The goal of the present studies was to investigate the relation between attitudes towards the self (i.e., self-esteem) and self-reported frequencies of negative thoughts about the self as measured by the Automatic Thoughts Questionnaire (ATQ; Hollon & Kendall, 1980). It is argued that participant responses on the ATQ are less likely to be guided, as is commonly thought, by memory retrieval than by reference to participant self-esteem. In Study 1, subliminal presentations of pairs of words were used to activate self-esteem. The presented pair of words consisted of one word referring to the self (i.e, Me ) and another word that was either a positive (e.g., Great ) or negative adjective (e.g., Ugly ). Self-esteem, frequency of negative thoughts about the self and private self-consciousness was assessed. Contrary to what was hypothesized, no difference was found in self-esteem or reported ideation across priming groups. Unexpectedly, participants presented with positive primes reported higher private self-consciousness that those presented with negative primes. In a replication, a control group was added in which participants were subliminally presented a string of random consonants as opposed to words. No significant differences were observed. The failure to replicate the significant difference for private self-consciousness across priming groups is discussed. Given that self-esteem was not affected, the central hypothesis of the study remains untested. Reasons for the failure of the subliminal priming method as well as alternative approaches for self-esteem activation are discussed.