The Harrington and Farias (2008) attempted to determine "why FMRI studies evaluating language related sex effects have been so inconsistent." Previous studies have failed to find consistent evidence for neural differences in language processing between men and women. While their study did find greater left hemisphere activations in men and greater right hemisphere activations in women, the main conclusion was that: "some methods . . . revealed sex differences while others methods did not, indicating a dependence on methodology." In general, however, Harrington and Farias did not indicate that sex differences in language processing were consistent enough to warrant separate groups for most FMRI studies.
Behavioral differences in language tasks seem to be fairly robust between men and women. However, they don't tend to appear in the sample size used in most fMRI studies (n = 40 in Harrington and Farias, and no behavioral difference on the language tasks was detected). I'm not surprised to hear that you can find sex differences by some methods and not by others; however, I do wonder about reporting bias. I try to write about "gender similarities" as often as possible, but they're not reported as often as "gender differences." They're not as sexy. How much evidence of similarity never gets published, do you think? As with this study, most similarities get marked down to "insufficient power" in the sample size. How many insufficiently powered similarities does it take to reach the conclusion that the null hypothesis (i.e. "no difference") is true?
Find out the day's topic before you read: follow diffblog on Twitter! Diffblog also available on LiveJournal.
Subscribe to:
Post Comments (Atom)
Posts
2 comments:
Not to mention the huge publishing bias. You generally can't publish null effects in any prestigious journals, so unless they are a side point to a different study, null effects don't tend to get reported.
... and, clicking "follow up".
Post a Comment