The aim of the paper is to investigate whether deviations from the double jeopardy pattern observed in brand buying (panel) data such as niche brands (brands with few users, but high levels of brand loyalty) and change-of-pace brands (brands with many users, but low levels of brand loyalty) correspond with deviations from the same pattern in brand image data (i.e. brand image associations gathered through consumer surveys). The analysis of a unique data set in three product categories including brand purchases and brand image associations from the same consumers and for the same brands shows that such deviations in brand buying data and in brand image data seldom correspond. Moreover, deviations from the double jeopardy pattern in brand image data are rare and imply the existence of an underlying impact of a brand's associative rate (i.e. the average number of associations a brand obtains in a brand image consumer survey) on brand loyalty (measured as the frequency of brand buying). These findings provide contribution to buyer behaviour theory and have practical implications for branding strategies. For instance, to obtain any impact on brand loyalty, marketers may need to develop branding strategies aimed at increasing a brand's associative rate, whilst maintaining or enlarging its level of market penetration.