1　Introduction

This section introduces the prominent role of relative clause structures in syntactic analysis and language acquisition research, noting the relatively fragmented state of relative pronoun studies and the prevalence of usage differences discussed mainly in grammatical works. Based on the Ten-thousand English Compositions Corpus of Chinese Students (TECCL) and the CLOB corpus, this study employs multifactorial analysis methods to explore the similarities and differences between Chinese English learners and native English speakers in their choice of the object-referring relative pronouns which and that, analyzing the influence of lexico-grammatical environments and providing theoretical interpretations of the linguistic choice differences between the two groups.

2　Literature Review

This section reviews the research trajectory of relative clauses and relative pronoun selection. Early studies predominantly focused on comparing the cognitive processing difficulty of relative clauses, with relative pronoun research remaining relatively scattered. Grammatical works by Jespersen (1933), Quirk et al. (1985), Biber et al. (1999), and Huddleston et al. (2002) systematically discussed the constraints on which and that selection, including antecedent complexity and definiteness, syntactic function of the relative clause, distance between antecedent and clause, stylistic features, and dialectal differences. Wiechmann (2015) applied multifactorial analysis methods to examine relative clause structural characteristics, finding that structural entrenchment affects relative pronoun omission. Fang Yinjie (2016) compared non-subject relative clause production between Chinese English learners and native English speakers, discovering that the former tend to use which while the latter prefer that. Existing research exhibits limitations including predominant reliance on introspective methods, unclear factor weightings, insufficient explanatory power, and lack of large-scale corpus validation. Based on this, the present study adopts multifactorial statistical analysis methods to investigate the common constraining factors and significant differences in object-referring relative pronoun selection between the two participant groups.

3　Research Design

The three core components of the research design: corpus selection, variable annotation, and statistical analysis. For corpus selection, approximately 1.8 million words of learner data from 2011–2015 were extracted from TECCL and 1 million words of native speaker data from 2008–2011 from the CLOB corpus. After manual screening, 5,650 restrictive relative clauses were obtained, ensuring inanimate antecedents with interchangeable that and which. For variable annotation, 14 predictive factors were determined based on previous research, with which and that selection as the response variable. For statistical analysis, the glm function in R software was used to construct binary logistic regression models, with stepwise regression optimization performed through the stepAIC function.

4　Results

Regression analysis was employed to examine the factors constraining object-referring relative pronoun selection for native English speakers and Chinese English learners respectively. For native speakers, main effect factors included antecedent complexity, clause length, and antecedent part of speech, with preference for that when the antecedent was a "determiner + head" structure or indefinite pronoun, and preference for which as clause length increased; interaction effects involved factors such as antecedent modifier number, definiteness, syntactic constituent, clause embedding type, distance, and predicate verb type. For learners, main effect factors included antecedent modifier number, clause voice, and clause length, with preference for which under higher cognitive processing load; learners showed greater sensitivity to increased cognitive processing load from passive voice, negative forms, separation between antecedent and clause, and center embedding, and the affirmative/negative factor significantly affected their selection through interaction effects.

5　Discussion

This section analyzes the common constraining factors and differential constraining factors in object-referring relative pronoun selection between Chinese English learners and native English speakers, provides interpretation from a construction grammar perspective, and offers pedagogical implications. Regarding commonalities, both groups' selections were influenced by multiple lexico-grammatical, semantic factors and their interaction effects, including antecedent complexity, definiteness, syntactic constituent, clause embedding type, length, voice, and main clause characteristics; cognitive processing load significantly affected both groups, with preference for the more explicit which in complex contexts, validating Rohdenburg's Complexity Principle. Differences manifested in three aspects: native speakers preferred that, especially in American English, while learners showed no significant preference; the two groups exhibited different selection tendencies at specific variable levels, such as when antecedents were "determiner + head noun" or indefinite pronouns, where native speakers preferred that and learners tended toward which; the same factors carried different weight, and learners were more sensitive to changes in cognitive processing load, preferring which and copular verbs to reduce processing burden in high-load contexts. The construction grammar interpretation supports the "No Synonymy Principle" and "Statistical Preemption Hypothesis," emphasizing the decisive role of context in near-synonymous construction selection and the balance between expressivity and efficiency. Pedagogical implications indicate that learners' construction awareness and context awareness should be cultivated, with authentic examples provided through corpus tools to help them form systematic linguistic knowledge structures and avoid treating near-synonymous structures as arbitrarily interchangeable equivalent forms.

6　Conclusion

Based on multifactorial analysis methods, this study systematically examined the commonalities and differences between Chinese English learners and native English speakers in object-referring relative pronoun selection. The findings reveal that both groups are jointly influenced by multiple lexico-grammatical and semantic factors with significant interaction effects, corroborating the decisive role of context; however, native speakers prefer that while learners show no stable tendency. Specific differences include: different selection tendencies regarding antecedent-related variables, varying factor weight (e.g., clause length ranks higher for native speakers), and learners' greater sensitivity to changes in cognitive processing load. The study proposes that future research could employ MuPDAR for more refined quantitative analysis and validate conclusions through psycholinguistic experiments, while acknowledging limitations regarding the comparability of corpus genres and registers.