Currently equivocal findings on personality similarity and attraction can be clarified if ‘personality’ and ‘acquaintance’ are recognized as essentially generic concepts rather than single entities. A ‘filtering’ process is proposed to account for longitudinal acquaintance development where ‘personality’ has several different and successive places and where different measures of personality are sequentially appropriate tests of the personality‐friendship relationship. A study is reported where 40 previously unacquainted subjects completed three different personality tests and then reported sociometric choices at different points of developing acquaintance (1 month; 3 months; 8 months). Whilst similarity on the CPI mediated sociometric choices at none of these points, similarity on Allport‐Vernon Study of Values predicted choices only at Time 2 (P < 0.01), and the Reptest only at Time 3 (P < 0.01). Results are discussed in terms of systematic test of models of partner's personality in acquaintance and it is suggested that a search not for the correct measure of personality but for the relative place of each measure in developing acquaintance will lead towards the resolution of several existing ambiguities in the attraction literature.