Negation is known to be more difficult to process, leading to more errors and longer reading and reaction times in comparison with affirmative sentences (Carpenter & Just, 1975; Clark & Chase, 1972; Fischler et al., 1983). The observed processing difficulty is correlated with the presentation of negative sentences out of context. However, when negation is embedded in a legitimizing context, its processing cost decreases and is even eliminated in some cases (for an overview, see Kaup & Dudschig, 2020). One such context is the context of plausible denial, by means of which the property to be denied is previously activated either explicitly or implicitly (Glenberg et al., 1999; Lüdtke & Kaup, 2006; Orenes et al., 2016; Wason, 1965). The processing difficulty is also eased when negation addresses a negative QUD, which can be underlined by grammatical structures like cleft sentences (Tian et al., 2010, 2016). Other pragmatic factors like informativity and polar QUDs are also correlated with the processing ease of negative sentences (Nordmeyer & Frank, 2014; Xiang et al., 2020).
The property being denied can be expressed via multiple modalities, namely it can be embedded in a verbal context by means of oral or written sentences, or in a non-verbal context by using pictures. Although there is evidence supporting the claim that the context of plausible denial leads to a facilitation in the processing of negative sentences, it is not entirely clear which factors are directly responsible for this effect. The current study aimed at providing more information concerning this issue. To that end, we contrasted the processing of affirmative and negative sentences in minimal and extended verbal contexts expressed by means of written sentences. In what follows, we will briefly present how the context of plausible denial was discussed in the literature and afterwards introduce our study.
Starting from the assumption that the function of negative statements is generally to emphasize that a fact was contrary to an expectation, Wason (1965) tested two hypotheses which differed in the way in which the negated property was coded. The exceptionality hypothesis emphasized an exceptional item and a residual class (Exactly one circle/seven circles is/are (not) …). The ratio hypothesis highlighted a smaller and a larger class, the dissimilar item being negated in terms of a property which would have been initially affirmed of a discrete class (Circle no. 7 is (not) ….). Participants were faster when the negative sentence described the exception item in the exceptionality condition than when they described a smaller or a larger class in the ratio condition. It was concluded that negation was not inherently more difficult to process than affirmation when presented in a legitimizing context such as the context of plausible denial, as it allowed participants to accommodate either false propositions or the expectations that were denied as necessary steps in the comprehension of negative sentences.
Assuming that the processing difficulty associated with negative sentences is directly related to their presentation out of context, Glenberg et al. (1999) investigated how negative sentences are processed when presented in linguistic contexts governed by pragmatic constraints. Based on the premise that negation is used to address or to counter presuppositions held by listeners, participants were presented with longer narratives in which the target sentence (The couch was (not) black) was embedded in either a supportive context, where the relevant dimension was mentioned (She wasn’t sure if a darkly colored couch would look the best or a lighter one) or in a non-supportive context, where another dimension was mentioned (She wasn’t sure what kind of material she wanted the couch to be made of). In order to make the stories sound natural, additional introductory, intermediate and final sentences were used. The results showed that the negated sentences in the supportive context were processed as easily as affirmative sentences whereas the negated sentences in the non-supportive context were more difficult to process in comparison with the corresponding affirmative sentences. Glenberg et al. highlighted the importance of context in the processing of negative sentences and suggested that mentioning the relevant property dimension in the prior context led to the elimination of the processing cost associated with negative sentences.
Lüdtke and Kaup (2006) explored other features that rendered a context adequate for the processing of negative sentences. Concretely, they investigated whether the processing of negation is facilitated in a context in which the negated proposition was either explicitly mentioned as a potential possibility in the prior context or could be inferred as a plausible possibility in the given context. Participants read longer narrative stories in which a specific attribute of a particular entity which was previously introduced (Danielle wondered whether the water would be warm/whether the water would be warm or cold/what the water would be like) was either affirmed or denied (The water was (not) warm). The results showed that the processing of negative sentences, if at all, was facilitated when the proposition to be denied was explicitly mentioned in comparison to the case when no particular attribute was mentioned (wondered what the water would be like). However, mentioning one (whether the water would be warm) or two attributes of the entity (whether the water would be warm or cold) did not appear to be a decisive factor.
The authors further investigated the case when the proposition to be denied was presented in the form of an inference that could be drawn from the given context. Following a similar design, participants read narrative stories in which the inference (e.g. that the active/shy boy’s T-shirt would be clean/dirty) was denied by means of an affirmative (The T-shirt was clean) or a negative sentence (The T-shirt was not dirty). A facilitation effect was found when the context strongly implicated the negated proposition. These results corroborated the importance of presenting negative sentences in an appropriate context and provided evidence that the activation of the proposition to be denied could be inferentially triggered in addition to being explicitly mentioned.
In the same vein with previous studies, Nieuwland and Kuperberg (2008) investigated the impact context and world knowledge have on the processing of negative sentences in an ERP study. Accordingly, they contrasted the processing of negative sentences in pragmatically licensed and unlicensed contexts. Negation was considered pragmatically licensed when it was presented in a context in which it was used to reject something that plausibly may have been true (With proper equipment, scuba-diving is very safe and often good fun). In contrast, a pragmatically unlicensed negation (Bulletproof vests are very safe and used worldwide for security) was said to violate pragmatic principles such as informativity. In comparison with previous research, instead of using longer narrative stories, shorter expressions that highlighted world knowledge were used (e.g. with proper equipment). No additional processing cost was found in the case of pragmatically licensed negative statements. The processing difficulty that emerged when negation was pragmatically unlicensed was considered to be the consequence of violating pragmatic communicational principles.
Additional evidence supporting the claim that the processing of negative sentences is facilitated in a context where the negated situation was previously activated is provided by Orenes et al. (2016). In a visual world paradigm, the authors tested affirmative (Her dad was rich) and negative sentences (Her dad was not poor) in a neutral (Her dad lived on the other side of town), consistent (She supposed her dad had enough savings) and inconsistent context (She supposed her dad had little savings). The processing of negation was facilitated in the inconsistent context, namely the context in which the negated situation (‘her dad was poor’) was activated before participants were presented the negative sentence. However, in comparison with the results from behavioral studies, negative sentences were always slower to process than affirmation.
Overall, these studies present evidence from different paradigms that negative sentences are facilitated in a context of plausible denial and that the salience of the property to be denied, in the form of expectations, assumptions or previous statements, represents the decisive factor in triggering this effect. In contrast, when negative sentences are presented in the absence of a legitimizing context to guide towards their intended interpretation, they become more difficult to accommodate, warranting additional inferences and increasing their processing cost.
In the current study, we investigated the conditions under which the processing difficulty associated with negation can be reduced, rendering the processing of negative sentences comparable to the processing of affirmative sentences. There are more ways in which the situation to be denied can be activated and made accessible to both the speaker and the hearer in a verbal discourse: by being explicitly mentioned in a previous context, by contextual inference or by means of linguistic structures (Larrivée, 2012). In our study, the context of plausible denial is represented by linguistic contexts generated by discourse markers, such as contrary to expectations and synonym expressions like unexpectedly, unpredictably, surprisingly (in the following ‘discourse markers’). They render the affirmative and negative sentences felicitous by providing the context of plausible denial, namely by creating a contrast between an alternative state of affairs and the actual state of affairs. To illustrate, in the case of a sentence like Contrary to expectations, John hasn’t eaten the soup, the marker implies an alternative state of affairs derived from an expectation (‘It was expected that John would eat the soup’) which stands in contrast to the actual situation (John hasn’t eaten the soup) and raises the issue ‘whether John has eaten the soup’. On the contrary, when the negative sentence is presented out of context more questions are at issue (‘What has John eaten?’, ‘Who has eaten the soup?’, ‘Has John eaten the soup?’). This presumably leads to a delay in comprehending the sentence because people potentially compute more questions. Evidence supporting this claim is provided by visual world paradigm studies (Tian et al., 2016), where participants, when hearing Bill has not opened his brother’s window, look at both the negated (open window) and the actual state of affairs (closed window) for a rather long time (until hearing window) whereas when hearing It was Bill who hasn’t opened his brother’s window they look significantly faster at the actual state of affairs (closed window). By and large, the discourse markers (contrary to expectations, surprisingly, unexpectedly, unpredictably) grant a context of plausible denial for the upcoming sentences (John has (not) eaten the soup), namely by making the situation to be denied salient (It was expected that John would (not) eat the soup) and by clarifying what the question at issue is (whether John has eaten the soup).
In six behavioral experiments, we investigated whether a verbal context of plausible denial always leads to a facilitation of the processing cost associated with negation. In these experiments, we contrasted the processing of affirmative and negative sentences in contexts of plausible denial (in the following ‘denial context’) and in non-denial contexts. For the purposes of our argument, the denial context was represented either by solely discourse markers (Experiments 1 to 4) or by extended contexts, where more contextual information was provided (Experiments 5 and 6). In Experiments 1 and 2, participants responded to a sensibility-judgement task (with the judgement times serving as dependent variables) and in Experiments 3–6, they read the sentences fragment by fragment (with fragment reading times serving as dependent variables). We predicted an interaction between the factors Context and Polarity with (a) significantly longer response times (RTs) for negative compared to affirmative sentences in the non-denial contexts and (b) similar RTs for affirmative and negative sentences in the denial contexts. In Experiment 1, we compared the processing of affirmative and negative sentences in minimal denial contexts (Contrary to expectations, John has/hasn’t eaten the soup) with those in non-denial contexts (John has/hasn‘t eaten the soup) by means of a sensibility judgement task. To rule out that the results were due to the length disparity between the sentences in the two context conditions, in Experiment 2, we introduced expressions with an identical number of syllables in the non-denial context ([Everybody is convinced that/ Everyone thinks that/ Based on what we know/ We believe that] John has/hasn’t eaten the soup). This led to a change in the type of non-denial context from providing no context to providing a type of context which reports somebody else’s words. Furthermore, to ensure that the observed effects were not specific to the task employed, in Experiment 3 and 4, we replaced the sensibility judgement paradigm with a fragment-by-fragment self-paced reading task (Contrary to expectations,//John has/hasn’t eaten the soup). Following the same bipartite division of non-denial contexts (namely providing a reporting context or no context at all), in Experiment 3, we employed markers similar in length with those used in the denial context (By all accounts,/Reportedly,/Apparently, Supposedly, John has/hasn’t eaten the soup) while in Experiment 4, we did not offer any additional contextual information, similar to Experiment 1. The same pattern of results emerged in all four experiments, with two robust main effects of context and polarity but no significant interaction, indicating that the two factors were independent of each other. However, when the markers were replaced with the explicit mentioning of the expectation to be denied (His parents expected John to eat the soup) a facilitation effect of negation emerged in the denial contexts (Experiment 5), replicating the findings from previous research. A facilitation effect was also found when negative sentences were used in combination with the discourse markers in extended contexts, indicating that the facilitation effect was most likely triggered by the interplay between pragmatic factors like informativity or relevance (Experiment 6).
All in all, although the minimal contexts render the negative sentences felicitous, we could not find any evidence suggesting that these contexts eliminate the processing difficulty associated with negative sentences. Instead, embedding the sentences in longer narratives led to a facilitation effect, making the processing of negative sentences similar to the processing of affirmative sentences. This pattern of results suggests that although the context of plausible denial renders negation felicitous, the determining factors in triggering the facilitation effect are informativity or relevance.