Three experimentsinvestigated the tendency of high-anxiety individuals to interpretambiguous information in a threatening fashion. Priming ambiguous sentences (concerned with ego-threat, physical-threat, or non-threat events) were presented, followed by a disambiguating sentence in which a target word either confirmed or disconfirmed the consequence implied by the priming context. The sentences were presented word-by-word at a predetermined pace. Subjects read the sentences and pronounced the target word (naming task), which appeared either 500 msec or 1,250 msec after the onset of the last word (pre-target word) in the priming context. Results indicated that high-anxiety subjects named target words confirming threats faster than low-anxiety subjects, relative to non-threat words. Furthermore, this interpretative bias is: (a) strategic, rather than automatic, as it occurred with a 1,250-msec SOA, but not with a 500-msec SOA; (b) temporary, as it was found under evaluative stress conditions increasing state anxiety, butnot withnon-stress; and (c) specificto ego-threats, as it happened with ambiguous information concerning self-esteem and social evaluation, rather than with physical-threat-related information.