Discourse-level Information Research Articles

Eye-tracking evidence from attachment structures favors a serial model of discourse–sentence interactivity

ABSTRACT Many psycholinguistic studies examine how people parse sentences in isolation; however, years of work in discourse processing have shown that sentence-level interpretations are influenced at some stage by discourse-level information. Evidence over the past 20 years remains mixed as to the temporal dynamics of such top-down interactions. In particular, dynamic accounts where readers use the discourse model to generate expectations for certain grammatical structures before and during parsing differ from serial accounts where an algorithmic first-pass processing mechanism precedes integration of sentence material into the discourse model. To test between these two theories, the current study investigates eye-movement behaviors when reading temporarily ambiguous attachment structures following discourses with biases either matching, mismatching, or neutral with respect to the attachment resolution. No evidence was found suggesting readers systematically use discourse information to generate structural expectations, in line with serial accounts of processing at the sentence–discourse interface. Scanpath analyses further highlight the confirmatory nature of rereading when participants encounter discourse continuations that do not fit with prior contexts.

Linking discourse-level information and the induction of bilingual discourse connective lexicons

The single biggest obstacle in performing comprehensive cross-lingual discourse analysis is the scarcity of multilingual resources. The existing resources are overwhelmingly monolingual, compelling researchers to infer the discourse-level information in the target languages through error-prone automatic means. The current paper aims to provide a more direct insight into the cross-lingual variations in discourse structures by linking the annotated relations of the TED-Multilingual Discourse Bank, which consists of independently annotated six TED talks in seven different languages. It is shown that the linguistic labels over the relations annotated in the texts of these languages can be automatically linked with English with high accuracy, as verified against the relations of three diverse languages semi-automatically linked with relations over English texts. The resulting corpus has a great potential to reveal the divergences in local discourse relations, as well as leading to new resources, as exemplified by the induction of bilingual discourse connective lexicons.

Variation in Aspect Usage in General-Factual Contexts: New Quantitative Data from Polish, Czech, and Russian

This study aims to account for the variation in aspect choices in factual imperfective contexts in Polish, Czech, and Russian. A series of online questionnaires were conducted wherein the native speakers of the tested languages were asked to fill in the missing verbs for two types of existential contexts (neutral and resultative) and four types of presuppositional factual contexts (weakly and strongly resultative with a focus on the initiator or the result state of the past event). We show that neutral existential factual contexts generally elicited significantly more imperfective choices than resultative existential factual contexts. Additionally, there was a trend towards a higher usage of imperfective in weakly resultative presuppositional contexts as compared to strongly resultative presuppositional contexts, suggesting that the less emphasis is placed on the result state the more likely the choice of imperfective aspect is for the expression of the temporal indefiniteness of factual contexts. Russian showed a significantly higher proportion of imperfective uses than Polish and Czech, with Czech being intermediate. We argue that these observations result from the fact that in all types of factual contexts (both existential and presuppositional) there is an interaction between two types of TEMPORAL (IN)DEFINITENESS of the past event: (i) temporal (in)definiteness at the micro-level (first phase syntax-vP) (depending on the position of the time variable within the temporal event of the past complex event) and (ii) (in)definiteness of the past event at the macro-level (second phase syntax–AspP and TP) (related to the position of the past event relative to the utterance time). We show that both discourse-level information and verb-level information interact in determining these two types of (in)definiteness, and they do it differently in Polish, Czech, and Russian.

The Role of Prosody and Morphology in the Mapping of Information Structure onto Syntax

The mapping of information structure onto morphology or intonation varies greatly crosslinguistically. Agglutinative languages, like Inuktitut or Quechua, have a rich morphological layer onto which discourse-level features are mapped but a limited use of intonation. Instead, English or Spanish lack grammaticalized morphemes that convey discourse-level information but use intonation to a relatively large extent. We propose that the difference found in these two pairs of languages follows from a division of labor across language modules, such that two extreme values of the continuum of possible interactions across modules are available as well as combinations of morphological and intonational markers. At one extreme, in languages such as Inuktitut and Quechua, a rich set of morphemes with scope over constituents convey sentence-level and discourse-level distinctions, making the alignment of intonational patterns and information structure apparently redundant. At the other extreme, as in English and to some extent Spanish, a series of consistent alignments of PF and syntactic structure are required to distinguish sentence types and to determine the information value of a constituent. This results in a complementary distribution of morphology and intonation in these languages. In contact situations, overlap between patterns of module interaction are attested. Evidence from Quechua–Spanish and Inuktitut–English bilinguals supports a bidirectionality of crosslinguistic influence; intonational patterns emerge in non-intonational languages to distinguish sentence types, whereas morphemes or discourse particles emerge in intonational languages to mark discourse-level features.

Discourse-Level Information Recall in Early and Late Bilinguals: Evidence From Single-Language and Cross-Linguistic Tasks.

Bilingualism research indicates that verbal memory skills are sensitive to age of second language (L2) acquisition (AoA). However, most tasks employ disconnected, decontextualized stimuli, undermining ecological validity. Here, we assessed whether AoA impacts the ability to recall information from naturalistic discourse in single-language and cross-linguistic tasks. Twenty-four early and 25 late Chinese-English bilinguals listened to real-life L2 newscasts and orally reproduced their information in English (Task 1) and Chinese (Task 2). Both groups were compared in terms of recalled information (presence and correctness of idea units) and key control measures (e.g., attentional skills, speech rate). Across both tasks, information completeness was higher for early than late bilinguals. This occurred irrespective of attentional speed, speech rate, and additional relevant factors. Such results bridge the gap between classical memory paradigms and ecological designs in bilingualism research, illuminating how particular language profiles shape information processing in daily communicative scenarios.

Principle B constrains the processing of cataphora: Evidence for syntactic and discourse predictions

We tested whether comprehenders can use Binding Principle B (Chomsky, 1981) to guide antecedent search during the processing of cataphoric pronouns. We ran two self-paced reading experiments using the gender mismatch paradigm (van Gompel & Liversedge, 2003) as an index of active prediction of coreference between a cataphor and a main clause subject. In both experiments, we find gender mismatch effects at the main clause subject when coreference with the cataphor is grammatically acceptable. We do not find comparable gender mismatch effects in conditions where coreference is ruled out by Principle B. Our results are broadly consistent with models in which grammatical constraints serve as early filters on anaphora resolution processes in comprehension. We illustrate how the parser can integrate syntactic and discourse-level information to achieve grammatical sensitivity during incremental referential processing.

Does hitting the window break it?: Investigating effects of discourse-level and verb-level information in guiding object state representations

ABSTRACT During language comprehension, comprehenders form mental representations of the described events. We investigate discourse-level and verb-level cues that guide this process. In particular, we investigate how comprehenders represent object states when events are described with manner verbs that do not entail change-of-state (e.g. hit, wash): a potential change-of-state of the object can be inferred but is not semantically required. We report two reaction-time-based experiments (Experiment 1: lexical decision, Experiment 2: self-paced reading) that investigated how rapidly comprehenders process linguistic material associated with potential change-of-state inferences, in contexts where the preceding discourse context and verb-level information are manipulated. In both experiments, we find an interaction between discourse-level information and verb-level information in guiding object state representations. We highlight the need to take into account discourse-level factors in theorising about the cognitive process of understanding the dynamics of event representation during language comprehension.

Limits on expectation-based processing: Use of grammatical aspect for co-reference in L2

AbstractThis study examines the use of discourse-level information to create expectations about reference in real-time processing, testing whether patterns previously observed among native speakers of English generalize to nonnative speakers. Findings from a visual-world eye-tracking experiment show that native (L1;N= 53) but not nonnative (L2;N= 52) listeners’ proactive coreference expectations are modulated by grammatical aspect in transfer-of-possession events. Results from an offline judgment task show these L2 participants did not differ from L1 speakers in their interpretation of aspect marking on transfer-of-possession predicates in English, indicating it is not lack of linguistic knowledge but utilization of this knowledge in real-time processing that distinguishes the groups. English proficiency, although varying substantially within the L2 group, did not modulate L2 listeners’ use of grammatical aspect for reference processing. These findings contribute to the broader endeavor of delineating the role of prediction in human language processing in general, and in the processing of discourse-level information among L2 users in particular.

Who are you talking to? The role of addressee identity in utterance comprehension

Experimental evidence suggests that speaker and addressee quickly adapt to each other from the earliest moments of sentence processing, and that interlocutor-related information is rapidly integrated with other sources of nonpragmatic information (e.g., semantic, morphosyntactic, etc.). These findings have been taken as support for one-step models of sentence comprehension. The results from the present event-related potential study challenge this theoretical framework providing a case where discourse level information is integrated only at a late stage of processing, when morphosyntactic analysis has been already initiated. We considered the case of Basque allocutive agreement, where information about addressee gender is encoded in verbal inflection. Two different types of Basque grammatical violations were presented together with the corresponding control conditions: one could be detected based on a morphosyntactic mismatch (person agreement violation), while the other could be detected only if the addressee's gender was considered (allocutive violation). Morphosyntactic violations elicited greater N400 effects followed by P600 effects, while allocutive violations elicited only P600 effects. These results provide new constraints to one-step accounts as they represent a case where speakers do not immediately adjust to the addressee's perspective. We propose that the relevance of discourse-level information might be a crucial variable to reconcile the dichotomy between one- and two-step models.

Comprehension of wh-questions in Turkish–German bilinguals with aphasia: A dual-case study

ABSTRACTThe aim of our study was to examine the extent to which linguistic approaches to sentence comprehension deficits in aphasia can account for differential impairment patterns in the comprehension of wh-questions in bilingual persons with aphasia (PWA). We investigated the comprehension of subject and object wh-questions in both Turkish, a wh-in-situ language, and German, a wh-fronting language, in two bilingual PWA using a sentence-to-picture matching task. Both PWA showed differential impairment patterns in their two languages. SK, an early bilingual PWA, had particular difficulty comprehending subject which-questions in Turkish but performed normal across all conditions in German. CT, a late bilingual PWA, performed more poorly for object which-questions in German than in all other conditions, whilst in Turkish his accuracy was at chance level across all conditions. We conclude that the observed patterns of selective cross-linguistic impairments cannot solely be attributed either to difficulty with wh-movement or to problems with the integration of discourse-level information. Instead our results suggest that differences between our PWA’s individual bilingualism profiles (e.g. onset of bilingualism, premorbid language dominance) considerably affected the nature and extent of their impairments.

The division of labor in explanations of verb phrase ellipsis

In this paper, we will argue that, of the various grammatical and discourse constraints that affect acceptability in verb phrase ellipsis (VPE), only the structural parallelism constraint is unique to VPE. We outline (previously noted) systematic problems that arise for classical structural accounts of VPE resolution, and discuss efforts in recent research on VPE to reduce explanations of acceptability in VPE to general well-formedness constraints at the level of information structure (e.g. Kehler in Linguist Philos 23(6):533–575, 2000; Coherence, reference and the theory of grammar, CSLI Publications, Stanford, 2002; Proceedings of semantics and linguistic theory, vol 25, 2015; Kertz in Language 89(3):390–428, 2013). In two magnitude estimation experiments, we show that—in line with Kehler’s predictions—degradation due to structural mismatch is modulated by coherence relation. On the other hand, we consistently find residual structural mismatch effects, suggesting that the interpretation of VPE is sensitive to structural features of the VPE antecedent. We propose that a structural constraint licenses VPE, but that sentences violating this constraint can nevertheless be interpreted. The variability in acceptability is accounted for not by additional constraints on VPE in the grammar, but by the numerous general biases that affect sentence and discourse well-formedness, such as information structural constraints (as proposed by Kertz 2013), discourse coherence relations (Kehler 2000), sensitivity to Question Under Discussion structure (e.g. Ginzburg and Sag in English interrogative constructions, CSLI Publications, Stanford, 2000; Kehler 2015), and thematic role bias at the lexical level (e.g. McRae et al. in J Mem Lang 38:283–312, 1998). We test the prediction that thematic role bias (Experiment 3) and QUD structure (Experiment 4) will influence both elliptical and non-elliptical sentences alike, while structural mismatch continues to degrade elliptical sentences alone. Our proposal differs from existing proposals in cutting the explanatory pie in a different way with respect to how variations in acceptability are accounted for. We suggest that degradation can result from at least two distinct and separable sources: violating construction-specific grammatical constraints, or from complexity differences in interpretation related to very general discourse level information.

On the role of discourse-level information in second-language sentence processing

Based on a detailed review of existing studies of high-proficiency second-language (L2) learners who acquired the L2 in adolescence/adulthood, Cunnings (Cunnings, 2016) argues that Sorace's (2011) Interface Hypothesis (IH) and Clahsen and Felser's (2006) Shallow Structure Hypothesis (SSH) do not explain the existing data as well as his memory-based approach which posits that memory-retrieval processes in the L1 and L2 do not pattern alike. Cunnings proposes that L1 and L2 processing differ in terms of comprehenders’ ability to retrieve from memory information constructed during sentence processing. He concludes that L2 processing is more susceptible to interference effects during retrieval, and, most relevantly for this commentary, that discourse-based cues to memory retrieval are more heavily weighted in L2 than L1 processing.

Integration and maintenance of gradient acoustic information in spoken language processing

Models of speech processing seek to explain how continuous acoustic input is mapped onto discrete symbols at various levels of representation, such as phonemes, words, and referents. While recent work has supported models that posit maintenance of fine-grained information, it is not clear how continuous, low-level information in the speech signal is integrated with discrete, higher-level linguistic information. To investigate this, we created acoustic continua between the pronouns “he” and “she” by manipulating the amplitude of frication in the initial phoneme. Using the visual world eye-tracking paradigm, listeners viewed scenes containing male and female referents and heard sentences containing a pronoun, which later disambiguated to a single referent. Measures of eye-gaze revealed immediate sensitivity to both graded acoustic information and discourse-level information. Moreover, when listeners made an initially incorrect interpretation of the referent, recovery time varied as a function of acoustic st...

Integrating meta-information into recurrent neural network language models

Due to their advantages over conventional n-gram language models, recurrent neural network language models (rnnlms) recently have attracted a fair amount of research attention in the speech recognition community. In this paper, we explore one advantage of rnnlms, namely, the ease with which they allow the integration of additional knowledge sources. We concentrate on features that provide complementary information w.r.t. the lexical identities of the words. We refer to such information as meta-information. We single out three cases and investigate their merits by means of N-best list re-scoring experiments on a challenging corpus of spoken Dutch (referred to as cgn) as well as on the English Wall Street Journal (wsj) corpus. First, we look at Parts of Speech (POS) tags and lemmas, two sources of word-level linguistic information that are known to make a contribution to the performance of conventional language models. We confirm that rnnlms can benefit from these sources as well. Second, we investigate socio-situational settings (ssss) and topics, two sources of discourse-level information that are also known to benefit language models. ssss are present in the cgn data, and can be seen as a proxy for the language register. For the purposes of our investigation, we assume that information on the sss can be captured at the moment at which speech is recorded. Topics, i.e., treatments of different subjects, are present in the wsj data. In order to predict POS, lemmas, sss and topic, a second rnnlm is coupled to the main rnnlm. We refer to this architecture as a recurrent neural network tandem language model (rnntlm). Our experimental findings show that if high-quality meta-information labels are available, both word-level and discourse-level information improve performance of language models. Third, we investigate sentence length and word length (i.e., token size), two sources of intrinsic information that are readily available for exploitation because they are known at the time of re-scoring. Intrinsic information has been largely overlooked by language modeling research. The results of both experiments on cgn data and wsj data show that integrating sentence length and word length can achieve improvement. rnnlms allow these features to be incorporated with ease, and obtain improved performance.

Reflexive anaphor resolution in spoken language comprehension: structural constraints and beyond.

We report results from an eye-tracking during listening study examining English-speaking adults’ online processing of reflexive pronouns, and specifically whether the search for an antecedent is restricted to syntactically appropriate positions. Participants listened to a short story where the recipient of an object was introduced with a reflexive, and were asked to identify the object recipient as quickly as possible. This allowed for the recording of participants’ offline interpretation of the reflexive, response times, and eye movements on hearing the reflexive. Whilst our offline results show that the ultimate interpretation for reflexives was constrained by binding principles, the response time, and eye-movement data revealed that during processing participants were temporarily distracted by a structurally inappropriate competitor antecedent when this was prominent in the discourse. These results indicate that in addition to binding principles, online referential decisions are also affected by discourse-level information.

Prosody and information structure in a tone language: an investigation of Mandarin Chinese

Prosody conveys discourse-level information, but the extent to which prosodic cues distinguish different kinds of information-structural concepts remains unclear. The prosodic encoding of information structure is even more complicated in tone languages, where acoustic cues such as F0, intensity and duration also distinguish lexical items (e.g. Mandarin). Prior work on Mandarin led to divergent findings regarding whether and what prosodic cues mark the distinctions between information-structural types. We conducted a production study on Mandarin to investigate whether (1) the presence/absence of corrective focus and (2) the distinction between new/given information are encoded prosodically. Our results show that correctiveness was reflected in all three acoustic parameters: corrective words had longer durations, larger F0 ranges and larger intensity ranges than non-corrective words. The new-given distinction was reflected only in lengthening and F0 range expansion, and only in the absence of correction (correctiveness-by-givenness interaction). This suggests that new information is encoded differently from corrective focus in Mandarin: only corrective focus is associated with intensity range expansion. Our results provide further evidence for the multi-functionality of acoustic-prosodic dimensions. Even in a language with lexical tones, which differ in F0, intensity and duration, all these dimensions also encode information structure. Furthermore, not only can prosodic cues indicate discourse importance, they also distinguish different types of information structure in Mandarin. Our findings highlight the fine-grained ability of the language production system to utilise different aspects of acoustic dimensions with great efficiency.

Children’s processing of reflexives and pronouns in English: Evidence from eye-movements during listening

This study examined how 6–9 year-old English-speaking children and adults establish anaphoric dependencies during auditory sentence comprehension. Using eye-movement monitoring during listening and a corresponding sentence–picture judgment task, we investigated both the ultimate interpretation and the online processing of reflexives in comparison to non-reflexive pronouns, focusing on how binding constraints interact with a competitor antecedent’s relative (discourse) prominence. Whilst our offline results show that the children’s ultimate interpretation for reflexives was constrained by binding principles in the same way as adults’, the eye-movement data revealed that during processing, children were temporarily more distracted than adults when multiple cues supported a prominent competitor antecedent. These results indicate that in addition to binding principles, children’s online referential decisions are also affected by discourse-level information. We suggest that the observed child/adult differences stem from children’s greater difficulty, compared to adults, in controlling multiple sources of information during sentence comprehension.

Referential context effects in L2 ambiguity resolution: Evidence from self-paced reading

This study investigates the degree to which referential context information influences structural ambiguity resolution preferences in non-native sentence comprehension, using both an off-line questionnaire and an on-line self-paced reading task. The critical target sentences contained prepositional phrases (PPs) modifying either the verb phrase (VP) or the preceding noun phrase (NP), as in Bill glanced at the customer with strong suspicion (with ripped jeans). These were embedded within short context paragraphs providing either one or two potential referents for the postverbal NP. The results showed that native Chinese-speaking learners of English and native English speakers were affected differently by referential information in the on-line task. The learners’ reading times of the critical PP were influenced significantly by the referential context, with VP-modifying items being read faster than NP-modifying ones in a VP-supporting context, and the reverse pattern seen in NP-supporting contexts. The native speakers’ ambiguity resolution preferences, on the other hand, were modulated by the referential context in the off-line task only. Our results indicate that non-native comprehenders are highly sensitive to extra-sentential discourse-level information during processing even at intermediate levels of proficiency, a finding that provides a challenge for ‘processing capacity limitation’ accounts for non-targetlike L2 performance.

The contributions of lexico-semantic and discourse information to the resolution of ambiguous categorical anaphors

The present studies employed event-related potentials (ERPs) to examine the time course for the integration of lexico-semantic and discourse information during the resolution of categorical anaphors. Scenarios were constructed to include three potential antecedents. Anaphors were semantically ambiguous in that two of the potential antecedents were exemplars of the anaphor. Final sentences resolved the anaphor with the correct (associatively related/contextually appropriate), incorrect (associatively related/contextually inappropriate), or control antecedent (associatively unrelated/contextually inappropriate). We examined the amplitude of the N400 component, which is thought to reflect the ease of semantic integration, at several points following the anaphor. The smallest N400 was evoked when the text referred back to a correct antecedent following an anaphor; an intermediate N400 was evoked by incorrect antecedents and the largest N400 was evoked by reinstating the control antecedent following an anaphor. Results demonstrated that, following an ambiguous anaphor, readers are able to use both lexico-semantic and discourse-level information to semantically integrate an antecedent into its larger discourse context.

Differential effects of saliency: An event-related brain potential study

We present event-related brain potential evidence from language comprehension that the N400-modulation during noun-phrase integration is a function of the type of referential dependency that is established (identity versus inference) and the saliency (in the following understood as the sum of factors that influence the degree of accessibility of an entity in the mental model) of the information unit that serves as an anchor for the dependency. Identity relations revealed a reduced N400 compared to inferential relations, confirming previous findings. More importantly, the investigation provides novel findings concerning the effect of saliency on noun-phrase integration: identity relations did not reveal a difference in the N400 as a function of the saliency manipulation. In contrast, inferential relations showed a more pronounced N400 with less salient anchors compared to inferential relations that involved a highly salient anchor. Moreover, no difference in N400-amplitude obtained between less salient inference-based entities and new information units. These data suggest that inferential processes are blocked in the absence of a highly accessible anchor. The findings generally indicate that the language system is not only sensitive to the kind of dependency that is established, but also to the saliency of the entity that serves as anchor for the dependency. The data further demonstrate that the processing system is responsive to semantic- and discourse-level information during the interpretation of noun-phrases.