My colleague Garett Jones is working on a book called Hive Mind: Why Your Nation’s IQ Matters So Much More Than Your Own.  I thought about his project while reading Robert Gordon’s article “Everyday Life as an Intelligence Test.”  One of the many fascinating questions in this piece: What happens when you test the IQ of a team?

Consider an experiment by Laughlin and Johnson (1966) that induced the local equivalent of random pairing. College students were administered the high level Tern-ran Concept Mastery Test (Tern-ran, 1956) and then assigned to low, medium, and high ability groups according to their scores. Students were paired up systematically to represent every combination of the three ability levels and readministered the same test working together, but some members of each level were left to repeat the test alone. 

Working with lower ability partners led generally to score improvements, but the more able the partner the greater was the gain. For present purposes, the most telling finding was that medium and high ability individuals improved slightly more working alone the second time than working with low ability partners. As these were college students, and the test was a difficult one, the low group would have scored well above many persons in the general population. [emphasis mine]

Here’s more on team dynamics, based on Schofield (1982):

Pupils preferred to exchange help with partners of similar achievement level, with whom they could reciprocate and who could reciprocate with them (p. 87). Poor students had little to offer as the brighter ones already knew the answers to easy problems, and so the poorer ones would often be ignored when seeking help from better students (p. 88). This left poor students to turn for help to one another (where they might well imbibe misinformation uncritically).  When help was offered to poor students by altruistic better students, the former often rejected it because they were embarrassed not to be able ever to reciprocate (pp. 89-90).

Gordon concludes:

Exchanges of direct, explicit help, not to mention true collaboration, are most likely to thrive between individuals unequal in g when they are separated by only small steps of IQ, so that gaps are not too big to permit adequate reciprocity. “I didn’t believe in working with anyone who wasn’t my equal,” a senior lawyer states… The constraint would impose a positive correlation on individuals between their own intelligence and the intelligence of those from whom they are most likely to receive help on a regular basis, even if some correlation were not already present for social structural reasons. The constraint may, for example, help explain assortative mating for IQ, the average correlation of about .36-.43 between spouses…

Question for discussion: The research Gordon discusses seems to focus on cognitive barter.  What do comparative advantage and market exchange imply about the broader significance of these results?