The aim of this research is to explore whether comprehension is impaired by test anxiety, whether the anxious person spends compensatory reading time, and which cognitive processes are affected. High- and low-anxiety students read texts word by word with the moving-window technique under test conditions. Multiple regression analyses on word-reading times were computed with a number of psycholinguistic variables--assumed to map onto specific processes--serving as predictors. Results indicated that anxiety did not impair comprehension, but increased word-reading times, which were affected interactively by anxiety and specific psycholinguistic variables (end of clause, serial position within the text, narrativity, and summary). These data reveal that anxious readers need to employ a greater amount of processing resources than their non-anxious counterparts to obtain a similar comprehension level. Furthermore, the interactive effects suggest that anxiety is selectively detrimental to the efficiency of text-level processes, such as those involved in integrating information across sentences. In contrast, anxiety does not impair low-level processes, such as encoding and lexical access.