Training improves discrimination of judgements of solvability, but not how well they predict later problem-solving success

Deciding whether a problem is solvable is an important step in the problem-solving process. Research suggests that rapid, initial judgements of solvability can discriminate solvable from unsolvable problems. Using solvable and unsolvable anagrams, we explored if training improves discrimination of judgements of solvability, and if judgements of solvability predict later problem-solving success. In a no-training condition, anagram presentation duration was 2 s in each of 4 blocks. In a training condition, anagram presentation duration started at 16 s and halved across blocks (16 s, 8 s, 4 s, 2 s). Judgements of solvability were discriminating in each block. Importantly, discrimination in the final 2 s block was more accurate after training. After making judgements of solvability participants attempted to solve the solvable anagrams. In both experiments, anagrams deemed solvable were more likely to be solved than not solved. However, for anagrams in the 2 s block, training did not make judgements of solvability more predictive of anagram-solving success. In sum, judgements of solvability were discriminating and trainable, and they predicted subsequent problem-solving success—but their predictiveness did not improve with training.