The
study collected brain scans of 324 people, and also tested those
people for their intelligence and general knowledge. This is a good
sample size, so I won't be able to make the complaint that I
frequently make about neuroscience studies: the complaint that the
sample size was so small there was a high chance of a false alarm.
The
scientists attempted to derive scores for several brain parameters,
including the following:
GMV:
grey matter volume in the brain.
WMV:
white matter volume in the brain.
“NETstruc:”
global efficiency of the brain's structural network.
“NETfunc:”
global efficiency of the brain's functional network.
There
is no generally accepted algorithm for computing the last two of
these things, so the numbers that were calculated were somewhat
arbitrary.
Below
are the correlation coefficients (r) that the study found:
First Parameter | Second Parameter | Correlation | Extent of Correlation (Using Rule of Thumb Cited Below) |
GMV (grey matter volume in brain ) | “Fluid intelligence” | R= .18 | Negligible |
GMV (grey matter volume in brain ) | General knowledge |
r=.12
|
Negligible |
WMV (white matter volume in brain) | “Fluid intelligence” | R
= .11 “We observed no significant correlation between fluid intelligence and WMV.” |
Negligible |
WMV (white matter volume in brain) | General knowledge |
r=.10
|
Negligible |
“NETstruc:” global efficiency of the brain's structural network. | “Fluid intelligence” | R=.13 “The association between fluid intelligence and NETstruc lost its statistical significance after we accounted for multiple comparisons.” |
Negligible |
“NETstruc:” global efficiency of the brain's structural network. | General knowledge | R=.19 | Negligible |
“NETfunc:” global efficiency of the brain's functional network. | “Fluid intelligence” | R=.15 | Negligible |
“NETfunc:” global efficiency of the brain's functional network. | General knowledge | R=
.09 “General knowledge was not significantly associated with ... NETfunc.” |
Negligible |
The scientists made
eight different comparison trying to find a correlation between some
general brain parameter on the one hand and either intelligence or
knowledge on the other hand. All of these comparisons found only correlations less than .2, all of which are negligible according to the rule of thumb cited below.
The
authors of the study apparently followed a policy that any
correlation coefficient of .15 or greater should be announced as a "significant" correlation. But correlation coefficients between .15 and .20 are more accurately described as "negligible." In the scientific paper entitled,
“A guide to appropriate use of Correlation coefficient in medical
research,” we read the following: “A correlation coefficient of
0.2 is considered to be negligible correlation while a correlation
coefficient of 0.3 is considered as low positive correlation.” Below is Table 1 from that paper, which has the heading of "Rule of Thumb for Interpreting the Size of a Correlation Coefficient."
If you do a Google image search for "correlation coefficient interpretation," you will find several tables or guidelines that list all correlation coefficients of 0.2 or less as either "negligible," "very poor," or "very weak," and some of them (like the table above) actually list all correlation coefficients of .3 or less as "negligible." Since all of the main eight correlations found by the "Neural Architecture of General Knowledge" paper (listed in the first table of this post) are all
correlations of less than .2, they should all be classified as
negligible correlations.
Size of Correlation | Interpretation |
.90 to 1.00 (−.90 to −1.00) | Very high positive (negative) correlation |
.70 to .90 (−.70 to −.90) | High positive (negative) correlation |
.50 to .70 (−.50 to −.70) | Moderate positive (negative) correlation |
.30 to .50 (−.30 to −.50) | Low positive (negative) correlation |
.00 to .30 (.00 to −.30) | negligible correlation |
Below is an example of a negligible correlation of only .19, the same as the highest correlation found in the table above listing results from the "Neural Architecture of General Knowledge" scientific paper in question. Using this page of the "Spurious Correlations" we site, you can graph many other equally negligible correlations.
Here is another example, showing a correlation of only .20:
The very tiny effects found in the study can easily be explained solely by differences in perceptual speed and manual dexterity of test-takers, differences having nothing to do with intelligence, knowledge or memory. Any paper-and-pencil written test of intelligence or general knowledge is at least 5% a test of perceptual speed and manual dexterity, which crucially affects the scores on written tests.
By finding only negligible correlations between brain parameters and intelligence, and by finding only negligible correlations between brain parameters and general knowledge, the scientific study in question is quite consistent with what I have long asserted: that the human brain is not the source of human intelligence, and that the human brain is not the storage place of human memories. See here for the many reasons I have for making such assertions.
Similar results came from a 2017 study that studied the correlations between grey matter volume in the brain and performance scores on six cognitive tests. The study was entitled, "Global associations between regional gray matter volume and diverse complex cognitive functions: evidence from a large sample study." For each of the six cognitive tests, more than 1000 subjects were used (as we can see in Table 1). Table 2 of the study shows that it found only negligible correlations between grey matter volume and the scores on these six tests. The correlations were only 0.032, 0.101, -0.080, 0.088, 0.074, and 0.032, all negligible.
There are some problems with the ""Neural Architecture of General Knowledge" scientific paper that anyone attempting a replication should try to avoid:
(1) The study paper makes no mention at all of any blinding protocol, meaning that the scientists analyzing brain parameters might have known about the intelligence scores and knowledge scores corresponding to the brains they were analyzing. Those who made such an analysis should have had no knowledge of such intelligence scores and knowledge scores, to avoid analysis bias (the meager effects reported might have been entirely due to such a lack of a blinding protocol).
(2) The authors should have done a pre-registered study in which they committed themselves (before data collection) to exact algorithms of measuring brain connectivity, white matter volume and grey matter volume (rather than having the freedom to adjust such methods until a slight correlation was obtained).
(3) The test subjects should have been tested for perceptual speed and manual dexterity, with the intelligence scores and knowledge scores adjusted to account for perceptual speed differences and manual dexterity differences that can affect paper-and-pencil test scores (something which may be the sole cause of the very slight effects reported).
Here is another example, showing a correlation of only .20:
The very tiny effects found in the study can easily be explained solely by differences in perceptual speed and manual dexterity of test-takers, differences having nothing to do with intelligence, knowledge or memory. Any paper-and-pencil written test of intelligence or general knowledge is at least 5% a test of perceptual speed and manual dexterity, which crucially affects the scores on written tests.
By finding only negligible correlations between brain parameters and intelligence, and by finding only negligible correlations between brain parameters and general knowledge, the scientific study in question is quite consistent with what I have long asserted: that the human brain is not the source of human intelligence, and that the human brain is not the storage place of human memories. See here for the many reasons I have for making such assertions.
Similar results came from a 2017 study that studied the correlations between grey matter volume in the brain and performance scores on six cognitive tests. The study was entitled, "Global associations between regional gray matter volume and diverse complex cognitive functions: evidence from a large sample study." For each of the six cognitive tests, more than 1000 subjects were used (as we can see in Table 1). Table 2 of the study shows that it found only negligible correlations between grey matter volume and the scores on these six tests. The correlations were only 0.032, 0.101, -0.080, 0.088, 0.074, and 0.032, all negligible.
There are some problems with the ""Neural Architecture of General Knowledge" scientific paper that anyone attempting a replication should try to avoid:
(1) The study paper makes no mention at all of any blinding protocol, meaning that the scientists analyzing brain parameters might have known about the intelligence scores and knowledge scores corresponding to the brains they were analyzing. Those who made such an analysis should have had no knowledge of such intelligence scores and knowledge scores, to avoid analysis bias (the meager effects reported might have been entirely due to such a lack of a blinding protocol).
(2) The authors should have done a pre-registered study in which they committed themselves (before data collection) to exact algorithms of measuring brain connectivity, white matter volume and grey matter volume (rather than having the freedom to adjust such methods until a slight correlation was obtained).
(3) The test subjects should have been tested for perceptual speed and manual dexterity, with the intelligence scores and knowledge scores adjusted to account for perceptual speed differences and manual dexterity differences that can affect paper-and-pencil test scores (something which may be the sole cause of the very slight effects reported).
No comments:
Post a Comment