ABSTRACT
The trait lists have been presented either simultaneously or sequentially, with appropriate changes in some of the models. In other words, most of the recent multiple cue work is self-consciously concerned with the testing of more or less formal models of how information is combined for decision-making purposes. Most models, therefore, predict that the responses are linear combinations of these input variables, to put it at its most general. However, bootstrapping does imply that a linear regression model derived from the subject's own responses performs better than the subject himself, and hence is very descriptive of that particular judge. Anderson has also included provision for an initial bias in the impression: one might, for example, be predisposed to like people even before meeting them or having them described. Finally, analysis of the subjects’ responses has depended heavily on the linear model, because it is central to the various conceptual and theoretical positions.
