Naval Postgraduate School students participated in a verbal discrimination experiment using three-word items of different frequency ratios. Half of the three-word items were composed of similar words and half, dissimilar words. Based on information theory, the words were grouped into two lists, both of equal lengths and approximately equal information. Performance by the subjects showed statistical significance for the trial and item-similarity interaction. The effect of using different frequency ratios within the separate word lists showed a facilitation of learning only for items with similar words. It was concluded that the differential ratios were not sufficiently apparent in the dissimilar items to permit the use of pre-experimental frequency differences as a cue to learning. Tables of data and references included. (Author/AL)