Unpacking the Alignment Effect: The Moderating Role of Input Text Complexity in L2 Continuation Writing

Abstract

Alignment is the key mechanism through which the continuation task promotes language acquisition, yet the effect of input text complexity on alignment has been examined only through binary simplified‐versus‐unsimplified comparisons, leaving it unclear how the relationship is shaped and whether different alignment layers respond differently. The present study established three complexity levels (low, medium, high) and measured alignment at three levels—lexical (LRR, LSS), phrasal (4‐gram borrowing), and syntactic (relative clauses, participial adverbials). Eighty‐seven Chinese university students who had passed the College English Test Band 4 (CET‐4) completed continuation tasks for three narrative texts in a within‐subjects repeated‐measures design; data were analyzed using repeated‐measures ANOVA/ANCOVA with comprehension as a covariate, generalized linear mixed‐effects models for low‐base‐rate syntactic indices, and orthogonal polynomial contrasts to test for non‐linearity. Complexity exerted a significant negative effect on all three alignment layers, with different shapes: lexical alignment was relatively flat across the low‐medium interval and dropped more steeply across the medium‐high interval; phrasal alignment declined throughout; syntactic alignment was stable until high. Polynomial analyses showed that the lexical layer deviated from a strictly linear decline, with a significant quadratic component, while the phrasal layer declined more linearly. No discrete threshold was found at any layer. A parallel rephrasing‐strategy analysis showed that lexical‐level decline at high complexity reflects both deeper processing and disengagement working alongside each other. The findings extend prior binary comparisons by characterizing the shape of decline and treating alignment as a layered construct, with implications for the selection of continuation task materials. Because one syntactic structure showed a high zero‐count rate, syntactic alignment is interpreted cautiously; and because each complexity level was represented by a single narrative passage, the findings are scoped to the three passages used here rather than treated as passage‐independent effects of complexity.