A number of studies have shown that children learn noun pairs faster when they have been presented in a noun-verb-noun context than when they are presented in a noun-conjunction-noun context. Suzuki and Rohwer (1969) hypothesized that the underlying strings of sentences correspond to memory units, making it easier to recall two nouns from the same underlying string than from two different underlying strings. They reported that data from ftfth graders supported this hypothesis but that data from college students did not. The experiments reported here were done to determine whether the hypothesis would be supported (a) if the test rate were faster than that used by Suzuki and Rohwer or (b) if sentences longer and more complex than those used by them were employed. The recall data accorded with some predictions derived from the hypothesis. However, data from Ss' posttraining reports on how they learned the pairs suggest that Ss, particularly adults, often made up their own strings rather than using those provided, and raise the question of the adequacy of Suzuki and Rohwer's procedure for testing this hypothesis, particularly in adults.A number of experiments Rohwer & Lynch, 1967) have shown that two nouns connected by a verb are learned faster as a pair than are two nouns connected by a conjunction. To account for this "verb facilitation effect," Suzuki and Rohwer (1969) proposed the deep-structure hypothesis, which asserts that the underlying strings of sentences correspond to the functional units of memory storage. Thus, two nouns in the same memory unit (e.g., when connected by a verb) will be learned more easily than two nouns in different memory units (e.g., when connected by a conjunction) because of a simpler retrieval process.In an experiment to evaluate this explanation, Suzuki and Rohwer (1969) used both simple and complex strings. On study trials, Ss saw each noun pair capitalized and underlined in a syntactical context; on test trials, the first noun of each pair was presented alone and S was instructed to provide the missing noun. Suzuki and Rohwer reported that with both simple and complex strings, fifth graders performed better when the two nouns were connected by a verb (same underlying string condition) than when connected by a conjunction (different underlying string condition), thus supporting the hypothesis. For college students, however, Ss in the *Experiment I was carried out by W.S.F. for his senior seminar project, and Experiment II is based on an MS thesis by the senior author (Allison,