Astrology vs. neuroscience

[JoeTheJuggler]

I don't mind if a mod cares to move this to the Science subforum.

Just read this article:
http://pagingdrgupta.blogs.cnn.com/2010/06/22/personality-shows-up-in-brain-structure/

about a study that found several correlations between brain structures (measured via MRI) and personality traits (measured by some sort of personality test).

Of course, this sort of thing is fascinating in itself, but it made me think that this is similar to something astrologers claim to be able to do. That is, they often claim to be able to discern personality based on natal charts or some such. They claim there is a correlation between the apparent positions of the planets at the time of birth and the personality of the individual. (The most popular sort of astrology, sun sign astrology, claims that your zodiac sign correlates with personality traits.)

Trouble is, they've never been able to show that such a correlation exists. This study shows exactly what such a test might look like. (In place of the MRI measures, you hypothesize consistent correlations between the natal chart, sun sign or whatever and the personality test results.)

So, any astrology bleevers care to explain why they have failed to substantiate their claims?

ETA: I should add that what such a correlation study does NOT look like is for an astrologer to do a reading and then a subject to volunteer a bunch of hits. This is just a demonstration of the Forer Effect. (Something that yields over an 80% hit rate when the reading is just a bunch of blank stuff that could apply to anyone. That is, people rate it over 4 out of 5 as uniquely suiting their personality.)

 

[Dancing David]

I do not know,
http://www.tc.umn.edu/~cdeyoung/Pubs/DeYoung_2010_Big_Five_brain_structure_PS.pdf
ETA
The initial effects size is so small .24, but then they use a Monte Carlo Method to further look at it. At this point I would say insignificant, I will reread and then look up the terms. So maybe the effects size is .48 after the adjustment, so I am not sure.

 

[JoeTheJuggler]

I do not know,
http://www.tc.umn.edu/~cdeyoung/Pubs/DeYoung_2010_Big_Five_brain_structure_PS.pdf
ETA
The initial effects size is so small .24, but then they use a Monte Carlo Method to further look at it. At this point I would say insignificant, I will reread and then look up the terms. So maybe the effects size is .48 after the adjustment, so I am not sure.

I confess I didn't read the study but just the CNN article describing it.

My main point is that this is the proper way to test the hypothesis that personality traits are correlated to something measurable about various brain structures. When you ask an astrology believer for proof that the apparent positions of celestial bodies influences or determines personality traits, they only offer something very similar to cold reading, with a lot of retrofitting and cherry picking.

I can't remember the source, but there was a video kicking around that was touted as strong evidence of an astrologer's ability. They had the astrologer do readings on a group of people, and the people then evaluated the readings. They claimed something close to an 80% hit rate.

The problem was that "hits" were counted when, for example, the reading said that a person was a professional dancer or martial artist, and the person (who was neither a professional dancer or martial artist) said that she once took a karate class and that she likes to dance. The other problem is that the 80% rate is no better than the Forer Effect, which only measures how willing people are to retrofit a generic "reading" to uniquely apply to themselves.

 

[JoeTheJuggler]

I do not know,
http://www.tc.umn.edu/~cdeyoung/Pubs/DeYoung_2010_Big_Five_brain_structure_PS.pdf
ETA
The initial effects size is so small .24, but then they use a Monte Carlo Method to further look at it. At this point I would say insignificant, I will reread and then look up the terms. So maybe the effects size is .48 after the adjustment, so I am not sure.

I'm not familiar with the Monte Carlo correction, but the authors claim the results are significant for 4 of the 5 traits. Until I hear otherwise, I'll consider their statistical analysis valid since this went through a decent peer review process. (But I'm certainly willing to hear otherwise, and happy to learn!)

The results of our analysis are reported in Table 1, which lists all clusters related to each of the Big Five traits, controlling for age, sex, and the other traits. The table also indicates which of these clusters reached the cluster-size threshold for significance (p < .05, corrected), in order to be considered ROIs. (Controlling for indices of registration quality did not substantively change results; see the Supplemental Material.) We found evidence to support our hypotheses about four of the Big Five. Specifically, for Extraversion, Neuroticism, Agreeableness, and Conscientiousness, ROIs were in the set of regions hypothesized to be related to each trait (see Table 1 and Fig. 1).

 

[Maia]

I read the original study, and I've got to say that I think that the researchers are really stretching here. Per their introduction:

descriptive taxonomies have not typically been linked to
empirical data from neuroscience research.

But I think that there are some very good reasons that there haven't been, one of which may very include that given the current state of our ability to define these taxonomies, they can't be. It isn't enough to just "link them with research", even with good research. The link itself has to be well justified.

However, the Big Five factors are descriptive rather than
explanatory constructs and do not inherently provide a theory
of the underlying forces that produce these five dimensions of
individual differences in personality.

Yes, that's exactly the case. But that's just why the link between theories of personality and neuroscience would have to be much, much stronger than it is in this study. A correlation study is not enough; the link itself would have be more strictly defined.

And then we get definitions such as:

Agreeableness appears to identify the collection of traits
related to altruism: one’s concern for the needs, desires, and
rights of others (as opposed to one’s enjoyment of others,
which appears to be related primarily to Extraversion). The
positive pole of Agreeableness describes prosocial traits, such
as cooperation, compassion, and politeness, whereas its negative
pole describes antisocial traits, such as callousness and
aggression.

And here, I think, is where we really have the problem. How on earth is it possible to get away from describing the personality factors in terms of... the personality factors? Given the way that they've already been defined, there seems to be an irreducible tautology. This may be one thing for personality tests, but we need something more than this to try to link it with neuroscience. It just doesn't really work.

I think I can see what they're trying to do here-- to make personality psychology more "scientific"-- and it's an interesting experiment, but I just don't think it's really working. The way that something sort of like this could be legitimately done, for instance (to go to a field where I actually do know something about it, because I've had to learn), is to study the brain functioning of people who've been diagnosed with ADHD through behavioral traits. Certain predictable neurological abnormalities show up over and over and over again. These abnormalities tend to correlate with heightened irritability,for instance. A person who has this abnormality shows a higher score on a scale that shows this trait. A person who does not have it will show a lower score.What makes this original study impossible to take seriously-- leaving aside anything else-- is that there's no control group, and the design really makes it impossible to have one. The theoretical ADHD study would not be at all comparable, come to think of it, because it would measure reactions to something identifiable that is happening in the brain (and that could potentially be changed with medication). This, I think, is why the original study resembles astrology. What is it studying? What is it measuring? Without a control group, we simply don't know.

I'm sorry to say this, but there's something weirdly woo-ish about the premises of the entire original research article. Ultimately, tt reminds me not even so much of astrology, but of glorified phrenology. I have to call them as I see them!

 

[Pup]

I don't know enough about the science or statistics to comment on that, but as far as a parallel to any kind of woo goes, the first thing that occurred to me is phrenology.

 

[Dancing David]

I'm not familiar with the Monte Carlo correction, but the authors claim the results are significant for 4 of the 5 traits. Until I hear otherwise, I'll consider their statistical analysis valid since this went through a decent peer review process. (But I'm certainly willing to hear otherwise, and happy to learn!)

The base correlation is r=.24 , which means the traits are 76% not related to the structures that they claim.

Valid does not mean meaningful.

ETA: I saw a guy riding a unicycle and juggling basket balls and thought of you this week, while in Oregon along RT. 20

 

[JoeTheJuggler]

Valid does not mean meaningful.

But if the statistical analysis the authors did was valid, and they concluded that there is a significant correlation, then it does. But again, I'm not savvy enough on this level of statistics, and it wouldn't be the first time a paper with conclusions not supported by the data was published in a peer reviewed journal, so I will certainly accept your position that this paper doesn't prove the correlation.

Again, though, the problems with astrology "evidence" goes far beyond this sort of problem.

ETA: I saw a guy riding a unicycle and juggling basket balls and thought of you this week, while in Oregon along RT. 20

 

[JoeTheJuggler]

A correlation study is not enough; the link itself would have be more strictly defined.

I agree, but my point is that correlation between personality traits and measures of brains structures (assuming these are statistically significant correlations) would be way more than what astrology offers!

 

[Dancing David]

I agree, but my point is that correlation between personality traits and measures of brains structures (assuming these are statistically significant correlations) would be way more than what astrology offers!

My issue Joe is that a correlation of r=.24 may be statiticaly valid with a p=.05, yet a correlation of .24 is not significant.

Back in the old days my psych profs said anything below .68 was not meaningful.

 

[JoeTheJuggler]

My issue Joe is that a correlation of r=.24 may be statiticaly valid with a p=.05, yet a correlation of .24 is not significant.

Back in the old days my psych profs said anything below .68 was not meaningful.

I understand. When I used "valid" I only meant it reference to their statistical analysis overall (not the specific meaning that the correlation was valid). In other words, if their statistical analysis was good or legit, and based on that they said, "We found evidence to support our hypotheses about four of the Big Five. Specifically, for Extraversion, Neuroticism, Agreeableness, and Conscientiousness, ROIs were in the set of regions hypothesized to be related to each trait", I can only suppose that the statistical analysis said that the correlations were significant. That is, I was relying on the peer-review process to trust that this conclusion was supported by the legitimate statistical analysis of the data. (Similarly, I wouldn't assume that the data collection was fudged--that there was bias in measuring the various structures on the MRIs.)

But again, I'm not savvy enough in statistics to critique their methodology. Your objections have made me skeptical that their conclusion is warranted, so I reserve judgement until someone tries to reproduce the results.

My overall point is that we can have this kind of discussion, and people with statistical knowledge can challenge the conclusion the authors presented. And other research groups can even try to reproduce the results, and if they find no evidence of the correlation this very same journal would likely publish their findings.

Not at all the sort of thing you get with astrology.

Really, one could make the same point comparing astrology "evidence" to any reasonably well done scientific study, but this one brought the contrast to mind because correlation with personality traits is one of the things astrology believers claim. They seem to have no idea what actual evidence of that correlation would even look like. They think playing the Forer Effect game is sufficient.