Saturday, 11 October 2014

Personality consommé: boiling off the non-believers


Dear Dr Harpending, Thank you for giving me a chance to comment on your first draft of “Plain and Simple: On a Novel Feature of Amish Personality”.

It is a pleasure to read. The second and third paragraphs are highly instructive, and would on their own probably make a good slim contemporary textbook.

From a psychological point of view, mating so as to match personalities and attitudes makes sense. For a marriage to last, similarities of opinion regarding marriage itself, the upbringing of children, duty to elders, the nature of society, and whether it is permissible to eat chocolates in bed are essential.

Michael Weight and Henry Harpending argue that since it is hard to enter the Amish sect, but easy to leave at 16-25 years when voluntary Baptism leads to a commitment to marrying within the faith, then those unsuited to the Amish lifestyle leave for the fleshpots of modernity, leaving the True Amish to become even more Amish with each passing generation. (In fact, about 10% of each generation leave, but only as a thoughtful move to neighbouring Mennonite groups). They measure this Amish quotient (degree of Amishness), and then set to work with the tools of population genetics to test an explanatory model. Unlike prose models, their model is testable.

In 1968 Paul Wittmer administered the Cattell 16 PF to 25 18-20 year old Amish young men and 25 young men of the same age who were not Amish, all from Davies County, Indiana. There are 180 questions on the form, from which are extracted 16 numbers supposed to represent the personality of the subject on 16 dimensions of personality.

Now, 25 is a small number, but a bigger number than most accounts of Amish personality, which in young children on the Myers-Briggs personality test is said to be “quiet, friendly, responsible and conscientious” (Amish Children: Education in the Family, School, and Community (Case Studies in Cultural Anthropology) Paperback – 1 Feb 1992 John A. Hostetler and Gertrude Enders Huntington ; cited in “Amish Society” John A. Hostetler, 1993). I have not been able to access this chapter, but it could be relevant.

Weight and Harpending do a Principal Components Analysis (which is factor analysis without an attitude problem, in that it is mathematically simpler and plainer and makes fewer assumptions) on the results for the young men. They then plot the results in factor space, and show that the Amish (letter A in red) are distinct from the Indiana boys (letter I in blue) the latter being closer to the UK mean for personality (large letter U).


Plotting out the cases in personality factor space shows a pretty clear discrimination between the Amish and Indiana men, (over 2 standard deviations) of the sort that would be apparent to a moderately attentive sheep dog. One bark and a sharp nip should bring the two errant AAs back into the fold.

The next generation of children will have an average AQ of 0:10 standard deviations greater than their parents did before emigration. The process of selective emigration repeats so that the mean Amish AQ increases by one tenth of a standard deviation per generation. With 25 years per generation, “Amishness" will increase by a full standard deviation in 10 generations or 250 years. This is substantial social evolution on a time scale of a few centuries.

Although the authors do not use the term, it would appear that the Amish are being bred up for old-fashioned civility.

They conclude: Our fit does not “confirm" the quantitative genetic model. Some other mechanism may fit the data as well or better. In particular purely cultural transmission might generate this pattern, with no genetic consequence at all. At this point the value of our approach becomes apparent since there is no such cultural transmission model, only prose about plausibilities and possibilities. In other words our model clarifies the mechanisms and magnitudes required of a purely cultural model. The difference between the two groups of subjects corresponds to the consequences of several centuries of mild selection by choice of membership in an otherwise closed group.

Four observations:

1 The sample size is small, and that is a limitation on the generalizability of the results. At the same time, if you have an interesting sample, from a Bayesian point of view it is better to publish than ignore the findings. Others then have some priors to improve upon in later work.

2 The Myers Briggs study on Amish children could be compared with population norms.

3 It would be good to look at any IQ results, since they would probably be likely to differ from the outside population.

4 Amish genetics are somewhat different, the 200 or so originators leading to a founder effect.

In summary, this is a lovely little study. It takes a small sample and lays out a plain and simple explanation, which leads to a plain and simple test. The 16PF, if given to a sample of contemporary Amish, should be one tenth of a standard deviation more Amish than that observed in 1968, 46 years ago. Indeed, it should be almost two tenths of a standard deviation more Amish. Who will be the first to test it?


Disclosure of interests

For 3 months of every year in childhood I lived an Amish lifestyle without knowing it, in an isolated cottage, ploughing land, filling kerosene lamps as illustrated in the Hostetler book, without electricity, television or radio, and often without running water. We rode horses, used a horse and cart, but in the spirit of absolute disclosure I must admit that there was also a Model T Ford and eventually a Jeep. All this may make me naturally disposed to see the Amish in a very favourable light, or it may incline me to the delights of urban living. Either way, you can be assured that I have an Agenda.


  1. PrincipAL component analysis (sorry for the pedantry, one of my bugbears...)

  2. Damn it! One of my bugbears as well, and I have fallen into the same trap.