Accuracy, latency, and confidence in abstract reasoning: the influence of fear of failure and gender

On a wide range of cognitive problems, people spend more time on the answers they get wrong than on the ones they get right. This pattern has a name, the false>correct phenomenon (F>C for short), and it has turned up again and again, in tests that adapt their difficulty to the test-taker and in ones that do not. One common way of explaining it points to a choice-accuracy heuristic. That idea draws some of its weight from a separate finding: how long someone takes to respond seems tied more closely to how sure they feel than to whether they are actually correct.

Franzis Preckel and Philipp Alexander Freund set out to connect several of these threads. Their central prediction was simple to state. If a person answers with high confidence, that answer should come faster than one given with low confidence. Response time, in other words, ought to track the subjective sense of being right.

They ran 103 students through a figural matrices test of abstract reasoning, delivered by computer and not adjusted to each person's ability. The twist was that after every item, participants rated how confident they were that their answer was correct. That gave the researchers two measures to line up against each other: the time taken, and the certainty reported.

The F>C effect showed up as expected. Wrong answers took longer than right ones. What did not appear was any sign that this differed by ability level; stronger and weaker reasoners behaved similarly on that score. The confidence prediction held as well. Answers delivered with confidence were quicker than those given hesitantly.

When the researchers held accuracy constant, confidence still accounted for a sliver of the variation in how long answers took, though only a small sliver. More striking was what fed into the confidence ratings themselves. Both gender and fear of failure predicted how confident participants reported being, and they did so independently of response time, of the actual score, and of the motivational variables the study measured.

That combination led Preckel and Freund to treat confidence less as a moment-to-moment reaction to a single item and more as something resembling a stable personality trait, one shaped by whether the answer happens to be correct, by the person's gender, and by how much they fear failing. The usable terms from the work: response latencies, the false>correct phenomenon, confidence ratings, abstract reasoning tasks, and gender differences.