This paper reports an eye-tracking experiment conducted to compare alternative representations of directories typically shown on web pages in search of a best representation for deaf persons. The experiment simulated a directory-based information search task to understand how it is performed when directories are represented in text, labeled-pictograms, or unlabeled-pictograms. Twenty-one deaf and 21 hearing participants were asked to select one of 27 directories represented in one of the three alternative formats for each of 38 queries. The eye movement data were analyzed by the parametric ANOVA to understand how the method of directory search adopted by the hearing group and the deaf group might be different under the influence of the differences in directory representations. The result demonstrated that only in the labeled-pictogram representation, the hearing group and the deaf group performed equally well in terms of the eye movement measures.