In reply to cb294:
> Arrggh, when will people start understanding statistics?
> There is no hard and fast rule as to how many measurements you need, once you pass some threshold required by the statistic test you chose beforehand (e.g., limits on how many of your data points are allowed to have a value of zero). Of course, if the effect turns out to be weak, you may miss it with a small sample size, even if it would have been significant with a larger sample.
> You can also calculate the required sample size beforehand if your either (roughly) know the size of your expected effect or the desired sensitivity of your study (e.g., there is normally no point in making ones statistic test more precise than the original measurements).
> 36 subjects is certainly enough to satisfy the mathematical constraints for most tests, and if it is sensitive enough for them to find a difference, then fine.
> Anyway, I don´t need to read the study, daily experience tells me this is true (sample size 2, me and my wife....).
> CB
36 subjects and what seems like 1-2 navigational exercises (that may be biased as they were probably designed by Carl [a man], the studie's lead author). That's by far nowhere near enough to develop an accurate answer.
Secondly it's a hugely well known issue that most university studies suffer from biases in their participants. They often recruit only university students, and thus you're already biasing the study by a magnificent amount (furthermore the students are often from the same department). To truly make a general statement about the sexes navigational differences you would need participants from all cultures, all levels of education, all ages, and a vast array of navigational exercises.
Furthermore the bar for statistical significant is often set very low (usually p < 0.01), which basically means 1/100 studies are _totally_ wrong. This is usually only taken into account by meta-analyses, which are probably the only trustworthy form of social studies.