Research paper
The N400 and Post-N400 positivity effect in Mandarin classifier-noun congruence: An ERP study

https://doi.org/10.1016/j.jneuroling.2020.100958Get rights and content

Highlights

  • A graded N400 response was elicited by manipulating semantic congruence.

  • Low-cloze plausible nouns elicited a larger frontal-central post-N400 positivity than high-cloze plausible nouns.

  • Implausible nouns elicited a larger central-parietal post-N400 positivity than plausible nouns.

Abstract

The current study employed electrophysiological measures to investigate semantic processing within a local classifier-noun phrase in Chinese. Toward this aim, semantic congruence between classifiers and nouns was manipulated, resulting in four conditions: (i) a strong-constraint/high-cloze plausible (SP) condition, (ii) a weak-constraint/low-cloze plausible (WP) condition, (iii) a strong-constraint/implausible (SI) condition, and (iv) a weak-constraint/implausible (WI) condition. Participants were asked to judge the acceptability of the classifier-noun pairs. Our results corroborated previous findings by showing that the processing of classifier-noun congruence is sensitive to manipulations of constraint and cloze probability, as indicated by a gradient N400 effect (SP < WP < SI/WI). In addition, WP (mean cloze score = 0.04) elicited a larger frontal-central positivity than SP (mean cloze score = 0.45) in the time window of 500–900 ms, whereas no such differences were observed between SI and WI. In the same time window, implausible nouns elicited a larger central-parietal positivity than plausible nouns. The results are discussed in terms of a functional dissociation between these two types of post-N400 positivity (PNP): a more frontally distributed PNP associated with successful post-lexical composition/integration, and a more centrally distributed PNP elicited by plausibility-driven reanalysis after failed integration attempts.

Introduction

Semantic processing is a fundamental aspect in human behavior. Investigation of the brain bases of meaning-related cognitive processes has obtained rich research results using the N400, a centroparietal negativity peaking around 400 ms after stimulus onset (Kutas & Federmeier, 2011; Kutas & Hillyard, 1980, 1984). A widespread generalization is that the amplitude of the N400 correlates with the degree of semantic fitness between a word and its preceding context, with larger N400s for a worse semantic fit than for a better semantic fit (Federmeier & Kutas, 1999; Federmeier, Wlotko, De Ochoa-Dewald, & Kutas, 2007).

The degree of semantic fitness between a word and its context can be manipulated by cloze probability and contextual constraint. Cloze probability reflects how predictable a word is in a given context, generally measured as the percentage of normal participants who choose a specific word to complete a sentence, such as “I drink my coffee with sugar and ”, Federmeier et al., 2007; Kutas & Federmeier, 2011; Kutas & Hillyard, 1984). Contextual constraint refers to the strength with which the context leads individuals to predict one particular completion versus a number of different possibilities (Federmeier & Kutas, 1999; Kutas & Federmeier, 2011). It is generally defined as the cloze probability of the highest-probability completion. Both cloze probability and contextual constraint can be measured in offline norming tasks, ranging from 0 to 1 in value. These two terms must be related, but it is unlikely that they are simply identical. For example, cloze probability and contextual constraint may not be dissociated for expected, high-cloze completions, since only strongly constraining contexts lead to high-cloze completions (Federmeier et al., 2007). However, both a strongly constraining and a weakly constraining context can accommodate low-cloze completions, and these completions can be matched for their cloze probability (see the next paragraph for examples).

It has been found that the N400 amplitude increases from target words with high cloze probability to target words with low cloze probability (Federmeier et al., 2007; Kutas & Hillyard, 1980, 1984; Van Berkum, Hagoort, & Brown, 1999; Van Berkum, Zwitserlood, Hagoort, & Brown, 2003), and from a strongly constraining context to a weakly constraining context, when the target word is expected (Federmeier et al., 2007; Kutas & Hillyard, 1980, 1984). In contrast, target words that are unexpected fail to elicit gradient N400 effects when the degrees of cloze probability between strongly and weakly constraining sentences are matched (e.g., an unexpected target word collection elicited equally large N400s to for a strongly constraining context like “He bought her a pearl necklace for her” and a weakly constraining context like “He looked worried because he might have broken his”; mean cloze score in both contexts = 3.1%; examples from Federmeier et al., 2007; critical words are in italics and underlined). Thus, it seems that the N400 difference between expected completions in strongly versus weakly constraining sentences is likely due to the cloze probability difference. Because the amplitude of the N400 does not increase with possible costs associated with the suppression of a strongly expected competitor, the N400 is considered to reflect the benefits of matching between semantic features that are predicted by a context and the semantic features associated with the upcoming target word (Kuperberg, Brothers, & Wlotko, 2020; see Van Petten & Luka, 2012, for a review).

The measure of contextual constraint is, however, limited because sentence contexts with a weak constraint might not only be those that yield a large number of alternative predictions (i.e., a weak constraint sentence frame like “It was time to hang the new” may elicit a set of completions with different frequencies, such as pictures, poster, banner, calendar, clock, etc.) but might also be those for which participants are simply unable to derive a specific prediction (e.g., in a weak constraint sentence frame like “They waited a long time to see the”, participants might not make predictions but rather adopt a “wait and see” strategy). In the latter sentence frame, constraint may affect not the strength of predictions but the likelihood of predictions (for more detailed discussion, see Van Petten & Luka, 2012, p. 180). Instead of calculating the probability of a word as the best completion, we can directly rate the strength of predictions by asking participants to estimate how many nouns could be followed by each context word based on a scale ranging from 1 (could not think of any nouns) to 5 (more than four nouns), as Chou, Huang, Lee, & Lee, 2014 did.

However, mismatched sentence completions elicit two distinct ERP components (Kuperberg et al., 2020): the posterior/parietal post-N400 positivity (p-PNP) and the frontal post-N400 positivity (f-PNP). The p-PNP, also labeled as late positive complexes (LPCs, Juottonen, Revonsuo, & Lang, 1996; Pijnacker, Geurts, van Lambalgen, Buitelaar, & Hagoort, 2010) or sometimes “semantic P600” in sentences that are well-formed but semantically abnormal (Kuperberg et al., 2020; Thornhill & Van Petten, 2012), is a positive-going shift that starts around 500 ms after stimulus onset (see Leckey & Federmeier, 2019; Van Petten & Luka, 2012, for reviews). It should be noted that although semantic P600 and LPC are both elicited by semantic plausibility, whether they share similar underlying mechanisms is still an ongoing debate (for more detailed discussions, see Kuperberg et al., 2020; Leckey & Federmeier, 2019; Van Petten & Luka, 2012). These late positive-going effects have been reported for implausible target words (Juottonen, Revonsuo, & Lang, 1996; Kuperberg, 2007; Van de Meerendonk, Kolk, Chwilla, & Vissers, 2009), with a posterior distribution (Diaz & Swaab, 2007; Kuperberg et al., 2020) and sometimes with a central-parietal distribution (Juottonen, Revonsuo, & Lang, 1996; Pijnacker, Geurts, van Lambalgen, Buitelaar, & Hagoort, 2010; Van Petten & Luka, 2012). In previous studies, the p-PNP was interpreted as reflecting an evaluation of the overall message-level plausibility of sentences (DeLong, Quante, & Kutas, 2014; Juottonen, Revonsuo, & Lang, 1996) or costs induced by recollection-based reanalysis (Van Petten & Luka, 2012). The f-PNP elicited by unexpected but plausible stimuli, on the other hand, is a positivity that appears in the same time window as the p-PNP and is most prominent at frontal electrode sites (DeLong et al., 2014; Delong, Urbach, Groppe, & Kutas, 2011; Federmeier et al., 2007; Federmeier, Kutas, & Schul, 2010; Kutas, 1993; Thornhill & Van Petten, 2012; Van Petten & Luka, 2012). In contrast to the p-PNP, less research has investigated the functional nature of the f-PNP. Nonetheless, different studies have reached inconsistent conclusions. The f-PNP is thought to be related to an inhibition or a revision of pre-activated but not presented completions when a plausible alternative is encountered (DeLong et al., 2014; Kutas, 1993; Thornhill & Van Petten, 2012; Van Petten & Luka, 2012; Wlotko & Federmeier, 2007) or a general, post-lexical revision process (Brothers, Swaab, & Traxler, 2015; Freunberger & Roehm, 2016). The different topographic distributions of the p-PNP and the f-PNP suggest that they may arise from different underlying processes (Thornhill & Van Petten, 2012).

However, previous ERP studies of language comprehension have focused primarily on the global integration of a target word into sentences/discourses. This raises the question of whether the three ERP components (N400, p-PNP, and f-PNP) are also sensitive to semantic manipulations within a simple phrase structure. This is important because phrases and sentences may be associated with different processing demands. Specifically, in contrast to within-phrase processing, in which the building of local structure is determinant, cross-phrase processing may require more cognitive resources to establish, integrate, and monitor the overall meaning of sentences with varied syntactic complexity (Friederici, 2004, 2011).

Previous findings have shown that an N400 effect could be robustly elicited across different levels of syntactic hierarchy, i.e., at the within-phrase level (Hagoort, 2003; Prior & Bentin, 2006) and at the cross-phrase level (Van Berkum et al., 1999, Van Berkum et al., 2003; Zhou et al., 2010; see Kutas & Federmeier, 2011, for a review). In contrast to the N400, ERP studies on the p-PNP that investigated semantic mismatches between constituents at different levels of the syntactic hierarchy have yielded inconsistent results. Some studies did not find a p-PNP effect in semantic-pragmatic anomalies (Kuperberg, Sitnikova, Caplan, & Holcomb, 2003), while other studies suggested biphasic N400/p-PNP effects elicited by deeply implausible target words that violated strong contextual constraints for an alternative interpretation (Friederici & Frisch, 2000; Kuperberg, Choi, Cohn, Paczynski, & Jackendoff, 2010; Van de Meerendonk, Kolk, Vissers, & Chwilla, 2010; Van Petten & Luka, 2006). Moreover, no p-PNP effect was recorded within a local phrase structure such as an adjective-noun mismatch (Hagoort, 2003; Prior & Bentin, 2006).

Like the p-PNP, most previous studies on the f-PNP have focused on cross-phrase processing (Brothers et al., 2015; Delong et al., 2011, 2014; DeLong, Groppe, Urbach, & Kutas, 2012; Federmeier et al., 2010; Thornhill & Van Petten, 2012; see Van Petten & Luka, 2012, for a review). Few studies have examined the f-PNP at the phrase level, except Federmeier et al. (2010), which focused on the semantic integration of single words into the context offered by noun phrases. In their study, Federmeier et al. (2010) built a constrained semantic context through phrases (e.g., “a kind of tree”) and manipulated the cloze probability of the category words with different degrees of typicality (high typicality, low typicality, and incongruence). A larger frontal positive-going amplitude was elicited for low-typicality but plausible target words (e.g., ash) than either high-typicality (e.g., oak) or incongruent (e.g., tin) target stimuli, indicating that the f-PNP is not specific to sentence processing but can appear in a linguistic design in which the context and the target word do not belong to a syntactic unit. This implies that f-PNP effects may not necessarily be related to consequences of additional cognitive processes for cross-phrase level interpretation. However, since Federmeier et al. (2010) examined only f-PNP effects between two separate noun phrases (e.g., between a noun phrase like “a kind of tree” and a noun phrase like ash), it is still unclear whether f-PNP effects could also be triggered at the phrasal level.

The present study aimed to investigate semantic processing within the local phrase structure and how semantic congruence would affect the lexico-semantic processing of target nouns within classifier-noun phrases in Chinese (see Fig. 1A), as reflected by the N400 and the late PNPs. To accomplish this, we examined the processing of Chinese classifier-noun phrases in an acceptability judgment task. Semantic congruence was manipulated within classifier-noun phrases, yielding the following four conditions: (i) a strong-constraint/high-cloze plausible (SP) condition, (ii) a weak-constraint/low-cloze plausible (WP) condition, (iii) a strong-constraint/implausible (SI) condition, and (iv) a weak-constraint/implausible (WI) condition.

On one hand, we examined whether manipulation of semantic congruence would reliably elicit gradient N400 effects. On the other hand, our design allowed us to investigate the neurocognitive mechanisms underlying the late PNPs within a noun phrase. Our hypotheses were as follows. If the f-PNP reflects additional processing associated with post-lexical revision following a constraint context, an f-PNP effect should be observed for low-cloze but plausible nouns (WP) relative to expected, plausible nouns (SP). In addition, if a p-PNP effect is always associated with a reanalysis process triggered by processing difficulties between phrases, it should not be observed for implausible target words within a phrase structure. Otherwise, an enhanced p-PNP was hypothesized for implausible conditions in relative to plausible conditions. Since a p-PNP effect is often elicited by implausible stimuli that violate strong contextual constraints for an alternative interpretation (see Kuperberg, 2007, for a review), we also examined possible p-PNP differences between SI and WI.

Section snippets

Participants

A total of 32 participants (mean age = 22 years, range: 18–30 years; 23 females) were enrolled. All were native Mandarin speakers, undergraduate or graduate students at Shanghai Jiao Tong University. All participants are right-handed, with no reading disabilities and with normal or correct-to-normal vision. Participants were paid 100 RMB for their participation. Signed informed consent was obtained from each participant before the experiment. The experimental protocol complied with the research

Acceptability judgments

With mean acceptability scores of 4.85 (SD = 0.50), 4.73 (SD = 0.70), 1.15 (SD = 0.27) and 1.20 (SD = 0.37) for SP, WP, SI, and WI, respectively, participants reliably accepted plausible classifier-noun pairs and rejected implausible classifier-noun pairs. Wilcoxon signed-rank tests were performed on context effect within the plausible conditions and the implausible conditions separately. There were significant differences between strong constraint and weak constraint in both the plausible (SP

Discussion

The aim of the present study was to understand the online semantic processing of Chinese classifier-noun congruence in four conditions: (i) a strong-constraint/high-cloze plausible (SP) condition, (ii) a weak-constraint/low-cloze plausible (WP) condition, (iii) a strong-constraint/implausible (SI) condition, and (iv) a weak-constraint/implausible (WI) condition. Behavioral measures revealed that WP was less acceptable than SP, and SI was less acceptable than WI. Since the plausibility between

Conclusions

Our results corroborate previous findings that the processing of classifier-noun phrases is sensitive to the manipulation of semantic congruence, showing a gradient N400 effect, i.e., SP < WP < SI/WI. This finding implies that the N400 amplitude can be determined primarily by whether or not lexico-semantic retrieval and integration could be conducted easily. In addition, we found a frontal-central positivity (f-PNP) for low-cloze but plausible target words and a central-parietal positivity

Author contributions

F. Li and X. Hong collaborated in specifying the experimental design. F. Li and Y. Wang collected data. F. Li and X. Hong participated in data analyses. F. Li, X. Hong and Y. Wang contributed to the writing of the article.

Ethics statement

This study was carried out in accordance with the recommendations of the Ethic Committee of School of Foreign Languages, Shanghai Jiao Tong University with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethic Committee of School of Foreign Languages, Shanghai Jiao Tong University.

CRediT authorship contribution statement

Fei Li: Writing - original draft. Xiangfei Hong: Conceptualization, Methodology, Software, Formal analysis, Data curation, Supervision. Yuxia Wang: Writing - review & editing.

Declaration of competing interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

This research was supported by Shanghai Pujiang Program (16PJC053), Shanghai Planning Office of Philosophy and Social Science (2019BYY014), Major Program of National Social Science Foundation of China (18ZDA293), Shanghai Jiao Tong University (YG2017MS44) and Shanghai Mental Health Center (2020-FX-05). We thank all the participants for their contribution to this study. We also thank Michelle M. Schoenecker for the help with language polishing of the manuscript.

References (63)

  • J.C.J. Hoeks et al.

    Seeing words in context: The interaction of lexical and sentence level information during reading

    Cognitive Brain Research

    (2004)
  • C.-R. Huang et al.

    Individuals, kinds and events: Classifier coercion of nouns

    Language Sciences

    (2003)
  • K. Juottonen et al.

    Dissimilar age influences on two ERP waveforms (LPC and N400) reflecting semantic context effects

    Cognitive Brain Research

    (1996)
  • A. Kim et al.

    The independence of combinatory semantic processing: Evidence from event-related potentials

    Journal of Memory and Language

    (2005)
  • H.H.J. Kolk et al.

    Structure and limited capacity in verbal working memory: A study with event-related potentials

    Brain and Language

    (2003)
  • G.R. Kuperberg

    Neural mechanisms of language comprehension: Challenges to syntax

    Brain Research

    (2007)
  • G.R. Kuperberg et al.

    Electrophysiological distinctions in processing conceptual relationships within simple sentences

    Cognitive Brain Research

    (2003)
  • B.J. Luka et al.

    Prospective and retrospective semantic processing: Prediction, time, and relationship strength in event-related potentials

    Brain and Language

    (2014)
  • J. Pijnacker et al.

    Exceptions and anomalies: An ERP study on context sensitivity in autism

    Neuropsychologia

    (2010)
  • K. Steinhauer et al.

    The priming of priming: Evidence that the N400 reflects context-dependent post-retrieval word integration in working memory

    Neuroscience Letters

    (2017)
  • L.A. Stowe et al.

    The sentence wrap-up dogma

    Cognition

    (2018)
  • D.E. Thornhill et al.

    Lexical versus conceptual anticipation during sentence processing: Frontal positivity and N400 ERP components

    International Journal of Psychophysiology: Official Journal of the International Organization of Psychophysiology

    (2012)
  • J.J.A. Van Berkum et al.

    When and how do listeners relate a sentence to the wider discourse? Evidence from the N400 effect

    Cognitive Brain Research

    (2003)
  • C. Van Petten et al.

    Neural localization of semantic context effects in electromagnetic and hemodynamic studies

    Brain and Language

    (2006)
  • C. Van Petten et al.

    Prediction during language comprehension: Benefits, costs, and ERP components

    International Journal of Psychophysiology

    (2012)
  • E.W. Wlotko et al.

    Finding the right word: Hemispheric asymmetries in the use of sentence context information

    Neuropsychologia

    (2007)
  • X. Zhou et al.

    Semantic integration processes at different levels of syntactic hierarchy during sentence comprehension: An ERP study

    Neuropsychologia

    (2010)
  • M. Zirnstein et al.

    Cognitive control ability mediates prediction costs in monolinguals and bilinguals

    Cognition

    (2018)
  • G. Baggio et al.

    Coercion and compositionality

    Journal of Cognitive Neuroscience

    (2010)
  • G. Baggio et al.

    The balance between memory and unification in semantics: A dynamic account of the N400

    Language & Cognitive Processes

    (2011)
  • T. Brothers et al.

    Distinct neural signatures of semantic retrieval and event updating during discourse comprehension

    (2020)
  • Cited by (5)

    • An ERP investigation of morpheme transposition in rapid serial visual presentation

      2022, International Journal of Psychophysiology
      Citation Excerpt :

      Hence, the modulation observed in our study in the P200 component suggested that part of the semantic memory network was activated since the early stages of visual word processing. The well-established N400 component was associated with integrating the semantic properties of the stimuli and mapping whole-word orthographic representations to semantic extraction (Berkum et al., 1999; Grainger and Holcomb, 2009; Li et al., 2021). Several researchers proposed that the smaller N400 reflected a greater degree of semantic activation.

    View full text