Search for rules for search

This paper presents the findings of some preliminary ‘laboratory’ investigations into ‘actual’ search behavior. Specifically we looked at situations in which searchers' initial information about the distributions was negligible. It seemed likely that simple ‘rules of thumb’ would be used in such situations. Our findings confirmed this view: we identified five such rules which between them ‘accounted’ for a high proportion of observed behaviour. Moreover, these rules although not ‘optimal’) appear to be reasonably good and may be fairly robust; jointly, they may constitute a better explanation (and predictor) of actual search behaviour than do the currently popular supposedly optimal rules.