Test-retest reliability and differential patterns of score change on the group embedded figures test

Retest reliability of the Group Embedded Figures Test (GEFT) was examined under three different intervals between test and retest, all with an interpolated cognitive task. Despite a significant increase in the group mean score from test to retest in all three experiments, retest reliability coefficients were high, .78-.92. There was also a suggestion that reliability increased with duration of delay. Examination of individual patterns of test-retest score change revealed four patterns: consistent field dependent (FD), consistent field independent (FI), unclassifiable (UNCL), and latent field independent (FIL), whose retest scores took them from the FD range to the FI range. The latter two patterns accounted for the significant retest improvement. Relative frequencies of subjects in each pattern were relatively constant over the three experiments. Subjects in the pattern categories also differed with respect to score on number series completion tests, suggesting that the four patterns reflect more general individual differences in analytic ability. It was concluded that the GEFT is a reliable test, but suggestions for a more error-free classification procedure based on a test-retest score pattern is proposed. Implications of the findings for intervention studies of cognitive style are also noted.