Classical measurement theory was used to investigate the measurement (psychometric) characteristics of both parts of the Finding Embedded Figures Test (FEFT) administered in either a "no guessing" supply format or a multiple-choice selection format to undergraduate college students or to middle school students. Three issues were addressed in this study: (1) how the alpha components for data from the FEFT compare across studies; (2) how test and item difficulty and discrimination coefficients compare across test administrations; and (3) how variables such as gender and grade levels of middle school students influence FEFT performance. Analyses were based on data from 69 undergraduates tested in a 1988 supply format study, 155 undergraduates who completed the multiple-choice format in 1989, and 1,528 middle school students who completed the multiple-choice format in this study. Coefficient alpha for the FEFT ranged between 0.81 and 0.90 across samples and administration formats. Items generally had desirable psychometric characteristics across studies. Construct validity analyses suggested that the measure is reasonably valid. Four tables present data from the study and three appendices give the FEFT composite scores. (SLD)