A meta-analysis examined a series of studies by F.M. Dwyer on the effect of illustrations on text comprehension. Principal component analysis was used to reduce the four posttests used by Dwyer to more fundamental factors of learning, followed by analyses of variance. All nine studies (involving secondary-school and college students) in which Dwyer provides the mean results for each experimental group and its control group for the identification, drawing, comprehension, and terminology tests were used. Results indicated that 2 factors (a vocabulary learning factor and a text-comprehension factor) accounted for most of the variance. Results also indicated a profound dichotomy between the vocabulary learning factor and the text comprehension factor results--the absence and presence of pictures, the degree of pictorial realism, and the absence and presence of color are all significant variables with respect to the vocabulary factor, but nonsignificant with respect to the text-comprehension factor. Findings suggest significant main effects only for the vocabulary factor, thus putting into question Dwyer's central hypothesis that realistic pictures accompanying text are significantly less effective than abstract ones. (Contains 87 references, 13 tables and 1 figure of data. Appendixes provide examples from the four posttests used in Dwyer's studies.) (RS)