The present study investigated the e}ects of the new context reinstatement procedure on recognition memory for faces[ All faces were portrayed in distinctive background contexts which were pretested to induce a negative emotion\ a positive emotion or no emotion at all[ These contexts were\ at test\ either the same "e[g[ the same concentration camp#\ changed "e[g[ concentration camp:road accident# or changed but of the same type "e[g[ {semantic context|] two di}erent concentration camps#[ Results provided evidence of considerable improvement of recognition performance with semantic context[ Moreover\ a context inducing a negative emotion was shown to impair face recognition\ but semantic context was found to counteract such an impairment[ The implications of these _ndings are discussed in terms of their relevance for theories of recognition memory and practical contributions to eyewitness identi_cation[ Copyright Γ 1990 John Wiley + Sons\ Ltd[ It is well known that human recognition performances are imprecise and in~uenced by situational constraints[ Eyewitness identi_cation is of uncertain accuracy at best and wrong and harmful at worst "Loftus\ 0868^Sporer et al[\ 0885^Wells and Loftus\ 0873#[ The identi_cation testimony of eyewitnesses\ however\ has an intuitive and commonsense appeal that makes its continued use highly likely "Ellison and Buckhout\ 0870^Wells et al[\ 0868#[ Thus\ it is appropriate to _nd means of strengthening techniques for obtaining\ evaluating\ and using eyewitness testimony so that accuracy can be enhanced and errors minimized or detected[ One line of research from cognitive psychology o}ers directions] context e}ects in recognition memory for faces[ In this framework\ McGeoch "0821#\ then Davies "0875# distinguished two dimensions of context] external environmental context "the physical environment in which learning takes place# and internal environmental context "the internal state of the learner#[ Concerning external environmental context\ the classic context reinstatement procedure is based on the {encoding speci_city| principle of Tulving and Thomson "0862# according to which recognition is a function of the degree to which the retrieval environment approximates the encoding environment[ In other words\ faces are better recognized when the context provided during study is reinstated during test[ This result was replicated by authors such as Watkins et al[ "0865#\ Davies and Milne "0871# or Memon and Bruce "0872#[ However\ if this procedure led to a higher hits rate\ it also gave rise to a higher false positive rate "for a review\ see Davies\ 0875\ 0877^Tiberghien\ 0878#[ Thus\ Baddeley and Woodhead Correspondence to] Dr N[ Rainis\ UFR de Psychologie\ Universite de Lille 2\ BP 038\ 48542 Villeneuve d|Ascq Ce dex\ France[ E!mail] rainisΓuniv!lille2[fr