Do comparisons of a food characteristic with ideal necessarily involve learning?

Untrained assessors graphically rated the saltiness of plain slices of white bread relative to a personally ideal level of saltiness, with each test session designed to minimize known contextual effects on ratings. Judgements were highly reliable from the first session. However, the mean of the residual variances of individuals' psychophysical functions was approximately halved by the fifth session. To test whether this was an accommodation to the method or to the product, assessors rated the same product by a different procedure in one further session and a different food by the same procedure in a final session. The mean residuals lay in the direction of an accommodation to the method, but this difference proved not to be significant. The initial fall in residuals might indicate some learning about the product which was not adequately tested in the experimental manipulation. Nevertheless, it is clear that the method can be immediately used with high precision by unpractised volunteers.