The consistency of hedonic responses to three brands of vanilla ice cream over 5 weeks was examined using two procedures. Respondents (n=40) came to the test after their usual lunch and rated their liking for ice creams side-by-side (three samples in one session) and after ad libitum consumption (one sample at a time). The quantity of ice cream consumed in ad libitum procedure was recorded. Respondents also assessed the appropriateness of the three samples of vanilla ice cream to ten usage situations, including ''as a dessert'' which was the context in the test situation. The consistency of hedonic responses was low over replicates (r=0.03 to 0.48). In the side-by-side condition, only 48% of the liking responses in replicates were within one point (not at all=1, extremely=9), and 63% in the after-consumption condition. Some respondents ate more of all brands than other subjects did in both replicates, unrelated to differences in liking. The appropriateness ratings of brands were similar but the ratings of appropriateness ''as a dessert'' differentiated the brands. In the after-consumption condition the ratings of liking and appropriateness, together with consumption measures, appeared to be slightly more consistent at the individual level than those made in the side-by-side situation. (C) 1995 Academic Press Limited