Four experiments were carried out to assess the validity of pause in speech as an index of unit boundaries in language. The criterion for the existence of a unit was resistance to fracturing. Verbal stimulus material learned by the Ss was fractured into parts in two different ways: at points where the Ss paused, and at points where there was no pause. The parts thus obtained were placed in a random order forming two new sets of verbal material which the Ss were required to learn. It was more difficult to learn the material constructed from the nonpause fractures. The same result was found for both narrative and for non-narrative material which was minimally syntactically organized. The result persisted when the nonpause material for one S was another S's pause-fractured material. This latter result indicates that pause is sensitive to individual differences in how verbal material is organized.