The expresser’s emotion and group membership of an unrelated (prime) stimulus can jointly affect evaluations in implicit paradigms. However, the two features seem differently weighted, with the assumption that task relevance determines their influence, at least in the evaluative priming paradigm (Craig, Lipp & Mallan, 2014). However, Gawronski et al. (2010) suggested, that several features are processed in parallel in the Affect Misattribution Procedure (AMP) without much weighting by task instructions. We pursued this line of research by examining whether and how task instructions influence the effect patterns of emotion and group features in the AMP. Primes consisting of ingroup and outgroup faces displaying happiness and fear were followed by ambiguous, neutral targets, consisting of masked, computer-generated faces. As a typical AMP instruction, in Experiment 1A, participants were instructed to report whether they found the target faces (more or less) pleasant or unpleasant than the average. In Experiment 1B, they were instructed to report whether or not they saw fear in the target faces, with the aim of increasing the attentional focus to the emotional expression as well as to a possible fearful emotional reaction towards an outgroup member. The emotion and group characteristics of primes interactively influenced target evaluations in Experiment 1A, consistent with the assumptions of Gawronski et al. (2010). In Experiment 1 B, however, neither emotion nor group characteristics modulated target evaluations, despite the task relevance of emotion was increased. The influence of contextual cues, such as instructions in implicit paradigms will be discussed.