Highlights
-
• Native speakers exhibited a larger left anterior N400 and posterior LPC.
-
• Non-native speakers exhibited a larger left-distributed LPC and delayed LPC.
-
• Non-native speaker’s exhibit delayed processing of indirect replies.
-
• The processing delay may be because of cognitive resource limitations.
1. Introduction
To achieve successful communication, listeners must decode not only the literal meaning of a statement but also the intended meaning conveyed by the speaker, a process requiring pragmatic inference. This process is essential for communication as the literal meaning of a statement might differ significantly from its intended meaning. Considering the following example—A: Is the teacher in his office now? B: The teacher’s car is outside the office. In this conversation, Speaker B’s reply is not directly relevant to Speaker A’s question. However, according to Grice’s theory of conversational implicature (Grice, Reference Grice, Cole and Morgan1975), interlocutors adhere to cooperative principles during conversation, including the principle of relevance maximization. Thus, Speaker A might assume that Speaker B is trying to give a relevant reply (i.e., the teacher is in his office now), and infer Speaker B’s intention beyond the literal meaning of the statement.
1.1. Processing of indirect replies
Relevance theory (Sperber & Wilson, Reference Sperber and Wilson1986) posits that recognition of a speaker’s intention is the core of human communication, and that inference plays a key role in understanding intention. To understand a speaker’s intended meaning, listeners must first construct the explicature (what is explicitly said) and then derive further implicatures (additional meanings implied by what is said). Both the construction of explicatures and the derivation of implicatures are driven by relevance principles, that is, searching for an optimally relevant interpretation in a given context, with as little processing effort as possible. Processing effort is related to contextual cues, that is, the number and intensity of contextual signals available for interpretation. Interpreting the intended meaning of an indirect reply requires contextual disambiguation and deriving implicit meanings conveyed by the speaker in a given context. When people interpret a statement, given the lexical retrieval from memory and contextual information, many different assumptions from various sources come to mind. From these assumptions, listeners select the most immediately relevant interpretation with the greatest contextual relevance and the least processing effort.
A range of diverse research designs have been used to study the processing of indirect replies. Using a priming paradigm, Holtgraves (Reference Holtgraves1999) investigated how listeners make sense of indirect replies, demonstrating that both literal and intended meanings were activated in conversation and that pragmatic inference was involved during processing. Neuroimaging work utilizing functional magnetic resonance imaging (fMRI) extends this finding, supporting that detection of literal irrelevance and inferring a speaker’s indirect meaning are also involved (Bašnáková et al., Reference Bašnáková, Weber, Petersson, van Berkum and Hagoort2014; Feng et al., Reference Feng, Wu, Jan, Yu, Jiang and Zhou2017; Feng et al., Reference Feng, Yu and Zhou2021; Jang et al., Reference Jang, Yoon, Lee, Park, Kim, Ko and Park2013; Shibata et al., Reference Shibata, Abe, Itoh, Shimada and Umeda2011; van Ackeren et al., Reference van Ackeren, Smaragdi and Rueschemeyer2016; Zhang, Xu et al., Reference Zhang, Xu, Yang and Yang2023). Several neurophysiological studies using electroencephalography (EEG) to measure event-related potential (ERP) during indirect reply processing reported indirect effects both during early and late processing stages (Guo et al., Reference Guo, Li and Yang2023; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021; Zhang, Pan et al., Reference Zhang, Pan, Yang and Yang2023). In their investigation of the role of working memory (WM) capacity during indirect reply processing, Zhang et al. (Reference Zhang, Zhang, Zhang, Yang and Yang2021) reported that, in high WM span readers, indirect replies elicited earlier ERP effects (i.e., P200 and P300), while in low span readers, only a delayed late positive component (LPC) effect was observed.
While the studies described above focused exclusively on processing indirect replies in a listener’s native language, processing in a second language (L2) has also garnered considerable attention. In a series of studies, Taugchi investigated L2 indirect reply processing, focusing on the relationship between contextual cues, processing load, and conventionality (e.g., Taguchi, Reference Taguchi2005, Reference Taguchi2008a, Reference Taguchi2008b, Reference Taguchi2011; Taguchi et al., Reference Taguchi, Gomez-Laich and Arrufat-Marques2016). Taugchi’s studies used offline comprehension judgment measures to investigate the comprehension of L2 indirect replies, which cannot provide insight into online processing. This limit can be addressed through the measurement of high temporal resolution ERPs, which can identify the time course of indirect reply comprehension during online processing. Additionally, while processing in L1 and L2 have both been investigated, little of this work has focused on how language influences the online processing of indirect replies.
1.2. Differences in pragmatic processing between native and non-native speakers
Both theoretical (Clahsen & Felser, Reference Clahsen and Felser2006a, Reference Clahsen and Felser2006b, Reference Clahsen and Felser2018; Morishima, Reference Morishima2013) and empirical studies on the related pragmatic phenomena (e.g., Foucart, Moreno et al., Reference Foucart, Moreno, Martin and Costa2015; Martin et al., Reference Martin, Thierry, Kuipers, Boutonnet, Foucart and Costa2013; Newman et al., Reference Newman, Tremblay, Nichols, Neville and Ullman2012; Pérez et al., Reference Pérez, Hansen and Bajo2019) have indicated that pragmatic processing differs between L1 and L2. Two hypotheses posit that pragmatic processing differs between native speakers and non-native speakers. The shallow parsing hypothesis states that differences can be attributed to the use of different parsing mechanisms. Relative to natives, non-native speakers employ a “shallow” parsing approach and are more sensitive to semantic and pragmatic information compared with syntactic information, making efficient use of various non-grammatical information for complex language comprehension (Clahsen & Felser, Reference Clahsen and Felser2006a, Reference Clahsen and Felser2006b, Reference Clahsen and Felser2018). The limited cognitive resources hypothesis assumes that L1/L2 processing is similar and that differences can be explained in terms of capacity-based limitations being exceeded during L2 processing (Morishima, Reference Morishima2013). For native speakers, lower-level language processes such as lexical access and parsing are automatized and are thought to consume few cognitive resources. Thus, native speakers are able to allocate their cognitive resources to high-level processing such as discourse comprehension (Harrington & Sawyer, Reference Harrington and Sawyer1992; Perfetti, Reference Perfetti1985). In contrast, for non-native speakers, the construction of surface form and text base are cognitively demanding. As a consequence, relatively fewer resources are left for the construction of higher discourse-level representations, such as semantic and pragmatic representations (Rai et al., Reference Rai, Loschky, Harris, Peck and Cook2011).
Findings from previous neurophysiological studies also provide mixed support for these two hypotheses. Results from studies on sentence and discourse comprehension have found that, for non-native speakers, pragmatic processing elicited earlier ERP effects (i.e., N400) (Foucart, Garcia et al., Reference Foucart, Garcia, Ayguasanosa, Thierry, Martin and Costa2015; Foucart et al., Reference Foucart, Martin, Moreno and Costa2014; Foucart et al., Reference Foucart, Romero-Rivas, Gort and Costa2016). This finding is thought to reflect non-native speakers’ increased sensitivity for and reliance on contextual cues for pragmatic processing. However, other studies reported a delayed or absent N400 effect during pragmatic processing in non-native speakers (Foucart, Moreno et al., Reference Foucart, Moreno, Martin and Costa2015; Martin et al., Reference Martin, Thierry, Kuipers, Boutonnet, Foucart and Costa2013; Newman et al., Reference Newman, Tremblay, Nichols, Neville and Ullman2012; Pérez et al., Reference Pérez, Hansen and Bajo2019), attributed to the reduced cognitive resources available for non-native speakers during L2 processing, biasing them toward a reliance on the less demanding reactive control mode.
1.3. The present study
While a number of studies have investigated differences between native and non-native speakers in pragmatic and discourse processing, whether differences can be identified during online indirect reply comprehension is an open question. Using EEG to measure ERPs, the present study aimed to investigate this question in a sample of native and non-native Mandarin speakers. Direct and indirect reply conversations were written in Chinese, and native and non-native speakers were asked to read each conversation while EEG signals were recorded.
Several relevant ERP components, including P200, N400, LPC, and late negativity, have been reported in previous studies investigating figurative and non-literal language comprehension. The P200 component is thought to reflect initial semantic analysis during pragmatic processing (Regel & Gunter, Reference Regel and Gunter2017; Regel et al., Reference Regel, Gunter and Friederici2011; Regel et al., Reference Regel, Meyer and Gunter2014), while the N400 has been associated with the difficulty of integrating current information with prior contexts (Coulson & Van Petten, Reference Coulson and Van Petten2002; Pynte et al., Reference Pynte, Besson, Robichon and Poli1996; Weiland et al., Reference Weiland, Bambini and Schumacher2014). Finally, both the LPC and late negativity have been related to pragmatic reanalysis of mismatched information (Arzouan et al., Reference Arzouan, Goldstein and Faust2007; Coulson & Van Petten, Reference Coulson and Van Petten2002; Goldstein et al., Reference Goldstein, Arzouan and Faust2012; Regel & Gunter, Reference Regel and Gunter2017; Regel et al., Reference Regel, Gunter and Friederici2011; Spotorno et al., Reference Spotorno, Cheylus, Van Der Henst and Noveck2013; Weiland et al., Reference Weiland, Bambini and Schumacher2014; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021; Zhang, Pan et al., Reference Zhang, Pan, Yang and Yang2023). Based on previous ERP studies investigating indirect reply processing (Guo et al., Reference Guo, Li and Yang2023; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021; Zhang, Pan et al., Reference Zhang, Pan, Yang and Yang2023), we predict that native speakers may proactively process indirect replies, thus both earlier (e.g., P200/N400) and later components (e.g., LPC/late negativity) may be elicited for indirect replies compared with direct replies. For non-native speakers, we predict that heightened sensitivity to pragmatic information, as supported by the shallow parsing hypothesis (Clahsen & Felser, Reference Clahsen and Felser2006a, Reference Clahsen and Felser2006b, Reference Clahsen and Felser2018), will lead to indirect reply effects in earlier ERP components (e.g., P200 and N400). Alternatively, based on the limited cognitive resources hypothesis (Morishima, Reference Morishima2013), indirect reply processing may be delayed because of the demands imposed by L2 use, presenting indirect reply effects in the LPC or late negativity.
2. Methods
2.1. Participants
Thirty-two native Mandarin speakers (16 females, Mage = 22.22, SDage = 2.21), and 30 non-native Mandarin speakers (14 females, Mage = 24.37, SDage = 3.47) participated in this experiment. Non-native speakers were proficient users of Mandarin who had begun studying the language as young adults (Mage = 19.55, SDage = 2.50), continued studying the language for an average of 57.83 months (SD = 26.27), and passed either the HSK 5 or HSK 6 examination. Additionally, non-native speakers had spent an average of 36.58 months living in China (SD = 25.84). Non-native Mandarin speakers reported native languages of Korean (9), Urdu (6), Bengali (4), Russian (3), Samoan (2), Spanish (2), Indonesian (1), Kazakh (1), Nepali (1), and Swahili (1). All participants were undergraduate or graduate students, right-handed, with normal or corrected-to-normal vision, reporting no reading, neurological, or psychiatric disorders. Written informed consent was collected from all participants before the start of the study in accordance with the Declaration of Helsinki. Participants received 100 RMB (~$14 USD) for their participation. Approval for this study was granted by the ethics committee of the Beijing Language and Culture University.
Mandarin proficiency of non-native speakers was assessed before the start of the experiment using a literacy test based on Jiang and Liu (Reference Jiang and Liu2004). During the test, 100 high-frequency Chinese characters were presented, and participants were asked to provide the correct pinyin for each character. Objective proficiency was operationalized as the total number of correct responses given out of 100. The average objective proficiency for non-native speakers was 68.64 (SD = 14.24). Participants were also asked to subjectively rate their Mandarin proficiency from 1 to 7 with 1 representing “not proficient at all” and 7 representing “very proficient.” The average subjective proficiency for non-native speakers was 4.77 (SD = 0.73).
2.2. Design and materials
A 2 × 2 mixed experimental design was adopted with Group (Mandarin L1 and Mandarin L2) as the between-subjects factor and Reply (Direct and Indirect) as the within-subjects factor. Seventy-four written conversations were selected through pre-experimental screening sessions (see below). Each conversation included two speakers (Speaker A and Speaker B), one question (asked by Speaker A), and one reply (given by Speaker B). Speaker B’s reply varied as either direct or indirect depending on Speaker A’s question, resulting in two conditions (see Table 1). All conversations conveyed neutral information and represented everyday topics commonly discussed within Chinese culture.
Table 1. Example of experimental stimuli
Note. Critical sentences (bolded) were divided into three parts with slashes, with the third part as the target segment (underlined). All materials were originally in Chinese. English versions shown above were translated word-by-word for illustrative purposes only.
2.2.1 Experimental material pretests
Experimental conversation stimuli were selected based on a series of pretests that assessed for the potentially confounding influence of indirectness, emotionality, cloze probability, and comprehensibility. Participants in each pretest did not take part in any other pretest or the formal experiment. Average scores or ratings for all measures of interest are reported in Table 2.
Table 2. Pretests results of the stimuli
Note. Data are expressed as mean (standard deviation). Comp: comprehension, Prob: probability
The first pretest was conducted to assess the level of indirectness for each written conversation to select the most appropriate experimental stimuli. Eighteen participants took part in the pretest. We generated 94 written conversation stimuli and divided them into two sets. Conversations in each set were rated by nine participants. During this pretest, participants were asked to read carefully and to rate how directly the given reply answered the question on a 7-point scale (1: most direct; 7: most indirect). If the reply was indirect (scores between 5 and 7), participants were asked to provide their interpretation of the speakers’ intended meaning. We coded participants’ interpretations as either 0 (did not identify what the speaker meant) or 1 (correctly identified what the speaker meant). Experimental conversations used in the formal study were selected according to the following criteria: ratings for direct replies ≤2.5; ratings for indirect replies >4; and, comprehension accuracy >77%. A total of 74 conversations were selected for the formal experiment. Indirect replies were rated as significantly more indirect than direct replies as revealed by a paired-samples t-test, t(73) = 46.8, p < .001.
To test for differences in stimulus emotionality, another pretest was conducted. Twenty participants were asked to rate the valence and arousal conveyed by each written conversation. Valence was rated on a scale from −3 (most negative) to 3 (most positive), and arousal was rated on a scale from 1 (least arousing) to 7 (most arousing). Paired t-tests revealed no significant differences between direct replies and indirect replies for both valence, t(73) = −0.248, p = .805, and arousal ratings, t(73) = 1.14, p = .261.
Next, we tested for differences in cloze probability in a sample of 18 participants. For all conversations, the third part of the target reply (underlined in Table 1) was removed, and participants were asked to complete the reply by providing the first word or phrase that came to their mind. No significant difference in cloze probability was identified between direct and indirect reply conditions, t(73) = −0.93, p = .356.
Finally, we tested for differences in the comprehensibility of Speaker A’s question. Questions were presented randomly and 16 participants were asked to rate the degree of comprehensibility on a 7-point scale (1: most incomprehensible; 7: most comprehensible). No significant difference in question comprehensibility was found between direct and indirect reply conditions, t(73) = 0.455, p = .650.
2.3. Procedure
Data collection occurred in a sound-attenuating, shielded room at Beijing Language and Culture University. Participants were seated in a comfortable chair 60 cm in front of a computer screen (43.18 cm) on which all experimental stimuli were presented. Experimental trials were presented using E-Prime 2.0 (Psychology Software Tools, Pittsburg, PA, USA). Each trial began with the presentation of a central fixation cross (+) for 1000 ms. Immediately after, Speaker A’s question was presented. Participants were asked to press the space bar once they had finished reading the question to advance to the next screen. Next, Speaker B’s reply (critical sentence) was presented as three individual segments (see Table 1), each on a separate screen. The first two segments were presented for 500 ms each, followed by a blank screen for 200 ms. The third segment (target segment) was presented for 600 ms, followed by a period (500 m), and then a blank screen for 200 ms. The third segment in the critical sentence was crucial for deriving the intended meaning of the indirect replies. Accurate comprehension required not only processing the new information contained in the third segment but also contextualizing it based on information contained in the two previous segments and Speaker A’s initial question. To enhance comprehension of the meanings conveyed in experimental conversations, the duration of the third segment was extended longer than the preceding two segments in accordance with practices reported in previous studies (Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021, Reference Zhang, Pan, Yang and Yang2023; Guo et al., Reference Guo, Li and Yang2023).
The 74 experimental conversations consisted of 37 different conversations written using both a direct or indirect reply, resulting in 74 conversations in total. Experimental conversations were divided into two sets ensuring that each conversation was only presented once per set, and sets were counterbalanced across participants and groups. Both sets contained 37 conversations for each condition. In the direct reply condition, Speaker A first provided a what/why/where question, while a yes/no question was given in the indirect condition. To balance the question type and to prevent participants from generating predictions of particular reply types based on Speaker A’s question, we added 60 filler conversations using the opposite question type, that is, yes/no questions followed by direct replies and what/why/where questions followed by indirect replies. This resulted in 134 total conversations for each set of stimuli. After reading each conversation, participants were asked to complete a comprehension judgment task (yes or no judgment based on whether they understood the written conversation). Participants were instructed to press “J” for “yes” and “F” for “no.” Trials on which participants indicated they did not understand the conversation were removed before analysis.
Experimental stimuli were divided into two blocks, each containing 67 conversations. Each block took around 10 minutes to complete, and participants were given a short break between blocks. Conversations were presented in a pseudo-random order to ensure that stimuli from a given condition were not presented more than three times in succession. Before the formal experiment, participants completed six trials as a practice to familiarize themselves with the procedure. Practice stimuli were different from those used in the formal experiment, but were identical in structure.
Accuracy and reaction time data from behavioral trials were analyzed using R software (version 4.2.0; R Core Team, 2023). To test for main effects and interactions, repeated measures ANOVA with Group (Mandarin L1 and Mandarin L2) as the between-subjects factor, Reply (Direct and Indirect) as the within-subjects factor, and the interaction between Group and Reply were used. Significant interactions were followed up with appropriate t-tests for pairwise comparison.
2.4. Electrophysiological recording and analysis
Electrophysiological data were recorded with 64 Ag/AgCl electrodes mounted on an elastic cap (10–20 system), using a SynAmps amplifier and Curry 7 Acquisition Software (Neuroscan Inc, USA). Data were collected at a sampling rate of 500 Hz with a bandpass filter of 0.05–100 Hz. Electrodes were referenced online to the left mastoid and re-referenced offline to the average of both left and right mastoids. An electrode between Fz and FPz electrodes served as the ground. Horizontal electrooculograms (HEOG) were monitored through two electrodes placed at the outer canthus of each eye. Vertical electrooculograms (VEOG) were recorded via two electrodes above and below the left eye. Electrode impedances were maintained below 5 kΩ during the whole experiment.
EEG data were preprocessed using EEGLAB (EEGLAB 14.1.1b, http://www.sccn.ucsd.edu/eeglab), an open-source MATLAB toolbox (Delorme & Makeig, Reference Delorme and Makeig2004). Continuous EEG data were band-pass filtered (0.05–30 Hz) and segmented from 200 ms (−200 ms) before the onset of the third segment of the critical sentence to 1300 ms after the onset of the third segment of the critical sentence. The mean amplitude from −200 to 0 ms served as a baseline for each trial. Independent component analysis (ICA) was applied to epoched data to isolate and remove components representing ocular or other artifacts. Trials contaminated with artifacts (excessive amplitude exceeding ±80 μV) were removed. For each participant, average ERPs were calculated for each trial condition at each electrode. Grand average ERPs for each group and condition were calculated using all included trials.
For ERP statistical analysis, cluster-based random permutation tests were conducted (implemented in FieldTrip, Maris & Oostenveld, Reference Maris and Oostenveld2007). This nonparametric statistical procedure uses a cluster method to correct for multiple comparisons, effectively controlling for the overall Type I error rate when no prior information is available to guide the choice of latency windows and electrode sites (Luck & Gaspelin, Reference Luck and Gaspelin2017). The permutation test was performed within 0–1300 ms post-the third segment onset (in 2-ms steps) over 60 electrodes (FP1/FP2, FPZ, AF3/AF4; F7/F8, F5/F6, F3/F4, F1/F2, FZ; FT7/FT8, FC5/FC6, FC3/FC4, FC1/FC2, FCZ; T7/T8, C5/C6, C3/C4, C2/C1, CZ; TP7/TP8, CP5/CP6, CP3/CP4, CP2/CP1, CPZ; P7/P8, P5/P6, P3/P4, P1/P2, PZ; PO7/PO8, PO5/PO6, PO3/PO4, POZ; O1/O2, OZ). A simple dependent t-test was performed on each data point (“electrode by time”) comparing two conditions (e.g., Direct vs. Indirect reply). All adjacent data points exceeding a preset significance level (p = .05) were grouped into clusters. Cluster-level statistics were calculated by taking the sum of the t-values for every cluster. The significance probability of the clusters was calculated using the Monte Carlo method with 1000 random draws.
3. Results
3.1. Behavioral results
Average comprehension rates and response times for direct and indirect replies are reported in Table 3. For the Mandarin L1 group, average comprehension was 96.45% (SD = 3.08%), indicating that experimental conversations were well understood. Average comprehension for the Mandarin L2 group was 84.36% (SD = 4.05%), indicating that participants understood most of the conversations used in the present study. Statistical analysis of comprehension data revealed significant main effects of Group [F(1, 56) = 103.14, p < .001] and Reply [F(1, 56) = 118.49, p < .001], as well as a significant interaction between Group and Reply [F(1, 56) = 58.70, p < .001]. Pairwise comparisons revealed that, while comprehension of indirect replies was lower than that of direct replies for both the Mandarin L1, t(31) = 7.72, p < .001, and Mandarin L2 groups, t(26) = 36.37, p < .001, overall comprehension was lower in the Mandarin L2 group (Indirect: t(57) = 10.44, p < .001; Direct: t(57) = 5.42, p < .001).
Table 3. Mean (SD) comprehension and reaction time (RT) for replies
Before analysis, reaction time data underwent a logarithmic transformation to address issues with non-normality. Analyses revealed significant main effects of Reply [F(1, 56) = 39.66, p < .001] and Group [F(1, 56) = 121.55, p < .001], indicating that participants’ responses to questions following indirect replies were faster compared with direct replies, and that the Mandarin L1 group faster than the Mandarin L2 group. The interaction between Reply and Group was not significant [F(1, 56) = 0.038, p = .846].
3.2. ERP results
Because of issues with EEG quality, data for one participant from the Mandarin L1 and three participants from the Mandarin L2 group were removed before analysis. This resulted in the inclusion of 31 participants for the Mandarin L1 group and 27 participants for the Mandarin L2 group. Pre-analysis trial removal for comprehension and EEG artifact contamination issues resulted in the inclusion of an average of 35 direct reply and 34 indirect reply condition trials for the Mandarin L1 group, and 30 direct reply and 24 indirect reply condition trials for the Mandarin L2 group. Figures 1 and 2 display grand average ERP waveforms generated by the critical segment across different conditions and groups. Topographic distributions of the indirect reply effect in both groups are shown in Figure 3.
Figure 1. Grand average waveforms elicited by direct (red line) and indirect reply (blue line) conditions in native Mandarin speakers. Nine electrodes are presented for nine regions: left-anterior (F5), left-central (C5), left-posterior (P5), middle-anterior (Fz), middle-central (Cz), middle-posterior (Pz), right-anterior (F6), right-central (C6), and right-posterior (P6).
Figure 2. Grand average waveforms elicited by direct (red line) and indirect reply (blue line) conditions in non-native Mandarin speakers. Nine electrodes are presented for nine regions: left-anterior (F5), left-central (C5), left-posterior (P5), middle-anterior (Fz), middle-central (Cz), middle-posterior (Pz), right-anterior (F6), right-central (C6), and right-posterior (P6).
Figure 3. Topographies of indirect reply effect (Indirect reply vs. Direct reply) in native (upper) and non-native (lower) Mandarin speakers.
For the Mandarin L1 group, compared with direct replies, indirect replies elicited a larger negativity from 314 to 500 ms after critical segment onset (negcluster, p = .007), and a larger positivity in the 692 to 1300 ms time window (poscluster, p = .047). The negativity effect was over anterior regions (electrodes: FP1, FPz, FP2, AF3, AF4, F7, F5, F3, F1, Fz, FT7, FC5, FC3, and FC1), with a left hemisphere dominance, whereas the positivity effect was over posterior regions (electrodes: P7, P5, P3, P1, Pz, P4, P6, P8, PO7, PO5, PO3, POz, PO4, PO6, PO8, O1, Oz, and O2).
For the Mandarin L2 group, indirect replies, relative to direct replies, elicited a larger positivity in the 522–966 ms time window (poscluster 1, p = .029), and in the time window 984–1176 ms (poscluster 2, p = .043). For the time window from 522 to 966 ms, the positivity effect was mainly distributed in central and posterior regions in the left hemisphere and the midline (electrodes: C1, C3, C5, Cz, T7, TP7, CP5, CP3, CPz, P7, P5, P3, P1, Pz, PO7, PO5, PO3, POz, O1, and Oz). For the time window from 984 to 1176 ms, the positivity effect was distributed across the whole cortex (electrodes: FP1, FPz, FP2, F4, FC3, FC5, FCz, C5, C3, C1, Cz, TP7, CP5, CP3, CP1, CPz, CP2, CP4, P7, P5, P3, P1, Pz, PO7, PO5, PO3, POz, and O1).
Given that the present study aimed to examine differences between native and non-native speakers during indirect reply comprehension, we conducted cluster-based permutation tests to examine whether Reply and Group interacted in the significant time windows identified during our initial ERP analysis. Accordingly, amplitudes at significant electrodes for each time window were averaged and 2 × 2 repeated-measure ANOVAs were performed, followed by pairwise comparisons to probe significant interactions and main effects.
For the time window from 314 to 500 ms (N400), a significant interaction between Group and Reply was identified [F(1, 56) = 5.86, p = .023, ηp2 = .184]. Pairwise comparisons revealed that, for the Mandarin L1 group, N400 amplitude of indirect replies was significantly higher than direct replies, t(30) = 3.64, p = .001. However, this finding was not observed in the Mandarin L2 group, t(26) = 1.09, p = .286.
For the time window from 522 to 966 ms (LPC 1), no significant interaction between Group and Reply was identified [F(1, 56) = 1.00, p = .327, ηp2 = .037]. Main effects of Group [F(1, 56) = 45.14, p < .001, ηp2 = .635] and Reply [F(1, 56) = 5.00, p = .034, ηp2 = .161] were significant. Pairwise comparisons showed that for the Mandarin L1 group, LPC amplitude for indirect and direct replies did not differ, t(30) = 0.49, p = .627. Conversely, LPC amplitude of indirect replies was higher than that of direct replies in the Mandarin L2 group, t(26) = 2.43, p = .022.
For the time window from 692 to 1300 ms (LPC 2), no significant interaction between Group and Reply was identified [F(1, 56) = 0.360, p = .554, ηp2 = .014]. However, main effects of Group [F(1, 56) = 4.51, p = .043, ηp2 = .148] and Reply [F(1, 56) = 7.67, p = .010, ηp2 = .228] were significant. Pairwise comparisons revealed that, for the Mandarin L1 group, LPC amplitude of indirect replies was higher than that of direct replies, t(30) = 2.10, p = .044. However, this finding was not observed in the Mandarin L2 group, t(26) = 1.50, p = .146.
Finally, for the time window from 984 to 1176 ms (delayed LPC), no significant interaction between Group and Reply was identified [F(1, 56) = 0.76, p = .391, ηp2 = .028]. The main effect of Group was significant [F(1, 56) = 29.39, p < .001, ηp2 = .531], but the main effect of Reply was not [F(1, 56) = 3.10, p = .090, ηp2 = .106]. Pairwise comparisons revealed that, for the Mandarin L1 group, the delayed LPC amplitude of indirect replies was not different from that of direct replies, t(30) = 0.53, p = .599. However, the delayed LPC amplitude of indirect replies was larger than that of direct replies in the Mandarin L2 group, t(26) = 2.12, p = .043.
In summary, we found that the interaction between Group and Reply was significant at the N400 time window, but not in the time windows of LPCs. Complete statistical test results can be found in Table 4.
Table 4. Statistical analysis of ERP data at the reported time windows
Note. “IN-DI” represents the difference between indirect replies (IN) and direct replies (DI). MND: Mandarin ** p < 0.01; * p < 0.05.
4. Discussion
In the present study, we investigated differences between native and non-native Mandarin speakers in the processing of indirect replies in written Mandarin conversations. Results showed that for native Mandarin speakers, indirect replies elicited both higher left anterior N400s and higher posterior LPCs than direct replies, whereas non-native speakers presented with higher globally distributed LPCs and higher delayed LPCs. These results support that native and non-native Mandarin speakers differ in the time course of indirect reply comprehension.
4.1. Processing of indirect replies by native Mandarin speakers
For native speakers, a left anterior-distributed N400 effect was found during indirect reply comprehension. In previous studies on figurative language comprehension, N400 effects have been associated with contextual integration of pragmatic meanings (Bambini et al., Reference Bambini, Canal, Resta and Grimaldi2019; Coulson & Van Petten, Reference Coulson and Van Petten2002; Lai & Curran, Reference Lai and Curran2013; Lai et al., Reference Lai, Curran and Menn2009; Pynte et al., Reference Pynte, Besson, Robichon and Poli1996; Weiland et al., Reference Weiland, Bambini and Schumacher2014), the retrieval of semantic information and construction of action meaning (Amoruso et al., Reference Amoruso, Gelormini, Aboitiz, Alvarez González, Manes, Cardona and Ibanez2013; van Elk et al., Reference van Elk, van Schie and Bekkering2010), and the degree to which WM is taxed (for reviews, see Coulson, Reference Coulson, Noveck and Sperber2004). In the present study, the anterior N400 effect for indirect replies could be associated with retrieving semantic information from memory and contextualizing it to immediately construct meaning. Thus, the anterior N400 effect of indirect replies could be linked to the immediate integration of the statement with context and world knowledge in a non-literal way for relevant interpretations.
The anterior-distributed negativity effect has also been established as an indicator of prediction generation and maintenance (Boux et al., Reference Boux, Tomasello, Grisoni and Pulvermüller2021; Grisoni et al., Reference Grisoni, Mohr and Pulvermüller2019; Li et al., Reference Li, Zhang, Xia and Swaab2017; Otten & Van Berkum, Reference Otten and Van Berkum2008). In the present study, the observed negativity effect might be related to prediction processing for indirect replies. Native speakers may proactively process upcoming information during conversations, generating predictions about what a speaker is likely to say (Grüter & Rohde, Reference Grüter and Rohde2013; Grüter et al., Reference Grüter, Rohde and Schafer2014; Pérez et al., Reference Pérez, Hansen and Bajo2019). During the processing of indirect replies, this tendency to make predictions may support more efficient detection of the literal irrelevance of a statement in a given context, consequently generating an early effect for indirect reply comprehension.
Following the N400 effect, an enhanced LPC was observed for indirect replies. Previous studies have linked the LPC effect to pragmatic analysis (Coulson & Van Petten, Reference Coulson and Van Petten2002; Regel et al., Reference Regel, Coulson and Gunter2010; Regel & Gunter, Reference Regel and Gunter2017; Regel et al., Reference Regel, Gunter and Friederici2011; Spotorno et al., Reference Spotorno, Cheylus, Van Der Henst and Noveck2013; Weiland et al., Reference Weiland, Bambini and Schumacher2014) and efforts in constructing the situation model (Coulson & Lovett, Reference Coulson and Lovett2010; Leuthold et al., Reference Leuthold, Filik, Murphy and Mackenzie2012). For indirect replies, readers may combine the upcoming information with previous contexts and world knowledge to make pragmatic inferences and reanalyze the intended meaning conveyed by a speaker. Thus, the late positivity effect might reflect that a detailed and elaborate situation model was built for indirect replies.
The biphasic N400-LPC pattern observed in the present study has been reported in previous investigations of figurative language in the comprehension of metaphors (Coulson & Van Petten, Reference Coulson and Van Petten2002; Weiland et al., Reference Weiland, Bambini and Schumacher2014), jokes (Coulson & Kutas, Reference Coulson and Kutas2001), or unfamiliar irony (Filik et al., Reference Filik, Leuthold, Wallington and Page2014). The N400 effect possibly indicates that non-literal language comprehension requires more semantic efforts than literal language comprehension. For non-literal language comprehension, it may be necessary to retrieve additional semantic information from contexts or long-term memory, leading to a larger N400 (Brouwer et al., Reference Brouwer, Fitz and Hoeks2012; Coulson & Van Petten, Reference Coulson and Van Petten2002). The LPC effect could also reflect pragmatic processing to derive non-literal meaning (Coulson & Van Petten, Reference Coulson and Van Petten2002; Filik et al., Reference Filik, Leuthold, Wallington and Page2014; Weiland et al., Reference Weiland, Bambini and Schumacher2014). For non-literal language comprehension, readers may expend greater efforts to integrate the retrieved information into mental representations than during literal language comprehension.
4.2. Processing of indirect replies by non-native Mandarin speakers
For non-native speakers, an enhanced LPC mainly distributed in left-central and left-posterior regions, and an enhanced delayed positivity were observed during indirect reply processing relative to direct reply processing. These findings support that indirect meanings were accessed by non-native speakers at a later stage. This finding aligns with previous studies indicating that non-native speakers have more difficulty during pragmatic processing than native speakers (e.g., Antoniou, Reference Antoniou and Taguchi2019), as enhanced LPC has been linked to pragmatic analysis (Coulson & Van Petten, Reference Coulson and Van Petten2002; Regel & Gunter, Reference Regel and Gunter2017; Regel et al., Reference Regel, Gunter and Friederici2011; Weiland et al., Reference Weiland, Bambini and Schumacher2014; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021; Zhang, Pan et al., Reference Zhang, Pan, Yang and Yang2023). Given the decreased cognitive resources available for processing intended meanings of indirect replies (Morishima, Reference Morishima2013), it is also reasonable that non-native speakers are slower in detecting differences between indirect and direct replies. Despite significant differences in time course, neurophysiological findings, together with our behavioral results, support that non-native Mandarin speakers are capable of deriving meaning from indirect replies, albeit at a later stage when compared with native speakers.
Interestingly, we observed that the LPC effect for non-native Mandarin speakers was mainly distributed in the left hemisphere. LPC effects observed in the central posterior area of the left hemisphere were previously reported (Peters & Daum, Reference Peters and Daum2009; Wilding & Rugg, Reference Wilding and Rugg1996; Zhang et al., Reference Zhang, Lu, Liang and Chen2020), which has been related to the encoding strength of memory (Paller & Wagner, Reference Paller and Wagner2002), retrieval of general (Peters & Daum, Reference Peters and Daum2009), and episodic memory (Rugg & Curran, Reference Rugg and Curran2007), and controlled semantic retrieval (Martin et al., Reference Martin, Dering, Thomas and Thierry2009). In the present study, non-native speakers may be applying more effort in general memory retrieval during online indirect reply processing. When target stimuli were presented, the information preceding it disappeared, requiring readers to retain the question asked by Speaker A in their WM and to retrieve world knowledge related to the conversation from long-term memory. These processes may be more challenging for non-native speakers but are still necessary for successful comprehension of indirect replies in a second language.
Additionally, for non-native speakers, we found an enhanced delayed positivity for indirect replies in comparison to direct replies. The enhanced delayed positivity has also been previously reported (Jiang & Pell, Reference Jiang and Pell2016a, Reference Jiang and Pell2016b; Jiang & Zhou, Reference Jiang and Zhou2015; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021). This electrophysiological result is thought to reflect attempts to infer other’s intentions and goals, as well as late pragmatic inference for deriving intended meanings. The higher delayed positivity elicited by indirect replies observed in the present study may be associated with pragmatic inference for deriving speakers’ meanings and communication intention.
The absence of an N400 effect for indirect replies is consistent with other ERP studies on sentence and discourse comprehension (Foucart, Moreno et al., Reference Foucart, Moreno, Martin and Costa2015; Martin et al., Reference Martin, Thierry, Kuipers, Boutonnet, Foucart and Costa2013; Newman et al., Reference Newman, Tremblay, Nichols, Neville and Ullman2012; Pérez et al., Reference Pérez, Hansen and Bajo2019). Given differences in cognitive resources available and prediction ability, non-native speakers might adopt a less demanding reactive processing strategy for conversation comprehension, unlike native speakers’ proactively processing strategy (Grüter & Rohde, Reference Grüter and Rohde2013; Grüter et al., Reference Grüter, Rohde and Schafer2014; Pérez et al., Reference Pérez, Hansen and Bajo2019). Consequently, the reduced processing efficiency of non-native speakers may present only in later-stage components. The absence of an N400 effect may also reflect heightened susceptibility to retrieval interference in non-native speakers (Cunnings, Reference Cunnings2017a, Reference Cunnings2017b). During the processing of indirect replies, both contextual cues and world knowledge stored in memory are important for identifying the most relevant interpretation. As non-native speakers are prone to retrieval interference, it is possible they could not efficiently use various information to access maximum relevant interpretations for indirect replies. This bears a resemblance to the processing of indirect replies in readers with low WM span (Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021). When comprehending indirect replies, low-span readers may not benefit from default knowledge, resulting in a delayed understanding of what a speaker is trying to say. In the present study, during indirect reply processing, non-native speakers may decode speakers’ meanings for coherent comprehension without initial analysis at an early stage.
4.3. Differences between native and non-native Mandarin speakers
Non-native Mandarin speakers performed worse than native speakers during online indirect reply processing, reflected by a significant interaction between Reply and Group on comprehension accuracy. These findings contrast with those in previous studies using offline multiple-choice tests which reported native-like processing of English direct replies in highly proficient non-native speakers who had relocated to the USA for an extended period of time (Bouton, Reference Bouton1992, Reference Bouton1994). Our findings also differ from those reported in more recent work using a self-paced reading paradigm (Zhang et al., Reference Zhang, Huang, Wang, Yang, Pan and Shenunder review), in which no significant interaction between Reply and Group on comprehension accuracy was identified. Under these conditions, which lack the time constraints present during online reading tasks like the one used in the present study, non-native speakers perform similarly to native speakers when reading indirect replies. These conflicting patterns of results support that differences between native and non-native Mandarin speakers may only manifest under certain conditions.
Results from our ERP analysis during indirect reply processing were similar to those reported in previous studies (Foucart et al., Reference Foucart, Garcia, Ayguasanosa, Thierry, Martin and Costa2015). Specifically, while both N400 and LPC effects were identified in native Mandarin speakers, only LPC effects were observed in non-native speakers. Taken together, these findings support that both native and non-native speakers are capable of online processing of indirect replies, but that the time course of this processing is delayed in non-native speakers.
Overall, the results of the present study support that high-proficiency non-native speakers behave like native speakers with low WM capacity or poor comprehension abilities (Boudewyn et al., Reference Boudewyn, Long and Swaab2013; Foucart et al., Reference Foucart, Romero-Rivas, Gort and Costa2016; Long & Chong, Reference Long and Chong2001; Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021). These findings are in alignment with the limited resources hypothesis. Pragmatic interpretation difficulties for L2 learners could be attributed to insufficient cognitive resources because of the more effortful nature of L2 language processing (e.g., Antoniou, Reference Antoniou and Taguchi2019). Decoding literal meanings of indirect replies is nearly automatic and requires very few cognitive resources for native speakers. In contrast, non-native speakers require more cognitive resources while using their L2, leaving fewer resources for deriving intended meanings from indirect replies. The inefficient allocation of cognitive resources in non-native speakers may also be behind other patterns of results observed during L2 processing including reduced speed (Dekydtspotter et al., Reference Dekydtspotter, Schwartz, Sprouse, O’Brien, Shea and Archibald2006; Hopp, Reference Hopp2010), decreased ability to generate predictions (Grüter & Rohde, Reference Grüter and Rohde2013; Grüter et al., Reference Grüter, Rohde and Schafer2014; Pérez et al., Reference Pérez, Hansen and Bajo2019), and susceptibility to retrieval interference (Cunnings, Reference Cunnings2017a, Reference Cunnings2017b).
Relevance Theory (Sperber & Wilson, Reference Sperber and Wilson1986) considers relevance-seeking as the cognitive mechanism behind comprehension of indirect replies. When a reply is presented, readers automatically seek relevance for the information by maximizing the use of available contextual cues, with the number of cues determining the strength of implicature. As Taguchi stated, relevance-seeking mechanisms can transfer from L1 to L2 (e.g., Taguchi & Yamaguchi, Reference Taguchi and Yamaguchi2019), thus non-native speakers can comprehend conversation implicature. However, for indirect reply comprehension, the number of relevant contextual cues available for native and non-native Mandarin speakers may differ. It is possible that relevant contextual cues are more quickly identified by native speakers, leading to processing differences for indirect replies at an earlier stage, as evidenced in our observed native speaker-specific N400 effect.
In the field of human communication, the principle of relevance maximization, as proposed by Relevance Theory (Sperber & Wilson, Reference Sperber and Wilson1986), has inspired a large number of linguistic and philosophical studies (e.g., Levinson, Reference Levinson2000; Sperber & Wilson, Reference Sperber and Wilson1986). More recently, this interest has manifested in neuropsychological studies (e.g., Hagoort, Reference Hagoort2019; Noveck & Reboul, Reference Noveck and Reboul2008), as well as computational investigations (e.g., Dale & Reiter, Reference Dale and Reiter1995; Goodman & Frank, Reference Goodman and Frank2016; Mi et al., Reference Mi, Wang, Camerer and Zhu2021). Informed by both Relevance Theory and psychological research methods, the present study examined whether non-native and native Mandarin speakers differed in online indirect reply processing. Our findings indicate that non-native Mandarin speakers can understand indirect replies in the absence of immediate pragmatic analysis, suggesting cross-language relevance maximization, albeit with delayed processing. These findings also align with the limited cognitive resources hypothesis for L2 pragmatic processing (Morishima, Reference Morishima2013). Together, findings from the present work have important implications for our understanding of human communication and L2 pragmatic processing.
Findings from the present study should be considered in light of a few limitations. First, we did not consider the potential influence of language proficiency and cultural differences in non-native Mandarin speakers. Given that language proficiency and cultural background may influence both linguistic (e.g., Momenian et al., Reference Momenian, Privitera and Weekes2024) and non-linguistic cognitive function (e.g., Privitera et al., Reference Privitera, Momenian and Weekes2023; Samuel et al., Reference Samuel, Roehr-Brackin, Pak and Kim2018), future studies should account for these differences. Additionally, we did not assess participants’ general cognitive ability including WM capacity or cognitive control. This prevented us from accounting for differences in domain-general cognitive function in native and non-native Mandarin speakers that may impact the processing of indirect replies (Zhang et al., Reference Zhang, Zhang, Zhang, Yang and Yang2021; Zhang et al., Reference Zhang, Wang, Yang and Yang2024). The ecological validity of the present study is also limited by the use of a paradigm that does not mimic the conditions of real-world reading. Future studies should consider combining EEG with eye tracking (e.g., Wei et al., Reference Wei, Tang and Privitera2023) or EEG-derived virtual eye tracking (Sun et al., Reference Sun, Cheng, Chan, Hsiao, Privitera, Gao and Tang2023) to test whether findings reported in the present study emerge under natural reading conditions. Finally, reliance on sensor space EEG measures prevented us from drawing conclusions regarding neural generators underlying our observed neurophysiological findings. Future work utilizing source space EEG measures (e.g., Pascual-Marqui, Reference Pascual-Marqui2002; Privitera & Tang, Reference Privitera and Tang2022) can provide insight into the brain’s role in the processing of indirect replies.
5. Conclusions
To summarize, the present study examined how native Mandarin speakers differed from non-native speakers in the online processing of indirect replies using both behavioral and neurophysiological measures. Our findings support that, like native speakers, highly proficient non-native speakers can comprehend indirect replies online, though the time course of processing is delayed. These findings contribute to our understanding of how native and non-native speakers differ in online language processing and cognitive resource limits associated with L2 use, and provide evidence in support of cross-language relevance maximization.
Supplementary material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1366728924000695.
Data availability statement
The datasets generated for this study are available on request to the corresponding author.
Acknowledgments
This work was supported by the Science Foundation of Beijing Language and Culture University (supported by “the Fundamental Research Funds for the Central Universities”) (Approval number 24QN47), the Fundamental Research Funds for the Central Universities, and the Research Funds of Beijing Language and Culture University (24YCX045), the Discipline Team Support Program of Beijing Language and Culture University (grant number 2023YGF07), the Research Project “the Construction of the Advanced Disciplines in Universities in Beijing,” and the National Natural Science Foundation of China (31800914).
Authorship contribution
Xiuping Zhang: the conception and design of the work, the interpretation of data for the work, Funding Acquisition, drafting and revising the work, and final approval of the version to be published. Xiaoxi Pan: the acquisition and analysis of data for the work, drafting and revising the work, and final approval of the version to be published. Yizhu Wang: drawing, the analysis of data for the work, revising the work, and final approval of the version to be published. Maoyao Xu: the analysis of data for the work, revising the work, and final approval of the version to be published. Adam John Privitera: revising the work, final approval of the version to be published.
Competing interests
The author(s) declare none.
Ethical standard
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. Approval for this study was granted by the ethics committee of the Beijing Language and Culture University.
Open access
