Letter to the Editor |
Size does count: a reply to Kitazawa and Kansaku
E-mail: i.sommer{at}azu.nl
Received January 13, 2005. Accepted .
Sir,
In their Letter to the Editor, Kitazawa and Kansaku responded to our meta-analysis on sex differences in language lateralization measured with functional imaging techniques (Sommer et al., 2004
). Though Kitazawa and Kansaku gave an interesting interpretation of the data, we do not agree with the alternative calculation of the effect sizes of two studies they suggested.
Kitazawa and Kansaku hypothesized that sex differences in language lateralization may be present only when subjects listen to stories, since two studies (Kansaku et al., 2000; Philips et al., 2001), applying story listening as an activation paradigm, both reported significant sex differences. However, both studies had relatively small samples sizes: Kansaku et al. (2000) included 14 females and 16 males, and Philips et al. (2001) included 10 males and 10 females.
One of the most remarkable findings of our meta-analysis is that sex differences are almost exclusively reported in studies with a small sample size. This is a strong suggestion that there is a publication bias for studies reporting a significant sex difference in language lateralization over studies that observed no difference. It also suggests that there may be many studies with small samples that found no sex difference, but have never been published. Rosenthal (1991) described this phenomenon as the ‘file drawer problem’. Thus, the interpretation of Kitazawa and Kansaku that a sex difference in lateralization exists for story listening is an interesting thought, but replication in a large sample is needed before we should reach a conclusion.
In the second part of their letter, Kitazawa and Kansaku suggested that the effect sizes of the studies of Kansaku et al. (2000) and Philips et al. (2001) should be larger than the sizes used in the meta-analysis, with Cohen's d of 1.77 and 3.0, respectively. In calculating these effect sizes, they restricted the analyses to the temporal lobe. However, this restriction is not correct, since the original studies defined more volumes of interest (also including frontal and parietal language areas), which all showed significant language activation. A post hoc restriction to only that area where the sex difference turned out to be significant (temporal lobe) is not justified. Furthermore, it is important to calculate the effect sizes of all studies in the same fashion. In our meta-analysis, the effect sizes were calculated from all language-related cortical areas of both hemispheres. By applying another method to calculate the effect sizes of two studies, Kitazawa and Kansaku invalidate the outcome of the meta-analysis.
We therefore reject the effect sizes as calculated by Kitazawa and Kansaku and stick to our conclusion that sex differences in language lateralization are not generally present at the population level.
References
Kansaku K, Yamaura A, Kitazawa S. Sex differences in lateralization revealed in the posterior language areas. Cereb Cortex 2000; 10: 866–72.
Phillips MD, Lowe MJ, Lurito JT, Dzemidzic M, Mathews VP. Temporal lobe activation demonstrates sex-based differences during passive listening. Radiology 2001; 220: 202–7.
Rosenthahl, R. Meta-analytic procedures for social research. London: Sage Publications; 1991.
Sommer IEC, Aleman A, Bouma A, Kahn RS. Do women really have more bilateral language representation than men? A meta-analysis of functional imaging studies. Brain 2004; 127: 1845–52.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||