A model for quadratic similarity judgements developed previously (Gregson, 1969, 1970) is extended to deal with a form of bias, and to assess the information taken up by the parameters in the model. It is shown that judgements collapse, in terms of the model, into near randomness under restricted information conditions, but that this collapse is generally explicable in form, although it exhibits marked inter‐individual differences. Implications for similarity scaling models are note