Content area
Social hierarchy as a social ranking system is conveyed verbally, nonverbally, or conjointly in communication. From a linguistic perspective, hierarchy concept is encoded via words, hierarchical or non-hierarchical when it is embedded in different contexts. However, few studies have explored how hierarchy is represented in the mental lexicon and processed neurocognitively. Backgrounded by this situation, we conducted two ERP experiments to investigate whether social hierarchy can be subcategorized into implicit and explicit forms and whether a word’s social hierarchy is processed similarly to its semantic knowledge relating to Who Does Whom neurologically. Experiment 1 compared the processing of three types of Chinese verbs with different degree of hierarchy (strong hierarchical verbs; weak hierarchical verbs; non-hierarchical verbs) in SVO sentences. Experiment 2 examined whether the dichotomy of hierarchical verbs was modulated by context type (neutral context or biased context) in processing. The results revealed three major findings: First, strong hierarchical verbs relative to non-hierarchical verbs elicited greater posterior-P600 at the verb and AN at the noun position; Second, weak hierarchical verbs relative to strong hierarchical verbs elicited enhanced AN and posterior-P600 effects at the verb and noun positions, while as compared to non-hierarchical verbs, weak hierarchical verbs elicited stronger P600 and AN effect at the verb and noun positions, respectively; Third, this hierarchy difference was much affected by context type. Specifically, in biased contexts, weak hierarchical verbs and strong hierarchical verbs became indistinguishable, while in neutral contexts, strong hierarchical verbs sentences were harder to process than weak hierarchical verbs, as indicated by the larger P600 effect at the verb position. These findings converge to suggest a unique neurocognitive mechanism underlying the processing of Chinese social hierarchy verbs and highlight the concept that a word’s social hierarchy is distinct from its lexical semantics. This study provides insights into how social hierarchy is decoded in language comprehension and offers implications for future research on linguistic structures and social cognition.
Introduction
Social hierarchy refers to the stratified organization of individuals or groups within a society. Schjelderup-Ebbe (1922) first discovered the existence of hierarchical relationships in the pecking order of poultry, which reduces violent conflicts and injuries, saves energy, and promotes social stability. Animals recognize social status of their conspecifics in nonverbal cues (e.g., body posture, vocal calls) to convey expectations about their territory, division of labor, or mating among others. Likewise, human hierarchy delineates relative social status of profession in a given social community, domestic and recreational settings, where they define implicit expectations and action dispositions that perform appropriate social behavior (Cummins, 2000). People with lower status tend to choose respectful ways towards those with higher status so as to obtain favorable alliances and avoid potential conflicts in social competition. In terms of communicating patterns, different social hierarchies are conveyed either by non-verbal cues (e.g., age, dress, etc.) (Karafin et al., 2004) or by verbal cues, or by both. In language, social hierarchy is either expressed explicitly through hierarchical nouns (boss vs. clerk), hierarchical verbs (raise/rear a child), honorifics (e.g., “您/nin” in Chinese and the suffix “-si” in Korean) (e.g., Agha, 2007; Jiang et al., 2013; Kwon and Sturt, 2024) or conveyed implicitly through metaphors (e.g., “the head of the department”), linguistic tense (e.g., formal and informal), verbal assignments, status-specific grammar (Park, 1991).
Human have developed and evolved complex neural mechanism for processing social hierarchy information in language comprehension. Neurophysiological studies have revealed that social hierarchy processing was related to N400, AN (Anterior Negativity, an anterior N400-like component), P600 (posterior), late positivity and late negativity. N400 was observed in hierarchy violation in the voice (e.g., using an upper-class accent to say “I have a tattoo on my back.”) (Van Berkum et al., 2008). AN effect was observed in thematic relations (i.e., “*The son raises up his father”) (Shi et al., 2022) and honorifics in over-respectful condition (Jiang et al., 2013; Ji et al., 2022; Tokimoto, et al., 2021) studies, which was often followed by a late positivity (e.g., Jiang et al., 2013) or late negativity (Ji et al., 2022) related to socio-pragmatic violation. Another relevant index P600 was found in the voice mismatch in Basque, manifesting the late integration of verbal inflection and speaker properties (Caffarra et al., 2020) or the re-integration process (Lattner and Friederici, 2003).
Despite the above research, it is controversial whether the concept of social hierarchy is different from general semantic knowledge. Some people believe that the social hierarchy information is a subsystem of semantic network, resulting from selective representation in the brain (Binder et al., 2016; Huth et al., 2016), evidenced by the similar brain structures: bilateral anterior temporal lobe, angular gyrus, medial prefrontal cortex, cingulate gyrus (Huth et al., 2016; Forgács et al., 2012; Graves et al., 2010). In contrast, other people claim that the concept of social hierarchy is thoroughly distinct from other semantic concepts, for the former is culture-specific and domain specific (Pexman et al., 2023) that would exert finer semantic constraints on verbs (e.g., pragmatic plausibility) than linguistic knowledge (e.g., animacy violation) (Altmann and Kamide, 1999; Li, et al., 2006; Paczynski and Kuperberg, 2012; Shi et al., 2022) and is commonly reflected as an anteriorly-distributed N400 (Shi et al., 2022; Jiang et al., 2013). To date, however, no direct study has been conducted on comparing social hierarchy with lexical semantics in sentence processing, electrophysiologically in particular, hence leaving the issue unknown whether a verb’s social hierarchy information is different from its semantic knowledge derived from its argument structure.
Social hierarchy is encoded in lexical forms (e.g., verbs, noun-pairs, pronouns) and realized via verb-based argument structure “Who Does Whom”. The semantic restriction of a verb is based on its thematic knowledge (event-based knowledge in nature). When a verb is recognized during sentence comprehension, its relevant thematic knowledge is automatically activated, helping participants to anticipate the upcoming words from top to bottom (Altmann and Kamide, 1999; Kamide et al., 2003), and so is the case for social hierarchical verbs (Shi et al., 2022) (e.g., The boss hired a bodyguard yesterday” vs “*The bodyguard hired the boss yesterday”). ERP studies show that inappropriate arguments elicited N400 amplitudes (Li et al., 2006; Paczynski and Kuperberg, 2012), particularly an anterior N400 (or AN) for the hierarchy concept (Shi et al., 2022), suggesting the importance of the thematic roles a verb assigns to the argument nouns in a sentence.
Social hierarchy is not only embedded in explicit vocabulary, but also used to convey implicit dynamic interpretations by context (Farrow et al., 2011). In some contexts, non-hierarchy verbs can also be used to transmit social hierarchical concepts temporarily. As an example, “bully” does not necessarily convey social hierarchy information but when combined with a hierarchical noun pair “master vs. apprentice”, the verb helps accentuate the “top-down” hierarchical relationship between the agent and the patient (e.g., “The master often bullies the apprentice.”). This suggests whether a word’s social hierarchy is not definitely binary but rather resembles a continuous system.
Based on the theory of verb’s semantics by Lan (1999), Zou (2009)1 and Williams (2015)2, we proposed that Chinese hierarchical verbs should be subcategorized into two types, strong hierarchical verbs and weak hierarchical verbs in terms of their attributes as [±hierarchy], [±action directionality], [±competence], and [±non-interchangeability between agent and patient]. Strong hierarchical verbs possess all of the positive attributes. Weak hierarchical verbs do not possess [+hierarchy], but satisfy [+directionality] and [+competence]. That is, this type of verbs requires their argument nouns (i.e., agents and patients) to be fixed in order, unlicensing the interchange of their positions (subject position or object position) in a sentence (e.g., *师傅经常贿赂徒弟/*The master often bribes the apprentice,). In addition, weak hierarchical verbs do not require their argument nouns to be necessarily hierarchical (c.f., “Jack often bribes his friend.” vs. “The clerk often bribes his boss.”). By contrast, non-hierarchical verbs like “看见/see” do not convey hierarchy information of their argument nouns and consequently the nouns can be position-interchanged without incurring semantically ill-formedness (Ma, 1997; Zhang, 1997; Zou, 2009; Yang, 2012). As a consequence, it is safe to argue that strong hierarchical verbs are less context dependent, weak hierarchical verbs are more context-dependent, and non-hierarchical verbs are free of context. In a nutshell, strong hierarchical verbs are used to convey explicit hierarchy, weak hierarchical verbs to convey implicit hierarchy and non-hierarchical verbs to convey no hierarchy (i.e., equality) in general.
The above literature suggests social hierarchy somewhat differs from lexical semantic knowledge in one’s mental lexicon. Nevertheless, no electrophysiological evidence has verified this distinction on sentence level. More importantly, a few studies were conducted on the explicit hierarchical words (verbs or nouns), leaving it unknown whether and how implicit hierarchy mediates people’s cognition in sentence processing. Given that Chinese is a language characterized by verbal social hierarchy and implicit semantics (i.e., context-dependent), it is interesting to explore the neurocognitive processing mechanism of social hierarchy embedded in Chinese sentences.
Backgrounded by the above gap, the current study adopted two ERP experiments to reveal whether a verb’s social hierarchy modulated Chinese sentence processing and how context types affected the processing pattern involving different hierarchical verbs. Specifically, we aimed to address the following two questions, namely, the first question (RQ1) for Experiment 1 and the second questions (RQ2) for Experiment 2.
RQ1: How do verb types (hierarchical or non-hierarchical) affect Chinese sentence processing?
RQ2: What effects do context types (biased or neutral) bring to the processing of sentences involving explicit or implicit hierarchical verbs?
In association with previous studies (e.g., Federmeier and Kutas, 2011; Schlesewsky and Schlesewsky, 2008; Jiang et al., 2013; Shi et al., 2024), we formulated two hypotheses corresponding to the two research questions:
The first hypothesis (H1): Strong hierarchical verbs and weak hierarchical verbs involved sentences are more difficult to process than non-hierarchical verbs involved sentence, indexing by an enhanced AN and P600 while strong hierarchical verbs and weak hierarchical verbs are not significantly differentiated in sentence processing.
The second hypothesis (H2): Biased context facilitates the processing of weak hierarchical verbs involved sentences more than the processing of strong hierarchical verbs involved sentences while neutral context brings small effects to the processing of the two verbs involved sentences.
Experiment 1
Method
Participants
We conducted a priori power analysis using GPower 3.1 (Erdfelder et al., 1996) before the experiment. For a repeated-measures ANOVA with within-subjects design involving six conditions (3 × 2), a minimum sample of 28 participants was required to detect a medium effect size (f = 0.25) with 80% power at α = 0.05. To ensure robust statistical power and account for potential exclusions, 46 college students were recruited from Qufu Normal University (15 females; age range: 20–24 years, M = 20, SD = 1.72). This sample size affords a statistical power > 0.99 for detecting medium or larger effects. All the participants were Chinese native speakers, right-handed as assessed by the Edinburgh Handedness Scale (Oldfield, 1971) with normal or corrected-to-normal vision. None of participants had a history of neurological or psychiatric disorders. Each of them was informed and signed on the written consent in accordance with the declaration of Helsinki prior to the experiment. This study was approved by the Ethics Committee of Qufu Normal University. One participant’s data were excluded from the analysis due to excessive artifacts over 25% of the data.
Design and materials
This experiment adopted a 3 (verb types: strong hierarchical verbs, weak hierarchical verbs, and non-hierarchical verbs) ×2 (semantic plausibility: plausible vs. implausible) within-subject design. All stimuli were composed of 120 sets of sentences, each following the pattern Subject-Adverb-Verb-Object, resulting in 720 stimuli in total (as illustrated in Table 1).
Table 1. Exemplar stimuli adopted in Experiment 1.
Verb type | Semantic plausibility | Examples |
|---|---|---|
Vss | Plausible | 老师 经常 体罚 学生。 The teacher often punishes students. |
Vss | Implausible | *老师 经常 请教 学生。 *The teacher often learns from the students. |
Vws | Plausible | 老师 经常 夸赞 学生。 The teacher often praises the students. |
Vws | Implausible | *老师 经常 贿赂 学生。 *The teacher often bribes the students. |
Vns | Plausible | 教练 经常 看见 学生。 The teacher often saw the students. |
Vns | Implausible | *教练 经常 种植 学生。 *The teacher often plants the students. |
Vss strong hierarchical verbs, Vws weak hierarchical verbs, Vns non-hierarchical verbs.
The selection of three types of verbs was based on the categorization criteria stated in the introduction part. Ninety verbs from each of the following categories were selected: strong hierarchical verbs (i.e., 养育/bring up; 雇佣/hire), weak hierarchical verbs (i.e., 贿赂/bribe; 咨询/consult), and non-hierarchical verbs (e.g., 看见/see; 想念/miss), with 30 verbs from each set repeated once, resulting in 120 semantically coherent sentences per category. Ninety strong hierarchical verbs and weak hierarchical verbs in implausible condition were chosen from those in plausible condition but imply reversed hierarchical relations, resulting in hierarchical violation. Another 90 non-hierarchical verbs were constructed by repeating 30 verbs (each repeated once) with restricted inanimate object arguments (e.g., 种植/plant; 打扫/sweep), resulting in 120 semantically implausible sentences. Verbs were combined with hierarchical noun pairs (i.e., with distinct social hierarchy) (e.g., “老师/ teacher” vs “学生/ student”) which involve family, society, job scenarios, implying top-down or bottom-up hierarchical relations. 120 noun pairs were constructed similarly for each type of verb. Number of strokes (F (5, 714) = 1.60, p = .188) and character frequency (F (5, 714) = 1.71, p = .392) were matched among the verb types.
Twenty participants, who were not involved in the formal experiment, rated the semantic acceptability of all sentences based on a 7-point Likert scale from 1 (fully unacceptable) to 7 (fully acceptable). The mean value of for strong hierarchical verbs, weak hierarchical verbs, non-hierarchical verbs in plausible sentences were 5.40 (SD = 0.44), 5.33 (SD = 0.43), 5.43(0.26) respectively and 2.384 (SD = 0.45), 2.43 (SD = 0.29), 2.40 (SD = 0.33) for implausible sentences. Repeated-measures ANOVA revealed an interaction between verb type and plausibility (F (1, 119) = 10226.24, p < .001, η2 = .998, large effect). There was also a main effect of plausibility (F (1, 119) = 4.01, p = .020, η2 = .033, small effect). However, no main effect of verb type was found (F (2, 238) = 0.22, p = .803, η2 = .002, small effect). To clarify the nature of the interaction, we conducted simple effects analyses, showing that the acceptability score of plausible sentence is significantly higher than implausible sentences across three verb types (Strong: Mdiff = 3.01, SE = 0.05, p < .001, d = 7.80, large effect, 95% CI [2.91, 3.12]; Weak: Mdiff = 2.89, SE = 0.05, p < .001, d = 7.44, large effect, 95% CI [2.79, 2.98]; Non-hierarchical: Mdiff = 3.06, SE = 0.04, p < .001, d = 9.88, large effect, 95% CI [2.98, 3.14]). However, there were no significant difference between any two of the three verb types either in plausibility (Strong -Weak: Mdiff = 0.07, SE = 0.05, p =.394, d = 0.15, small effect, 95% CI [−0.24, 0.11]; Weak -Non-hierarchical: Mdiff = −0.11, SE = 0.05, p = .085, d = 0.11, small effect, 95% CI [−0.09, 0.12]; Strong - Non-hierarchical: Mdiff = −0.04, SE = 0.05, p = .878, d = 0.03, negligible effect, 95% CI [−0.16, 0.09]) or implausibility condition (Strong -Weak: Mdiff = −0.05, SE = 0.05, p =.581, d = 0.06, small effect, 95% CI [−0.16, 0.06]; Weak -Non-hierarchical: Mdiff = 0.07, SE = 0.04, p = .218, d = 0.07, small effect, 95% CI [−0.02, 0.16]; Strong - Non-hierarchical: Mdiff = 0.02, SE = 0.04, p = .979, d = 0.01, negligible effect, 95% CI [−0.09, 0.12]).
As a result, 120 sets of sentences (720 in total) were divided into four lists/runs, counterbalanced across participants and conditions. For each participant, verbs and noun pairs were not repeated. 180 fillers with same sentence structure “Name + Verb + Verb + Noun / Name” with half plausible and half implausible, such as “李兰负责打扫垃圾/*王强, Li Lan is responsible for cleaning rubbishes/*Wang Qiang.”) were created for repetition in each run.
Procedure
Each participant was randomly assigned to one of four lists and was given twelve practice trials at the beginning of the experiment.
All stimuli were visually displayed using E-Prime software (3.0) in font size 28. Each sentence appeared word by word at the center of the screen in a rapid serial visual presentation (RSVP) mode. Each trial began with a 500 ms cross fixation, followed in turn by each word for 400 ms and a 200 ms blank screen. At the end of each sentence, there appeared question marks “???” on the screen for 2000 ms, during which each participant was required to make a semantic acceptability judgment of the presented sentence. Afterward, there came a 200 ms interval blank and 1500 ms inter-stimulus interval (ISI) until a next trial was to occur. Each trail was presented randomly. Participants were instructed to press the button “1” for semantically plausible sentences, while “3” for semantically implausible sentences. Response keys were counterbalanced across participants to control for potential spatial-response bias. One participant’s experiment lasted around 35 minutes on average, and each participant was paid for participation after the experiment.
EEG recording and data analysis
EEG data were collected using a standard 64-channel cap (10/20 system) from Neuroscan and using 1000 Hz sampling rate. Data were referenced online to the left mastoid (M1). Electrode impedance was kept below 5 kΩ. All electrodes recordings underwent manual inspection prior to re-referencing. The EEG data were re-referenced offline to the global average of all 64 channels, bandpass-filtered to 0.1-30 Hz and were continuously re-sampled at 500 Hz. Following visual inspection, bad channels were removed and then interpolated using EEGLAB’s spherical interpolation function (M = 0.18, Max = 2, Min = 0). Artefactual independent components (ICs) related to eye blinks, eye movements, and muscle activity were identified and removed using Independent Component Analysis (ICA, Makeig et al., 1997) in EEGLAB 14.1.1 (Makeig et al., 1997; Delorme and Makeig, 2004). An average of [2.14] ± [0.65] ICs were rejected per participant. Subsequently, the data were epoched, and each epoch was visually inspected to exclude any remaining trials containing noise. Epochs in which amplitudes exceeded ±90 μV were rejected. Finally, 22.5% of the trails were rejected. For plausible sentences, there were on average 23.58 (Max = 28, Min = 17), 23.39 (Max = 30, Min = 16) and 22.91 (Max = 29, Min = 17) of trials for strong hierarchical verbs, weak hierarchical verbs and non-hierarchical verbs respectively; for implausible sentences, there were 22.64 (Max = 29, Min = 17), 23.78 (Max = 29, Min = 16) and 23.07 (Max = 30, Min = 17) of trials for the three conditions, respectively. The grand average waveform of correct trials was computed for each condition.
EEG data were processed by EEGLAB 14.1.1 (Delorme and Makeig, 2004) in MATLAB 2013b (MathWorks). To start with, EEG data were segmented epochs from -200ms prior to the stimulus onset and 1000 ms post stimulus, with the -200 ms to 0 ms interval serving as the baseline for the verb position. Next, the time window of 0 ms to 100 ms was selected as baseline for the noun position after the verb to avoid the late effect of the former verbs according to Luck (2014). In light of the classic time window of N400 and P600 effect, we selected 300–500 ms window to capture the AN effect at the verb and noun position and 500–700 for the P600 effect at the noun position. To avoid contamination from nouns appearing after 600 ms post-verb onset, we selected the 500–650 ms window at the verb position for the P600 effect with barely no impact of the 0–50 milliseconds after the noun onset.
Based on our hypotheses, repeated ANOVA was conducted to test the anterior negativity and posterior P600 effects across verb types (strong hierarchical verbs, weak hierarchical verbs, and non-hierarchical verbs), plausibility (plausible vs implausible). Two regions of interest (ROIs) were selected for anterior negativity, i.e., anterior-left (F1, F3, F5, FC1, FC3, FC5) and anterior-right (F2, F4, F6, FC4, FC6), and for posterior P600, i.e., posterior-left (P1, P3, P5, CP1, CP3, CP5) and posterior-right (P2, P4, P6, CP2, CP4, CP6) according to Jiang et al. (2013) and Shi et al. (2022). All p values were adjusted with the Greenhouse–Geisser correction for non-sphericity when the degree of freedom in the numerator was more than one (Greenhouse and Geisser, 1959). Eta squared (η²) was reported for overall ANOVA effects (Olejnik and Algina, 2003). Cohen’s d (d) was used to quantify the effect size with 95% confidence intervals (CI) for pairwise comparisons within simple effects (Cohen, 1988) which were adjusted by the Bonferroni correction for multiple condition comparisons to interpret interaction effects. ERP waveforms and topographies were generated via ERPLAB4.0.2.3 toolbox (Lopez-Calderon and Luck, 2014) in MATLAB and finalized in Canvas 11.
Results
Behavioral results
Accuracy (M ± SD) of strong hierarchical verbs, weak hierarchical verbs and non-hierarchical verbs for plausible sentences were 94.4% (4.9), 92.8% (6.6), 90.3% (9.5) respectively, and 91.3% (6.9), 81.3% (12.6), 93.4% (5.6) for implausible sentences (as shown in Fig. 1). Repeated ANOVAs revealed a significant interaction (F (1, 41) = 10.17, p < .01, η2 = .25, large effect) between verb type and plausibility. There were significant main effects of verb type (F (1, 41) = 18.48, p < .01, η2 = .37, large effect) and plausibility (F (1, 41) = 67.54, p < .01, η2 = .69, large effect). No significant difference between any two of the three verb types in plausible sentences (Strong -Weak: Mdiff = 0.03, SE = 0.02, p = 0.160, d = 0.35, small effect, 95% CI [−0.01, 0.07]; Weak-Non-hierarchical: Mdiff = −0.01, SE = 0.02, p = 0.780, d = −0.16, negligible effect, 95% CI [−0.06, 0.03]; Strong - Non-hierarchical: Mdiff = 0.02, SE = 0.01, p = 0.565, d = 0.22, small effect, 95% CI [−0.02, 0.05]). However, in implausible conditions, non-hierarchical verbs were significantly higher than weak hierarchical verbs and strong hierarchical verbs (ps < .001) (Non-hierarchical - Weak: Mdiff = 0.12, SE = 0.02, p < .001, d = 1.32, large effect, 95% CI [0.08, 0.18]; Non-hierarchical - Strong: Mdiff = 0.02, SE = 0.12, p < .001, d = 1.16, large effect, 95% CI [0.07, 0.17]). No difference was found between strong hierarchical verbs and weak hierarchical verbs (Strong -Weak: Mdiff = 1.00, SE = 0.03, p = .998, d = 1.16, large effect, 95% CI [0.07, 0.17]).
[See PDF for image]
Fig. 1
Accuracy rate and reaction time for three verb types in two plausible conditions in Experiment 1.
Error RTs and those exceeding ±2.5 SD were removed (20.2%). The M (SD) of the accuracy rate of strong hierarchical verbs, weak hierarchical verbs and non-hierarchical verbs in plausible sentences were 744.76 ms (78.20), 758.93 ms (82.85), 779.27 ms (70.66) respectively, and 757.42 ms (84.80), 776.19 ms (92.25), 719.65 ms (73.52) for implausible sentences (as shown in Fig. 1). Then statistical analyses revealed a main effect of verb (F (2, 88) = 4.33, p = .016, η2 = 0.09, medium effect) and interaction between verb and plausibility (F (2, 88) = 27.56, p < 0.001, η2 = 0.39, large effect) but no main effect of plausibility (F (1, 44) = 1.71, p = 0.198, η2 = 0.04, small effect). Simple effect analyses show that strong hierarchical verbs (744.76 ms) showed shorter RTs than non-hierarchical verbs (779.270 ms) (Strong-Non-hierarchical: Mdiff = −34.51, SE = 10.78, p = 0.008, d = −0.47, small effect, 95% CI [−61.27, −7.74]). There were no differences between strong hierarchical verbs and weak hierarchical verbs (Strong -Weak: Mdiff = −14.17, SE = 6.65, p = 0.112, d = −0.32, small effect, 95% CI [−30.68, 2.34]) or between weak hierarchical verbs and non-hierarchical verbs (Weak - Non-hierarchical: Mdiff = -20.34, SE = 11.91, p = 0.258, d = −0.25, small effect, 95% CI [−49.892, 9.211]) in plausible condition. In implausible condition, the RTs for three types of verb from quick to slow is: non-hierarchical verbs > strong hierarchical verbs > weak hierarchical verbs (Strong -Weak: Mdiff = −18.77, SE = 6.57, p = 0.019, d = −0.43, small effect, 95% CI [−35.08, −2.47]; Weak - Non-hierarchical verbs: Mdiff = 56.55, SE = 8.58, p < 0.001, d = 0.98, large effect, 95% CI [35.26, 77.83]; Strong - Non-hierarchical: Mdiff = 37.77, SE = 7.84, p < .001, d = 0.72, medium effect, 95% CI [18.33, 57.22]). The RTs of implausible sentences and plausible sentence were not significantly different both in strong hierarchical verbs (Implausible -Plausible: Mdiff = 12.66, SE = 9.58, p = 0.193, d = 0.20, small effect, 95% CI [−0.65, −31.98]) and weak hierarchical verbs (Implausible -Plausible: Mdiff = 17.27, SE = 10.26, p = 0.100, d = 0.25, small effect, 95% CI [3.42, 37.95]) conditions, but the RTs were significantly longer for plausible sentences than for implausible sentences in non-hierarchical verbs condition (Plausible -Implausible: Mdiff = 59.62, SE = 10.50, p < .001, d = 0.85, large effect, 95% CI [38.46, 80.78]).
ERP results
Verbs
300–500 ms
As expected, a larger anterior negativity (AN) was shown during 300–500 ms across six conditions, distinct from posterior-N400 (as indicated in Fig. 2). A repeated-measures ANOVA revealed a larger interaction effect between verb type and plausibility (F (2, 514) = 32.68, p < 0.001, η2 = 0.11, large effect). There were significant main effects of verb type and plausibility (Verb: F (2, 514) = 43.52, p < 0.001, η2 = 0.15, large effect; Plausibility: F (1, 257) = 9.14, p = .003, η2 = .034, small effect). To decompose the interaction, simple effects analyses were performed. In the semantically plausible condition, there was no significant difference between weak hierarchical and non-hierarchical verbs (Weak - Non-hierarchical: Mdiff = 0.16μV, SE = 0.18, p =0.751, d = 0.06, negligible effect, 95% CI [−0.60, 0.28]). However, both weak hierarchical verbs and non-hierarchical verbs elicited significantly greater negativity than strong hierarchical verbs (see Fig. 2) (Weak- Strong: Mdiff = −0.54μV, SE = 0.10, p < 0.001, d = −0.35, small effect, 95% CI [0.35, 0.72]; Non-hierarchical -Strong: Mdiff = −0.70μV, SE = 0.20, p = 0.001, d = −0.22, small effect, 95% CI [−1.18, 0.23]). The pattern differed in the semantically implausible condition. Here, non-hierarchical verbs elicited the largest negativity, followed in turn by strong hierarchical verbs and weak hierarchical verbs (Non-hierarchical -Strong: Mdiff = −1.33μV, SE = 0.17, p < 0.001, d = −0.48, small-to-medium effect, 95% CI [−1.74, −0.91]; Non-hierarchical- Weak: Mdiff = −1.91μV, SE = 0.19, p < 0.001, d = −0.62, medium effect, 95% CI [−2.37, −1.45]; Strong - Weak: Mdiff = −0.59μV, SE = 0.14, p < 0.001, d = −0.26, small effect, 95% CI [−0.92, −0.25]). As the verb condition was set, implausible sentences elicited larger AN for both strong hierarchical verbs (Mdiff = −0.44μV, SE = 0.11, p < .001, d = −0.25, small effect, (95% CI [−0.66, −0.22]) and non-hierarchical verbs (Mdiff = −1.07μV, SE = 0.19, p < .001, d = −0.35, small effect, 95% CI [−1.44, −0.69]). In contrast, weak hierarchical verbs showed the opposite pattern, with implausible sentences eliciting less negativity than plausible ones (Mdiff = 0.68μV, SE = 0.16, p < .001, d = 0.27, small effect, 95% CI [0.37, 0.99]).
[See PDF for image]
Fig. 2
Grand-average event-related potentials (ERPs) and topographic maps showing the verb plausibility effect at the verb position in Experiment 1.
Waveforms depict ERPs linear derivation of all the electrodes used for analysis (F1, F3, F5, FC1, FC3, FC5, F2, F4, F6, FC4, FC6, P1, P3, P5, CP1, CP3, CP5, P2, P4, P6, CP2, CP4, CP6). The shaded areas around the waveforms represent the standard error of the mean (SEM). Topographic maps illustrate the voltage distribution across the scalp during the 300–500 ms (AN component) and 500–650 ms (P600 component) time windows. Vss = strong hierarchical verbs; Vws = weak hierarchical verbs; Vns = non-hierarchical verbs.
500–650 ms
For the 500–650 ms time window, there was a significant interaction between verb and plausibility (F (2, 514) = 16.50, p < .001, η2 = .06, medium effect). Significant main effects were also found for verb type and plausibility (Verb: F (2, 514) = 68.60, p < .001, η2 = .21, large effect; Plausibility: F (1, 257) = 33.28, p = .001, η2 = .12, medium effect). Simple effects analyses were conducted to decompose the interaction. First, the comparisons between verb types showed that in plausible conditions, strong hierarchical verbs evoked larger positive amplitudes than both weak hierarchical verbs and non-hierarchical verbs (Strong -Weak: Mdiff = 0.63μV, SE = 0.13, p < .001, d = 0.31, small effect, 95% CI [0.32, 0.94]; Strong - Non-hierarchical: Mdiff = 1.12μV, SE = 0.18, p < .001, d = 0.38, small effect, 95% CI [0.68, 1.56]). Weak hierarchical verbs also evoked larger amplitudes than non-hierarchical verbs in plausible conditions (weak -non-hierarchical: Mdiff = 0.49μV, SE = 0.17, p = .016, d = 0.17, negligible effect, 95% CI [0.07, 0.91]). This pattern trend was similar but more pronounced in implausible conditions, where both strong and weak hierarchical verbs differed significantly from non-hierarchical verbs (Strong - Non-hierarchical: Mdiff = 1.52μV, SE = 0.18, p < .001, d = 0.52, medium effect, 95% CI [1.08, 1.95]; Weak - Non-hierarchical: Mdiff = 1.36μV, SE = 0.19, p < .001, d = 0.44, small effect, 95% CI [0.89, 1.83]). However, the difference between strong and weak hierarchical verbs was not significant in implausible conditions (Strong - Weak: Mdiff = 0.16μV, SE = 0.15, p = .670, d = 0.06, negligible effect, 95% CI [−0.52, 0.21]). Additionally, the effect of plausibility was modulated by verb type. Implausible sentences generated more positive waveforms than plausible ones for both strong hierarchical verbs (Mdiff = 0.46μV, SE = 0.17, p < .001, d = 0.18, negligible effect, 95% CI [0.14, 0.79]) and weak hierarchical verbs (Mdiff = 0.94μV, SE = 0.15, p < .001, d = 0.40, small effect, 95% CI [0.65, 1.23]) but not for non-hierarchical verbs (Mdiff = 0.06μV, SE = 0.24, p = .786, d = 0.02, small effect, 95% CI [0.40, 0.53]).
Nouns
300–500 ms
The statistical analysis revealed a significant interaction between verb type and plausibility (F (2, 514) = 20.358, p < .001, η2 = .08, medium effect). Although a main effect of verb type was observed (F (2, 514) = 32.16, p < .001, η2 = .11, medium effect), the main effect of plausibility was not significant (F (1, 257) = 0.41, p = .52, η2 = .08, medium effect). In plausible conditions, strong hierarchical verbs elicited the largest amplitudes, followed in turn by weak hierarchical verbs and non-hierarchical verbs. (as shown in Fig. 3) (Strong - Weak: Mdiff = −0.67μV, SE = 0.19, p = 0.001, d = 0.23, small effect, 95% CI [0.26, 1.22]; Strong - Non-hierarchical: Mdiff = −1.86μV, SE = 0.21, p < .001, d = −0.15, negligible effect, 95% CI [−0.94, −0.00]; Weak - Non-hierarchical: Mdiff = −1.19μV, SE = 0.19, p < .001, d = −0.09, negligible effect, 95% CI [−0.72, −0.19]). For implausible condition, strong hierarchical verbs induced larger AN effect than weak hierarchical verbs and non-hierarchical verbs respectively (Strong - Weak: Mdiff = −0.74μV, SE = 0.20, p = .001, d = −0.22, small effect, 95% CI [−1.12, −0.22]; Strong - Non-hierarchical: Mdiff = −0.47μV, SE = 0.20, p = .049, d = −0.55, medium effect, 95% CI [-2.37, −1.36]). However, the difference between weak and non-hierarchical verbs was not significant (Mdiff = −0.27μV, SE = 0.19, p = .412, d = −0.40, small effect, 95% CI [−1.64, −0.74]). Finally, plausibility comparison within each verb type showed that implausible sentences evoked greater negativity than plausible sentences for both strong hierarchical verbs and weak hierarchical verbs (Strong: Mdiff = −0.38μV, SE = 0.19, p = .044, d = −0.13, negligible effect, 95% CI [−0.74, −0.01,]; Weak: Mdiff = −0.44μV, SE = 0.16, p < .007, d = −0.17, negligible effect, 95% CI [−0.77, −0.12]). Conversely, for non-hierarchical verbs, plausible sentences evoked significantly greater negativity than implausible ones (Mdiff = −1.02μV, SE = 0.19, p < .001, d = −0.33, small effect, 95% CI [−1.39, −0.64]).
[See PDF for image]
Fig. 3
Grand-average event-related potentials (ERPs) and topographic maps showing the verb plausibility effect at the noun position in Experiment 1.
Waveforms depict ERPs linear derivation of all the electrodes used for analysis (F1, F3, F5, FC1, FC3, FC5, F2, F4, F6, FC4, FC6, P1, P3, P5, CP1, CP3, CP5, P2, P4, P6, CP2, CP4, CP6). The shaded areas around the waveforms represent the standard error of the mean (SEM). Topographic maps illustrate the voltage distribution across the scalp during the 300–500 ms (AN component) and 500–700 ms (P600 component) time windows. Vss = strong hierarchical verbs; Vws = weak hierarchical verbs; Vns = non-hierarchical verbs.
500–700ms
A late time window analysis revealed a significant interaction between verb type and plausibility (F (2, 514) = 52.11 p < .001, η2 = .17, large effect). Given this interaction, the significant main effects of verb type and plausibility were not interpreted independently (Verb: F (2, 514) = 156.93 p < .001, η2 = .39, large effect; Plausibility: F (1, 257) = 233.12, p < .001, η2 = .48, large effect). The greatest AN was anteriorly distributed for strong hierarchical verbs, while greatest P600 was posteriorly dominant for non-hierarchical verbs (Fig. 3). Simple effects analyses showed that the positive amplitudes elicited by non-hierarchical verbs were the largest, followed by weak hierarchical verbs, and then strong hierarchical verbs, irrespective of plausiblity (as shown in Fig. 3). In plausible conditions, this pattern was supported by a significant amplitude difference between weak and strong hierarchical verbs (Mdiff = 0.82μV, SE = 0.20, p < .001, d = 0.60, medium effect, 95% CI [0.33, 1.31]) and a larger difference between non-hierarchical and strong hierarchical verbs (Mdiff = 1.30μV, SE = 0.18, p < .001, d = 0.95, large effect, 95% CI [0.87, 1.73]). The difference between non-hierarchical and weak verbs was marginally significant (Mdiff = 0.48μV, SE = 0.21, p = .059, d = 0.15, negligible effect, 95% CI [0.01, 0.98]). In implausible conditions, all the pairwise comparisons had markedly larger effects (Non-hierarchical > Strong: Mdiff = 3.52μV, SE = 0.22, p < 0.001, d = 2.58, large effect, 95% CI [3.01, 4.04]; Non-hierarchical > Weak: Mdiff = 3.06μV, SE = 0.20, p < .001, d = 2.24, large effect, 95% CI [2.58, 3.54]; Weak > Strong: Mdiff = 0.464μV, SE = 0.177, p = 0.027, d = 0.34, small effect, 95% CI [0.04, 0.89]). Implausible sentences evoked stronger positivity than plausible ones across verbs (Strong: Mdiff = 1.16μV, SE = 0.16, p < .001, d = 0.46, small effect, 95% CI [0.85, 1.47]); Weak: Mdiff = 0.81μV, SE = 0.18, p < .001, d = 0.28, small effect, 95% CI [0.45, 1.16]; Non-hierarchical: Mdiff = 3.38μV, SE = 0.24, p < .001, d = 0.87, large effect, 95% CI [2.90, 3.86]).
Discussion
Experiment 1 revealed significant processing differences across verb hierarchy and plausibility.
As expected, implausible sentences were harder to comprehend than plausible sentences across most verb types. For strong hierarchical verbs, implausible sentences induced larger AN and P600 at the verbs and nouns, suggesting the difficulty in semantic activation and integration for hierarchy violation. Unexpectedly, plausible sentences for weak hierarchical verbs and non-hierarchical verbs showed enhanced AN at verbs and nouns respectively. For weak hierarchical verbs, plausible sentences involved semantic competition (e.g., “The teacher often praises/asks/calls…”), requiring deeper scrutiny than low-frequency implausible sentences (“? The teacher often consults the student…”)3, thus triggering N400-like AN. However, this processing difficulty was resolved with the appearance of object. This also holds true for non-hierarchical verbs, in which implausible sentences involved animacy violation and induced weaker AN but greater P600 at the noun position.
Verb type comparisons revealed processing differences. In plausible sentence, weak hierarchical verbs relative to strong hierarchical verbs induced greater AN and P600 effects at the verb and noun position respectively, suggesting weak hierarchical verbs’ processing greater difficulties in predicting and integrating with the object noun, probably due to their low semantic restriction. However, strong hierarchical verbs evoked greater P600 at verbs and AN at nouns, demonstrating strong hierarchical verbs’ harder integration with subject and difficult activation of hierarchical noun predicted by strong hierarchical verbs, irrespective of plausibility. These difficulties extended to the early time of the noun position. Similar results were found for implausible sentences with the only difference that strong hierarchical verbs induced larger AN but vacant P600 at verbs, probably due to abnormality and low-expectation of strong hierarchical verbs in implausible conditions (e.g., ? The student often punishes…). As times passed by, weak hierarchical verbs became harder to integrate with hierarchical nouns, reflected by an enhanced P600 at the noun position, which is likely the so-called semantic P600 (Schlesewsky and Schlesewsky, 2008). In addition, the “Strong - Non-hierarchical” amplitudes in plausible and implausible conditions basically take the same tendency as the “Strong - Weak” amplitudes in plausible conditions.
Experiment 2
Experiment 1 revealed strong hierarchical verbs and weak hierarchical verbs were processed quite differently at noun position in both semantically plausible and implausible sentences. However, it remains unknown whether the difference should be attributed to the verbs alone or the conjoint influence of verbs and nouns in sentences, for both the subject and the object in each sentence are hierarchical nouns. In order to further detach the effect of hierarchical verbs from that of hierarchical nouns in sentence processing, Experiment 2 manipulated the context type (biased or neutral), in which strong hierarchical verbs and weak hierarchical verbs co-occurred with non-hierarchical nouns (proper names). By this fashion, we could exclusively investigate different hierarchical verbs’ impact on sentence processing for one thing, and compare distinct contexts’ role in sentence processing for the other, given the logic that context modulates the construal of a sentence (Breton et al., 2014).
Method
Participants
Following the same power analysis procedure as in Experiment 1, the priori calculation for the present 2 × 2 within-subjects design showed that a minimum of 24 participants were required to detect a medium-sized effect (f = 0.25) with 80% power. The same 46 participants from Experiment 1 took part in this experiment, resulting in a statistical power exceeding 0.99. To mitigate the potential repeating effect due to similar stimuli, we counterbalance the order of Experiment 1 and 2 across participants. In addition, Experiment 1 and 2 were conducted at different time to minimize fatigue effects of the participants. After the experiment, the EEG data of three participants were discarded due to the EEG equipment’s malfunctioning.
Design and materials
Experiment 2 adopted a 2 (verb type: strong hierarchical verbs vs weak hierarchical verbs) ×2 (context type: biased context vs. neutral context) within-subjects design, comprising four conditions: strong hierarchical verbs in biased context, weak hierarchical verbs in biased context, strong hierarchical verbs in neutral context and weak hierarchical verbs in neutral context. Each stimulus consisted of two short clauses: the first clause served as the context using the pattern “N1 (Name1)+是 (is)+N2 (Name2)+的 (de) + N3 (identity word)”, which is to clarify the relations between N1 and N2, and the second clause was the target sentence involving either strong hierarchical verbs or weak hierarchical verbs, using the pattern “Adv1 + Adv2/V1 + V2 + N2 (Name2)”, which is to convey the hierarchy information explicitly or implicitly. N3 in biased context denoted a top-down or bottom-up hierarchical relation, whereas N3 in neutral context implicated a relative equal relation (e.g., 同学, classmate), between N1 and N2. As exemplified in Table 2, all the verbs in the second clauses were either strong hierarchical verbs or weak hierarchical verbs used in Experiment 1. The N3 in the first clause was evenly balanced between hierarchical nouns and non-hierarchical nouns, and their number of strokes (t (1, 238) = 0.88, p = 0.379) and character frequency (t (1, 238) = 1.44, p = 0.153) were matched.
Table 2. Exemplar stimuli adopted in Experiment 2.
Verb type | Context type | Examples |
|---|---|---|
Vss | Biased context | 李丹是钱枫的老师, 最近经常体罚钱枫。 Li Dan is Qian Feng’s teacher and often punished him recently. |
Vss | Neutral context | *李丹是钱枫的同学, 最近经常顶撞钱枫。 Li Dan is Qian Feng’s classmate and often defied him recently. |
Vws | Biased context | 李丹是钱枫的老师, 最近经常夸赞钱枫。 Li Dan is Qian Feng’s teacher and often praised him recently. |
Vws | Neutral context | 李丹是钱枫的同学, 最近经常贿赂钱枫。 Li Dan is Qian Feng’s classmate and often bribed him recently. |
Vss strong hierarchical verbs, Vws weak hierarchical verbs, Vns non-hierarchical verbs.
The semantic plausibility of all sentences was evaluated based on a 7-point Likert scale by twenty undergraduates, who did not participate in the current experiment. Descriptive analysis shows that the mean value of the plausibility for strong hierarchical verbs in biased context and neutral context were 5.53 (SD = 0.61) and 2.68 (SD = 0.49) respectively and for weak hierarchical verbs in biased context and neutral context were 5.460 (SD = 0.51) and 5.50 (SD = 0.43). Repeated-measures ANOVA revealed an interaction between verb type and context (F (1, 119) = 1848.08, p < .001, η2 = 0.94, large effect). There were also main effects of context (F (1, 119) = 1264.01, p < .001, η2 = .91, large effect) and verb type (F (1, 119) = 744.90, p < .001, η2 = 0.86, large effect). Simple effects analyses showed that the acceptability score of weak hierarchical verbs is significantly higher than strong hierarchical verbs in neutral context (Strong-Weak: Mdiff = -2.83, SE = 0.06, p < .001, d = 6.08, large effect, 95% CI [−2.94, −2.72]). Strong hierarchical verbs are significantly larger in biased context than in neutral context (Biased- Neutral: Mdiff = 2.85, SE = 0.06, p < .001, d = 6.13, large effect, 95% CI [2.735, 2.97]. However, there were no significant difference between the two verb types in biased context (Strong-Weak: Mdiff = 0.07, SE = 0.07, p = .283, d = 0.13, negligible effect, 95% CI [−0.06, 0.20]). Likewise, weak hierarchical verbs in biased context did not differ from those in neutral context (Biased- Neutral: Mdiff = −0.04, SE = 0.04, p = .326, d = 0.13, negligible effect, 95% CI [−0.13, 0.04).
As a result, 120 sets of sentences (480 in total) were divided into four lists, counterbalanced across participants and conditions. 120 fillers with same sentence pattern but different words were created for repeating in each run, with 30 plausible sentences (e.g., 鸡蛋有益身体健康, 但是不能食用过量。/Eggs are good for your health, but you cannot eat too much.) and 90 implausible sentences with similar structure (e.g., *鸡蛋有益身体健康, 但是不能食用鼻子。/*Eggs are good for your health, but you cannot eat nose). To avoid participants seeing the verbs used in Experiment 1, materials were rearranged in sequence of order and presented randomly.
Procedure, EEG recording and data analysis
The procedure, EEG recording, preprocessing and data analysis of Experiment 2 were identical to those in Experiment 1, except that the repeated ANOVA was conducted to test the effect among verb types, context type. The mean number of bad channels interpolated per participant in this experiment was 0.19 (Max = 2, Min = 0). An average of [2.10] ± [0.62] ICs were rejected per participant during Independent Component Analysis. After the data processing, 22.3% of the trails were rejected. For biased context condition, there were on average 23.43 (Max = 29, Min = 17), 23.76 (Max = 30, Min = 17) trials for strong hierarchical verbs and weak hierarchical verbs respectively; for neutral context, there were 23.52 (Max = 29, Min = 17) and 23.07 (Max = 28, Min = 16) trials for the two conditions, respectively.
Results
Behavioral result
The M (SD) of the accuracy scores of Vss, weak hierarchical verbs in biased context were 87.4% (5.4), 93.2% (5.9) respectively, and 85.9% (9.0), 82.1% (7.9) for neutral context (as shown in Fig. 4). Repeated ANOVAs revealed an interaction between verb and plausibility (F (1, 41) = 11.557, p = .002, η2 = 0.32, large effect). Given this interaction, the main effects were superseded and not interpreted independently. Nevertheless, a main effect of context type was observed (F (1, 41) = 18.48, p < .001, η2 = 0.59, large effect), but no main effect of verb type was found (F (1, 41) = 0.31, p = .582, η2 = 0.01, small effect). To decompose the interaction, simple effects analyses were conducted. These analyses showed that the effect of plausibility was modulated by verb type. Critically, there was no significant difference between strong hierarchical verbs and weak hierarchical verbs when presented in a neutral context (F (1, 41) = 0.040, p = .137, d = 0.30, small effect, 95% CI [0.82, 0.90]), which is in line with our hypothesis. However, the accuracy rate of weak hierarchical verbs was significantly higher than strong hierarchical verbs in biased context (Mdiff = 0.057, p < .001, d = 0.08, negligible effect, 95% CI [0.85, 0.89]). In weak hierarchical verbs condition, the accuracy rate of biased context was significantly higher than neutral context (Mdiff = 0.112, SE = 0.018, p < .001, d = 1.22, large effect, 95% CI [0.08, 0.15]). However, this difference disappeared in strong hierarchical verbs condition (Mdiff = 0.02, SE = .02, p = .385, d = 0.17, negligible effect, 95% CI [0.02, 0.05]).
[See PDF for image]
Fig. 4
Accuracy rate and reaction time for two verb types in two context conditions in Experiment 2.
Errors as well as RTs exceeding ±2.5 SD were rejected from the analysis (19.2%). The M (SD) of the accuracy of strong hierarchical verbs, weak hierarchical verbs in biased context were 711.07 ms (70.62), 758.93 ms (82.85) respectively, and 704.28 ms (79.15), 724.05 ms (82.17) for neutral context (as shown in Fig. 4). There was no interaction between verb and context (F (1, 42) = 2.77, p = .104, η2 = .06, medium effect). Furthermore, there were no significant main effects of verb type (verb effect: F (1, 42) = 3.33, p = .075, η2 = .07, medium effect) or context type (F (1, 42) = 0.32, p = .57, η2 = .01, small effect).
ERP results
Verbs
300–500 ms
There was no interaction between verb and context (F (1, 239) = 0.58, p = .448, η2 = .002, small effect) or main effect of verb type (F (1, 239) = 0.50, p = .481, η2 = .16, large effect). However, there appeared a main effect of context (F (1, 239) = 9.55, p = .002, η2 = .17, large effect), in which neutral context induced larger negativity amplitudes, irrespective of verb type (as indicated in Fig. 5).
[See PDF for image]
Fig. 5
Grand-average event-related potentials (ERPs) and topographic maps showing the verb plausibility effect at the verb position in Experiment 2.
Waveforms depict ERPs linear derivation of all the electrodes used for analysis (F1, F3, F5, FC1, FC3, FC5, F2, F4, F6, FC4, FC6, P1, P3, P5, CP1, CP3, CP5, P2, P4, P6, CP2, CP4, CP6). The shaded areas around the waveforms represent the standard error of the mean (SEM). Topographic maps illustrate the voltage distribution across the scalp during the 300–500 ms (AN component) and 500–650 ms (P600 component) time windows. Vss = strong hierarchical verbs; Vws = weak hierarchical verbs; Vns = non-hierarchical verbs.
500–650 ms
The analysis for the 500–650 ms window revealed a significant interaction between verb type and context type (F (1, 239) = 119.85, p < .001, η2 = .33, large effect). Given this interaction, the significant main effects of verb type (F (1, 239) = 33.24, p < .001, η2 = .17, large effect) and context type (F (1, 239) = 8.08, p = .005, η2 = .06, medium effect) were not interpreted independently. Simple effects analyses showed that strong hierarchical verbs induced larger positive amplitudes than weak hierarchical verbs in neutral context (as shown in Fig. 5)(Strong - Weak: Mdiff = 2.04μV, SE = 0.17, p < .001, d = 2.26, large effect, 95% CI [1.72, 2.37]), however the two verb types did not differ statistically in biased context (Strong - Weak: Mdiff = 0.34μV, SE = 0.20, p = .09, d = −0.31, small effect, 95% CI [−0.74, 0.06]). Weak hierarchical verbs were more positive-going in biased context than in neutral context (Biased-Neutral: Mdiff = 0.81μV, SE = 0.16, p <.001, d = 0.31, small effect, 95% CI [0.49, 1.14]). Conversely, the amplitudes of strong hierarchical verbs were larger in neutral context than in biased context (Neutral-Biased: Mdiff = 1.57μV, SE = 0.18, p < .001, d = −0.55, medium effect, 95% CI [−1.93, −1.22]).
Nouns
300–500 ms
ERP effects at the noun position (300–500 ms) were similar to those in 500–750 ms time window at the verb position. There was a significant interaction between verb and context (F (1, 239) = 113.138, p < .001, η2 = .316, large effect). Given this interaction, the significant main effects of verb (F (1, 239) = 104.09, p < .001, η2 = .30, large effect) and context (F (1, 239) = 8.50, p = .004, η2 = .08, medium effect) were not interpreted independently. Simple effects analysis revealed that, although there was no difference between strong hierarchical verbs and weak hierarchical verbs in biased context (Strong - Weak: Mdiff = −0.20μV, SE = 0.17, p = .223, d = −0.11, negligible effect, 95% CI [−0.53, 0.12]), strong hierarchical verbs evoked greater negativity than weak ones in the neutral context (as shown in Fig. 6) (Strong - Weak: Mdiff = -2.38μV, SE = 0.16, p < .001, d = −1.31, large effect, 95% CI [-2.70, -2.07]). Weak hierarchical verbs exhibited more negative amplitudes in biased context versus neutral context (Biased-Neutral: Mdiff = −0.74μV, SE = 0.15, p < .001, d = −0.45, small effect, 95% CI [−1.02, −0.45]).In contrast, strong hierarchical verbs showed the reverse pattern (Biased-Neutral: Mdiff = 1.45μV, SE = 0.17, p < .001, d = 0.83, large effect, 95% CI [1.11, 1.78]).
[See PDF for image]
Fig. 6
Grand-average event-related potentials (ERPs) and topographic maps showing the verb plausibility effect at the noun position in Experiment 2.
Waveforms depict ERPs linear derivation of all the electrodes used for analysis (F1, F3, F5, FC1, FC3, FC5, F2, F4, F6, FC4, FC6, P1, P3, P5, CP1, CP3, CP5, P2, P4, P6, CP2, CP4, CP6). The shaded areas around the waveforms represent the standard error of the mean (SEM). Topographic maps illustrate the voltage distribution across the scalp during the 300–500 ms (AN component) time window. Vss = strong hierarchical verbs; Vws = weak hierarchical verbs; Vns = non-hierarchical verbs.
Discussion
The results from experiment 2 are complimentary to the findings of experiment 1, further confirming the processing difference between strong hierarchical verbs and weak hierarchical verbs when they were placed in different contexts. As predicted, context exerts an influence on the processing of hierarchical verbs.
Different from neutral context, biased context enhanced the processing of sentences with strong hierarchical verbs and weak hierarchical verbs. Lexically, strong hierarchical verbs elicited greater P600 and AN in neutral context than that in biased context at the verb and noun positions. Similarly, weak hierarchical verbs evoked an enhanced AN effect at the verb position only in neutral context. Nevertheless, weak hierarchical verbs became easier to process in neutral context at the noun position, with diminished AN and P600. These findings suggest that biased context facilities the processing of hierarchical verbs, while neutral context supports the integration of weak hierarchical verbs with nouns, highlighting the distinct impact of context on verbs endowed with explicit or implicit hierarchical meaning during Chinese sentence comprehension.
Contexts exert more impacts on weak hierarchical verbs than on strong hierarchical verbs. Specifically, biased context seems to weaken the processing difference between strong hierarchical verbs and weak hierarchical verbs as indicated by the vacant AN and P600 at the verb and noun positions. But the pattern changed in neutral contexts. In the early time of verbs’ processing, no AN effect was observed, suggesting the activations of strong hierarchical verbs and weak hierarchical verbs are similar. Conversely, in the later time, strong hierarchical verbs evoked greater P600 at verb position and N400 at noun position, demonstrating that strong hierarchical verbs were harder to process than weak hierarchical verbs under neutral context, probably due to the unexpected heterogenous hierarchical information between context and strong hierarchical verbs.
General discussion
This study leveraged two ERP experiments (based on the SVO structure mapping Who Does Whom) to investigate how verb types (hierarchical or non-hierarchical) affect Chinese sentence processing and how context types (biased or neutral) modulate the processing of sentences involving explicit or implicit hierarchical verbs. The results reveal three findings: firstly, strong hierarchical verbs elicited greater P600 at verb position and AN at noun position than non-hierarchical verbs, indicating more difficult processing of hierarchy relative to general lexical information; secondly, weak hierarchical verbs relative to strong hierarchical verbs evoked greater AN at verb position and greater P600 at the noun position, suggesting implicit hierarchy’s more demands in early verb activation and object integration; thirdly, biased contexts removed the inherent differences between strong and weak hierarchical verbs while neutral contexts amplified the differences between which strong hierarchical elicited larger P600 at verbs and AN at nouns than weak ones. These results converge to indicate that verbs’ hierarchy influences sentence processing and the processing pattern varies with context.
Social hierarchy distinct from lexical semantic knowledge
The most fundamental finding across our experiments is that verbs encoding social hierarchy, particularly strong hierarchical verbs, are processed differently from those conveying general lexical semantics.
These ERP patterns reflect variation in cognitive load during lexical activation and thematic integration, arising from differences in processing demands. Specifically, a participant’s task at the verb position involved verb’s lexical meaning retrieval and subject-verb (SV) integration, whereas the task at noun position included noun’s lexical meaning retrieval and subject-verb-object (SVO) integration so as to access Who does Whom successfully. In Experiment 1, strong hierarchical verbs evoked larger AN than non-hierarchical verbs. This N400-like component was related to lexical-semantic access difficulty (Kutas and Federmeier, 2011; Lau et al., 2009), which has been discovered in Chinese social hierarchy violation (Mu et al., 2015; Shi et al., 2022). Strong hierarchical verbs form natural sentences with hierarchical nouns, aligning with expectations and collocation patterns. This prediction pre-activated the upcoming noun and reduced cognitive activation load as a result, which contradicts our hypothesis (H1). Conversely, non-hierarchical verbs paired with hierarchical nouns created heterogeneous semantics, complicating expectation (larger AN at the verbs) and formulating tricky integration (larger P600 at the noun). In this case, the objects of non-hierarchical verbs were perceived more easily (reduced AN at the nouns) due to lack of hierarchical restrictions but demanded greater resources for thematic integration.
These differences were also modulated by semantic abstraction and reversibility. Strong hierarchical verbs, endowed with abstract concept, facilitated processing compared to concrete non-hierarchical verbs, which was consistent with findings that concrete verbs typically elicit larger N400 amplitudes than abstract ones (Tsai et al., 2009; Xia et al., 2012). Reversibility further differentiated noun-position processing: reversible sentences (e.g., non-hierarchical verbs in “Mom misses her son”) are harder than irreversible (e.g., strong hierarchical verbs in “The teacher punishes students”), inducing P600 for integration difficulties (Richardson et al., 2010; Vercesi et al., 2020; Kuperberg et al., 2007). Put another way, the SVO (mapped onto Who Does Whom) contains a strong hierarchical verb and thus becomes an irreversible structure, enhancing predictability and thematic role assignment, as indicated by the reduced P600 at the nouns.
Crucially, the violation of hierarchical expectations (e.g., “The student often punishes…”) was detected earlier (at the verbs) than general animacy violations, as indicated by increased P600 at verbs. This indicates hierarchical verbs’ unique capacity for schema-based violation prediction, establishing preliminary relations at verbs under hierarchical subject influence, which extends Shi et al. (2022) and aligns with Rayner et al. (2004) and Shi et al. (2024) on semantic pre-processing. Conversely, animacy violation encoded in non-hierarchical verbs is more obvious, thus creating a rapid breakdown of semantic integration that push participants to abandon the original prediction and shift to syntactic reanalysis, thus evoked the attenuated AN and enhanced P600. This keeps concord with the dynamic competition between automatic conflict detection (weaker N400) and controlled reanalysis (greater P600) in high-stakes incoherence (Brouwer et al., 2012). These results were in accordance with our hypothesis (H1) and presumably reflect the distinct neural representations: general lexical knowledge exists in the core mental lexicon, while social hierarchy is culture-specific and appears to reside as an extra and unique part in the lexicon.
Implicit hierarchy vs. explicit hierarchy
Our data also reveal that even a weak hierarchical verb can profile implicit (albeit relatively small) hierarchy in appropriately triggered condition. Unlike prior studies, we selected weak hierarchical verbs, where subject (S) and the object (O) are not fixed and anticipated but become hierarchical and irreversible upon the introduction of an object, contrasting strong hierarchical verbs’ explicit lexical hierarchy.
This fundamental difference explains the divergent ERP patterns. Strong hierarchical verbs, endowed with explicit hierarchical meaning, elicited effects indicative of immediate access and integration of hierarchical information. For instance, a reduced AN emerged earlier for strong hierarchical verbs at the verbs, implying comprehenders prioritize access to explicit hierarchical cues as soon as the verb is encountered. Crucially, both verb types showed comparable suppression of the P600 effect in implausible condition during subject-verb integration, suggesting implicit hierarchical relations were established similarly to explicit hierarchical relations during subject-verb integration, compatible with our hypothesis (H1). For example, in a sentence such as “*老师经常请教/咨询,The teacher often consult…”4, the subject-verb relation is similarly salient for both explicit and implicit hierarchy. However, this effect disappears in plausible condition, likely due to overlapping lexical properties.
The two types of hierarchical representation also differentially influence expectations about the upcoming object. The weaker AN observed for weak hierarchical verbs at the nouns supports our theoretical classification that weak hierarchical verbs have lower restrictions on the object than strong hierarchical verbs, thereby facilitating its activation. Notably, the enhanced P600 effect of weak hierarchical verbs at the nouns was absent in Shi et al.’s (2022) study, which likely reflects thematic integration, termed “semantic P600” (Schlesewsky and Schlesewsky, 2008) and semantic-pragmatic reasoning as observed in implicit hierarchy like horrifies (e.g., Jiang et al., 2013; Breton et al., 2014; Breton et al., 2019; Miao et al., 2022) and non-literal language studies (Coulson and Van Petten, 2007; Regal et al., 2010). Because weak hierarchical verbs involve dynamic and context-dependent role assignments. They require additional inferential processing once a hierarchical noun is encountered, thus yielding larger noun-position P600. In contrast, strong hierarchical verbs’ fixed roles eased decoding and led to a reduced P600. Overall, these results align with explicit (Shi et al., 2022) and implicit hierarchy research (Breton et al., 2014; Breton et al., 2019; Miao et al., 2022), contradicting to our hypothesis (H1).
Context modulating the processing pattern of different hierarchical verbs involved sentences
As expected, context (biased context vs neutral context) modulates sentence comprehension, which varies with the type of hierarchical information encoded in verbs (strong hierarchical verbs vs. weak hierarchical verbs).
Biased context, rich with hierarchical cues, reduced the processing difference between two hierarchical verbs (Experiment 2) by reducing weak hierarchical verbs’ activation difficulty (cf. the 300–500 ms data at the verb position in Experiment 1 and Experiment 2). This changed pattern is consistent with Ferretti et al. (2007) wherein attenuated N400 was found in semantic congruent or predicable context. By contrast, strong hierarchical verbs sentences look insensitive to biased context at the early time, probably relying on inherent hierarchy to predict patients. In the late stage of verbs’ processing, biased context promotes strong hierarchical verbs’ integration with subject (cf. the 500–650 ms data at the verb position in experiment 1 and experiment 2), probably due to the hierarchical scenarios implied by identity nouns (e.g., teacher) in biased context. This also explains why biased context did not differentiate between strong hierarchical verbs and weak hierarchical verbs.
In contrast, neutral context lacked such schematic support, amplifying the differences between verb types. The larger P600 for strong hierarchical verbs is particularly telling at the verb position, as unlocked in both Experiment 1 and Experiment 2. This suggests that without contextual support, the integration of their strong, directive hierarchical relations would become a more effortful process, presumably because it introduces a salient social structure into a neutral setting, creating a mild pragmatic mismatch (e.g., “Zhang San is Li Si’s friend and he often punishes …”). To the opposite, neutral context seems to diminish weak hierarchical verbs’ activation difficulty (cf. Experiment 1’s larger AN vs. Experiment 2’s vacant effect). This contradicts our hypothesis (H2), which can be explained by the context congruency effect (Grzybowski et al., 2014) whereby a mismatch between context and target elicits N400 (e.g., Cacioppo et al., 1996; Calvo et al., 2006; Kissler and Koessler, 2011), amplified by lower predictability (Delogu et al., 2019). In experiment 2, neutral context used non-hierarchical nouns (e.g., 同学/classmate), which was more typical or predicable for non-hierarchical verbs, making weak hierarchical verbs more semantically congruent and predicable than strong hierarchical verbs, thus eliciting weaker anterior N400. Strong hierarchical verbs evoked larger AN at nouns, marking the difficulty in resolving conflicts between the verb’s directional hierarchical relations and neutral context’s unbiased or equal status. This conforms to AN’s role in Shi et al.’s study (2022). Violations elicited expectancy-related N400 (Kutas and Hillyard, 1980; Kutas and Federmeier, 2011), thus larger AN at nouns was evoked.
Overall, context does not merely facilitate processing in a general way. It provides a hierarchical schema that pre-activates social roles and relations, consequently assisting comprehenders in establishing the match or mismatch between this contextual frame and the hierarchical information encoded by the verbs.
Conclusion
This ERP study was conducted to investigate how Chinese SVO sentences are comprehended by Chinese native speakers when the sentences involved verbs of different social hierarchy. Results showed that verbs’ hierarchy influenced sentence processing and the processing was modulated by context type. Compared with non-hierarchy, implicit hierarchy (by weak hierarchical verbs) caused greater difficulty in lexical activation and syntactic-semantic integration, while explicit hierarchy (by strong hierarchical verbs) incurred the load of hierarchy establishment initially. In our experiments, strong hierarchical verbs elicited larger posterior-P600 at verbs and AN at nouns while weak hierarchical verbs evoked larger AN at verbs and posterior-P600 at nouns. However, this difference vanished in biased context (indexed by the vacant AN in Experiment 2), suggesting that biased context alleviates the processing difficulty of weak hierarchical verbs. By comparison, neutral contexts amplified the differences between which strong hierarchical elicited larger P600 at verbs and AN at nouns, spotlighting a congruency effect between context and hierarchy concept.
These findings converge to demonstrate that social hierarchy is lexically inherent in Chinese hierarchical verbs, making hierarchical sentences more difficult to process for they require more cognitive resources for words’ retrieval, inference, and thematic integration. On this basis, lexical meaning and thematic relations jointly contribute to decoding hierarchy information, with congruency established between verbs and arguments as well as between context and verbs. The findings also justify the dichotomy of Chinese hierarchical verbs we propose and provide novel insights into the neural basis of social hierarchy in language comprehension.
Despite the above advancements, there are some limitations in this study. Firstly, specific brain regions relating to Chinese verbs’ social hierarchy remains unclear. Secondly, the target word was designed in the final position of the sentences, potentially causing a wrap-up effect. Thirdly, hierarchical verbs imply affective denotations (e.g., bribe) that may influence participants’ response, though all verbs include positive, negative and neutral semantic domains. Future research should integrate fMRI or fNIRS with varied structures to map spatiotemporal neural correlates and enhance ecological validity.
Acknowledgements
This study was sponsored by a grant from the Major Project of National Social Science Foundation of China (22&ZD298). Warmest thanks for all professors and students at school of linguistic sciences and arts at Jiangsu Normal University for support in experimental discussion and experimental operation. Also, genuine thanks to all the participants who took part in the ERP experiments and questionnaires test.
Author contributions
YL, JG, TZ, ZZ conceived the study. YL, JG and ZZ performed the experiments. YL and WZ collated and analyzed the data. YL drafted the first manuscript, which TZ and YX revised. All authors edited the final version of the manuscript and have approved it for publication.YL, JG, and TZ contributed equally to the work and hence share the first authorship.
Data availability
The data and research materials of two experiments are available at: https://figshare.com/s/28a6163b91a3e244d691.
Competing interests
The authors declare no competing interests.
Ethical approval
This study was approved by the Ethics Committee of Qufu Normal University (Approval ID: 2023128; Date of Approval: 26 September 2023). All procedures involving human participants were performed in accordance with the ethical standards of the University’s research ethics committee and with the 1964 Helsinki Declaration and its later amendments. The ethical approval required voluntary participation and informed consent. The approval encompasses the entire research protocol for this study, including its aim, participant recruitment, experimental procedures, and data management.
Informed consent
Informed consent was obtained in written form from all individual participants prior to the commencement of data collection on 1 November 2023. The written consent included the nature and purpose of the research, the procedures involved, the assurance of anonymity, the use of data for academic purposes, and the potential for publication. Participants were informed that their participation was entirely voluntary and that they had the right to withdraw from the study at any stage without penalty. The study posed no foreseeable risks of physical, emotional, or psychological harm, and no privacy concerns were identified.
Lan (1999) and Zou (2009) believed that the spatial concepts in Chinese can be mapped onto the social hierarchy domain through spatial metaphors, that is, “up” indicates higher social hierarchy, while “down” implies lower social hierarchy.
2According to Williams (2015), certain verbs impose strict requirements on the capacity differences between the subject (Agent) and the object (Patient), for instance: “rescue” highlights the agent’s significantly greater capacity (physical or otherwise) than the patient, emphasizing the agent’s superior position in providing aid.
3In Chinese, the verb “咨询” (to consult) typically implies that the subject (the one obtain consultations from others) has less knowledge or lower status than the object (the one being consulted). Its subjects are therefore usually students, trainees, or those lower in the hierarchy of [CONSULTING] event. Consequently, pairing the subject like “老师” (teacher) with the object like “学生” (the student) is highly uncommon in practice, making a phrase like “The teacher often consults the student…” sound unnatural, though not strictly grammatically incorrect.
4In Chinese, “请教” and “咨询” share the meaning of “consult”. The difference is that the former requires the subject to have a lower social hierarchical status than the object. However, this restriction is not necessarily compulsory for “咨询”.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
Agha A (2007) Language and social relations. Cambridge University Press
Altmann, GTM; Kamide, Y. Incremental interpretation at verbs: Restricting the domain of subsequent reference. Cognition; 1999; 73, pp. 247-264.1:STN:280:DC%2BD3c%2FlsVajtg%3D%3D [DOI: https://dx.doi.org/10.1016/S0010-0277(99)00059-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10585516]
Binder, JR; Conant, LL; Humphries, CJ et al. Toward a brain-based componential semantic representation. Cogn Neuropsychol; 2016; 33,
Breton, A; Jerbi, K; Henaff, MA et al. Face the hierarchy: ERP and oscillatory brain responses in social rank processing. PLoS One; 2014; 9,
Breton, A; Ligneul, R; Jerbi, K et al. How occupational status influences the processing of faces: an EEG study. Neuropsychologia; 2019; 122, pp. 125-135. [DOI: https://dx.doi.org/10.1016/j.neuropsychologia.2018.09.010] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30244000]
Brouwer, H; Fitz, H; Hoeks, J. Getting real about semantic illusions: Rethinking the functional role of the P600 in language comprehension. Psychophysiology; 2012; 49,
Cacioppo, JT; Crites, SL; Gardner, WL. Attitudes to the right: evaluative processing is associated with lateralized late positive event-related brain potentials. Pers Soc Psychol Bull; 1996; 22, pp. 1205-1219. [DOI: https://dx.doi.org/10.1177/01461672962212002]
Caffarra, S; Wolpert, M; Scarinci, D et al. Who are you talking to? The role of addressee identity in utterance comprehension. Psychophysiology; 2020; 57,
Calvo, MG; Dolores Castillo, M; Fuentes, LJ. Processing of “unattended” threat-related information: Role of emotional content and context. Cognit Emotion; 2006; 20, pp. 1049-1074. [DOI: https://dx.doi.org/10.1080/02699930500298351]
Cohen J (1988). Statistical power analysis for the behavioral sciences (2nd ed). Lawrence Erlbaum Associates
Coulson, S; Van Petten, C. A special role for the right hemisphere in metaphor comprehension? ERP evidence from hemifield presentation. Brain Res; 2007; 1146, pp. 128-145.1:CAS:528:DC%2BD2sXkt1ejtr4%3D [DOI: https://dx.doi.org/10.1016/j.brainres.2007.03.008] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17433892]
Cummins, DD. How the social environment shaped the evolution of mind. Synthese; 2000; 122, pp. 3-28. [DOI: https://dx.doi.org/10.1023/A:1005263825428]
Delogu, F; Brouwer, H; Crocker, MW. Event-related potentials index lexical retrieval (N400) and integration (P600) during language comprehension. Brain Lang; 2019; 193, pp. 58-69. [DOI: https://dx.doi.org/10.1016/j.bandl.2019.04.008]
Delorme, A; Makeig, S. EEGLAB: an open-source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods; 2004; 134,
Erdfelder, E; Faul, F; Buchner, A. GPOWER: A general power analysis program[J]. Behav Res Methods Instrum Comput; 1996; 28,
Farrow, TFD; Jones, SC; Kaylor-Hughes, CJ et al. Higher or lower? The functional anatomy of perceived allocentric social hierarchies. Neuroimage; 2011; 57,
Ferretti, TR; Kutas, M; Mcrae, K. Verb aspect and the activation of event knowledge. J Exp Psychol; 2007; 33,
Forgács, B; Bohrn, I; Baudewig, J et al. Neural correlates of combinatorial semantic processing of literal and figurative noun-noun compound words. Neuroimage; 2012; 63,
Graves, WW; Binder, JR; Desai, RH et al. Neural correlates of implicit and explicit combinatorial semantic processing. Neuroimage; 2010; 53,
Greenhouse, SW; Geisser, S. On methods in the analysis of profile data. Psychometrika; 1959; 24, pp. 95-112.103783 [DOI: https://dx.doi.org/10.1007/BF02289823]
Grzybowski, SJ; Wyczesany, M; Kaiser, J. The influence of context on the processing of emotional and neutral adjectives – an ERP study. Biol Psychol; 2014; 99, pp. 137-149. [DOI: https://dx.doi.org/10.1016/j.biopsycho.2014.01.002] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24495849]
Huth, AG; de Heer, WA; Griffiths, TL et al. Natural speech reveals the semantic maps that tile human cerebral cortex. Nature; 2016; 532,
Ji, L; Cai, L; Ji, A. Respectfulness-processing revisited: An ERP study of Chinese sentence reading. PLoS One; 2022; 17,
Jiang, X; Li, Y; Zhou, X. Is it over-respectful or disrespectful? Differential patterns of brain activity in perceiving pragmatic violation of social status information during utterance comprehension. Neuropsychologia; 2013; 51,
Kamide, Y; Altmann, GTM; Haywood, SL. The time-course of prediction in incremental sentence processing: Evidence from anticipatory eye movements. J Mem Lang; 2003; 49,
Karafin, MS; Tranel, D; Adolphs, R. Dominance attributions following damage to the ventromedial prefrontal cortex. J Cogn Neurosci; 2004; 16, pp. 1796-1804. [DOI: https://dx.doi.org/10.1162/0898929042947856] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15701229]
Kissler, J; Koessler, S. Emotionally positive stimuli facilitate lexical decisions—An ERP study. Biol Psychol; 2011; 86, pp. 254-264. [DOI: https://dx.doi.org/10.1016/j.biopsycho.2010.12.006] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21184799]
Kuperberg, GR. Neural mechanisms of language comprehension: Challenges to syntax. Brain Res; 2007; 1146, pp. 23-49.1:CAS:528:DC%2BD2sXkt1ejtrY%3D [DOI: https://dx.doi.org/10.1016/j.brainres.2006.12.063] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17400197]
Kutas, M; Federmeier, KD. Thirty years and counting: finding meaning in the N400 component of the event-related brain potential (ERP). Annu Rev Psychol; 2011; 62, pp. 621-647. [DOI: https://dx.doi.org/10.1146/annurev.psych.093008.131123] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20809790][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4052444]
Kutas, M; Hillyard, SA. Reading senseless sentences: brain potentials reflect semantic incongruity. Science; 1980; 207,
Kwon, N; Sturt, P. When social hierarchy matters grammatically: Investigation of the processing of honorifics in Korean. Cognition; 2024; 251, [DOI: https://dx.doi.org/10.1016/j.cognition.2024.105912] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39116506]105912.
Lan, C. Cong renzhi jiaodu kan Hanyu de kongjian yinyu (A cognitive perspective on spatial metaphors in Chinese). Foreign Lang Teach Res; 1999; 4, pp. 7-15.
Lattner, S; Friederici, AD. Talker’s voice and gender stereotype in human auditory sentence processing - evidence from event-related brain potentials. Neurosci Lett; 2003; 339,
Lau, EF; Almeida, D; Hines, PC et al. A lexical basis for N400 context effects: Evidence from MEG. Brain Lang; 2009; 111,
Li, X; Shu, H; Liu, Y et al. Mental representation of verb meaning: Behavioral and electrophysiological evidence. J Cogn Neurosci; 2006; 18, pp. 1774-1787. [DOI: https://dx.doi.org/10.1162/jocn.2006.18.10.1774] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17014380]
Lopez-Calderon, J; Luck, SJ. ERPLAB: a toolbox for the analysis of event-related potentials. Front Hum Neurosci; 2014; 8, 213. [DOI: https://dx.doi.org/10.3389/fnhum.2014.00213] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24782741][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3995046]
Luck SJ (2014) An introduction to the event-related potential technique (2nd ed.). MIT Press
Ma, QZ. V lai/qu” yu xiandai hanyu dongci de zhuguan fanchou (The directional complements “lái”/“qù” and the subjective category of modern Chinese verbs). Chin Res; 1997; 3, 8.
Makeig, S; Jung, TP; Bell, AJ et al. Blind separation of auditory event-related brain responses into independent components. Proc Natl Acad Sci USA; 1997; 94,
Miao, C; Li, X; Derrington, E et al. ERPs responses to dominance features from human faces. Sci Rep; 2022; 12,
Mu, Y; Kitayama, S; Han, S et al. How culture gets embrained: Cultural differences in event-related potentials of social norm violations. Proc Natl Acad Sci USA; 2015; 112,
Olejnik, S; James, A. Generalized Eta and Omega squared statistics: measures of effect size for some common research designs. Psychol Methods; 2003; 8,
Oldfield, RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia; 1971; 9,
Paczynski, M; Kuperberg, GR. Multiple influences of semantic memory on sentence processing: Distinct effects of semantic relatedness on violations of real-world event/state knowledge and animacy selection restrictions. J Mem Lang; 2012; 67,
Pexman, PM; Diveica, V; Binney, RJ. Social semantics: the organization and grounding of abstract concepts. Philos Trans R Soc B; 2023; 378,
Park MR (1991) Conflict Avoidance in Social Interaction: A Sociolinguistic Comparison of the Korean and Japanese Honorific Systems. In: Hoji, H (ed) Japanese/Korean Linguistics, 1, CSLI Publications Stanford, p 111–128
Rayner, K; Warren, T; Juhasz, BJ et al. The effect of plausibility on eye movements in reading. J Exp Psychol Learn Mem Cogn; 2004; 30,
Richardson, FM; Thomas, MS; Price, CJ. Neuronal activation for semantically reversible sentences. J Cogn Neurosci; 2010; 22, pp. 1283-1298. [DOI: https://dx.doi.org/10.1162/jocn.2009.21277] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19445603][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2860570]
Regal, S; Coulson, S; Gunter, T. The communicative style of an addresser can affect language comprehension? ERP evidence from the comprehension of irony. Brain Res; 2010; 1311, pp. 121-135.1:CAS:528:DC%2BC3cXot1ak [DOI: https://dx.doi.org/10.1016/j.brainres.2009.10.077]
Schjelderup-Ebbe, T. Beiträge zur sozialpsychologie des haushuhns. Z Psychol; 1922; 88, pp. 225-252.
Schlesewsky, M; Schlesewsky, M. An alternative perspective on “semantic P600” effects in language comprehension. Brain Res Rev; 2008; 59,
Shi, Y; Zhou, T; Zhu, Z et al. The social hierarchical restrictions of Chinese verbs rapidly guide online thematic role assignment in comprehension. Brain Lang; 2022; 232, [DOI: https://dx.doi.org/10.1016/j.bandl.2022.105161] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35863276]105161.
Shi YJ, Zhou TQ, Zhao SM et al. (2024) The immediate integration of semantic selectional restrictions of Chinese social hierarchical verbs with extralinguistic social hierarchy information in comprehension. Lang Cogn 1–15. https://doi.org/10.1017/langcog.2024.11
Tokimoto, S; Miyaoka, Y; Tokimoto, N. An EEG analysis of honorification in Japanese: Human hierarchical relationships coded in language. Front Psychol; 2021; 12, [DOI: https://dx.doi.org/10.3389/fpsyg.2021.549839] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33762986][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7982684]549839.
Tsai, PS; Yu, BHY; Lee, CY et al. An event-related potential study of the concreteness effect between Chinese nouns and verbs. Brain Res; 2009; 1253, pp. 149-160.1:CAS:528:DC%2BD1MXhtlSmu70%3D [DOI: https://dx.doi.org/10.1016/j.brainres.2008.10.080] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19059223]
Van Berkum, JJ; Van den Brink, D; Tesink, C et al. The neural integration of addresser and message. J Cogn Neurosci; 2008; 20, pp. 580-591. [DOI: https://dx.doi.org/10.1162/jocn.2008.20054] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18052777]
Vercesi, L; Sabnis, P; Finocchiaro, C et al. The role of the l-IPS in the comprehension of reversible and irreversible sentences: an rTMS study. Brain Struct Funct; 2020; 225, pp. 2403-2414. [DOI: https://dx.doi.org/10.1007/s00429-020-02130-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32844277][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7544754]
Williams A (2015) Agent and patient. In: Peter KA, JJ Moravcsik (eds) Arguments in syntax and semantics. Cambridge University Press, Cambridge, p 141–162
Xia QS, Lv Y, Shi F (2012) Hanyu mingci he dongci de yuyi jiagong zhong jutixing xiaoying he cilei xiaoying de ERP yanjiu (An ERP study on the concreteness effect and grammatical class effect on the semantic processing of Chinese noun and verb). Nankai Linguist (1): 110–119
Yang X (2012) Xiandai Hanyu Shehui Dengji Guanxi Dongci Yanjiu (A study on modern Chinese social hierarchical verbs). Beijing Language and Culture University, Beijing
Zhang, GX. “V_shuang + N_shuang” duanyu de lijie yinsu (The understanding factors of the phrase “V_double + N_double”). Chin Lang; 1997; 3, pp. 176-186.1997sref.book...Z
Zou WJ (2009) Xiandai Hanyu dongci de fangxiang fenlei ji jieshi (Direction classification and interpretation of modern Chinese verbs). Shanghai International Studies University, Shanghai
© The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.