Motivational Test Instructions and Conditional Structure of Cognitive Disengagement in Low-Stakes Assessments

Motivational test instructions (MTIs) are widely used in low-stakes assessments, but evidence for their effectiveness is mixed. Prior work has evaluated them almost exclusively in aggregate terms, asking whether they reduce the frequency of disengaged responding. We argue that MTIs may instead change the conditions under which students disengage, and we test this using data from a randomized low-stakes digital mathematics assessment ($N = 536$ eighth-graders), theory-based indicators of cognitive disengagement, and joint response--engagement models within an adapted Mislevy--Wu framework. Disengagement rates were similar across conditions, as was the association between latent ability and engagement propensity, which was strong in both groups. What differed was the extent to which disengagement was conditioned on would-be success. This dependence was weaker under MTIs than in the control condition, although the evidence for the contrast was suggestive rather than conclusive. An exploratory moderation analysis locates this difference on items of low cognitive demand. MTIs may therefore change how effort is allocated across an item pool rather than how much is withheld overall.