The effects of test anxiety and evaluative stress on reading speed, ar
ticulatory rehearsal, reading regressions, and comprehension were exam
ined. High- and low-test-anxiety subjects read texts under conditions
of stress (Studies 1, 2 and 3) or non-stress (Study 4). Texts were pre
sented either with concurrent irrelevant speech (heard), an articulato
ry suppression task, or no concurrent task. Measures of working memory
span and prior vacabulary knowledge were collected under non-stress c
onditions (Study 5). There were no differences in comprehension perfor
mance as a function of anxiety, but high anxious subjects were less ef
ficient than low-anxious subjects, as the former employed more reading
time and regressions, though not more articulation, than the latter t
o obtain an equivalent comprehension score. Reading regressions emerge
d as the most discriminating compensatory strategy associated with anx
iety. This reduced efficiency is partly dependent on a basic deficit i
n vocabulary knowledge-but not in memory span-, and it is partly trans
itory and dependent on the presence of external evaluative stress.