Relative clauses (RCs), while ubiquitous, present processing challenges. A well-established asymmetry exists: subject relative clauses (SRCs) are generally easier to process than object relative clauses (ORCs). Several theories attempt to explain this. Structure-based accounts, like the Noun Phrase Accessibility Hierarchy (NPAH), emphasize syntactic factors, suggesting a universal preference for subject relativization. Conversely, working memory-based approaches, such as Dependency Locality Theory (DLT), focus on the cognitive load imposed by the processing of the relative clause, suggesting that ORCs require more working memory resources than SRCs. However, neither syntactic nor memory-based accounts fully explain the observed variability in processing difficulty depending on factors like animacy and verb type. Some studies have shown that ORCs with inanimate heads are easier to process than those with animate heads. Experience-based approaches like the Production-Distribution-Comprehension (PDC) framework propose that production mechanisms influence distributional patterns in language, which in turn affect comprehension. This study investigates whether the PDC account, which links production, distribution, and comprehension, can explain the processing asymmetry of SRCs and ORCs among Chinese EFL learners, examining the influence of animacy and verb type on both production and comprehension difficulties.

Existing research demonstrates a processing advantage for SRCs over ORCs, with various theories proposed to explain this asymmetry. Structure-based accounts, such as the NPAH, posit a universal preference for subject relativization across languages. Conversely, working memory-based theories, like the DLT, highlight the cognitive load associated with processing non-local dependencies in ORCs. However, these theories don't fully account for the influence of factors like animacy and verb type. Studies have shown that animacy plays a role, with ORCs containing inanimate heads being processed more easily. The PDC framework provides an alternative, suggesting that comprehension difficulty is linked to the frequency of specific structures in language production and their resulting distributional patterns. While this approach has been explored with L1 speakers, less research examines its applicability to L2 learners, particularly Chinese speakers learning English, whose relative clause structure differs significantly from English.

The study employed two experiments. Experiment 1 used a sentence completion task to investigate the production patterns of SRCs and ORCs among 35 Chinese EFL undergraduates. Participants completed 48 incomplete sentences, varying animacy (animate vs. inanimate head nouns) and verb type (agent-theme vs. theme-experiencer verbs) in a within-subjects design. The frequency of different RC types was analyzed. Experiment 2 involved 38 participants completing a self-paced reading task with 44 sentence pairs (SRCs and ORCs) also manipulating animacy and verb type. Reading times for different sentence regions and overall comprehension accuracy were measured. The self-paced reading task employed a word-by-word moving-window paradigm. Comprehension questions followed each sentence to ensure engagement and to filter out data from trials with incorrect comprehension.

Experiment 1 revealed a significant preference for SRCs over ORCs in production. This asymmetry was modulated by animacy and verb type. Animate head nouns strongly favored SRC production, while inanimate head nouns showed a higher frequency of ORCs. Agent-theme verbs were associated with more SRCs, while theme-experiencer verbs led to a greater proportion of passive relative clauses. Experiment 2 demonstrated that SRCs were generally read faster than ORCs, confirming the processing asymmetry. This asymmetry was also influenced by animacy. For animate head nouns, SRCs were processed significantly faster than ORCs. However, with inanimate head nouns, the processing difficulty was reversed, with ORCs processed faster than SRCs. This finding aligns with the distributional patterns observed in Experiment 1. The analysis of verb type in Experiment 2 showed that ORCs with agent-theme verbs were processed faster than those with theme-experiencer verbs, while the difference in SRC processing speed between the two verb types was not significant.

The findings strongly support the PDC account. The observed comprehension difficulties mirror the distributional patterns found in production. The higher frequency of SRCs with animate head nouns and agent-theme verbs corresponds to their easier processing in comprehension. Conversely, the lower frequency of ORCs with animate head nouns aligns with their greater processing difficulty. The reversal of this pattern with inanimate head nouns is also consistent with the production data. The interaction between animacy and verb type further underscores the importance of considering these factors together in explaining processing difficulty. These results demonstrate that the comprehension process is shaped by the learners' prior experience with different sentence structures, indicating that frequency of exposure is a significant determinant of processing ease.

This study confirms the processing asymmetry between SRCs and ORCs in Chinese EFL learners, demonstrating that this asymmetry is influenced by animacy and verb type. The results strongly support the PDC account, showing a clear link between production frequency, distributional patterns, and comprehension difficulty. Future research could explore additional factors, such as discourse context and sentence complexity, and utilize more advanced methodologies like eye-tracking or ERP to gain a deeper understanding of the underlying cognitive processes.

While this study provides robust evidence supporting the PDC account, some limitations exist. The study focuses solely on animacy and verb type, and other factors such as discourse context, sentence complexity, and individual differences in working memory capacity could influence processing. The sample size, while sufficient for the statistical analyses, could be expanded to enhance generalizability. The use of a specific English proficiency test (CET-4) might limit the extent to which the findings can be generalized to other proficiency levels.