How alliteration enhances conceptual–attentional interactions in reading

Abstract

In linguistics, the relationship between phonological word form and meaning is mostly considered arbitrary. Why, then, do literary authors traditionally craft sound relationships between words? We set out to characterise how dynamic interactions between word form and meaning may account for this literary practice. Here, we show that alliteration influences both meaning integration and attentional engagement during reading. We presented participants with adjective-noun phrases, having manipulated semantic relatedness (congruent, incongruent) and form repetition (alliterating, non-alliterating) orthogonally, as in “dazzling-diamond”; “sparkling-diamond”; “dangerous-diamond”; and “creepy-diamond”. Using simultaneous recording of event-related brain potentials and pupil dilation (PD), we establish that, whilst semantic incongruency increased N400 amplitude as expected, it reduced PD, an index of attentional engagement. Second, alliteration affected semantic evaluation of word pairs, since it reduced N400 amplitude even in the case of unrelated items (e.g., “dangerous-diamond”). Third, alliteration specifically boosted attentional engagement for related words (e.g., “dazzling-diamond”), as shown by a sustained negative correlation between N400 amplitudes and PD change after the window of lexical integration. Thus, alliteration strategically arouses attention during reading and when comprehension is challenged, phonological information helps readers link concepts beyond the level of literal semantics. Overall, our findings provide a tentative mechanism for the empowering effect of sound repetition in literary constructs.

Introduction

The question as to whether the phonological form of a word has any bearing on its meaning has intrigued scholars for millennia (cf. Plato's Cratylus). Mainstream opinion in the language sciences advocates no such relationship, claiming instead that phonological forms are arbitrarily associated with semantic concepts (Saussure, 1916, Gasser, 2004, Lupyan and Winter, 2018). Nevertheless, proponents of sound symbolism (i.e., iconicity in natural language) have advocated that a word's phonology can and does reflect some of its semantic features, particularly in non-indoeuropean languages (cf. Perniss et al., 2010, Monaghan et al., 2014, Kovic et al., 2010, Asano et al., 2015; see also; Culler, 1975, Jakobson, 1960), as is the case for onomatopoeic words (e.g., bang, pop, splash; Perniss & Vigliocco, 2014).

A natural extension to this question, then, is whether interactions between phonology and semantics extend beyond the level of intra-lexical iconicity, i.e., affect relationships between words at the phrasal level. In spoken language, comprehenders use prosodic patterns to structure the input and parse the speech stream which has an immediate impact on semantic processing (e.g., Breen et al., 2014, Brown et al., 2015). But it is unclear whether and how phonological information derived during (silent) reading affects comprehension. Recent work in neurocognitive poetics suggests that stylistic prosodic features in phrases, such as phonological repetition, attract more attentional resources and that their neural representations are more strongly activated than those of neutral, declarative forms, as shown in behavioural (e.g. Carminati et al., 2006, Yaron, 2002, Tillmann and Dowling, 2007, Hanauer, 1998) and in event-related potential (ERP; Chen et al., 2016, Obermeier et al., 2013; Vaughan-Evans et al., 2016) studies.

Here, we measured event-related brain potentials (ERPs) and pupil dilation (PD) changes in native English speakers reading adjective-noun phrases manipulated orthogonally for semantic relatedness and alliteration. We chose to study form-meaning relationships at the most elementary level of word combination in reading, i.e., two-word phrases, in order to (a) have full experimental control via counterbalancing of stimuli across conditions, and (b) remove an inherently ‘poetic’ attribute from the potentially biasing context of verse, thus providing an evaluation of form-meaning relationship as it occurs in natural language. Our choice of methods also allowed us to examine semantic processing in the context of attentional engagement: In ERP research, increased negativity in mean amplitudes of the N400 wave is associated with increased difficulty in accessing the meaning of a stimulus (Chwilla et al., 1995, Kutas and Federmeier, 2011). On the other hand, increased PD indexes the recruitment of attentional resources and task-related uncertainty (Geng et al., 2015, Kang et al., 2014, Mathôt, 2018). Early dilation (<1000 msec) is associated with attentional orienting, relating to stimulus saliency or novelty, whereas later dilation (>1000 msec) is thought to reflect autonomic arousal, linked with mental effort or interest (Mathôt, 2018, Wang and Munoz, 2015, Wetzel et al., 2016).

We anticipated that semantic processing would be more difficult (and thus elicit greater N400 amplitudes) for incongruent adjective-noun pairs, and that alliteration would interact with semantic processing. We also expected that attention allocation would be boosted (and thus increase PD) in response to more effortful semantic processing, analogous to the N400 (Kuipers and Thierry, 2011, Wetzel et al., 2016), and that alliteration would also attenuate this response. We further expected these effects to occur during the later phase of pupil dilation (Mathôt et al., 2017, Wetzel et al., 2016). Moreover, potential correlations between ERP amplitude and PD index offered an opportunity to empirically describe dynamic links between semantic integration and attentional engagement (cf. Kuipers and Thierry, 2011, Kuipers and Thierry, 2013). In order to investigate whether semantic integration (occurring ~400 msec) further relates to early or later phases of attentional engagement, we examined the relationship between mean ERP amplitude in the classical N400 time window (300–5000 msec) and pupil dilation over the entire sequence of a trial.