In a later study, Garrison (1987) sought to test Boshier's congruence model with 110 learners in a high school completion program (Grade 10 math classes). He found that dropout, in this group had less self/other incongruence than persisters. This was the reverse of one of Boshier's findings. Garrison uses this anomalous finding to stress the need for limiting generalization to specific populations. In this study, Garrison was able to correctly predict 93 percent of the persisters, but only 20.8 percent of the dropouts, leading him to conclude that "reasons for dropout are many and complex, thus making prediction of dropout very difficult," (p. 219).

Darkenwald and Gavin (1987) used social environment theory "to determine the relation of dropout behavior to the social ecology of the classroom" (p. 152). Their sample consisted of 77 adults enrolled in GED preparation classes. They found a significant difference between persisters and dropouts in the item of Affiliation. Dropouts expected a classroom environment characterized by less social involvement with other students. The authors remark that this finding has more implications for research and theory-building than for practice! It has been established that the goal of obtaining a GED is a great motivator in adult basic education (Bosma, 1988; Diekhoff and Diekhoff, 1984; Valentine and Darkenwald, 1986). Adults in a GED preparation class are likely highly motivated and goal-oriented. One explanation of the Darkenwald and Gavin finding could be that the social atmosphere detracts from the work-orientation and purposefulness of some adults in this group.

Diekhoff and Diekhoff (1984) discovered in their analysis of 66 students enrolled in literacy training that dropouts tended to be (a) young, (b) Hispanic, (c) not seeking GED certification, (d) unemployed but available for work, and (e) from illiterate families. Using these variables, the authors were able to successfully assign 82 percent (36 out of 44) of the students to a "persister" or "dropout" group. An attempt to cross-validate the study the following year with a new sample of 48 students was not successful. However, the authors explained this discrepancy by indicating that program procedures had been altered between the two program years. In an attempt to reduce the attrition rate of 53 percent in the first year, new intake and screening procedures were introduced. These included a student waiting-period of up to one month before being assigned a tutor after the initial interview. This procedure became a self-screening device, so that students most likely to drop out did so, in effect, before actually entering the program. This dropout group evidently closely matched the dropout profile in the original sample. The program's attrition rate was considerably lower the second year.