October 21, 2005

Miers dementia unlikely

A reader suggested that I check out the flap at the Volokh Conspiracy over Harriet Miers' commas. Jim Lindgren, the responsible conspirator, flagged the corpus delicti with boldface in this quote from p. 50 of Miers' response to the Senate questionnaire:

My experience on the City Council helps me understand the interplay between serving on a policy making board and serving as a judge. An example, of this distinction can be seen in a vote of the council to ban flag burning. The Council was free to state its policy position, we were against flag burning. The Supreme Court’s role was to determine whether our Constitution allows such a ban. The City Council was anxious to encourage minority and women-owned businesses, but our processes had to conform to equal protection requirements, as well.

But I'd be a hypocrite to join in abusing Miers for misdemeanors of proofreading. Geoff Pullum has put so much time into fixing the errors in my Language Log posts that UCSC gave him a sabbatical as compensation*, and just yesterday, Chris Waigl caught me misspelling my own senator's name as "Spector".

On the other hand, Americans look for a higher level of qualification in SCOTUS nominees than "it's not so bad, really" and "everybody makes mistakes", and in the case of Harriet Miers, they're not getting much help. [The most coherent case for Miers comes from Michael Bérubé, of all people!] So I'm going to buck the storm surge that's been sloshing from right to left across the political spectrum, and say something positive about her. Not only that, but my encomium will make testable predictions based on on well-established science. Ready for it? OK, here goes: her prose style indicates that Harriet Miers is unlikely to get Alzheimer's. Surely this is an important factor in the case of a lifetime judicial appointment.

I base this prediction on the results of the famous "Nun Study". Briefly,

Two measures of linguistic ability in early life, idea density and grammatical complexity, were derived from autobiographies written at a mean age of 22 years. Approximately 58 years later, the women who wrote these autobiographies participated in an assessment of cognitive function, and those who subsequently died were evaluated neuropathologically. [...] Low idea density and low grammatical complexity in autobiographies written in early life were associated with low cognitive test scores in late life. Low idea density in early life had stronger and more consistent associations with poor cognitive function than did low grammatical complexity. Among the 14 sisters who died, neuropathologically confirmed Alzheimer's disease was present in all of those with low idea density in early life and in none of those with high idea density. CONCLUSIONS--Low linguistic ability in early life was a strong predictor of poor cognitive function and Alzheimer's disease in late life.
[D. A. Snowdon et al., Linguistic ability in early life and cognitive function and Alzheimer's disease in late life. JAMA 275 (7), 1996.]

Unfortunately, we don't have a sample of Miers' writing at the age of 22. But we can extrapolate from measurements taken now, because according to Kemper et al., "Language decline across the life span: findings from the Nun Study", Psychol Aging, 16(2):227-39 (2001),

Idea density averaged 5.35 propositions per 10 words initially for participants who did not meet criteria for dementia and declined an average of .03 units per year, whereas idea density averaged 4.34 propositions per 10 words initially for participants who met criteria for dementia and declined .02 units per year.

I measured "idea density" in a sample of sentences from p. 50 of Harriet Miers' response to the Senate Questionnaire, and the result was 4.77 "ideas" per 10 words. My (possibly faulty) understanding of the estimation method came from reading the cited sources, which are Kintsch, W. (1972) "Notes on the structure of semantic memory", in E. Tulving and W. Donaldson (eds) Organization of Memory, pp. 247–308. New York: Academic Press; and Kintsch, W. and Keenan, J. (1973) "Reading rate and retention as a function of the number of propositions in the base structure of sentences", Cognitive Psychology 5, 257–74. I also looked carefully at the examples given in papers by Snowdon and others.

The participants in the "Nun Study" were originally tested at a mean age of 22; Harriet Miers is 60; so if she were in the non-demented group, her score at age 22 should have been 4.77 + .03*38 = 5.91. Even adjusted by the .02 units per year found in the demented group, her age-22 score should have been 4.77 + .02*38 = 5.53. In either case, this is well above the (unadjusted) mean of the non-demented group.

But wait, there's more. Miers' "idea density" score also predicts that she'll be good at solving "unstructured problems", according to R. Davidson et al., "Using linguistic performance to measure problem-solving", Accounting Education, 9(1) 53-66 (2000) They define "a structured problem as one for which solution procedures are known, objectives are clearly defined, and there is usually an identifiable single correct answer", while "an unstructured problem ... is likely to require heuristic solution methods as well as intuition and experience, since there may be multiple objectives that are not clear-cut and the ‘goodness’ of the results is difficult to evaluate". The results?

We found that idea density does appear to be related to the ability to solve unstructured problems, but not to the ability to solve structured problems, even after controlling for the effects of other variables. We found that grammatical complexity was not significantly related to either problem-solving ability.

Some of this study's results were surprising, at least to me:

... we found a significant positive correlation ... between idea density and unstructured problem-solving, but no significant correlation with structured problem-solving ... We found no significant correlation between grammatical complexity and structured or unstructured problem-solving ... In addition, there was a significant negative correlation ... between measures of grammatical complexity and idea density ... The two measures of problem-solving performance were significantly correlated...

The mean "idea density" of the (prose samples written by the) 90 college-age participants in the Anderson study was 4.17 , with a s.d. of 0.39, so Miers' age-60 score of 4.77 was more than 1.5 standard deviations above the mean. Her extrapolated age-22 score of 5.91 would be almost 4.5 standard deviations above the mean.

I need to say at this point that you should not trust me on this at all. My sample of Miers' writing was small -- less than 200 words. I have a lot of questions about how to apply the "idea density" method, and I have no confidence that my practice was the same as that of Snowdon et al. or Anderson et al. When I was unsure, I gave Miers' credit for an extra "idea", a term that I'm putting in scare quotes because I feel that it's inexcusably tendentious: as far as I can tell, the factor most strongly influencing "idea density" is the frequency of modification and phrasal conjunction. The correlations in the Anderson study were not very high -- r = 0.3 for the relationship between "idea density" and "unstructured problem solving". And it's clear that different styles and genres of text from the same person will have systematically different "idea density" measures.

But still, if you're looking for the bright side of Miers' nomination, I'm here to help.


Posted by Mark Liberman at October 21, 2005 04:47 PM