This paper analyses the role of information in the search process. Ibuild a simple model of a good with two random attributes with somejoint probability distribution. I consider seemingly unimportant changesin this distribution, i.e. changes which neither affect expected utility norits variance. These changes have a great impact on the search behaviour:the customer may start to search the characteristics and buy the goodeven if she did not do so before. The optimal search rule is derived andthe model is generalized to multiple objects.