Running head: NEGATION AND SITUATIONAL PRESENCE

Two experiments examined the impact of different kinds of representations on the accessibility of negated text information. Participants read narratives containing a color term that was mentioned either within the scope of an explicit negative or not, and with the described situation being such that the corresponding color was either present or not. Accessibility of the color term was measured by means of a probe-recognition task either 500 ms (Experiment 1) or 1500 ms (Experiment 2) after participants had read the sentence mentioning the color term. After the 500 ms delay, the accessibility of the color term was influenced mainly by the structure of the sentence. After a 1500 ms delay, the accessibility was influenced only by the content of the described situation. These results are consistent with the view that comprehenders construct two kinds of representations, a linguistic representation of the text and a situation model in which only those properties are represented that are present in the described situation. An alternative account, according to which comprehenders only construct a perceptual simulation of the referent situation will be discussed. Negation and Situational Presence 3 In language comprehension research, negation is usually considered an operator that shifts the discourse focus away from the information mentioned within its scope and thereby reduces the accessibility of this information (e.g., Lea & Mulligan, 2001; MacDonald & Just, 1989; see also Paterson, Sanford, Moxey, & Dawydiak, 1998; Sanford, Moxey & Paterson, 1996). For instance, when readers are presented with sentences such as (1), and immediately afterwards make a word recognition or naming response, they are quicker to respond to the probe word bread than to the probe word cookies, which seems to suggest that the negation marker no reduces the accessibility of bread, a noun mentioned within the operator’s scope (MacDonald & Just, 1989). (1) Every weekend, Mary bakes bread but no cookies for the children. Independent of the validity of this claim, however, there is another potential explanation for this result. It is conceivable that it is not the negation operator that is responsible for the reduced accessibility of the probe word bread, but rather a situational variable. In the situations described in (1), there is bread present but not a single cookie. Assuming that language comprehension involves the construction of a representation in which only those entities that are present in the described state of affairs are represented, the accessibility difference after reading sentences such as (1) can be attributed to the fact that this representation contains a token for bread but no token for cookies (Kaup, 1997, 2001). Distinguishing empirically between these two explanations is important because they are based on very different assumptions about the kind of representations involved in language comprehension. The first explanation, according to which the scope of the negation operator is the relevant variable, rests on the assumption that readers construct a linguistic representation of the sentence, for instance a propositional representation. In a propositional representation, negation is an explicitly represented operator that takes a whole proposition in its scope. Thus, the sentence Negation and Situational Presence 4 Mary bakes bread but no cookies is represented as “bakes[Mary, bread] and NOT[bakes[Mary, cookies]]”, and it is conceivable that cookies is less accessible than bread after reading this sentence, simply because it is encapsulated by the negation operator (MacDonald & Just, 1989). The second explanation, according to which the situational content is the relevant variable, rests on the assumption that comprehension is tantamount to the construction of a situation model (Van Dijk & Kintsch, 1983). A situation model is a representation of the situation described by the linguistic input, and thereby differs in important ways from a propositional representation. Whereas a propositional representation is a mental description of the state of affairs under consideration, a situation model represents the state of affairs itself. In other words, the components of a situation model are not propositions describing particular aspects of the state of affairs, but the entities, properties, and relations that make up the state of affairs (Zwaan & Radvansky, 1998). Thus, a situation model contains tokens for entities and properties only that are actually present in the described state of affairs (Anderson, Garrod, & Sanford, 1983; Carreiras, Carriedo, Anlonso, & Fernandez, 1997; Glenberg, Meyer, and Lindem, 1987; Zwaan, Madden, & Whitten, 2000). Accordingly, the situation model constructed for (1) contains a token for bread but no token for cookies, which might well be the cause of the observed difference in accessibility. To evaluate these two explanations, we conducted two experiments. Participants read narrative texts that contained a color word in the penultimate sentence [e.g., (2)-(5)]. This color word was either mentioned within an affirmative [(2) and (4)] or within a negative phrase [(3) and (5)]. Moreover, the corresponding color was either present in the described situation [(2) and (3)] or absent from it [(4) and (5)]. The accessibility of the color term was measured by means of a proberecognition task after the end of the sentence. Thus, in the examples below, the probe word would be pink. Negation and Situational Presence 5 (2) Sam was relieved that Laura was wearing her pink dress. (3) Sam wished that Laura was not wearing her pink dress. (4) Sam wished that Laura was wearing her pink dress. (5) Sam was relieved that Laura was not wearing her pink dress. According to the propositional explanation, negation functions as an accessibility-reducing mechanism, and thus we would expect to find longer latencies for the color word after reading (3) and (5) than after reading (2) and (4). In contrast, according to the situation-model explanation, the relevant variable is whether or not the color is present in the described situation, and thus we would expect to find longer latencies after reading (4) and (5) than after reading (2) and (3). Most researchers investigating situation-model construction during language comprehension assume that comprehenders construct a situation model in addition to a propositional representation of the text (for an overview, see Fletcher, 1994). On the basis of these multi-level accounts, it could be proposed that both variables have an impact on the accessibility of text information, but at different levels of representation. In fact, the results of an earlier study, in which negation and situational presence were varied orthogonally to each other, are consistent with this view (Kaup, 2001). Participants were presented with narratives that contained negation sentences mentioning either constructing or destroying activities (e.g., John was building the castle but not the church; Sarah was burning the letters but not the photographs, respectively). The negation effect proved to be significantly larger for the passages with constructing activities than for the passages with destroying activities. This differential result was interpreted as reflecting the fact that negation and situational presence reinforce each other for passages with constructing activities (the non-negated entity is present in the resulting situation, the negated entity is not), whereas the two variables are counteracting for passages with destroying activities (it is the negated Negation and Situational Presence 6 entity that is present in the resulting situation). Thus, the findings of this study are consistent with the view that comprehenders construct a propositional representation in which negation is represented explicitly as well as a situation model in which only those entities and properties are represented that are present in the described situation. This multi-level explanation, however, was clearly post-hoc in this study. Studies explicitly dealing with the different levels of representation constructed during comprehension have provided evidence that propositional representations are available earlier than situation models in the comprehension process (e.g., Schmalhofer & Glavanov, 1986). Thus, if indeed both variables have an impact on the accessibility of text information, their impact should vary with the delay at which accessibility of text information is measured. The impact of the negation operator should be stronger after a short than after a long delay, and the impact of the situational content should be stronger after a long than after a short delay. Accordingly, we varied the delay with which the probe was presented after the penultimate sentence of the narratives. In Experiment 1, the probe was presented with a 500 ms delay, in Experiment 2 with a 1500 ms delay.