In the course of four initial studies a new stereotype measure is developed which captures ingroup serving generalisations. The new implicit ratio technique draws on experiences gained from using existing instruments to assess stereotyping. It combines the probabilistic approach of diagnostic ratios (McCauley & Stitt, 1978) with elements introduced by Esses et al. (1993), using perceiver-generated idiosyncratic stereotypic material which is qualified for meaning. The implicit ratio assesses individual levels of generalisation. The trait frequency processing involved in generalisation was found to be consistently biased in favour of the ingroup. Trait frequency bias tends to be spontaneous, automatic and largely independent of specific content or context. Since mere categorization into arbitrary groups already elicits trait frequency bias, four further studies examine perceivers' level of control over trait frequency processing. Changing attributional perspective towards the ingroup as a whole accentuated bias, whereas heightened accountability and direct instructions to suppress stereotyping reduced it. However, significant inhibition of bias as result of suppression instructions were restricted to within subjects designs. After these instructions are relaxed trait frequency bias recovers in strength. The differential effect of category primes on content valence and trait frequency bias suggests a direct effect of manipulations, merely mediated by content dilution. Differences between semantic suppression and inhibition of trait frequency bias are discussed. The discussion extends to the usefulness of the implicit ratio as an experimental tool and to wider implications of the trait frequency bias approach to stereotype measurement