A study was conducted to discover whether college students could predict which linguistic subunits of expository prose would be recalled. Comparison was made between the predicted subunits and the actual subunits recalled. Subjects were presented with 650 or 810 word prose passages and informed that they would be tested on their recall of passages at some time in the future. Half the subjects were tested immediately after reading the passages and half were tested after a seven-day interval. Two trained raters made judgements as to which linguistic subunits would be recalled. An independent group of 48 raters were given details of the experiment and asked to predict which phrase units they would recall had they been subjects in the experiment. The textual passages were ranked according to their predicted recalls and then divided into four groups ranging from highest to lowest predicted recall. The results showed that the predictions of which subunits would be recalled were generally accurate. It was suggested that raters could discriminate differences in meaningfulness among the units and could perceive differences in structural importance of linguistic subunits. References and tables are included. (AL)