“Aha” is right: insight solutions are more likely to be correct than are analytic solutions

When people attempt to solve a problem, the solution might be achieved through methodical analysis, or through a sudden insight, often accompanied by an “Aha!" or "Eureka!" experience. During a Compound Remote Associate (CRA) paradigm (Bowden & Jung-Beeman, 2003b), participants view three problem words (crab, pine, sauce) and must generate one solution word (apple) that can form a compound word or common two words phrase with each problem word (crab apple, pineapple, apple sauce). Solvers can achieve the solution to a CRA either with a moment of insight, achieved primarily without conscious deliberation, or with analysis, achieved primarily with conscious deliberation. Behavioral (Beeman & Bowden, 2000; Bowden & Jung-Beeman, 2003a) and neuroimaging (Jung-Beeman et al., 2004; Kounios et al., 2006) studies have demonstrated that, compared to analytic solutions, achieving solution via insight involves a stronger contribution from a neural network, including right anterior temporal lobe and the anterior cingulated cortex (ACC). We investigated how problem-solving style affected response accuracy in two experiments. Participants attempted to solve CRA problems in 15 seconds each, 90 problems in the first experiment and 120 in the second. After each problem, participants reported how they achieved the solution– via insight or analysis. Participants were given no feedback regarding whether the solution they reported was correct. We analyzed both problems solved correctly and “errors,” namely incorrect responses assumed by participants to be correct. In Experiment 1, of all responses participants (n=40) labeled as insight, 91% were correct and 9% were errors; of all responses labeled as analytic, 81 % were correct and 19% were errors. A repeated measures 2 x 2 ANOVA comparing problem solving style and accuracy revealed a reliable interaction. That is, participants reported generating their correct solutions more often with insight than analysis, but showed the opposite pattern when they generated incorrect responses. Experiment 2 replicated this interaction: 95% of insight responses were correct, and only 85 % of analytic responses were correct. Experiment 2 further examined solution accuracy across two time windows to better isolate unbiased responses. Even excluding errors produced before 2 seconds (i.e. when participants were likely to have immediately recognized the solution) and after 10 seconds (i.e. when participants were likely to have guessed as the time limit was ending) incorrect responses were produced more by analysis than by insight. The result demonstrates that problems solved via insight are more likely to be correct than problems solved via analysis. Insight may be more likely to yield correct solutions because it depends on integration of multiple weak associations; i.e., weak associations from all three problem words to the solution must summate to boost activation of the solution concept into consciousness. When the solution does emerge, it necessarily fits all three words. Alternatively, the result might be explained by the lower probability to make an error in a one-step insight process compared to a multi-step analytic strategy, with each step prone to error.