Divergent thinking (DT) tasks are frequently used to estimate the potential for creative thinking. They are usually scored for 3 ideational outcomes: fluency, flexibility, and originality. The focus here is on flexibility, which is quantified in 2 ways. One flexibility score is based on the number of conceptual categories. The second is based on the number of switches between conceptual categories. In both cases, scoring is manual and depends on trained raters. The objective of this study was to test the possibility that category switches could be predicted by the latency between consecutive ideas that were generated during think-aloud sessions. To this end, 4 different DT tests were employed and 1,408 responses were produced while thinking aloud about them. The ideas were transcribed and the latency (in seconds) was calculated as the difference between 2 consecutive responses. A 3-level hierarchical linear modeling analysis was used. This allowed identification of the variance attributable to the DT tasks and to individual differences. Most important was that it allowed an examination of the relationship between latency and categorical switch, after controlling for DT test and individual differences. Results indicated that latency was approximately 5 s longer when the ideational category was switched than when a new idea was found within the same category. In addition, latency was 2.5 s higher in figural tasks than verbal tasks. Implications of the findings are discussed for creativity assessment.