The paper that "Science" referred to is :
Willerman, et. al., "In Vivo Brain Size and Intelligence", Intelligence 15,
223-228 (1991).
Some details :
The sample was 40 right-handed Anglo introductory psychology students,
about 19 years old. These were drawn from a larger pool of introductory
psychology students with SAT scores either >= 1350 or =< 940. "With
prior approval of the University's research review board, students selected
for MRI were required to obtain prorated full-scale IQs of >=130 ( M=136.4,
SD = 3.95) or <= 103 (M=90.5, SD = 8.12), and were equally divided by sex
and IQ classification."
"Using the lowest margin of the cerebellum in a midsagittal view to align
the first axial (horizontal) MR slice", 18 images were taken, each separated
by 2.5mm and 5mm thick. Each image was 256 x 256 pixels with 256 levels of
gray. Within the boundary of the scalp, all gray scale intensities of < 96
were converted to zero to delete the skull, meninges and interhemispheric
fissure; other brain coverings were deleted manually with a cursor. The
non-zero pixels were then counted, their summed value serving as the index
of overall brain size.
"Brain tissue was present on only 17 of the 18 slices for 12 of the subjects
( 9 women). "
Some peculiarities :
"Average-IQ men were taller than high-IQ men (p < 0.05)."
"After controlling for sex, correlations of height and weight with brain
size were only r = 0.09 and r = 0.10" (more on why I consider this peculiar
later on.)
Some of the results :
"Among men, IQ scores as a continuous variable correlated with brain size,
before and after adjusting for body size, r = 0.51 (p < 0.05) and r = 0.65
(p< 0.01). Corresponding correlations for women were r = 0.33 and = 0.35
both n.s."
[does n.s. mean non-significant ? If it does, is there any sense to the
following line : ]
"With sexes pooled, the IQ-adjusted brain-size correlation was r = 0.51
( p<0.1 )"
"This correlation is higher than expected for the general population because
of selection of extreme IQ groups. Applying a statistical correction (Guilford
and Fruchter, 1973) predicted a correlation of r = 0.35 for a more
representative sample."
"Not all brain levels contributed equally to the brain-size-IQ correlation;
[figure 1 : the graph that is displayed in "Science". ]
size differences were greatest for ventricular-level slices in men. These
levels include language circuits, association fibers, and association cortex."
**
The figure caption :
As mentioned earlier, the average sizes of 16 slices ( labelled from -9,-8,
-7,..., +6) for each of the four groups is plotted on a plot of slice level
vs. brain size standard score that ranges from -1.0 to +1.0. The largest
differences are between average-IQ and high-IQ males, around slices -3 to +2,
with the high around brain size score 1.0 and the low around brain size
score -0.5.
The caption reads : Brain area in standard scores ( M = 0 +- SD = 1) by
high versus average IQ and sex, adjusted for weight and height. Squares
refer to men and circles to women, empty symbols refer to high IQ, filled
symbols to average IQ. Slice 0 is at the mid-ventricular level as shown
in the insert [ the insert is a line drawn through a sketch of the brain ]
5 mm thick MRI images are separated by 2.5 mm. Large brain area -IQ
differences in men near the ventricles (slices -3 below to +1 above 0)
include neural substrates of language and association. A repeated-measures
ANOVA on ventricle size across slices -3 to +1 reveals no effect of sex,
IQ or their interaction (all ps > 0.20) so ventricles are included in area
measurements. Tissue above or below graphed levels are excluded because all
subjects could not be represented.
***
If you take out the ventricular area, isn't the area of brain slice in
the mid-ventricle slices is likely to be much smaller than other brain slices ?
If so, then if the areas of each of the slices are plotted on the same scale,
then these differences will appear to be small relative the sizes of the
other slices, is it not ? That is, there are large differences in small areas
of the brain -- is that correct ?
***
Coming to what I mentioned as peculiarities :
(1) "Average-IQ men were taller than high-IQ men (p < 0.05)."
Gould, "The Mismeasure of Man", page 108,
"...Jensen cites an average correlation of 0.25 between IQ and physical
stature" [ Jensen, 1979, "Bias in Mental Testing",pp 361-362].
(2) "After controlling for sex, correlations of height and weight with brain
size were only r = 0.09 and r = 0.10"
"Analysis of Brain Weight", Ho, et. al., Arch. Pathol. Lab. Med., Vol 104,
Dec. 1980, pages 635, and 640 : finds a correlations of 0.24 and 0.23
(p < 0.01) between body weight and brain weight for white males and females
respectively. They find correlations of 0.20 and 0.24 (p<0.01) between body
height and brain weight for white males and white females, respectively.
These numbers are from post-mortem results of people without any obvious
brain pathology. They also write that "The reported coefficients for
brain weight and body height relationships vary from 0.17 to 0.24 form men
and from 0.09 to 0.23 for women; these are comparable to our results".
[The results vary by race and gender : for example the correlation between
brain weight and
body height body weight body surface area
white male 0.20 (p<0.01) 0.24 (p<0.01) 0.27 (p<0.05)
black male 0.20 (p<0.01) 0.15 (p<0.05) 0.20 (p<0.01)
white female 0.24 (p<0.01) 0.23 (p<0.01) 0.29 (p<0.01)
black female 0.15 (p<0.05) 0.10 (p<0.05) 0.14 (p<0.05)
(body surface area in square meteres equals
(weight in grams)**0.425 x (height in centimeters)**0.725 x 0.007184
]
(3) Another peculiarity, visible from the graph is that the women have
brain slice areas less than men in the graph; the graph presumably is with
height- and weight-corrected figures. Ho et. al, have "White women have
more brain mass than white men when adjusted for body weight and surface
area, less when adjusted for height."
***
Finally, from the ISI citation index in the sciences (SCI; SSCI is said to
have more citations), the citations listed for Willerman, et. al., from 91
to May-June 94 are : (of these, Andrease NC is another MRI study like
Willerman , the "Nature" stuff is a debate about P. Rushton; I didn't check
Rushton, Vernon, Jensen or Sells,)
Author Journal Vol page year
Brand CR Nature 359 768 92
Rushton JP Nature 358 532 92
Anderson B Neurosci L 153 98 93
Andrease NC Am. J. Psy. 150 130 93 (Psy = Psychiatry)
Heller J. Affect D 29 129 93
Schultz R Behav. Genet 23 565 93
Vernon PA Behav. Genet 23 568 93
Jensen AR Ann NY Acad 702 103 93
Anderson B Brain Res 641 160 94
Schultz R Ann Neurol 35 732 94
Ratcliff Develop Med 36 533 94
Sells CJ J Pediat 125 S 9 94
***
arun gupta
Brain size and IQ