Congruency (the availability of a direct first language translation) and level of proficiency have been reported among the most important determinants of second language collocation processing. However, only very few studies looked at the interaction between the two determinants, and none of these directly compared untimed collocation recognition assessed through traditional tests to timed recognition evident in psycholinguistic tasks. The current study administered both types of form recognition measures to 228 female Saudi English as a foreign language learners in two separate experiments: a traditional multiple-choice test (Experiment 1) and a timed acceptability judgment task (Experiment 2). Experiment 2 also tested 37 native speakers of English as a baseline for comparison. Congruency, estimated proficiency (vocabulary test scores), and the interaction between the two were evaluated as predictors of untimed and timed recognition through mixed-effects modeling. Results showed that congruency and estimated proficiency had a clear effect on untimed and timed recognition. More interesting, the effect of proficiency was clearer on timed recognition with a gradual decrease in the first language effect as proficiency increased getting closer to nativelike collocation processing. Results have implications for second language collocation learning and testing.