The false > correct phenomenon and subjective confidence: two separate forces behind response times in psychological testing

Across a wide assortment of psychological tasks, one finding keeps turning up: people answer correctly faster than they answer incorrectly. Researchers call this the false > correct phenomenon. Correct responses come with shorter latencies, wrong ones take longer, and the pattern holds up well enough across different kinds of tests that it has been treated as something close to a general law of response behavior.

Stefan Troche and Thomas Rammsayer wanted to pull apart what actually drives that speed gap. Is it the objective correctness of the answer that matters, or is it how sure the person feels when they respond? Those two things usually travel together, so teasing them apart takes a task designed to separate them.

They recruited 87 participants and gave each of them two visual discrimination tasks. In both, the job was simple to state: look at two lines and decide which one is longer. The difference lay in how answers were recorded. The first task used a plain two-alternative forced choice, longer line on the left or longer line on the right. The second task added a confidence dimension, asking people to rate their answers on four levels of certainty rather than just committing to one option.

The results split along the lines you might hope for if correctness and confidence are doing different jobs. In the forced-choice task, correct responses were reliably quicker than incorrect ones, the familiar effect. In the confidence-judgment task, though, that advantage for correct answers disappeared. What stayed strong was something else entirely: responses given with high confidence were much faster than responses given with low confidence, regardless of whether they happened to be right.

So the speed of a response seems to depend on two factors that can be peeled apart. One is whether the answer is objectively correct. The other is how confident the person feels while giving it. Both can shift latency, and they do not reduce to each other.

A second takeaway concerns how universal the false > correct phenomenon really is. It did not hold in the confidence-rating version of the task, which suggests the effect has limits. It shows up under some testing conditions and fades under others rather than applying everywhere response times are measured.