Isn’t it Ironic? An Analysis on the Elaboration of Ironic Sentences with ERPs

Although frequent in our everyday conversations, irony is a complex pragmatic phenomenon involving specific linguistic, communicative and cognitive abilities in order to be fully understood. In this study we examined the pragmatic comprehension of ironical and non ironical language by analysing event-related potentials (ERPs) of irony decoding process. We asked 12 subjects to listen to 240 sentences with a counterfactual vs. non-counterfactual content and spoken with ironical vs. neutral prosody. ERPs morphological analysis showed a negative deflection peaking in central-frontal and parietal areas at about 460ms post stimulus onset (N400) for all the conditions. Statistical analyses applied to peak amplitudes showed no statistically significant differences between the conditions as a function of the type of sentence (ironical vs. non ironical) and the content of ironical sentences (counterfactual vs. non counterfactual). An increase of N400 related to ironical sentences was nonetheless observed. The absence of an N400 effect may indicate that irony is not treated as a semantic anomaly, although, the observed differences in amplitude could be probably attributed to a higher requirement for the cognitive system in order to integrate contrasting and complex lexical, prosodic and contextual cues. INTRODUCTION: IRONY BETWEEN PRAGMATICS AND PSYCHOLINGUISTICS Although frequent in our everyday conversations, irony remains a complex communicative and pragmatic phenomenon whose correct decoding requires specific linguistic, communicative and cognitive abilities. While trying to explain how irony is used by speakers, linguistics and pragmatics elaborated different theories exploring the nature of ironic communication and of its production and comprehension processes. Within these frameworks, irony has been considered a form of semantic anomaly [1, 2] or as a pragmatic construct involving forms of pragmatic insincerity [3], pretense [4], echoic elements [5] or context-inappropriateness [6]; or again it has been considered as a form of indirect negation [7, 8]. On a cognitive level it has been conceptualized as a form of thought [9-11] involving different grades of contrast between linguistic representation and the reality domain it refers to [12] Finally, with a communicative approach, we can consider irony not as a semantic or pragmatic anomaly, but as a form of communication involving different levels of representation and complex communicative intentions ([13, 14], for analytic review of principal theories on irony). Generally irony can be defined as a pragmatic phenomenon where a gap exists between what is said and what is meant by the speaker. This gap is not necessarily a form of counterfactuality, in fact, Kreuz demonstrated [15] that saying something patently false is only one cue of irony, but not a sufficient one. Some examples of non counterfactual irony are represented by the case of understatements and hyperboles, those particular forms of language where a situation is *Address correspondence to this author at the Center for Communication Psychology, Department of Psychology, Catholic University, Largo Gemelli, 1, 20123 Milan, Italy; Tel: +39-2-7234.2586; Fax: +39-27234.2769/2280; E-mail: [email protected] described in intermediate terms between its opposite and its reality [16] or, again, the case of true statements that can be appropriate but not fully relevant in the situation [17, 6]. These two cases are frequent between ironic remarks. Futhrmore, some comments such as questions, offerings, overpolite requests or expressives, could be ironic even if they cannot be evaluated in terms of truth conditions. In order to explain the ironicity conveyed by these expressions, Kumon-Nakamura and colleagues [3] appealed to the construct of pragmatic insincerity that is an expression that violates felicity conditions more than truth conditions [2, 18, 19]. However, pragmatic insincerity cannot be considered a sufficient cue for irony as well. Other cues have been identified in the callbacks to expectations, whether explicit or implicit [4, 20, 21], norms or shared beliefs [17, 10]. Finally extralinguistic cues, such as, vocal and prosodic profile or facial mimics, have been considered important on a communicative level [15, 22, 23]. In order to decode these cues and consequently understand the communicative intention involved in such a complex pragmatic phenomenon, it is necessary to apply different meta-representational, meta-linguistic and pragmatic abilities. Thence the complexity of irony is apparent and it is not surprising that different theories have been elaborated to explain what irony is and how it is understood by decoders. This last question – relative to irony comprehension -, in particular, has been the focus of psycholinguistics investigations. In this perspective, three main models of language processing tried to explain irony comprehension processes formulating three different hypotheses on the basis of the relation between lexical access and contextual information of ironic statements: the sequential model elaborated by the Standard Pragmatic View [1, 2]; the Parallel Access Models [24, 25]; and finally, the Direct Access View [9, 26]. According to the Standard Pragmatic View, irony can be considered a violation of conversational maxims, in particu10 The Open Applied Linguistics Journal, 2008, Volume 1 Balconi and Amenta lar of the quality (truthfulness) one. It is thus necessary, in order to understand an ironic meaning to process, at first, the literal meaning of the ironic statement, then test this meaning against the context and, whether the violation of conversational maxims is detected, to look for an alternative – nonliteral – meaning. The process to retrieve the intended meaning involves extra-inferential processes named “conversational implicatures”. Only at the end of this process the incoherence is solved and the intended meaning reconstructed [27, 19]. This process is supposed to require an extra-effort, therefore extra-time, to reach the right interpretation. This standard model has been actually disconfirmed by different empirical studies which found that, in particular cases, response time (RT) to ironies are not slower than RT to literal statements [9, 26, 28-31]. The Parallel Access Model, on the contrary, suggests that both the literal and the nonliteral interpretation of an utterance are always processed. Two main models can be brought back to this general model: the Parallel Race Model [24] and the Graded Salience Hypothesis [25]. Long and Graesser [24] assumed that, in ironic communication, a discrepancy exists between what is expected and what actually occurs. In particular, within an ironic statement, what is expected is the intended ironic meaning while what occurs is the literal meaning. Following this hypothesis, the context (background, linguistic, conversational and social) plays a crucial role in irony comprehension processes either allowing comprehension to occur after some discrepancy is recognized or biasing the interpretation early on. Either way both literal and ironic meaning are simultaneously accessed and concurrently processed, and they both participate in ironic meaning construction. Dews and Winner [32, 33] tested this model showing that deriving an ironic interpretation involves some recognition (conscious or unconscious) of the discrepancy between literal and ironic meaning. If one only recognizes the intended meaning of an ironic utterance without noticing at some level what was literally said, one has not fully understood irony. Anyway Dews and Winner concludes that the entire literal meaning need not be processed before the intended meaning is accessed, as was the case of standard pragmatics, since multiple meanings can be processed simultaneously. On the other side, Giora [25, 29] suggests that the initial elaboration of literal or nonliteral meanings is linked to meaning salience. To be salient a meaning should be coded in the mental lexicon and that happens when it is familiar, frequent, conventional and prototypical. When two or more meanings are salient, they should be accessed in parallel. According to this model, irony elaboration is a matter of salience more than of context [34]. Giora distinguishes between familiar (conventional) and non familiar (non conventional) ironies on the basis of salience and aptness and supposes that non lexical contextual information should not affect initial meaning activation: salient lexical meaning (literal) of non familiar ironies is the only one instantly activated also in ironically biased context, even though it is incompatible with contextual information. On the contrary, the salient literal meaning of familiar ironies is available in ironically biased contexts in spite of a mismatch with contextual information. In sum, irony comprehension seems to be a function of meaning salience and not of context [34]. A third approach to irony comprehension proposes that, under some particular circumstances, ironic meaning is directly accessed [26, 35]. Gibbs proposes that in a context offering enough ironic cues, non literal interpretation of a statement is direct and automatic with no need to compute the literal inconsistent interpretation. That could happen because in a high constraining context, a nonliteral interpretation is conventional, so the elaboration of nonconventional literal meaning is optional and the listener does not have to fully elaborate the literal meaning and its incongruity. This hypothesis has been tested through reading time paradigms, whose basic assumption is that reading times of sentences could be suggestive of the initial comprehension processes. Reading time of ironic non conventional utterances showed they require extra effort to elaborate in comparison to their literal conventional interpretation, whereas this is not the case of conventional ironies [30, 33, 36]. As we have seen, empirical findings of behavioural studies, supporting or opposing these models, are controversial, therefore the question about how irony is processed remains open. More recently, neuropsychological technologies have been applied to figurative language comprehension. In the following paragraph we will examine some of the findings these technologies provided when applied to irony studies. EMPIRICAL FINDINGS FROM ELECTROPHYSIOLOGY: APPLYING ERPS TO IRONY INVESTIGATION Event-related potentials (ERPs) have been successfully used in psycholinguistic studies on literal and figurative language ([37, 38] for a review) with interesting results that, together with reading time and lexical decision paradigms, helped to understand complex pragmatic phenomenon such as metaphor [39-42] or humour [42], and only recently have been applied to irony studies [43-45]. ERPs are a measure characterized by high temporal definition and provide intrinsic indexes of cognitive processes [46]. The most known index of semantic integration is N400, a negative, central-parietal deflection, that peaks around 400 ms from stimulus onset [47]. N400 is an amodal index elicited by any stimulus with a semantic valence and which amplitude is directly proportional to the effort requested by the integration process of the stimulus it is related to. It is generally considered as a measure of the difficulty for the retrieval and integration of semantic information related to an incongruous or ambiguous word in a context [47, 48] N400 has been proved to be sensitive to ambiguous words in context, to semantic and pragmatics anomalies and to unexpected words or sentence finals [48-53]. In figurative language studies, N400 has been used as an index to study qualitative and quantitative differences between literal and non literal language. Metaphor studies found generally an increase of the N400 amplitude associated with metaphoric expression in comparison to literal sentences. Coulson and Van Petten [39] compared literal sentences to a literal mapping condition and to metaphoric sentences finding a progressive increasing of N400 peak amplitudes from literal to metaphoric expressions. The authors interpreted these results as a gradient of complexity in sentence decoding in the three conditions. Coulson and Van Petten results seem to confirm Pynte et collaborators [49] ERP Analysis of Ironic Sentences Elaboration The Open Applied Linguistics Journal, 2008, Volume 1 11 findings, who previously reported progressive variations in N400 intensity across literal sentences, conventional metaphors and novel metaphors. Anyway, the identification by Coulson and Van Petten of an intermediate condition, namely the literal mapping, between literal and metaphoric sentences seemed to indicate that elaboration processes literal and metaphoric sentences do not qualitatively differ, since the processes seem to share the same mechanism of conceptual integration. In addition, Pynte, suggested other factors could concur in N400 modulation and these are context and familiarity, the first being fundamental since it could frame the sentence biasing the interpretation process. N400 sensitivity to contextual information, renders it a valid index to explore the process of irony comprehension and to test the pragmatic models of irony processing. A recent study [43] used ERPs to assess the effects of cognitive elaboration strategies (analytic versus holistic) on irony comprehension. The results of the study show that in the holistic condition, where subjects were asked to look for the global sense of the proposed sentence, each category tested (literal vs. ironical vs. nonsensical) generate a negativity, analogue to the N400 component, mainly in the left frontal-central zone. Significant differences were observed in the literal condition when compared to both the ironical category and the nonsensical category. Authors concluded that the differences observed in the N400 for both the nonsensical and the ironical condition compared to the literal condition could be due, in the first case to the semantic incongruity of the nonsensical stimulus, instead, in the second case, the increase of the N400 could be attributed to the lack of contextual information favouring the ironical comprehension. An opposite result was found by another study [54] that tried to assess the influence of prosody on ironic comprehension by manipulating both the context and the paralinguistic components of ironic expressions. Results of this study showed no difference in the N400 amplitude due to prosody. These findings seem to confirm that, when sufficient contextual information is present, ironical elaboration is easier at least in the first phases (no N400 effect appeared in a prosodically enriched context). OBJECTIVES AND HYPOTHESES Main aim of the present study is to explain how ironic sentences are processed, testing pragmatic models of irony comprehension through the ERP methodology. In particular, in the present study we intend to explore the influence of the prosodic cues on the initial phases of irony comprehension process where little contextual information is given. We are interested in better understanding the cognitive and neuropsychological mechanisms involved in early phases of the comprehension process of ironic statements. Therefore our analyses will focus on the 300-400 time window. We hypothesize that if irony is perceived and categorized as a semantic incongruity, therefore we should observe an ampler N400 (from now hence: N400 effect). On the contrary, if irony is immediately recognized (that is the ironic remark is not understood as semantically incongruous), therefore no N400 effect should be present. Tracking it down to a classical pragmatic hypothesis we Fig. (1). Pragmatic models and expected ERPs effects. ! " # $ $ $