Some personality test scores improve on retest. Three studies investigated the role of meaning change in producing this phenomenon. In Study 1 multiple versions of a Manifest Anxiety Scale were administered, with counterbalanced item orders. It was found that measurement-induced improvement (a) occurred within a test as well as between test and retest, (b) was unaffected by participants' anxiety scores, and (c) occurred even when the retest contained different items than the first test. Studies 2 and 3 found that as respondents experience more of a test, they are better able to discern its meaning and to use that meaning to interpret an item. These findings indicate that mean shifts in answers from test to retest also occur within a test along with context-induced shifts in meaning.