and 2) when 100 Visual concepts impede negation negation involves a non-visual word and the information is wrong (e.g., John said that the boy was awake and he was not brave. John was wrong. The boy was awake and he was brave). The next conditions are similar to Experiment 1, people had to change to the alternative: 3) when negation involves a visual word and the information is right (e.g., John said that the boy was brave and he was not asleep. John was right. The boy was brave and he was awake); and 4) when negation involves a non-visual word and the information is right (e.g., John said that the boy was awake and he was not cowardly. John was right. The boy was awake and he was brave). Each sentence appeared successively according to the reader’s own pace. Also, 34 filler sentences were presented. The only task given to subjects was to read the sentences and press the button as soon as possible. Some comprehension questions also appeared to check that people were actually reading the sentences. The participants completed the experiment in an individual cubicle and the instructions and trials were presented to them via the program Presentation 12.1 in a PC with CRT screen. 7.4.2. Results and discussion We removed ten subjects whose reaction times were more than 2.5 standard deviations from the mean, thus we analyzed 82 participants. 91% of the comprehension questions were responded to correctly, thus indicating that participants were reading and comprehending the sentences. There was interaction between degree of visualization of the argument that was negated in the first sentence (visual and non-visual) and the message of the second sentence, wrong or right (F(1,81)=4.612, p=.035). There were no main effects. Post hoc t-tests revealed that participants took longer to read the sentence 101 Visual concepts impede negation that appeared after the wrong description with negation of a visual argument (e.g., John said that the boy was brave and he was not awake. John was wrong. The boy was brave and he was awake; 1933 ms; sd=456) than the sentence that appeared after the wrong description with negation of a non-visual argument (e.g., John said that the boy was awake and he was not cowardly. John was wrong. The boy was awake and he was brave; 1858 ms; sd=437); t(81)=2.542, p=.013). This finding rejects the possibility that this effect is due to a cancellation of the negated concept. In other words, as a linguistic operator, negation would operate better with arguments involving non-visual words (verbal representations) than with visual words. The iconic representation related to visual words seems to slow down the processing of negation. There was no difference between the two sentences that appeared after the right description for the visual condition (e.g., John said that the boy was brave and he was not asleep. John was right. The boy was brave and he was awake; 1853 ms; sd=418) and the non-visual condition (e.g., John said that the boy was awake and he was not cowardly. John was right. The boy was awake and he was brave; 1859 ms; sd=436; t(81)=-.209, p=.835). These two conditions were similar to those presented in Experiment 1; that is to say, individuals needed to change to the alternative in the target sentence. However, we found no differences in Experiment 2 similar to those found in Experiment 1. One possible reason would be that the difference between the visual and non-visual conditions found in Experiment 1 is small (41 ms), and it could vanish with the introduction of the second sentence (e.g., John was right) in Experiment 2. However, the second sentence: John was wrong introduces another negation and it can make comprehension more difficult and the studied effect is maintained as described above. 102 Visual concepts impede negation 7.5. General discussion Negation is a syntactic operator that takes an argument and reverses its truth value. A proposition can be true or false. If a proposition is true, its negation is false and vice versa. However, images cannot be true or false, and as result, negation cannot apply directly to them. From this definition of negation, it could be inferred that the negation of arguments involving visual words (related to imagery representations) could be slower than for arguments involving non-visual words. The results of Experiments 1 and 2 corroborated this novel hypothesis. Sentences that appeared after the negation of an argument involving visual words took longer to be read than those appearing after non-visual words, even though visual words took less time to read than non-visual words in a lexical decision task. Both results constitute converging evidence of a different representation of visual and non-visual words. These results fit well with theories that distinguish between a visual code and a non-visual code, such as the dual coding theory (Paivio, 1983) or the mental model theory (Johnson-Laird, 1983). Also, Barsalou (2008) pointed out that the simulation system is closely integrated with the linguistic system. From all these points of view, the role of imagery is crucial in language, memory, problem-solving and reasoning in order to improve performance, although this advantage could be reversed in certain cases, such as negation. Finally, it is important to note that our results may be not restricted to negation. Knauff and Johnson-Laird (2002) found that visual relations slowed down the process of reasoning (relational inference and conditional reasoning) in comparison with control relations, visuospatial and spatial relations. The authors concluded that visual images 103 Visual concepts impede negation are not critical for deductive reasoning and may even interfere with the process, whereas spatial representations help individuals to reason deductively. The authors reported studies that have showed an advantage for visual representation; however it seems that the material used in those studies has more spatial than visual properties (cf. De Soto, London, & Handel, 1965). Also, several studies have failed to detect any effect of imageability on reasoning (cf. Sternberg, 1980). Those results that have been corroborating in neuroimaging studies showing that only visual relations elicited additional activity in visual context while all types of relations evoked activity in the parietal cortex related to reasoning (Knauff, Fangmeier, Ruff, & Johnson-Laird, 2003). This hypothesis has been tested with participants who were blind from birth, and they were not affected by the visual-impedance effect (Knauff & May, 2006). Taken together, we could conclude that when image representations yields information relevant to a processing, as it does with mental rotation (Shepard & Cooper, 1982) then it would improve the processing and performance. But if visual images containing details that are irrelevant to a processing should impede and takes longer the processing. It leads to a more general and remarkable conclusion: negation and certain types of reasoning must apply to the concepts related to abstraction and not to the image related to figuration and concreteness, details that may not be relevant for those processing. 104 Inconsistencies make negation plausible 8. Inconsistencies make negation plausible: Evidence from the VWP 8.1. Abstract Communicative language usually refers to what is true or what has happened. Consequently, negation is typically harder to understand than affirmation. However, when context makes negation plausible it could be as easy as affirmation. One example is when the negative sentence denies a false belief. We explored this hypothesis using the visual world paradigm. In the present study, we presented three types of context followed by affirmative or negative sentences while presenting images of the affirmed or denied entities. In a neutral context the participants looked at the affirmed entities prior to the denied entities. This effect was magnified when the context corresponded to a contextually true belief. However, people looked at the object that corresponded to the negation within the same time frame as they looked at the object that corresponded to the affirmation when the sentence referred to a contextually false belief. Our results suggest that negation is easier to understand within a context that presents a false argument. 8.2. Introduction The comprehension of language is context-dependent: its content depends not only on the syntactic and semantic properties of the types of expressions used, but also on facts about the situation in which the expressions are used (Stalnaker, 1998). Negative sentences do not point to the actual situation but to what is not the case. As a 105 Inconsistencies make negation plausible consequence, in their comprehension we should figure out what the speaker is referring to, and pragmatics becomes crucial. If someone tells us that their car is not red there should have been a reason to presume (or speculate) that it is red. Otherwise the utterer would have referred to the actual color of the car. In fact, individuals seem to use affirmation to communicate as default, but negation is often used in some specific contexts, and outwith these contexts negation is difficult to understand (Khemlani, Orenes, & Johnson-Laird, 2012). The comprehension of negation is relatively easy in a binary context (also called complementary) because the actual situation is available. For instance, if people are speaking about the United States elections in 2012 and they say: ‘eventually Romney did not win the election’, hearers are likely to understand that Obama won because only one of them could win, even though many parties were involved. It is assumed that the comprehension of negative sentences with binary predicates such as the above example would be accomplished through two representational steps (Kaup, Lüdtke & Zwaan, 2006). The first step is represented via a perceptual simulation of the negated situation that corresponds to its affirmative counterpart (e.g., Romney won the election); in a second step, this perceptual simulation would be changed to the actual situation implied by the negative (e.g., Obama won the election). The aim of the present study was to explore how this change of attention from the negated situation toward the actual situation is modulated by context. Literature on negation has reported that negative sentences are more difficult to comprehend than their corresponding affirmatives: people take longer to process them and they make more mistakes (see Carpenter & Just, 1975; Clark & Chase, 1972; Mayo, Schul, & Burnstein, 2004; Trabasso, Rollins, & Shaughnessy, 1971; Wason & Johnson106 Inconsistencies make negation plausible Laird, 1972; Wason & Jones, 1963). This finding would indicate that negation adds semantic richness to the argument denied (Miller, 1962). However, recently many studies have also shown that when negation is presented in an adequate context, it could be similar to affirmation (Dale & Duran, 2011; Givon, 1978; Johnson-Laird & Tridgell, 1972; Nieuwland & Kuperberg, 2008; Strawson, 1952; Tian, Breheny, & Ferguson, 2010; Wason, 1972). Thus, it is relevant to study the contextual factors that make the unfolding processing of negation easier to understand. Wason (1965) was the first psychologist who showed how pragmatics could benefit the comprehension of negation. This author presented seven red circles and one blue circle while participants were instructed to describe the stimulus aloud. The results showed that participants took longer when using negation to describe the red circles compared to the corresponding affirmative, while this complexity of negation was less pronounced when using negation to describe the blue circle. This finding supports the exceptionality hypothesis which claims that the effect of negation is reduced when used to refer to a distinctive or unusual attribute. Glenberg and collaborators (1999) found that people took longer to read negative sentences rather than affirmative sentences (e.g., the couch (was/was not) black) into a non-supportive context which presented a different attribute dimension to the affirmative or negative sentences (e.g., She wasn´t sure what kind of material she wanted the couch to be made of), whereas they read similarly both types of sentences into a supportive context which highlighted the relevancy of the attribute dimension that was referred to the affirmative and negative sentences (e.g., she wasn’t sure if a darkly colored couch would look the best or a lighter color). However, Lüdtke and Kaup (2006) pointed out that a context that explicitly mentions the negated concept (e.g., whether the water would be warm or 107 Inconsistencies make negation plausible whether the water would be warm or cold) is a better context for the processing of a negative sentence (e.g., the water was not warm) than a context that merely activates the relevant attribute dimension (e.g., what the water would be like). They also showed that negation (e.g., The T-shirt was not dirty) is as easy as affirmation (e.g., The T-shirt was clean) when the negated concept was highly plausible (e.g., that a boy´s T-shirt is dirty after he played outside in the backyard). It is interesting to note that all these studies described above used binary (or complementary) predicates, thus the fact that the actual situation was available seems important to facilitating comprehension of negative sentences (see Beltrán, Orenes, & Santamaría, 2008). In the current study, we presented three different types of context followed by affirmative and negative sentences while presenting images of the affirmed or denied entities. As negation is focused on what is not the case, we predicted that a context that presents the argument that is going to be negated first (inconsistent context) should make negation easier to understand while a context that presents the actual situation (consistent context) should make it harder to understand. An intermediate case would be a neutral context. We explored this hypothesis using the visual world paradigm. Until now, most studies of negation have used RT and ERP methodologies, but the visual world paradigm presents some advantages. In a typical visual world task verbal and visual inputs are presented simultaneously while the participants’ eye movements are recorded. Eye movements provide an index of real time processing that most other methodologies do not, and they are sensitive to subtle aspects of language, attention and memory (Allopenna, Magnuson, & Tanenhaus, 1998; Duñabeitia et al, 2009; Rayner, 1998). From the literature, it is expected that when something is heard, it is processed and attended to automatically. At the same time, if the object is visible, the eyes begin 108 Inconsistencies make negation plausible to move towards the corresponding object (Cooper, 1974; Tanenhaus et al., 1995). In other words, there is a correspondence between what is visually attended and what is been processing (Eye-mind Hypothesis; Just & Carpenter, 1976). Also, this methodology allows us to see the unfolding process without the intervention of a researcher. That is, the images, related or not to the information that participants hear, are on the screen the whole time, and participants look at them as a consequence of their comprehension. In most other studies, images are presented after the verbal or written information, so it is not possible to have an online measure of processing, but this is not the case with the visual world paradigm. 8.3. Method Participants. Fifty native Spanish speakers from the University of La Laguna, Tenerife (Spain), participated in the experiment in exchange for course credits. All of them had uncorrected vision or wore soft contact lenses or glasses. Materials. Forty-eight experimental vignettes of simple events were heard. The first two sentences per vignette described a general situation (e.g., Veronica needed a new car for work. She wondered whether her dad could help her financially). The last sentence in each vignette presented a context. The context could be consistent (e.g., She supposed that her dad had enough savings), inconsistent (e.g., She supposed that her dad had little savings) or neutral (e.g., her dad lived on the other side of town) with respect to the target part of the story that consisted of an affirmative or a negative sentence (e.g., her dad was not poor). 109 Inconsistencies make negation plausible Figure 1. Example of a display with two images, one corresponds to a rich man and another to a poor man. For each trial, a different display with two images appeared on the screen. For example, an image of a rich man on the left and a poor man on the right (see Figure 1). The position of each image was counterbalanced. The order of the materials was randomized, and 12 versions of the full set of materials were prepared because each context (consistent, inconsistent, and neutral) was combined with the two types of sentences (affirmative and negative) and the two poles of the adjective contraries (e.g., poor-rich), so that only one of the 12 possible versions of each experimental item occurred in each presentation (see Table 1). In total, there were 8 experimental trials in each of six conditions: consistent-affirmative, consistent-negative, inconsistentaffirmative, inconsistent-negative, neutral-affirmative, and neutral-negative. For all trials, the target word was the adjective word mentioned in the last sentence of the story. This target word established the onset from which the time course of fixations was analyzed. 110 Inconsistencies make negation plausible Table 1. Examples of experimental sentences Veronica needed a new car for work. She wondered whether her dad could help her financially. Consistent-Affirmative: She supposed that her dad had little savings. Her dad was poor. Version 1. She supposed that her dad had enough savings. Her dad was rich. Version 2. Consistent-Negative: She supposed that her dad had enough savings. Her dad was not poor. Version 3. She supposed that her dad had little savings. Her dad was not rich. Version 4. Inconsistent-Affirmative: She supposed that her dad had enough savings. Her dad was poor. Version 5. She supposed that her dad had little savings. Her dad was rich. Version 6. Inconsistent-Negative: She supposed that her dad had little savings. Her dad was not poor. Version 7. She supposed that her dad had enough savings. Her dad was not rich. Version 8. Neutral-Affirmative: Her dad lived on the other side of town. Her dad was poor. Version 9. Her dad lived on the other side of town. Her dad was rich. Version 10. Neutral-Negative: Her dad lived on the other side of town. Her dad was not poor. Version 11. Her dad lived on the other side of town. Her dad was not rich. Version 12. Apparatus and Procedure. Participants’ eye movements were recorded at a rate of 500 Hz using an SR Research EyeLink II head-mounted eye-tracker connected to a 21-inch color CRT for visual stimulus presentation. Procedures were implemented in SR Research Experiment Builder. Calibration and validation procedures were carried out at the beginning of the experiment and repeated several times per session. Trials started with the presentation of a central fixation dot for drift correction while participants listened to the general situation and the context (consistent, inconsistent, or neutral). After that, a display with two images appeared for 2 seconds and then the story finished 111 Inconsistencies make negation plausible with the presentation of an affirmative or negative sentence that was the target. The trial concluded with the appearance of a written question (e.g. did Veronica need a new car?) which participants had to answer by pressing either a “yes” or a “no” button. There was a practice block of four trials before the experiment proper started. The entire experiment lasted approximately 30 minutes. 8.4. Results The eye-movement data generated by the EyeLink system were analyzed as follows. First, bitmap templates were created for identifying regions of interest in each display (e.g., a rich and a poor man). The object regions were defined in terms of rectangles containing the relevant objects; fixations landing within the perimeters of these rectangles were coded as fixations on the relevant objects. The output of the eye-tracker included the x- and y-coordinates of participant fixations, which were converted into region of interest codes using the templates. Fixations shorter than 80 ms were pooled with preceding or following fixations if they were within 0.5 degrees of visual angle. Times for blinks were added to the immediately preceding fixations. The analyzed time period was from the onset of the target word to 1500 ms after the target word (as established on a by-trial basis). This time window was chosen to guarantee there was enough time to comprehend negations (e.g., Kaup et. al., 2006; Lüdtke et al., 2008). This period was further divided into 50 ms time slots. For each time slot, the number of fixations on each image was counted and converted into fixation probabilities obtained from the proportion of one of them and the sum of both of them fixation proportions. 112 Inconsistencies make negation plausible Figure 2 depicts probabilities of fixations on the target region of interest as a function of time and context. The target region of interest or ‘the mentioned image’ was the image that was literally mentioned. For example, poor man when the target word was ‘poor’ or ‘not poor’. Be aware that when people heard ‘poor’, they increased their attention on the poor man while when they heard ‘not poor’, they decreased their attention on the mentioned image for attending the rich man. Remember that the proportion of fixation on the mentioned image (e.g., the poor man) was computed by counting, the proportion of fixations on the poor man relative to the sum of probability of fixation on the poor man and the rich man. The sum of both proportions is 1. This means that the decrease of one is the increase of another. Figure 2.1 corresponds to the neutral context, where people heard a context that was not related to the critical sentence (or target). For example, ‘Her dad lived on the other side of town’. In this case the context does not say anything about the images that are shown on the screen, therefore individuals look at both images equally often. When participants listened to the affirmative sentence, e.g. ‘Her dad was poor’, fixation probabilities on the mentioned image increased (e.g. the poor man) and decreased during negative sentence: ‘Her dad was not poor’ on the mentioned image (e.g. the poor man). The latter indicates that there was an increase of fixation probabilities on the alternative image (e.g., the rich man). This simple effect of polarity (affirmative versus negative) in neutral contexts started at around 500 ms after target word onset. Importantly, before this point in time, there was no clear difference between affirmative and negative sentences. 113 Inconsistencies make negation plausible Figure 2.1. Probabilities of fixations on the (mentioned) target image over time, separately for affirmative and negative sentences within neutral contexts. Before the onset of the target word (t < 0), participants focused on both images about equally frequently because the context was neutral. From about 500 ms after target word onset, probabilities of fixations on the mentioned image started to increase for affirmative sentences while they decreased for negative sentences. Error bars represent 95% confidence intervals by subjects such that no overlap between conditions indicates a significant difference. Figure 2.2 shows the data for the inconsistent context condition, where participants heard the critical affirmative or negative sentence following a context that was inconsistent with what was mentioned in the critical sentence (e.g., She supposed that her dad had enough savings -> Her dad was poor/not rich. OR She supposed that her dad had little savings -> Her dad was rich/not poor). When the critical sentence was inconsistent affirmative (e.g., Her dad was poor after She supposed that her dad had enough savings), participants shifted their attention from the image not mentioned in the critical sentence (rich man) toward the image mentioned in the critical sentence 114 Inconsistencies make negation plausible (poor man). When the critical sentence was inconsistent negative (e.g., Her dad was not poor after She supposed that her dad had little savings), they shifted their attention from the image of the poor man (inferred from the context) toward the image of the rich man (inferred from the target sentence). As can be seen, from ca. -500 ms to +550 ms relative to the target word onset, the mentioned referent was more likely to be fixated in negative than in affirmative sentences. This effect reversed from ca. 600 ms onwards. Figure 2.2. Probabilities of fixations on the (mentioned) target image over time, separately for affirmative and negative sentences within inconsistent contexts. In the case of negation, we observed that participants looked at the mentioned image before its onset, and when it was negated, they decreased their attention on it and focused on the alternative image. In the case of affirmation, we observed that participants looked at the alternative image and then they changed toward the mentioned image when it was heard. Error bars represent 95% confidence intervals by subjects such that no overlap between conditions indicates a significant difference. Figure 2.3 shows the data for the consistent context condition, where participants heard a context (e.g., She supposed that her dad had enough savings OR She supposed 115 Inconsistencies make negation plausible that her dad had little savings) that was consistent with the critical sentence (e.g., her dad was rich/not poor OR her dad was poor/not rich). When the critical sentence was affirmative (e.g., her dad was poor after She supposed that her dad had little savings), individuals kept active the same concept (the poor man), the opposite happens for negative sentence (e.g., Her dad was not poor after She supposed that her dad had enough savings). Figure 2.3. Probabilities of fixations on the (mentioned) target image over time, separately for affirmative and negative sentences within consistent contexts. In the affirmative condition, probabilities of fixations on the mentioned image were reliably above 50% across the entire time course, but there was a notable increase in those probabilities from around 700 ms after the onset of the target word. In the negative condition, the mentioned image was fixated reliably less than 50% of the time, but we observed a slight increase in those probabilities when the critical word was heard (peaking around 500 ms after critical word onset). Error bars represent 95% confidence intervals by subjects such that no overlap between conditions indicates a significant difference. 116 Inconsistencies make negation plausible To explore the statistical significance of the observed pattern, mean fixation probabilities were first calculated for 100 ms time-regions from 100 to 1500 ms after critical word onset. Given that 180-200 ms are usually assumed to account for saccade programming (Martin, Shao, & Boff, 1993), the mean of the first time-region (0 - 100 ms) was considered to be the baseline and used to conduct statistical comparisons against means on all others 100 ms time-regions (for a similar method, Huettig & Altmann, 2010). This correction to baseline allowed us to control for any bias in the pattern of fixations on figures caused by the type of context (see Table 2). T-tests and ANOVAs were also used for additional comparisons between conditions. We eliminated one subject due to problems in the recording of her eye movements. We performed 3 (context) * 2 (polarity) factorial ANOVAs and found an interaction between context and polarity starting at 300 ms (F1(2,47)=3.212, p=.049; F2(2,46)=5.288, p=.009). Post hoc t-tests revealed that differences between affirmative and negative sentences were found starting at 300 ms (t1(48)=-2.070, p=.044; t2(47)=2.524, p=.015) for the inconsistent context. There was an increase in fixations on the mentioned image for affirmative sentences (2%) and a decrease in fixations on the mentioned image for negative sentences (-3%). This pattern shows that both types of sentence started to be processed at that time, but in opposite directions, an increase on the mentioned image for affirmative sentences and a decrease for negative sentences. 117 Inconsistencies make negation plausible Table 2. Summary of the difference of probabilities of fixation between each region and the baseline for the Mentioned image for affirmative and negative sentences into contexts: neutral (NEU), inconsistent (INC) and consistent (CON). Asterisks indicate significant differences (p<.05) with respect to baseline and the signal minus ‘-‘ indicates a decrease of the proportion of fixation with respect to the baseline. 100-200 100-300 100-400 100-500 100-600 100-700 100-800 100-900 100-1000 100-1100 100-1200 100-1300 100-1400 100-1500 AF-NEU NE-NEU AF_INC NE-INC AF-CON NEG-CON -.0035 -.0074 .0036 .0581 .1345* .1681* .2079* .2341* .2643* .2687* .2671* .2558* .2640* .2622* -.0105 -.0087 -.0030 -.0151 -.0300 -.0455 -.0866* -.1221* -.1694* -.2028* -.2109* -.2246* -.2424* -.2696* .0064 .0156 .0337 .0750* .1671* .2757* .3457* .4003* .4342* .4591* .4684* .4725* .4762* .4725* -.0162 -.0327 -.0439 -.0536* -.0824* -.1262* -.1967* -.2755* -.3359* -.3830* -.4113* -.4302* -.4335* -.4458* -.0177 -.0137 -.0044 .0056 .0400 .0949* .1261* .1342* .1373* .1481* .1564* .1574* .1531* .1516* .0076 .0165 .0388 .0621* .0814* .0967* .0808* .0665 .0551 .0251 -.0083 -.0172 -.0356 -.0466 There was also a main effect of polarity (F1(1,48)=5.308, p=.026; F2(1,47)=9.051, p=.004) in addition to the interaction (F1(2,47)=8.851, p=.001; F2(1,46)=12.802, p<.001) starting at 500 ms. This main effect of polarity shows that affirmative and negative sentences are different across all contexts. In the neutral context, we found differences between both types of sentences (t1(48)=-2.071, p=.044; t2(47)=-2.132, p=.038), an increase in fixations on the mentioned image for affirmative sentences (6%) and a decrease in fixations on the mentioned image for negative sentences (-2%). From Table 2, there was an increase in fixation probabilities on the mentioned image starting at 600 ms for affirmative sentences (t1(48)=4.344, p<.001; t2(47)=3.900, p<.001) while there was a decrease on the mentioned image starting at 118 Inconsistencies make negation plausible 800 ms for negative sentences (t1(48)=-3.014, p=.004; t2(47)=-2.620, p=.012) compared to the baseline. This is the typical effect of negation found in most of the studies when negation is presented out of context: negation is harder than affirmation. In this case, when the critical word is negated, the differences from the baseline appeared later than when the critical word is affirmative. In the inconsistent context, the pattern of results was similar but appeared much earlier. There was an increase in fixation probability on the mentioned image for affirmative sentences (t1(48)=2.430, p=.019; t2(47)=3.455, p=.001) and a decrease on the mentioned image for negative sentences (t1(48)=-2.044, p=.046; t2(47)=-1.859, p=.069) compared to the baseline starting at 500 ms (see Table 1). This result proves that negation is processed as easily as affirmation when it is presented in an appropriate context, that is, for example, when negation is used to deny a belief, as in the inconsistent context condition. Finally, there were differences between affirmative and negative sentences at 500 ms (t1(48)=2.103, p=.041; t2(47)=1.834, p=.073) in the consistent context, an increase in fixations on the mentioned image for negative sentences (6%) and no change for affirmative sentences (1%). This finding would indicate that when participants heard the affirmative sentence ‘her dad was poor’ after a context with the same information, the processing of affirmative sentences was delayed (700 ms compared to the baseline; t1(48)=3.518, p=.001; t2(47)=3.009, p=.004); however when participants were focusing on the image of the rich man and heard ‘not poor’, they needed to change their attention to the poor man in processing “not poor” (500 ms compared to the baseline; t1(48)=2.736, p=.009; t2(47)=2.706, p=.009). Finally, participants seems to recognize that ‘not poor’ means ‘rich’, and there was a decrease in fixations on the mentioned image (poor) to attend the alternative (rich) again. In sum, 119 Inconsistencies make negation plausible for negative sentences in the consistent context, individuals started looking at the rich man after context, they increased their visual attention on the poor man after hearing ‘not poor’ and finally, they realized that ‘not poor’ is rich and decreased their visual attention on poor man to attend the rich man again. All those inferences are timeconsuming. In sum, as mentioned in the introduction the comprehension of negation supposes a change of attention from the mentioned word toward the alternative, and results showed that the shift of attention occurs earlier for inconsistent contexts compared to neutral contexts and consistent context. These findings suggest that the comprehension of negation is as fast as affirmation in the inconsistent context, that is, when negation is used to deny a false belief. 8.5. General discussion The goal of this study was to explore how the unfolding processing of negation is modulated by context. The results corroborated the classical effect of negation in a neutral context with the visual world paradigm. There was an increase in fixation on the mentioned concept for affirmative sentences starting at 600 ms but a decrease in fixation on the mentioned concept for negative sentences starting at 800 ms. This result shows that the processing of negation is more difficult than that of affirmation (see Carpenter & Just, 1975; Clark & Chase, 1972; Mayo et al., 2004; Trabasso et al., 1971; Wason & Johnson-Laird, 1972; Wason & Jones, 1963) and our paradigm was sensitive enough to detect it. Our eye-movement data also corroborates that the effect of negation is removed when negation is used to deny a belief. We found a difference compared to 120 Inconsistencies make negation plausible the baseline starting at 500 ms for both, affirmative and negative sentences in the inconsistent context. The advantage of using the visual world paradigm is that this method allows us to see online, within each time window, what people do when they process negation in those contexts. For example, when the participants were focusing on the mentioned concept (poor) in the inconsistent context, and heard ‘he was not poor’, they changed toward the alternative immediately. However, when the participants were focusing on both images equally often in the neutral context, and heard one of them (he was not poor), they changed toward the alternative with a delay compared to the inconsistent context. In the first case, the mentioned word was already processed and in the second case, more attention is needed to process the mentioned word. These results are consistent with those of a recent study in which affirmative and negative sentences (e.g., the figure is (not) red) were presented into a specific context (binary context: e.g., the figure could be red or green) while four figures appeared on the screen and eye-movements were monitored. The results showed that people shift their attention to the actual situation after 900 ms (Chapter 6). In the present study, participants looked at the actual situation starting at 800 ms in the neutral context. In this sense the present study is a generalization of the previous results to everyday contexts. However, together the studies show the relevant role of the availability of situational knowledge (in a binary context) to process negation regardless of the content of language (colors or others features). Finally, the consistent context was quite different from the other contexts, probably because it would seem odd. The increase of fixation on the mentioned word for affirmative sentences started at 700 ms. This delay in comparison to other contexts may be due to the target not being very informative, that is, it repeats information. As 121 Inconsistencies make negation plausible the theory of relevance purports, language or communication must be relevant (Grice 1961; Wilson & Sperber, 2004), however in the consistent context, after listening to for example, Veronica needed a new car for work. She wondered whether her dad could help her financially. She supposed that her dad had little savings, individuals identified that her dad was poor, thus when they heard, for example, Her dad was poor this information is less relevant in the sense that participants already knew it. In summary, the information that is repeated is less informative or relevant, and as result, processing is delayed. Also, it is odd to hear the negative sentence Her dad was not poor after hearing She supposed that her dad had enough savings as people are more likely to say: her dad was rich. In this case, participants needed to change their attention from the image of the rich man (after hearing the context) toward the image of the poor man (after hearing negation) to process the mentioned word around 500 ms. And then, they inferred that not poor is rich, so they needed to change their attention from the poor man toward the rich man again at a later stage. These two inferences take extra time compare to the previous contexts (see Johnson-Laird & Tridgell, 1972). The advantage of using the visual world paradigm is that processing is registered online and it is possible to observe each step of processing in each time window. In conclusion, although, in general, negation is more difficult to understand than affirmation, a context that presents the negated argument first makes negation as easy to process as affirmation (Givon, 1978; Khemlani et al., 2012). However, in a context that presents what is the case (as in the consistent condition), where affirmation would normally be used, negation is difficult to process, as extra processing is required to work out what is the case. 122 General Conclusion 9. General Conclusion Throughout this dissertation, we have studied how individuals understand negation and the resulting mental representations. Also, we have shown that its understanding may improve when either the negated argument is symbolic (or non-visual) or when it is presented in a denial context. These results corroborate the three main predictions of the present dissertation. We started with one of the most consistent results in literature: negation is more difficult (larger latencies to response and higher error rates) to understand than affirmation. Among others explanations, this result has been accounted for on the basis of a requirement of two representational steps to understand negation (e.g., the figure is not red), the negated situation: a red figure; and the actual situation: a green figure, while in affirmation (e.g., the figure is red), the actual situation is directly available: a red figure (Kaup, Lüdtke & Zwaan, 2006). The first goal of the present thesis was to study whether this two-step processing is necessary to understand negation, or just one possibility. Also, this is an important issue in cognitive science because processing is related to representation. It is assumed that if people represent the actual situation (e.g., a green figure) to understand negation, the resulting representation is iconic because negation is not represented explicitly, whereas if they represent just the negated situation, this representation should be symbolic, or at least not purely iconic, because a symbolic marker is needed to represent negation without swapping to the actual situation. In other words, it can only be done by keeping ‘not red’ in mind. Therefore, this topic is also important to be able to differentiate between theories of representation. 123 General Conclusion 1. Individuals are able to represent symbolic information. Results described in the chapter 6 showed that people could well understand negation by either representing the actual situation or by building up a symbolic representation of the negated situation. Whether one or the other kind of representation takes place in the mind seems to depend on the availability of the actual situation. Experiment 1 showed that when the actual situation is available, such as in a binary context (e.g., the figure could be red or green), participants understood negation (e.g., the figure was not red) by changing their visual attention from the negated situation (e.g., a red figure) to the actual situation (e.g., a green figure). This shift is time-consuming to envisage (around 900 ms) and proves that, when the actual situation is available, people tend to cancel negation by a representation of the alternative affirmation. As Johnson-Laird (2001) pointed out, negation can be stored as ‘mental footnotes’, but people soon lose track of these footnotes because they are hard to remember. The alternative affirmation, though non-negated in form, helps to further recall of the meaning of the negated sentence, since it is easy to store and corresponds to the actual situation being described in the sentence. Thus, negation processing instills the actual situation (if available) while taking out the situation that is within its scope. Other studies have also shown that negated concepts are less accessible than affirmative concepts, and hence somehow displaced from the mind by negation (Fillenbaum 1966; Lea & Mulligan, 2002; MacDonald & Just, 1989). However, there are cases, such as the multary context, in which the actual situation is not available and therefore many alternatives are possible to describe what the actual situation is. Results of Experiment 1 also showed that participants focused on the figure corresponding to the negated situation in the multary context. This finding is 124 General Conclusion consistent with the possibility that negation is represented symbolically (Clark & Chase, 1972; Giora, 2006; Khemlani, Orenes, & Johnson-Laird, 2012; Mayo, Schul, & Burnstein, 2004; Schul, 2011). This would result as consequence of the limitations of working memory: multiple models can overload its processing capacity and lead to errors. Johnson-Laird (2001) pointed out that the fewer the number of models needed for an inference, the easier the inference should be by the effect of limit memory. In short, it is likely that individuals change one negated model with one affirmative model (alternative) because negation is hard to remember. However, people would prefer to maintain negation rather than two or more possible affirmative models. This latter finding was corroborated in Experiments 2 and 3, in which negative sentences were presented without any context using two different tasks: recognition and verification. It is important to consider the stable pattern observed in the understanding of negation; individuals increased their attention on the negated situation from 400 ms onwards in all experiments. This timing presents an interesting correspondence with the timing for brain semantic processing according to event-related potentials (ERP) research. The N400 component, which unfolds between 200-500 ms, has been repeatedly associated with the integration of word meaning into sentence context (Kutas & Federmeier, 2011), and therefore could be taken as an indicator of comprehension. Although there is not much evidence in the case of negation linking its comprehension to N400 response (the only exception being Niewland & Kuperberg, 2008), future research should examine this possible relation using EEG and Eye-Tracking coregistration. Experiments 2 and 3 revealed an interesting similarity in the temporal course fixations for the verification (e.g., could the figure be red?) and the recognition task 125 General Conclusion (e.g., were the figures circles?). There is a debate in cognitive science as to whether verification is an automatic comprehension process or not. For instance, Fischler and colleagues (1983) found that both affirmative sentences (e.g., a robin is a bird/tree) and negative sentences (e.g., a robin is not a bird/tree) elicited identical ERP effects in sentence category verification tasks, with no systematic differences observed between true and false sentences. Specifically, they observed that in both affirmative and negative sentences, object nouns elicited a larger N400 component when they were categorically unrelated to the subject noun (robin-tree) than when they were related (robin-bird). This seems to suggest that the validation of the truth of a sentence is not a case of automatic processing. However, Richter, Schroeder, and Wöhrmann (2009) showed that reading true sentences (e.g., perfume contains scents) as compared to false sentences (e.g., soft soap is edible) activated more correct responses in an orthographic task, even though a truth validation was not required. Wiswede et al. (2012) presented affirmative and negative sentences (e.g., Saturn is a planet; Saturn is not a planet; Saturn is a continent; Saturn is not a continent) which participants were instructed to read. After that, either verification or identification tasks were conducted. The results showed the classical interaction between polarity and truth value for the verification task. Participants identified the probe words ‘true’ and ‘false’ faster when they matched the truth content (e.g., Saturn is (not) a planet) than when they did not, and it was also slower for negative sentences than for affirmative sentences. No difference was observed between conditions in the identification task, where participants had to indicate if the sentence were identical to a probe sentence or not. This would appear to indicate that the validation of the truth value of the sentences was not an automatic process. Thus, these findings support theories postulating that the processing of truth 126 General Conclusion value is goal-dependent (e.g., Dodd & Bradshaw, 1980; Green & Brock, 2000; Schul, Mayo, & Burnstein, 2004). Apparently, evaluating the truth value of a proposition does not automatically occur whenever a sentence is read and understood, but must be regarded as a conditionally automatic process, the occurrence of which depends on the requirements of the task at hand and the resulting mindset. Carpenter and collaborators (1999) presented a sentence-picture verification task using functional magnetic resonance imaging, and the results showed higher activation in the left posterior temporal gyrus related to language comprehension and higher activation in the left and right parietal regions related to the visuospatial processing. The higher activation for negative rather than affirmative assertions was related to the increased processing difficulty. At the same time, Tettamanti and collaborators (2008) found that negation deactivated cortical areas and the left pallidum, which could be interpreted as a reduction in access to mental representations of the concept being negated. These two studies would seem to be contradictory, but the presentation format of the sentences and the tasks could explain these differences. When people listened to the sentences in Tettamanti’s study, access to the negated word was reduced, while when people were asked to decide whether an image corresponded or not to the previous negated sentence in the verification task in Carpenter’s study, the comparison of the sentence and the image must have been more difficult (higher activation) for negation than for affirmation, because there was no image matching the negation (Khemlani et al., 2012). For example, when participants read a negative sentence: ‘It is not true that the star is above the plus’, they needed to infer that ‘the star is below the plus’ or ‘the plus is above the star’; however when they read an affirmative sentence ‘It is true that the star is above the plus’, there was a correspondence between the sentence 127 General Conclusion and their representation, that is, ‘the star is above the plus’. Therefore, both studies are relevant because they allow us to dissociate between comprehension, related to the reduction of activation for negation as compared to affirmation, and verification, related to a higher activation due to the comparison between the sentence and the picture. This dissociation allows us to be aware about the role of the task in the comprehension of negation. In general, our findings in Experiments 1, 2 and 3 are opposed to the mandatory view of two-step theory, which predicts that individuals will always understand negation by representing the actual situation (e.g., representing the door is not open by a closed door). This form of processing negation fits well with the embodiment theory, one of the dominant frameworks in cognitive science that holds that all representations, including those relating to abstract concepts, must be iconic to the world (Barsalou, 1999; 2005; 2012; Glenberg et al., 1987; Lakoff & Johnson, 1980; Zwaan & Radvansky, 1998). Our results were, in contrast, consistent with the mental model theory, which, while holding that most representations are iconic, also allows for the use of combinations with symbolic representations (Johnson-Laird, 1983; 2006). In summary, individuals are able to understand negation with just one step, and consequently, negation could be symbolically represented. At this point, we can reject the hypothesis that the difficulty of negation comes from the two-step processing because it is not always necessary. This opens a new possible explanation: negation is more difficult because it has a symbolic representation. A number of studies have shown the advantages of iconic representation. What we did in the next study was to find one exception to the supposed iconic advantage. We predicted that an abstract 128 General Conclusion (non-iconic) argument would make negation easier to understand than an iconic argument. 2. Visual concepts could impede negation. Chapter 7 showed that the negation of arguments involving visual words (related to iconic representation; e.g., the boy was brave and he was not asleep) was slower than negation involving non-visual words (related to verbal or symbolic representation; e.g., the boy was awake and he was not cowardly). This finding was corroborated in two experiments. In Experiment 1, we presented the previous sentences followed by a target sentence with the same meaning as the previous one, but with a replacement of the previously negated word with its contrary (e.g., the boy was brave and he was awake). The target sentence was similar for both conditions, but participants took longer to read the target sentence after the negation of the visual argument than the non-visual argument, even thought visual concepts were easier than non-visual concepts in a lexical decision task previously carried out. In Experiment 2, we found the same results. In this case, we presented the same sentence as the previous one but without negation (e.g., the boy was brave and he was asleep). Both experiments showed that iconic representation could impede some symbolic operations such as negation. This finding is very relevant because other authors have also found that visual relations slowed down the process of reasoning (relational inference and conditional reasoning) compared to control relations, visuospatial and spatial relations (Knauff , 2009; 2013; Knauff & Johnson-Laird, 2002). Taken together, we could conclude that although iconic representation has enormous advantages in cognition, its semantic richness could be unnecessary in certain cognitive processing such as negation. For example, it could be easier to negate a symbolic representation of a cowardly child than an iconic representation of a sleeping child. That 129 General Conclusion is, the negation of an image could contain details that are irrelevant to the processing, and consequently it could be slower. It leads to a more general and remarkable conclusion: negation and certain types of reasoning are better applied to the concepts related to abstraction than to the images related to figuration and concreteness. In short, symbolic representation is essential in cognition because it reduces the processing load on working memory and allows more complex operations to proceed without the interference of unnecessary representational elements. Before finishing this part, I would like to relate our results with some theories of representation. The context availability theory holds that concrete concepts are more difficult than abstract concepts by effect of context (Kieras, 1978; Schwanenflugel & Shoben, 1983). According to this theory, concrete concepts can activate more semantic information in isolation than abstract concepts. This increase of activation would provoke the typical concreteness effect, but it would vanish when both types of concepts are presented in equally supporting sentence contexts. In Experiment 1 (chapter 7), we presented a supportive context (e.g., the boy was brave and he was not asleep) and the second sentence (the target) was highly predictable (e.g., the boy was brave and he was awake); however the results were not similar for visual and nonvisual conditions, as the context availability theory would predict. A similar case occurs with the framework that defends qualitative differences in the organization of concrete and abstract concepts in the mental lexicon, that is, concrete concepts would be primarily organized following a semantic similarity principle, whereas abstract concepts would be mainly organized by their association with other concepts (e.g., Crutch, 2006; Crutch & Warrington, 2005; Crutch, Ridha & Warrington, 2006; Duñabeitia et al, 2009; Warrington & Crutch, 2007). Experiment 1 130 General Conclusion (chapter 7) shows that the negation of non-visual concepts was faster than visual concepts. This could prove this framework because non-visual words would activate their associates, as happening in the second sentence of Experiment 1 (e.g., not cowardly-brave). However, we found the same result in Experiment 2 when the same word (e.g., not cowardly- cowardly) and not its associate was presented. Thus, the associate is not the factor that explains the differences between both types of words. Recently, Kousta, Vigliocco and Del campo (2011) have found an advantage for abstract words rather than concrete words in a lexical decision when they match in imageability and the two sorts of words differ only on emotionality (see also, Barber et al., 2013). We found the opposite results in a lexical decision experiment where imageability was manipulated. Both results could indicate that imagery is a necessary and sufficient factor for the concreteness effect because when this factor is present, we found the effect as literature has showed in several studies; however, when this variable is absent, the effect disappeared. To control the emotionality of our eighty words in both experiments in chapter 7, we asked 55 participants to rate the two dimensions of emotionality: the valence, whose range extends from unpleasant to pleasant (1-9), and the activation, extending from calm to excitation (1-9), by the Self-Assessment Manikin (SAM) using the program E-PRIME 2.0. The procedure was the same as that of Redondo and collaborators (2005). The results showed higher valence for visual words (5.43; sd=.47) than non-visual words (5.15; sd=.41; F(1,54)=21.713, p<.001). An average around 5 would indicate that the words were neutral. However, the activation was higher for non-visual (4.74; sd=1.55) than visual words (3.73; sd=1.54; F(1,54)=92.692, p<.001). Therefore, our results do not seem to be due to the grade of emotionality. One relevant question for future research would be a meta-analysis that 131 General Conclusion could point out the effect of the percentage that is explained in each case by the grade of visualization, the emotionality, the frequency, the type of task and other variables on the processing of words (see for example, Zdrazilova & Pexman, 2013). 3. Inconsistency contexts make negation more plausible. Results described in chapter 7 allow rejection of the possibility that the difficulty of negation is related to its representation. Our hypothesis is semantic (Miller, 1962). That is, negation has a more complex semantic meaning than the corresponding affirmation. Negation (e.g., non-red) adds semantic richness to the argument that it takes (e.g., red). If this hypothesis is true, then when we use an affirmation (e.g., her dad was poor) in a neutral context (e.g., her dad lived on the other side of town), it should be easier than the negation (e.g., her dad was not poor) which adds semantic information as we said above. However, if affirmation and negation are used to deny an argument, then both of them would add semantic information and no differences between then are expected. That is, we hold that negation is more difficult than affirmation because it denies an argument and adds semantic information, therefore if affirmation is used similarly, then both of them should be equally easy. In other words, if we presented affirmative sentences to deny a false belief: She supposed that her dad had enough savings and thereafter it is said: her dad was poor, it should be equally easy as the negation (e.g., her dad was not poor followed by She supposed that her dad had little savings). Results of chapter 8 corroborated our hypothesis using the visual world paradigm. It is interesting to find that affirmative and negative sentences were easier when they were used to deny a false belief than in a neutral context because in the first case individuals needed to change a belief and literature suggests that once beliefs are held, they are difficult to change. This means that people tend to maintain their beliefs even 132 General Conclusion when faced with evidence that is contrary or inconsistent (Khemlani & Johnson-Laird, 2012; Orenes et al., 2012). From this premise, one could predict that when people listen to a context that is inconsistent with the target, this latter must be understood slowly (see Ferguson, Sanford, & Leuthold, 2008). However, our results show that both affirmative and negative sentences are processed faster in the inconsistent context (when they are used to deny or change a belief) than in a neutral context maybe due to a priming effect. In our experiment, we presented a context that predicted if her dad was rich or poor. For example, ‘She supposed that her dad had enough savings’ and then it was rejected: ‘Her dad was poor/her dad was not rich’. This inconsistent context may make one think of rich or poor man rather than the neutral context which does not activate any of these concepts: Her dad lived on the other side of town. As Khemlani and collaborators (2012) point out, individuals should find easier to understand a negation if they have already constructed the models of the corresponding affirmative assertion. In summary, both of the sentences, affirmative and negative, could be processed similarly when they are use to deny a false belief as we have shown in our previous study. This could indicate the importance of falsity on negation. I believe that they must be quite similar in language production, but there are not many studies about that, because negation is used to falsify. That is, we can say ‘no’ or ‘it is false’. However, many comprehension tasks, such as verification, have found that false sentences are easier than negative sentences. For example, if people read ‘the figure is red’ and then a green figure appeared. They compared both representations, a red figure and a green figure, and easily found a mismatch. However, negation is not easy to match with an image; therefore this processing could be more complex. For example, if people read ‘the figure is not red’ and then a green figure appeared. They needed to infer that ‘the 133 General Conclusion figure is not red’ is a green figure, and this step, as we have studied along the dissertation, takes time, therefore it could need more time than the first comparison. Our general conclusion from the present thesis would be that the syntactic operator of negation is difficult because it adds semantic information to the argument that it takes, specifically reverses the truth value of the argument, and this function could only be represented symbolically. Individuals are able to understand negation symbolically and this type of representation also improves its computation and comprehension. Furthermore, when negation is used to deny a belief, it could be as easy as affirmation. To sum up, negation has an essential linguistic function, to falsify an argument or to correct a misconception, and this should be fundamental to the cognitive system, since it is presented in natural and artificial languages and from an early age. 134 References References Allopenna, P., Magnuson, J.S. & Tanenhaus, M.K. (1998) Tracking the time course of spoken word recognition using eye-movements: evidence for continuous mapping models. Journal of Memory and Language, 38(4), 419-439. Altmann, G.T.M. (2011). The mediation of eye movements by spoken language. In S.P. Liversedge, I.D. Gilchrist, & S. Everling (Eds.), The Oxford Handbook of Eye Movements (pp. 979-1003). Oxford: OUP. Altmann, G.T.M., & Kamide, Y. (1999). Incremental interpretation at verbs: restricting the domain of subsequent reference. Cognition 73, 247-264. Altmann, G.T.M., & Kamide, Y. (2004). Now you see it, now you don't: Mediating the mapping between language and visual world. In J.Henderson & F. Ferreira (Eds.), The interface of language, vision, and action: Eye movements and the visual world. New York: Psychology Press. Anderson, S., Huette, S., Matlock, T., & Spivey, M. (2010). On the temporal dynamics of negated perceptual simulations. In F. Parrill, V. Tobin, & M. Turner (Eds.), Meaning Form and Body (pp. 1-20). Stanford University, California: CSLI Publications. Aristotle. (1984). The complete works of Aristotle, Vols. 1 and 2. In J. Barnes (Eds.), Princeton, NJ: Princeton University Press. Barber, H.A., Otten, L.J., Kousta, S.T., & Vigliocco, G. (2013). Concreteness in word processing: ERP and behavioural effects in a lexical decision task. Brain & Language. 125, 47-53. 135 References Barclay, J.R. (1973). The role of comprehension in remembering sentences. Cognitive Psychology, 4, 229-254. Barr D.J. (2008) Analyzing ‘visual world’ eyetracking data using multilevel logistic regression. Journal of Memory and Language, Special Issue: Emerging Data Analysis, 59, 457-474. Barsalou, L.W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22, 577-660. Barsalou, L.W. (2005). Continuity of the conceptual system across species. Trends in Cognitive Science, 9, 309-311. Barsalou, L.W. (2008). Grounded Cognition. Annual Review of Psychology, 59, 617645. Barsalou, L.W. (2012). The Human conceptual system. In M.J. Spivey, K. Mcrae, & M.F. Joanisse (Eds.), The Cambridge Handbook of Psycholinguistics, Cambridge (pp. 239-258). University Press. Beltrán, D., Orenes, I., & Santamaría, C. (2008). Context effects on the spontaneous production of negation. Intercultural Pragmatics, 5-4. Bransford, J.D., Barclay, J.R., & Franks, J.J. (1972). Sentence memory: a constructive versus interpretive approach. Cognitive Psychology, 3, 193-209. Bruner, J.S., Goodnow, J.J., & Austin, G.A. (1956). A Study of Thinking. New York: Wiley. Byrne, R.M.J. (2005). The rational imagination: How people create alternatives to reality. CA: MIT Press. Carpenter, P.A., & Just, M.A. (1975). Sentence comprehension: A psycholinguistic processing model of verification. Psychological Review, 82, 45-73. 136 References Carpenter, P.A., Just, M.A., Keller, T.A., Eddy, W.F., & Thulborn, K.R. (1999). Time course of fMRI-Activation in language and spatial networks during sentence comprehension. NeuroImage, 10, 216-224. Clark, H.H. (1969). Linguistic processes in deductive reasoning. Psychological Review, 76, 387-404. Clark, H.H. (1974). Semantics and comprehension. In T.A. Sebeok (Eds.), Current trends in linguistics, Vol. 12: Linguistics and adjacent arts (pp. 1291-1428). The Hague: Mouton. Clark, H. H., & Chase, W.G. (1972). On the process of comparing sentences against pictures. Cognitive Psychology, 3, 472-517. Cooper, R.M. (1974). The control of eye fixation by the meaning of spoken language: a new methodology for the real-time investigation of speech perception, memory, and language processing. Cognitive Psychology, 6, 84-107. Craik, F.I.M. & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268-94. Crutch, S.J. & Warrington, E.K. (2005). Abstract and concrete concepts have structurally different representational frameworks. Brain, 128, 615-627. Crutch, S.J. (2006). Qualitatively different semantic representations for abstract and concrete words: further evidence from the semantic reading errors of deep dyslexic patients. Neurocase, 12, 91-97. Crutch, S.J., Ridha, B.H., & Warrington, E.K. (2006). The different frameworks underlying abstract and concrete knowledge: evidence from a bilingual patient with a semantic refractory access dysphasia. Neurocase, 12, 151-163. 137 References Dale, R., & Duran, N.D. (2011). The cognitive dynamics of negated sentence verification. Cognitive Science, 35, 983-996. Davis, C.J., & Perea, M. (2005). BuscaPalabras: A program for deriving orthographic and phonological neighborhood statistics and other psycholinguistic indices in Spanish. Behavior Research Methods, 37, 665-671. Denis, M. (1989). Image et cognition. Paris: Presses Universitaires de France. De Soto, C.B., London, M., & Handel, S. (1965). Social reasoning and spatial paralogic. Journal of personality and social psychology, 2, 513-521. Dodd, D.H., & Bradshaw, J.M. (1980). Leading questions and memory-pragmatic constraints. Journal of Verbal Learning and Verbal Behavior, 19(6), 695-704. Donders, F.C. (1969). On the speed of mental processes. In W.G. Koster (Eds.), Attention and Performance II, vol 30 (pp. 412-431). Amsterdam: North-Holland. Duñabeitia, J.A., Avilés, A., Afonso, O., Scheepers, C. & Carreiras, M. (2009). Qualitative differences in the representation of abstract versus concrete words: Evidence from the visual-world paradigm. Cognition, 110, 284-292. Edwards, S., & Pratt, S. (2009). Rationality in collective decision-making by ant colonies. Proceedings of the Royal Society B: Biological Sciences. Eifermann, R.R. (1961). Negation: a linguistic variable. Acta Psychologica, 18, 258273. Evans, J.St.B.T. (1972). Interpretation and matching bias in a reasoning task. British Journal of Psychology, 24, 193-199. Ferguson, H.J., Sanford, A.J., & Leuthold, H. (2008). Eye-movements and ERPs reveal the time course of processing negation and remitting counterfactual worlds. Brain Research, 1236, 113-125. 138 References Ferguson, H.J., Scheepers, C., & Sanford A.J. (2010). Expectations in counterfactual and theory of mind reasoning. Language and Cognitive Processes, 25(3), 297346. Fernández, A., Díez, E., Alonso, M.A., & Beato, M.S. (2004). Free-association norms for the Spanish names of the Snodgrass and Valderwart pictures. Behavior Research Methods, Instruments, & Computers, 36, 577-583. Fillenbaum, S. (1966). Memory for gist: Some relevant variables. Language and Speech, 9, 217-227. Fischler, I., Bloom, P., Childers, D., Roucos, S., & Perry, N. (1983). Brain potentials related to stages of sentence verification. Psychophysiology, 20, 400-409. Freud, S. (1925). Negation. In J. Strachey (Eds.), Complete psychological works of Sigmund Freud, Vol. 19: The ego and the id and other works (pp. 181-185). London: Hogarth. Giora, R. (2006). Anything negatives can do affirmatives can do just as well, except for some metaphors. Journal of Pragmatics 38, 981-1014. Giora, R., Fein, O., Aschkenazi, K., & Alkabets-Zlozover, I. (2007). Negation in Context: A functional approach to suppression. Discourse processes, 43(2), 153172. Giora, R., Heruti, V., Metuki, N., & Fein, O. (2009). “When we say no we mean no” Interpreting negation in vision and language. Journal of Pragmatics, 41, 22222239. Givón, T. (1978). Negation in language: Pragmatics, function, ontology. Syntax and Semantics, 9, 69-105. 139 References Glenberg, A.M., Meyer, M., & Lindem, K. (1987). Mental models contribute to foregrounding during text comprehension. Journal of Memory and Language, 26, 69-83. Glenberg, A.M., Robertson, D.A., Jansen, J.L., & Johnson-Glenberg, M.C. (1999). Not propositions. Journal of Cognitive Systems Research, 1, 19-33. Green, M.C., & Brock, T.C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701-21. Grice, H.P. (1961). The causal theory of perception. Proceedings of the Aristotelian Society, Supp. Vol. XXXV, 121-153. Hasson, U., & Glucksberg, S. (2006). Does understanding negation entail affirmation? An examination of negated metaphors. Journal of Pragmatics, 38, 1015-1032. Hayhoe M., & Ballard, D. (2005) Eye movements in natural behavior. TRENDS in Cognitive Sciences, 9(4), 188-193. Heller, D., Grodner, D., & Tanenhaus, M.K. (2008). The role of perspective in identifying domains of references. Cognition, 108, 811-836. Horn, L.R. (1989). A natural history of negation. Chicago: Chicago University Press. Huettig, F., & Altmann, G.T. (2010). ‘Looking at anything that is green when hearing “frog”: How object surface colour and stored object colour knowledge influence language-mediated overt attention. The Quarterly journal of Experimental psychology, I, 1-24. Huettig, F., Rommers, J., & Meyer, A.S. (2011). Using the visual world paradigm to study language processing: A review Psychologica, 137, 151-171. 140 and critical evaluation. Acta References Jaeger, T.F. (2008). Categorical Data Analysis: away from ANOVAs (transformation or not) and towards Logit Mixed Models. Journal of Memory and Language, 59, 434-446. Johnson-Laird, P.N. (1983). Mental models: Towards a cognitive science of language, inference, and consciousness. Cambridge, MA: Harvard University Press. Johnson-Laird, P.N. (2001). Mental models and deduction. TRENDS in Cognitive Sciences, 5(10), 434-442. Johnson-Laird, P.N. (2006). How We Reason. Oxford: Oxford University Press. Johnson-Laird, P.N., & Bethell-Fox, C. (1978). Memory for questions and amount of processing. Memory and Cognition, 6, 496-501. Johnson-Laird, P.N., & Tridgell, J. (1972). When negation is easier than affirmation. Quarterly Journal of Experimental Psychology, 24, 87-91. Jones, S. (1966). The effect of a negative qualifier in an instruction. Journal of Verbal Learning and Verbal Behavior, 5, 497-501. Jones, S. (1968). Instructions, self-instructions and performance. Quarterly Journal of Experimental Psychology, 20, 74-781. Just, M.A., & Carpenter, P.A. (1976). Eye fixations and cognitive processes. Cognitive psychology, 8, 441-480. Kaup, B., (2001). Negation and its impact on the accessibility of text information. Memory & Cognition, 20, 960-967. Kaup, B., Lüdtke, J., & Zwaan, R.A. (2006). Processing negated sentences with contradictory predicates: Is a door that is not open mentally closed?. Journal of Pragmatics, 38, 1033-1050. 141 References Kaup, B., Zwaan, R.A., & Lüdtke, J. (2007). The experiential view of language comprehension: How is negated text information represented?. In F. Schmalhofer, & C.A. Perfetti (Eds.), Higher level language processes in the brain: Inference and comprehension processes (pp. 255-288). Mahwah, NJ: Erlbaum. Khemlani, S., & Johnson-Laird, P.N. (2012). Hidden conflicts: Explanations make inconsistencies harder to detect. Acta Psychologica, 139, 486-491. Khemlani, S., Orenes, I., & Johnson-Laird, P.N. (2012). Negation: A Theory of its Meaning, Representation, and Use. Journal of Cognitive Psychology, 24, 541-559. Kieras, D. (1978). Beyond pictures and words: Alternative information processing models for imagery effects in verbal memory. Psychological Bulletin, 85, 532554. Klima, E.S. (1964). Negation in English. In J.A. Fodor & J.J. Katz (Eds.), The structure of Language (pp. 246-323). Englewood Cliffs, NJ: Prentice Hall. Knauff, M. (2009). A neuro-cognitive theory of deductive relational reasoning with mental models and visual images. Spatial Cognition & Computation, 9(2), 109137. Knauff, M. (2013). Space to reason: A spatial theory of human thought. Cambridge, MA: MIT Press. Knauff, M., Fangmeier, T., Ruff, C.C., & Johnson-Laird, P.N. (2003). Reasoning, models, and images: behavioral measures and cortical activity. Journal of Cognitive Neuroscience, 15, 559-573. Knauff, M., & Johnson-Laird, P.N. (2002). Imagery can impede inference. Memory & Cognition, 30, 363-371. 142 References Knauff, M., & May, E. (2006). Mental imagery, reasoning, and blindness. Quarterly Journal of Experimental Psychology, 59, 161-177. Kousta, S.T., Vigliocco, G., Vinson, D.P., Andrews, M., & Del Campo, E. (2011). The representation of abstract words: Why emotion matters. Journal of Experimental Psychology: General, 140(1), 14–34. Kutas, M. & Federmeier, K.D. (2011). Thirty years and counting: Finding meaning in the N400 component of the event-related brain potentials. Annual Review of Psychology, 62, 621-647. Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago: University of Chicago Press. Lawrence, B.M., Myerson, J., & Abrams, R.A. (2004). Interference with spatial working memory: An eye movement is more than a shift of attention. Psychonomic Bulletin & Review, 11(3), 488-494. Lea, R.B., & Mulligan, E.J. (2002). The Effect of Negation on Deductive Inferences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28(2), 303-317. Liversedge, S.P., & Findlay, J.M. (2000). Eye movements reflect cognitive processes. Trends in Cognitive Science 4(1), 6-14. Lüdtke, J., Friedrich, C.K., De Filippis, M., & Kaup, B. (2008). ERP Correlates of Negation in a Sentence-Picture-Verification Paradigm. Journal of Cognitive Neuroscience, 20(8), 1355-70. Lüdtke, J., & Kaup, B. (2006). Context effects when reading negative and affirmative sentences (poster). 28th Annual Conference of the Cognitive Science Society. 143 References MacDonald, M.C., & Just, M.A. (1989). Changes in activation levels with negation. Journal of Experimental Psychology: Learning, Memory and Cognition, 15(4), 633-642. Martin, E., Shao, K.C., & Boff, K.R. (1993). Saccadic overhead –Informationprocessing time with and without saccades. Perception & Psychophysics, 53, 372380. Mayo, R., Schul, Y., & Burnstein, E. (2004). “I am guilty” vs. “I am innocent”: Successful negation may depend on the schema used for its encoding. Journal of Experimental Social Psychology, 40, 443-449. Miller, G.A. (1962). Some psychological studies of grammar. American Psychologist, 17, 748-762. Mirman, D., Dixon, J.A., & Magnuson, J.S. (2008). Statistical and computational models of the visual world paradigm: Growth curves and individual differences. Journal of memory and language, 59, 475-494. Morris, B., & Hasson, U. (2010). Multiple sources of competence underlying the comprehension of inconsistencies: A developmental investigation. JEP: Learning, Memory, and Cognition, 36, 277-287. Moxey, L.M., & Sanford, A.J. (1987). Quantifiers and focus. Journal of Semantics, 5, 189-206. Nieuwland, M.S., & Kuperberg, G.R. (2008). When the truth is not too hard to handle: An Event-Related Potential study on the Pragmatics of Negation. Psychological Science, 19(12), 1213-1218. 144 References Nieuwland, M.S., & Martin, A.E. (2011). If the real world were irrelevant, so to speak: The role of propositional truth-value in counterfactual sentence comprehension. Cognition, 122(1), 102-109. Orenes, I., Navarrete, G., Beltrán, D., & Santamaría, C. (2012). Schizotypal people stick their first choices. Psychiatry Research, 200, 620-628. Paivio, A. (1983). The empirical case for dual coding. In J. C. Yuille (Eds.), Imagery, memory and cognition: Essays in honor of Allan Paivio (pp.307-332). Hillsdale, NJ: Erlbaum. Paivio, A. (1986). Mental representations: a dual coding approach. Oxford, England: Oxford University Press. Paivio, A. (1991).Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45(3), 255-287. Paivio, A. (2007). Mind and its evolution: A dual coding theoretical approach. Mahwah, NJ: Erlbaum. Paivio, A. (2013). Dual coding theory, word abstractness, and emotion: A critical review of Kousta et al. (2011). Journal of Experimental Psychology: General, 142, 282-287. Peirce, C.S. (1931-1958). Collected papers of Charles Sanders Peirce. In C. Hartshorne, P. Weiss, & A. Burks (Eds.), 8 Vols. Cambridge, MA: Harvard University Press. Posner, M.I. (1980). Orienting of attention. The quarterly journal of experimental psychology, 32(1), 3-25. Rayner, K. (1978). Eye movements in reading and information processing. Psychological Bulletin, 85, 618-660. 145 References Rayner, K. (1998) Eye Movements in Reading and Information Processing: 20 Years of Research. Psychological Bulletin, 24 (3), 372-422. Redondo, J., Fraga, I., Montserrat, C., & Perea, M. (2005). Estudio normativo del valor afectivo de 478 palabras españolas. Psicológica, 26, 317-326. Richter, T., Schroeder, S., & Wöhrmann, B. (2009). You don´t have to believe everything you read: Background knowledge permits fast and efficient validation of information. Journal of Personality and Social Psychology, 96(3), 538-58. Rorty, R. (1982). Method, Social Science, Social Hope. In R. Rorty (Eds.), Consequence of Pragmatism (Essays: 1972-1980; pp. 191-210), Minneapolis: University of Minnesota Press. Russell, B.A.W. (1948). Human Knowledge: Its Scope and Limits. London: George Allen & Unwin. Salverda, A.P., & Altmann, G.T.M. (2011). Attentional capture of objects referred to by spoken language. Journal of Experimental Psychology: Human Perception and Performance. 37, 1122-1133. Salverda, A.P., Brown, M., & Tanenhaus, M.K. (2011). A goal-based perspective on eye movements in visual world studies. Acta Psychologica, 137(2), 172-180. Scheepers, C. (2003). Syntactic priming of relative clause attachments: Persistence of structural configuration in sentence production. Cognition, 89, 179-205. Scheepers, C., Keller, F., & Lapata, M. (2008) Evidence for serial coercion: A time course analysis using the visual-world paradigm. Cognitive Psychology, 56, 1-29. Schul, Y. (2011). Alive or not dead: Implications for framing from research on negations. In G. Keren (Eds.), Perspective on Framing (pp. 157-176). NY: Psychology Press. 146 References Schul, Y., Mayo, R., & Burnstein, E. (2004). Encoding under trust and distrust: The spontaneous activation of incongruent cognitions. Journal of Personality and Social Psychology, 86(5), 668-79. Schwanenflugel, P.J, & Shoben, E.J. (1983). Differential context effects in the comprehension of abstract and concrete verbal materials. Journal of Experimental Psychology: Learning, Memory, & Cognition, 9, 82-102. Soto, D., & Humphreys, G.W. (2007). Automatic guidance of visual attention from verbal working memory. Journal of Experimental Psychology: Human Perception and Performance, 33(3), 730-757. Stalnaker, R. (1998). On the representation of context. Journal of Logic, Language, and Information, 7, 3-19. Sternberg, R.J. (1980). Representation and process in linear syllogistic reasoning. Journal of Experimental Psychology: General, 109, 119-159. Strawson, P. (1952). Introduction to logical Theory. London: Methuen and CO. Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J. C. (1995). Integration of visual and linguistic information in spoken language comprehension. Science, 268(5217), 1632-1634. Tettamanti, M., Manenti, R., Della Rosa, P.A., Falini, A., Perani, D., Cappa, S.F., & Moro, A. (2008). Negation in the brain: Modulating action representations. NeuroImage 43, 358-367. Tian, Y., Breheny, R., & Ferguson, H.J. (2010). Why we stimulate negated information: A dynamic pragmatic account. Quarterly Journal of Experimental Psychology, 63, 2305-2312. 147 References Trabasso, T., Rollins, H., & Shaughnessy, E. (1971). Storage and verification stages in processing concepts. Cognitive Psychology, 2, 239-289. Von Eckardt, B. (1993). What is Cognitive Science?. Cambridge, MA: The MIT Press. Wang, J., Conder, J.A., Blitzer, D.N., & Shinkareva, S.V. (2010). Neural representation of abstract and concrete concepts: a meta-analysis of neuroimaging studies. Human brain mapping, 31, 1459-1468. Warrington, E. K. & Crutch, S. J. (2007). Semantic refractory access disorders. In J.J.Hart & M. A. Kraut (Eds.), The Neural Bases of Semantic Memory (pp. 3-27). Cambridge: Cambridge University Press. Wason, P.C. (1959). The processing of positive and negative information. Quarterly Journal of Experiment Psychology, 11, 92-107. Wason, P.C. (1961). Response to affirmative and negative binary statements. British Journal of Psychology, 52, 133-142. Wason, P.C. (1965). The contexts of plausible denial. Journal of Verbal Learning and Verbal Behavior, 4, 7-11. Wason, P.C. (1972). In real life negatives are false. Logique et Analyse 15, 17-38. Wason, P. C., & Johnson-Laird, P. N. (1972). Psychology of reasoning: Structure and content. Cambridge, MA: Harvard University Press. Wason, P.C., Jones, S. (1963). Negatives: denotation and connation. British Journal of Psychology, 54, 299-307. Watson, J. M. (1979). Referential description by children in negative form. British Journal of Psychology 70, 199-204. Wilson, D. & Sperber, D. (2004). Relevance theory. In L.R. Horn, & G. Ward (Eds.), The Handbook of pragmatics (pp. 607-632). Oxford: Blackwell. 148 References Wiswede, D., Koranyi, N., Müller, F., Langner, O., & Rothermund, K. (2012). Validating the truth of propositions: behavioral and ERP indicators of truth evaluation processes. Scan, 1-7. Wittgenstein, L. (1953). Philosophical investigations. New York: The MacMillan Company. Young, R., & Chase, W.G. (1971). Additive stages in the comparison of sentences and pictures. Paper presented at Midwestern Psychological Association meetings, Chicago. Zdrazilova, L., & Pexman, P. M. (2013). Grasping the invisible: Semantic processing of abstract words. Psychonomic Bulletin & Review. Zwaan, R.A., & Radvansky, G.A. (1998). Situation models in language comprehension and memory. Psychological Bulletin, 123, 162-185. Zwaan, R.A., Stanfield, R.A., & Yaxley, R.H. (2002). Language comprehenders mentally represent the shapes of objects. Psychological Science, 13, 168-171. 149