For second language learners, language learning is not simply a matter of vocabulary building, especially when it comes to the richness and variety of prepositions, which are subject to bias. This study investigates Chinese learners’ misuse of Spanish prepositions, explains the transfer between native cognitive schemas and target language structures from a cognitive linguistic perspective, and provides a new approach to understanding the cognitive mechanisms of second language acquisition. A statistical linguistics-based text retrieval model (SLM-IR), including probabilistic model and language generation model, is introduced and optimized. On the other hand, rooted in the prototype and schema theories of cognitive linguistics, it is argued that the bias arises from the learner’s attempt to understand and apply the Spanish prepositional system with the conceptual categories inherent in the native language. Experiments confirmed that a statistical language model using Bayesian smoothing techniques performed best in distinguishing prepositional positives and negatives, with a word perplexity of 472 and a MAP = 75.34%. Meanwhile, the TF-IDF weighting strategy achieves an optimal balance between both model size retrieval performance, with a model size of only 6.96 megatrigram when the word perplexity is 500, achieving a MAP of 75.15%. The usage of Spanish prepositions by Chinese learners, influenced by the Chinese word “de”, overuses “de”, with its usage frequency accounting for as high as 27.47%. The error rates of “en” and “para” are prominent, being 12.99% and 15.77% respectively. The alternative category of bias accounts for 61.89% of the total bias, suggesting that the learner difficulty lies in the misalignment of conceptual mappings.
