Experimental investigation on quantifier scope in Chinese relative clauses

1 Introduction

Unlike English, Mandarin Chinese (henceforth Chinese) has been widely claimed to exhibit scope rigidity (Aoun and Li 1989; Huang 1982; Lee 1986, and thereafter), in that it only permits the surface scope interpretation determined by the surface structural relationship between an existential quantifier (e.g., a/an) and universal quantifiers (e.g., every) in simple active transitive clauses; example (1) shows the crosslinguistic difference between English and Chinese. Inverse scope interpretation, for which the scope relation between quantificational phrases (QPs) is neither directly read off their linear word order nor their surface syntactic positions, is possible in English but not in Chinese.^[1]

While there is empirical support for scope rigidity in Chinese simple active transitive clauses (Chien and Wexler 1989; Fang 2023; Scontras et al. 2014, 2017]; but see Zhou and Gao 2009), explanations vary for the lack of inverse scope, often couched in different theoretical frameworks (e.g., Aoun and Li 1989, 1993]; Huang 1981, 1982]; Larson and Wu 2018; Wu et al. 2018; Wu 2019; Wu and Larson 2023). For instance, Aoun and Li (1989, 1993] postulated that scope rigidity in sentences like (1b) can be accounted for by the Scope Principle:

In sentence (1b), where the subject phrase ‘a child’ c-commands the object phrase ‘every tree’ but not the other way round, the Scope Principle only yields the surface scope interpretation.

However, beyond simple active transitive sentences, various syntactic contexts have been observed to show scope ambiguity with both surface and inverse scope readings available in Chinese. Such syntactic contexts include ditransitives (Gan and Tsai 2020; Liu and Wu 2016), passives (Yang and Wu 2020), relative clauses (Aoun and Li 2003; Chen 2020, 2024]; Huang 1982; Wu et al. 2018), thetic sentences (Wu 2023b), and adverbial clauses (Wu 2023a). Nevertheless, much of the research on the extent of scope rigidity in complex clauses lacks extensive empirical support and often remains confined to theoretical debates. This paper focuses on investigating the presence (or lack thereof) of scope ambiguities in Chinese relative clauses (RCs), with the goal of providing the first of this kind of large-scale experimental evidence and in-depth discussion on related theoretical debates.

Consider the RCs in (3). As a note, scope relations in this paper are defined in terms of the c-command structural relationship between universal and existential quantifier phrases. For example, every > 3 is the surface scope reading for (3a) because ‘every book’, as the head of the relative clause, is structurally higher than ‘three men’. Unlike English, where the relative clause (RC) follows the head it modifies, the RC precedes the head in Chinese. There are two types of Chinese RCs, differentiated by whether the head of the RC is the grammatical subject or the object of the RC. Example (3a) illustrates an object relative clause (ORC), in which the head quantifier phrase of the RC is co-indexed with a corresponding gap in the RC’s object position. Conversely, example (3b) demonstrates a subject relative clause (SRC), where the head quantifier phrase of the RC corresponds to a gap in the RC’s subject position.

For quantifier scope, Chinese relative clauses seem to exhibit greater scope flexibility, with the embedded and external QPs able to take scope over each other in both SRCs and ORCs, contrary to the scope rigidity claim for Chinese simple transitive sentences. Take the SRC (3b) as an example. The surface scope reading (three > every) is straightforward as the RC head c-commands the RC subject (Figure 1), making the three > every reading easier to access and process (Anderson 2004). As for the every > three reading of (3b), there have been many debates over how the reading is derived. Example (3b) is structurally represented in Figure 1. Some studies argue for long-distance quantifier raising of the RC object ‘every language’ out of the RC to c-command the QP head ‘three students’ (Huang 1982; Wang 2023). However, it is not fully clear why long-distance QR is only available to relative clauses but not simple active transitives in Chinese. Aoun and Li (2003) propose a movement of the RC to a higher position, based on their observation that the scope ambiguity is not available when dou ‘all’ occurs after the embedded RC subject, but this crucial observation is challenged by the experimental results of Chen (2020, 2024]. Another proposal to derive the every > three reading in (3b) is proposed by Wu et al. (2018) and Larson and Wu (2018). They argue that Chinese relative clauses have a smaller clausal size than matrix simple transitive clauses, and the head QP of the RC ‘three students’ is first reconstructed back to the RC and then goes through a quantifier raising over ‘every language’ within the relative clause.

Figure 1:

The syntactic structure of the Chinese subject relative clause in (3b).

However, Lee (2019) argues against Larson and Wu’s (2018) account. He argues that the three > every reading in (3a) is a clause-bounded branching (scope-independent) reading of the numeral phrase rather than an inverse scope reading. Meanwhile, he also argues that the inverse scope reading in (3b) is not available.

As we can see from existing studies, researchers debate the availability of inverse scope in doubly quantified relative clauses in Chinese, but with little experimental investigation into this question. We are therefore left with a research gap and uncertainty. Chen (2020) and Chen (2024) are the only studies, to the best of our knowledge, that experimentally investigate the availability of inverse scope in Chinese RCs. Chen (2020) conducted an experiment using a binary-choice picture-matching truth-value judgment task with 30 Chinese speakers. The findings indicate that when ORCs feature a plural existential quantifier as the RC head and a universal quantifier as the RC subject, as in (4), Chinese speakers accept these sentences as true under picture contexts that favor inverse scope readings. Although the stimuli and experimental procedures used by Chen (2024) differ from those in Chen (2020), the results are similar, showing that Chinese speakers accept ORCs with a plural existential quantifier as the RC head and a universal quantifier as the RC subject under the inverse scope reading. However, Chen’s (2020, 2024] research focuses on Chinese ORCs like (4) only, without examining SRCs.

Examples (4) and (3a) both contain ORCs, but the structural order of the QPs is different. It has been noted in theoretical and experimental studies on the quantifier scope in matrix transitives that the c-commanding relationship between quantifiers has been shown to play a significant role in quantifier scope interpretation (Musolino and Lidz 2006; Scontras et al. 2014, 2017], and papers cited there).

To this end, we conducted a large-scale experiment using an untimed truth-value judgment task (TVJT) to experimentally investigate whether Chinese speakers accept inverse scope in Chinese RCs and what factors affect the scope interpretations, aiming to provide a comprehensive and first-of-its-kind experimental study on quantifier scope in Chinese RCs. In the stimuli, we manipulated the type of RC – SRC versus ORC – and the structural order of quantifiers – EQF (i.e., existential quantifier structurally higher as the head of the RC) versus UQF (i.e., universal quantifier structurally higher as the head of the RC).

However, it should be noted that there is an entailment issue associated with UQF sentences, where the universal quantifier is structurally higher as the head of the RC (Mayr and Spector 2010; Reinhart 1997; Scontras et al. 2017). Specifically, the inverse scope reading of (3a) entails the surface scope reading, meaning that if it is true that I bought every book written by a set group of three authors, it must be true that I bought every three-authored book. Therefore, following Scontras et al. (2017), UQF would not be an ideal test case for the availability of the inverse scope reading, unlike EQF, which does not have such an entailment issue. That said, we included both types of structure to examine whether there was any structural order effect for relative clauses in a way comparable to Scontras et al.’s (2017) experiment on quantifier scope in simple structures. However, the conclusion about whether Chinese relative clauses permit inverse scope was largely drawn based on EQF structures, which effectively circumvents the entailment problem.

Another factor we manipulated in this study is the number of the existential quantifiers, such as singular (‘one book’) versus plural (‘three books’). In simple transitives, it has been argued that the quantificational phrase ‘one + classifier + NP,’ especially in the subject position, tends to be interpreted as specific and referential, while plural numerals do not exhibit this interpretive tendency (Chien 1994; Lee 1986; Su 2001; Wu 2023a).

Besides investigating the presence of scope ambiguities in Chinese RCs as a whole, our study also analyzes both inter-participant and within-participant variations in ratings to assess potential differences in native grammar with respect to scope interpretations among participants.

2 The present study

The aim of this study is to examine the prevalence of scope ambiguity in Chinese doubly quantified relative clauses and what factors influence scope interpretations in Chinese native speakers. It seeks to provide empirical evidence for the ongoing debate regarding the extent of scope rigidity as a property of Chinese grammar, extending the investigation from simple transitive clauses to complex clauses.

2.3 Procedure

The experiment was untimed and conducted via the internet using the Qualtrics software. The survey links were distributed through social media. Before the judgment task, participants were given information about the study and were asked for consent. Subsequently, they completed a block of two practice items to familiarize themselves with the task. Figure 2 is a screenshot of the online survey. Participants were instructed to rate on a 7-point Likert scale the extent to which the given sentence matched the story context in terms of semantic interpretation (1 = “not matched at all”, 7 = “completely matched”). Notably, graded response scales were chosen over binary response scales for detecting semantic ambiguities in TVJT, because graded response scales were found to excel in revealing interpretations that would otherwise remain concealed due to a strong preference for a more accessible interpretation (Marty et al. 2013, 2015]). Before concluding the experiment, participants completed a brief background questionnaire asking about their age, gender, and language experience.

Figure 2:

A sample of the truth-value judgment task in Qualtrics.

2.4 Data analysis and results

For each participant, the ratings for all items were transformed into z scores to mitigate potential scale bias, a practice common in judgment studies (Schütze and Sprouse 2013). A z score of 0 indicates that a participant has assessed a specific item as being of average rating in relation to others. Participants rated unambiguously matched dummy items highly (M = 5.62, SD = 1.83) and unambiguously mismatched dummy items low (M = 2.96, SD = 2.37), on average.^[3] This is a good indicator that participants understood and performed the task as expected.

Data were visualized and analyzed in R (version 4.3.2; R Core Team 2023). Statistical analyses on z scores were performed using linear mixed-effects models with the lme4 package (Bates et al. 2015). The model included the three factors and their interactions as fixed effects. In line with the recommendation of Schad et al. (2020), independent variables included in the fixed-effects structures were contrast-coded and centered around the grand mean. Based on Barr et al. (2013), the random-effects structure initially included random intercepts for both subjects and items and by-participant random slopes for all fixed effects and their interaction. We simplified the random-effects structures until model convergence was achieved. Moreover, when p values are greater than 0.05, the statistical results were corroborated in a Bayesian framework to mitigate the risk that the observed lack of significance stemmed from insufficient statistical power (Fukuda et al. 2022).

Group results may mask important individual differences, especially in the domain of scope interpretation for which there is no sufficient input (Han et al. 2016). As such, we also report individual results to examine whether between- or within-participant variability underlies the observed patterns. As a note, the results of all conditions (both EQF and UQF) at a group level are reported here, to make a comparison with Scontras et al. (2017) regarding the effect of structural order on scope interpretations, and to check if there is any interaction effect between structural order and other factors like clause type and number. However, due to the entailment issue associated with the scope interpretations of UQF sentences, the conclusion on whether inverse scope reading is available in Chinese relative clauses is solely based on the results from EQF sentences.

2.4.1 Group results

Table 1 shows the mean ratings of all conditions. Overall, the mean ratings for all conditions exceeded 4 out of 7, except for SRCs (with a mean rating of 3.88) in which the existential quantifier precedes the universal quantifier at the synaptic level and the existential quantifier is a singular noun phrase. An average rating of 4, we claim, characterizes the availability of inverse scope interpretations for given conditions.

Table 1:

Mean ratings of all the target conditions.

Clause type	Plural		Singular
	EQF	UQF	EQF	UQF
SRC	4.45	5.22	3.88	4.77
ORC	4.84	5.11	4.22	4.69

We employed a mean rating of 4 as a cutoff point in our study for several reasons, integrating insights from both current and prior research. First, on a 7-point Likert scale, a rating above 4 is a reliable indicator of the presence of the inverse scope reading. Our choice of 4, while slightly lower than the 4.46 – interpreted as inverse scope being available – in Scontras et al. (2017) in their study of English simple transitive clauses, is justified by the greater complexity of the clauses we tested. Wu (2023a) similarly investigated the inverse scope reading in Chinese ambiguous PP datives and reported a mean rating of 4.33 out of 7 (adjusted rating of 3.33 on a 6-point scale). The complex nature of the relative clauses in our experiment naturally leads to lower acceptability ratings, consistent with findings by Sprouse and Schütze (2019) that increased complexity and higher processing demands can decrease acceptability. Moreover, our decision is compromised by the advantage of graded judgments as opposed to a binary judgment, as advocated by Marty et al. (2015), which allows for the detection of dispreferred readings. A tradeoff to consider between graded and binary scales, however, is that, while there are advantages to graded judgments, they also raise the question of how to select an appropriate baseline. In our analysis, we did not opt for a very high cutoff such as 6 or 7, recognizing the processing cost associated with inverse scope interpretations as argued by Anderson (2004). Similarly, Meyer and Sauerland (2009) noted that participants might assign lower ratings in contexts where only the less preferred interpretation is plausible. This nuanced understanding influenced our choice of a baseline rating of 4.

We chose not to set the cutoff point lower, like 2 or 3, because such ratings are below the midpoint of the scale and suggest a tendency towards “unacceptable”. According to Scontras et al. (2017), the average rating for inverse scope in the Chinese unambiguous simple transitives is 1.56, similar to the findings of Scontras et al. (2014). Wu (2023a) reports that the ratings for Chinese unambiguous doubly quantified double object constructions and unambiguous simple actives are 1.59 and 1.27, respectively, on a scale of 0–6, which roughly translates to 2.59 and 2.27 on a 1-to-7 scale.^[4] Although the specific participant groups surveyed in these studies may vary, both suggest that average ratings in the range of 2–3 are likely indicative of the unavailability of inverse scope readings. Despite these considerations, we acknowledge that a low rating does not necessarily preclude the possibility of an inverse scope interpretation. This recognition led us to analyze the distribution of ratings at individual levels and normalize them for statistical analysis to mitigate scale bias.

With other factors aggregated, participants exhibited a marginal preference for ORCs, with an average rating of 4.72, compared to SRCs, which had an average rating of 4.58. However, this difference was not statistically significant, as revealed in the mixed-effects model analysis (β = −0.06, p = 0.5035). To further corroborate the null difference, we calculated the Bayes factor using the BayesFactor package (Morey and Rouder 2018). The resulting value was 0.0094, which is less than 1/3. This suggests substantial evidence in favor of the null hypothesis (Jeffreys 1998), and hence no statistical difference. A significant main effect of number was found (β = −0.23, p = 0.0085), indicating that sentences featuring a plural existential quantifier phrase received higher ratings than those with a singular existential quantifier phrase. There was also a main effect of structural order of quantifiers (β = 0.27, p = 0.0023), induced by higher ratings given for sentences with universal quantifiers structurally preceding the existential quantifier. No interactions between any factors were observed. The results of linear mixed-effects models are summarized in Table 2.

Table 2:

Results of linear mixed-effects models based on z scores.

Predictor	β	SE	p
Clause type	−0.06	0.09	0.50346
Number	−0.23	0.08	0.00845
Structural order of quantifiers	0.27	0.08	0.00233
Clause type*number	0.01	0.17	0.94450
Clause type*structural order of quantifiers	0.20	0.17	0.24050
Number*structural order of quantifiers	0.08	0.17	0.62976
Clause type*number*structural order of quantifiers	−0.05	0.33	0.87947

To pinpoint the locus of the effects of number and structural order of quantifiers in the ratings of ORCs and SRCs, we conducted by-structure analyses. For both types of structures, the models included number and structural order of quantifiers and their interaction as fixed effects, along with appropriate random-effects structures. As seen in Table 3, the model for ORCs returned a marginal main effect of number only (β = −0.23, p = 0.092).

Table 3:

By-structure results based on z scores.

	β	SE	p
ORCs
Number	−0.23	0.13	0.0920
Structural order of quantifiers	0.17	0.14	0.2235
Number*structural order of quantifiers	0.11	0.27	0.6792
SRCs
Number	−0.22	0.09	0.02902
Structural order of quantifiers	0.36	0.09	0.00091
Number*structural order of quantifiers	0.05	0.19	0.77836

A Bayes factor of 0.0703 suggests that ratings did not differ by number. For the modeling of SRCs, we observed a significant main effect of number (β = −0.22, p = 0.029). This indicates that sentences with a plural existential quantifier phrase received higher ratings compared to those with a singular existential quantifier phrase. Additionally, there was a main effect of structural order of quantifier (β = −0.23, p < 0.001), driven by higher ratings for sentences with a structural order of universal quantifier preceding the existential quantifier. This finding, which is observed in the doubly quantified relative clauses, echoes the structural order effect found in Scontras et al. (2017), which was for Chinese simple transitives. The observation that UQF structures generally received higher ratings than EQF structures for inverse scope readings across different syntactic environments may result from the fact that the inverse scope reading of UQF entails the surface scope reading while EQF has no such entailment relation between the two readings. Nevertheless, our conclusions regarding whether Chinese permits inverse scope in relative clauses are based solely on the results from EQF structures. In the following section, we, therefore, only reported data from these EQF structures.

2.4.2 Individual results

As mentioned earlier, group results may mask important individual differences, especially in the domain of scope interpretation (Han et al. 2016). As such, we also report individual results to examine whether between- or within-participant variability underlies the observed patterns. While Chinese speakers generally lean towards accepting inverse scope interpretations, a bimodal distribution of ratings emerged for each condition, illustrated in Figure 3 (histogram, using raw ratings) and Figure 4 (density plot using normalized z ratings). Note that these figures pertain exclusively to EQF structures. This distribution implies a population split: slightly more than half of Chinese speakers endorse inverse scope interpretations for relative clauses, while the other half do not.

Figure 3:

The histogram for raw ratings on xistential quantifier first (EQF) conditions. (ORC = object relative clause; SRC = subject relative clause; Plu = plural existential quantifier phrase; Sig = singular existential quantifier phrase.)

Figure 4:

The density plot for normalized z ratings on existential quantifier first (EQF) conditions. (ORC = object relative clause; SRC = subject relative clause; Plu = plural existential quantifier phrase; Sig = singular existential quantifier phrase.)

Differences in how individual participants interpreted inverse scope across various conditions were also evident when we looked at the percentage of participants who gave different ratings to these conditions. To capture this better, we focused on the average ratings that participants gave for inverse scope interpretations under one specific factor, while aggregating other factors. An individual fell into one of the four patterns depending on whether the individual mean ratings of the target conditions were equal to or greater than 4. Interestingly, as shown in Figures 5 and 6, around 50 % of the participants fully accepted the inverse scope interpretations, regardless of factors.

Figure 5:

Mean percentage of judgments in different relative clauses for existential quantifier first (EQF) conditions.

Figure 6:

Mean percentage of judgments in relative clauses with singular versus plural existential quantifier phrase for existential quantifier first (EQF) conditions.

3 Discussion

Three main results emerge: (a) inverse scope interpretations for Chinese relative clauses are generally available; (b) the number of QPs and the structural order of quantifiers affect scope interpretation in the context of RCs, and RC types are not found to statistically affect scope interpretations; (c) Chinese speakers exhibit variability regarding scope ambiguity as a property of grammar. We discuss these findings and return to how these results fit with theoretical approaches to quantifier scope interpretation in Chinese.

In a study using a 7-point scale, Scontras et al. (2017) report that English native speakers gave inverse scope readings in English simple transitive clauses an average rating of 4.46. In this study, we obtained mean ratings ranging from 3.88 to 4.84 across different EQF conditions, suggesting that Chinese speakers can access inverse scope readings in relative clauses. In contrast, when Chinese native speakers were tested on Chinese simple transitive clauses in Fang (2023) using a TVJT, the mean ratings for doubly quantified matrix transitive sentences with an existential quantifier preceding a universal quantifier for inverse scope interpretations were as low as 1.76 out of 7. This rating is significantly lower than the range of ratings observed across conditions in the present study, indicating that the presence of relative clauses greatly enhances the access to inverse scope interpretations. Therefore, this finding contributes to the literature by extending the investigation of quantifier scope from simple clauses to complex clauses, particularly those manifested as relative clauses, in addition to adverbial clauses (Wu 2023a). Going beyond Chen (2020, 2024], which only focuses on ORC EQF sentences, we provide novel empirical evidence for the availability of inverse scope readings in both ORCs and SRCs across different factors. Moreover, despite the different experimental methods – Chen (2020, 2024] employed a binary-choice picture-matching truth-value judgment task, while our study utilized a graded judgment task – the results of both studies align, supporting the presence of inverse scope readings in ORC EQF sentences. This convergence offers robust evidence that inverse scope presence holds across tasks at least in ORCs where a universal quantifier is in the RC subject position and a plural existential quantifier is in the RC head position.

This empirical evidence challenges the scope rigidity widely claimed for Chinese. There has been a long-standing theoretical debate regarding the explanations for the potential lack of scope ambiguity in Chinese. Various theoretical approaches, spanning surface c-command relation (Huang 1981), syntactic movement (Aoun and Li 1989, 2003]; Huang 1982), and semantic (Lee 1986; Xu and Lee 1989) perspectives, have been proposed. However, none of these early approaches seem satisfactory in explaining the existence of scope ambiguity in relative clauses, with inverse scopes readily available to Chinese speakers, as demonstrated in this study. For instance, Aoun and Li’s (1989, 1993] Scope Principle, as illustrated in (2), predicts only one available interpretation due to an isomorphic mapping between surface syntactic structure and semantic interpretation, thereby limiting the scope of interpretation. This account appears to fail to accommodate the empirical data in our study. Among more recent accounts, a head-raising analysis and long-distance quantifier raising have been proposed as plausible explanations for the presence of scope ambiguity in Chinese relative clauses (Larson and Wu 2018; Wang 2023; Wu 2019; Wu et al. 2018). However, a comprehensive account for the presence of the inverse scope reading in the Mandarin relative clauses is beyond the scope of this paper. That said, the crucial point we wish to make here is that the conception of scope rigidity as a linguistic feature unique to languages like Mandarin Chinese needs to be revisited in light of the empirical evidence.

Linguists have utilized judgments to investigate scope interactions between existential and universal quantifiers in Chinese. However, the reported judgments often lack consensus and occasionally even contradict one another (Chien and Wexler 1989; Scontras et al. 2014, 2017]; Zhou and Gao 2009). This inconsistency may arise from inter-speaker variability in making judgments. For example, in Chen (2020, 2024], when examining ORCs for the presence of scope ambiguities, it was observed that not all participants consistently accepted the test sentence under pictures endorsing the inverse scope reading. Such discrepancies are not unusual, as adult monolingual native speakers of the same language do not necessarily share identical mental grammar (Dąbrowska 2012; Han et al. 2007). Individual analyses reveal significant variability in scope interpretations. This discovery is consistent with the two-grammar hypothesis posited by Han et al. (2011), based on their study of negation interpretation in Korean. Following this hypothesis, Chinese speakers can be divided into two distinct groups, each exhibiting a unique preference for scope interpretations. Consequently, an interpretation favored by one group may not be accessible to the other.

The variation in the ability of Chinese speakers to allow inverse scope reading in relative clauses can be attributed to insufficient and unclear language input in the environment. This ambiguity prevents a definitive interpretation of relative clauses involving quantifier scope. A similar argument was made by Han et al. (2016), who postulated that Korean-speaking children lack clear evidence regarding the syntactic positioning of verbs, leading to significant variability in their interpretation of negation and object-quantified phrases in adulthood. Similarly, in early childhood, Chinese speakers are exposed to two possible interpretations made available by universal grammar, as evidenced in Zhou and Crain’s (2009) study on the interpretation of scope relationships between universal quantifiers and negation among Chinese children. In the case of the doubly quantified RCs under investigation, the scarcity of clear input may lead children to prematurely settle on one interpretation over another, a choice that persists into adulthood. Consequently, without adequate positive and negative evidence, adults may not recognize that only one grammatical interpretation is viable.

In fact, variability in linguistic interpretations, often manifesting as bimodality, is not uncommon. Monolingual speakers from the same community have shown considerable individual differences in grammatical representations, as evidenced by Dąbrowska (2012), where complex constructions like passives and quantifications were examined. Besides language experience derived from input, other factors are believed to contribute to the phenomenon where speakers fall into distinct groups that assign different grammatical interpretations to the same linguistic phenomena. These factors include working memory, task demands, educational background, and willingness to cooperate with experimental protocols. Consequently, more research is necessary to corroborate the influence of these factors and to establish the bimodality following particular statistical procedures (Houghton and Kapatsinski 2024). If a speaker consistently adheres to one grammatical interpretation, it is then crucial to verify if such patterns of interpretation preference are stable over time. Therefore, future studies are also needed to determine whether the same patterns of interpretative behavior persist over time in the same speaker. In a nutshell, the individual analysis results offer additional evidence against the scope rigidity claim in Chinese.

In the literature, nonstructural factors such as lexical semantics have also been found to affect scope interpretation (Beghelli and Stowell 1997; Lee et al. 1999). This study also observed the influence of number in SRCs. Our findings indicate that sentences containing a plural existential quantifier phrase received higher ratings within SRCs compared to those featuring a singular existential quantifier phrase. This aligns with prior research on quantifier phrase effects (Scontras et al. 2017; Wu 2023a; Yang and Wu 2020). In particular, phrases in the pattern of “yi ‘one/a’ + classifier + NP” in Chinese tend to receive referential readings, thereby predisposing against inverse scope interpretations, as observed in our study. The existence of the effect of number in SRCs but not ORCs is an interesting observation but will warrant further study to find out what results in the asymmetry of the number effect in the scope interpretations in relative clauses.

4 Conclusion and future work

Based on group and individual analyses of EQF structures, our study confirms that Chinese complex clause structures, both SRCs and ORCs, predominantly allow for inverse scope interpretation. The observed variability in scope judgments may suggest a two-grammar hypothesis: Chinese speakers’ differing preferences for interpreting scope may stem from the lack of clear evidence supporting or refuting specific quantifier scope interpretations in their linguistic input. Consequently, an interpretation favored by one group may not be accessible to the other. As a result, an interpretation accepted by one group (e.g., inverse scope reading) might not be acceptable to the other. Additionally, nonstructural factors such as the lexical properties of quantifier phrases have been found to modulate scope interpretations.

As a reviewer pointed out, it is not an optimal practice to assert the availability of inverse scope based solely on a rating of 4 without a baseline group. In future research, we aim to further compare the availability of inverse scope reading in Mandarin relative clauses with that in unambiguous simple transitives which do not permit inverse scope. A more robust conclusion about the availability of inverse scope in relative clauses could be drawn if its ratings were high and statistically different from those on simple transitives. Meanwhile, a triangulation of methods could be employed, incorporating diverse methods such as tasks matching sentences to pictures and act-out stories, which can help us test if methodology could affect the elicitation of inverse scope readings. Moreover, future research could also investigate additional factors affecting scope interpretation in Chinese, such as the presence or absence of aspectual markers for main verbs, as noted by Wang (2023). A timed experimental method would also help us to gain further insights into the cognitive processes involved in interpreting quantifier scope in complex clauses such as relative clauses and would hence allow us to explore potential convergences or divergences between offline and online results.