The compound Remote Associates Test (RAT) is a classic measure of creativity. Participants are shown three cue words (sore-shoulder-sweat) and asked to generate a word that connects them (cold). Theoretical views of RAT performance differ in the degree to which they conceptualize performance as depending on automatic spreading activation across semantic networks, strategic generation of bi-associations, and other analytical processes (e.g. executive processes that support fluid intelligence). We tested these views by adding a fourth cue word to determine whether it impaired RAT accuracy (e.g. generation of bi-associations), impaired response times (analytic processes), or improved RAT accuracy without changing response times (e.g. spreading activation). Across four experiments, 551 adults completed 3- and 4-item RAT trials that were matched on linguistic and semantic metrics. Across experiments, adding the fourth word improved accuracy by 27.91%. This performance gain occurred with either modest or no changes to response times or ratings of insight/strategy use. Interestingly, the fourth word predominantly benefited accuracy and response times on difficult trials; on easy trials, the fourth word impaired or did not change performance. The findings suggest that both automatic and strategic/analytical processes contribute to successful RAT performance, with relative dependence on these processes dynamically adapting to the demands of the individual trial.