Purpose This paper investigates how test-takers switch use of test-taking strategies in an English reading comprehension test as they tackle tasks with increasing difficulty. It further investigates whether/how test-takers’ ability moderates transition of strategy use.

Method Test-takers (N=189) reported their use of ten pre-specified strategies on a 5-point Likert-scale immediately after completing each one of the four reading tasks of the CELPIP-General Test- a test used to screen immigrants to Canada. These strategies were shown to represent three types of test-taking strategies: ‘comprehending meaning,’ ‘score-maximizing’ (maximizing the score relying on partial comprehension), and ‘test-wiseness’ (Wu & Stone, 2015).

Results Latent class analyses identified three distinct groups. Class one test-takers (C1, 32%) tended to use all three types of strategies; Class two (C2, 41%) tended to use ‘comprehending meaning’ and ‘score-maximizing’ strategies; Class three (C3, 27%) tended to use only ‘comprehending meaning’ strategies. Results of the 3-Step latent transition analysis (Muthen & Asparouhov, 2011, 2013) showed that, comparing strategy use from the easiest task to the most difficult task, 45% changed their strategy use as test difficulty increased. About 22.3% switched from C2 to C1 (by augmenting the strategy of test-wiseness). About 7.9% switched from C3 to C2 (augmenting one strategy of score-maximizing) and about 9.3% switched from C3 to C1 (augmenting two strategies of score-maximizing and test-wiseness). Furthermore, the transition of strategy use was found to be moderated by test-takers’ ability. For example, the probability of transition from C3 to C1 decreased as the proficiency level increased. In contrast, the probability of staying in C3 increased as the proficiency level increased.

Conclusions The findings suggest that use of test-taking strategies is not a trait-like phenomenon. Rather, it may change with the task difficulty, vary with test-taker ability, and react to the intersection of the two. These findings confirmed the conjecture in Wu & Stone (2015) and findings suggested in Purpura (1998).