One of the few established cognitive differences between men and women lies in spatial ability. But in recent years, this ‘fact’ has been shaken by evidence that training can close the gap between the genders. In this new study, 545 students were given a standard 3D mental rotation task, while at the same time manipulating their confidence levels.
In the first experiment, 70 students were asked to rate their confidence in each answer. They could also choose not to answer. Confidence level was significantly correlated with performance both between and within genders.
On the face of it, these findings could be explained, of course, by the ability of people to be reliable predictors of their own performance. However, the researchers claim that regression analysis shows clearly that when the effect of confidence was taken into account, gender differences were eliminated. Moreover, gender significantly predicted confidence.
But of course this is still just indicative.
In the next experiment, however, the researchers tried to reduce the effect of confidence. One group of 87 students followed the same procedure as in the first experiment (“omission” group), except they were not asked to give confidence ratings. Another group of 87 students was not permitted to miss out any questions (“commission” group). The idea here was that confidence underlay the choice of whether or not to answer a question, so while the first group should perform similarly to those in the first experiment, the second group should be less affected by their confidence level.
This is indeed what was found: men significantly outperformed women in the first condition, but didn’t in the second condition. In other words, it appears that the mere possibility of not answering makes confidence an important factor.
In the third experiment, 148 students replicated the commission condition of the second experiment with the additional benefit of being allowed unlimited time. Half of the students were required to give confidence ratings.
The advantage of unlimited time improved performance overall. More importantly, the results confirmed those produced earlier: confidence ratings produced significant gender differences; there were no gender differences in the absence of such ratings.
In the final experiment, 153 students were required to complete an intentionally difficult line judgment task, which men and women both carried out at near chance levels. They were then randomly informed that their performance had been either above average (‘high confidence’) or below average (‘low confidence’). Having manipulated their confidence, the students were then given the standard mental rotation task (omission version).
As expected (remember this is the omission procedure, where subjects could miss out answers), significant gender differences were found. But there was also a significant difference between the high and low confidence groups. That is, telling people they had performed well (or badly) on the first task affected how well they did on the second. Importantly, women in the high confidence group performed as well as men in the low confidence group.