There is this claim sometimes seen when discussing highly gifted people. One interested reader emailed me to ask where it was that Arthur Jensen made this claim:
Anyway, what I am after is a reference to where Art first presented his notion of “zone of tolerance”. The only text I have been able to find is the document I emailed about. I would be grateful if you can help me here.
I did some digging. And it is true, that one can find various sources talking about Jensen mentioning this idea. For instance, in the book Misdiagnosis and Dual Diagnoses of Gifted Children and Adults:
Gifted and talented persons, in general, are in the upper 3% to 5% of the population. If highly gifted, an individual might be in the upper 1% or .5% of the population. It now becomes more understandable that finding a person of approximately the same intellectual ability could pose some difficulties. Probably 80% of the general population would not provide sufficient intellectual stimulation or would not have the kinds of interests that would allow one to want to spend extended time with them or to have a long-term intimate relationship with them. Some people have stated it even more harshly. Well-known researcher Arthur Jensen (2004) said that, for each person, there is a “zone of tolerance” of plus or minus 20 IQ points.
Unfortunately, checking the 2004 citation (above) just leads to a “personal communication”. In other words, the author or someone else emailed Jensen to ask, and he replied with that. Not really evidence. Looking at Google books, we can find a few more that reference Arthur Jensen for this claim, none of which provide a proper citation. Back in 2017, the blogger Neuroskeptic offered a debunking. He quotes some people on Quora spreading this claim. True enough, I looked on Quora and you can find claims like:
Since I am the one who has popularized the concept of communication range on Quora, perhaps I am the best to answer.
The communication range of +/- 2 standard deviations or 30 points (60 points total) has originally been suggested by researchist Leta Hollingworth and popularized by Grady M. Towers, Michael J. Ferguson and Darryl Miyaguchi. It denotes the range of difference on intelligence between two individuals within which meaningful human interaction is possible. Note the word “meaningful”. Not all human interaction is “meaningful”.
This post is by a Susanna Viljanen and has been read a staggering 250M times, at least according to the site.
What evidence is offered? Not much, but some links are offered. These are links to one of the ‘super-Mensa’ writers, Grady M. Towers, who wrote an influential article in 1987 (The Outsiders), which included:
Observation shows that there is a direct ratio between the intelligence of the leader and that of the led. To be a leader of his contemporaries a child must be more intelligent but not too much more intelligent than those to be led… But generally speaking, a leadership pattern will not form–or it will break up–when a discrepancy of more than about 30 points of IQ comes to exist between leader and led [3, p. 287].
The reference is to Leta Hollingworth‘s 1942 book Children above 180 IQ Stanford-Binet; origin and development (she died in 1939 so the book was based on her notes). The book was described as:
12 very exceptional children were known and studied, some for as long as 23 years. The author at the time of her death in 1939 had been bringing a survey of them up to date. Part I, on general orientation, contains chapters on the concept of intellectual genius, on the early scientific study of eminent adults, and on published reports of tested children. Part II describes the 12 cases new to the literature, 7 reports having been formulated by H. L. Hollingworth from data left in the files of the author. Information besides that of repeated testings includes, whenever possible, facts concerning heredity, educational and social information, character traits, physical measurements, creative activities, test records of brothers and sisters, etc. Talking and reading most clearly differentiated the records of these children from the norms. Difficult educational problems were presented from school entrance. 1/3 showed notable signs of creativeness. As these children matured, they maintained their high initial intellectual status. Part III, giving selections from or complete reproductions of published papers, presents the author’s views concerning general principles and about the social and educational implications resulting from her study of children of high intelligence.
Her book includes these parts:
Where the gifted child drifts in the school unrecognized, working chronically far below his capacity (even though young for his grade), he receives daily practice in habits of idleness and daydreaming. His abilities never receive the stimulus of genuine challenge, and the situation tends to form in him the expectation of an effortless existence. Children with IQ’s up to 150 get along in the ordinary course of school life quite well, achieving excellent marks without serious effort. But children above this mental status become almost intolerably bored with school work if kept in lockstep with unselected pupils of their own age. Children who rise above 170 IQ are liable to regard school with indifference or with positive dislike, for they find nothing in the work to absorb their interest. This condition of affairs, coupled with the supervision of unseeing and unsympathetic teachers, has sometimes led even to truancy on the part of gifted children.
…
This boy had qualities of political leadership. This limiting adjective opens the large subject of the different kinds of leaders. Leaders of whom, and for what ends? Observation of children suggests that there is a direct ratio between the intelligence of the leader and that of the led. To be a leader of his contemporaries, a child must be more intelligent, but not too much more intelligent, than those who are to be led. There are rare exceptions to this principle, as in the case we have cited. But, generally speaking, a leadership pattern will not form—or it will break up—when a discrepancy of more than about 30 points of IQ comes to exist between the leader and the led.
So we see that her comment were made in the discussion of one particular child. I don’t know if someone wrote about this topic before Hollingworth, but I doubt it, because she truly was a pioneer in the study of gifted children. Based on this, we can probably say that Hollingworth was the originator of the idea. The other writings by these Mensa types don’t really have anything else to offer in terms of actual evidence, and we can’t be satisfied with some observations published 80 years ago based on observations of 12 children.
What about the empirical claim? It is patently untrue that people who are distant in intelligence cannot talk meaningfully together. This kind of communication occurs all the time. For instance, when leading politicians talk to voters, psychiatrists talk to their patients, judges talk to juries, police investigators talk to witnesses and suspects. But what Towers and his types are getting at — and was Hollingworth was getting at — is that people find it hard to bond with people who are very distant in intelligence. As such, their claim when interpreted in a charitable manner really is just a restatement of the general finding of social homophily for intelligence. People who are good friends tend to be similar in intelligence, as well as age, political ideology, interests and so on. People tend to marry others similar in intelligence (and other traits, which we call assortative mating), so much so, in fact, that their spouses are about as similar to them in terms of intelligence as are their close relatives (correlation about .40 to .50).
But what about the related part of the claim that people who are too high intelligence don’t make good leaders because they are too much smarter than their subordinates? Well, Dean K. Simonton published a theoretical paper about this in 1985, and finally in 2017, he and colleagues managed to get a dataset to test the predictions:
Although researchers predominately test for linear relationships between variables, at times there may be theoretical and even empirical reasons for expecting nonlinear functions. We examined if the relation between intelligence (IQ) and perceived leadership might be more accurately described by a curvilinear single-peaked function. Following Simonton’s (1985) theory, we tested a specific model, indicating that the optimal IQ for perceived leadership will appear at about 1.2 standard deviations above the mean IQ of the group membership. The sample consisted of midlevel leaders from multinational private-sector companies. We used the leaders’ scores on the Wonderlic Personnel Test (WPT)—a measure of IQ—to predict how they would be perceived on prototypically effective leadership (i.e., transformational and instrumental leadership). Accounting for the effects of leader personality, gender, age, as well as company, country, and time fixed effects, analyses indicated that perceptions of leadership followed a curvilinear inverted-U function of intelligence. The peak of this function was at an IQ score of about 120, which did not depart significantly from the value predicted by the theory. As the first direct empirical test of a precise curvilinear model of the intelligence-leadership relation, the results have important implications for future research on how leaders are perceived in the workplace.
They assembled a hard to get dataset:
We studied 379 leaders (26.39% women; mean age of the leaders 38.34 years, SD 6.39) on whom we obtained ratings on leadership as well as several individual differences predictors. The sample of midlevel leaders were drawn from nine different groups composed of seven multinational private- sector companies (n 351) and two cohorts of working leaders (n 28) attending an executive education course. The leaders were distributed across 30 countries, mostly from Switzerland (n 139), the Netherlands (n 37), United Kingdom (n 27), France (n 23), Germany (n 23), Sweden (n 24), Greece (n 14), Ireland (n 12), and the United States of America (n 12). These data overlap with data published by Antonakis and House (2014, see Study 4), who examined a different phenomenon.
We collected the data on the leaders over a course of 6 years. To avoid selection effects and hence biased ratings, we requested the human resources office of the companies in which the leaders were employed to provide us with the contact details of about 12 raters per leader from mostly their subordinates (i.e., n 6 – 8), but also from some peers (i.e., n 3– 4), and their supervisor; we asked that the raters must be representative (in this way participants leaders could not select those individuals from whom they would expect to receive good ratings).
We obtained ratings of leadership from 2,905 raters (i.e., 7.66 raters per leader; note, because of a very small degree of missing data on leadership ratings, the total raters on the leadership scales ranged from 2,896 to 2,905). To reduce the likelihood of rating leniency (Antonioni, 1994) the raters participated anonymously and no rater identifiers were recorded. Participating companies (and percentage of participant leaders) included firms from banking (6.33%), insurance (38.79%), food manufacturing (26.65%), telecommunications and high technology (13.46%), hospitality and retail (7.39%), and other (7.39%).
Visually:
The statistics, however, are not so convincing:
As for intelligence, the main effect on the 10 active constructive leadership styles (i.e., transformational, contingent reward, and instrumental leadership) was positive. As concerns the significance of the quadratic term on the effective leader styles (i.e., transformational, contingent reward, and instrumental leadership), it was negatively predictive and significant eight of 10 times (seven at p .05 and one at p .10), thus demonstrating incremental validity. In terms of incremental validity across all the leadership styles, the main effect of intelligence only added, on average, .0054 to the prediction beyond the rest of the individual difference factors. The quadratic effect added much more, beyond the main effect of intelligence to the R2 , that is, on average, .0343 (for the full sample the increase was, on average, .0543). We did not find significant effects of IQ on transactional or laissez-faire leadership. With respect to the curvilinearity arguments we made, these results mostly support H1 (regarding the prototypically good leadership styles) but not H2 (the prototypically bad leadership styles).
The article has some pretty strange analytic decisions. Since they had missing data, they decided to use list-wise deletion (complete cases only, n = 171). This is the worst method, as it loses the most data and does so in a biased way (since cases with incomplete data are usually different from cases with full data). But they also showed their regression results for the full dataset (n = 379), which showed smaller p values for the quadratic terms of IQ:
As you can see, it’s a bit of a mess. There are 13 leadership scales, and of these, 6 show some evidence of quadratic effects (including with p < .01), but the rest do not. So what does this mean? Do we really have to develop theories about 13 different aspects of leadership and why intelligence may have nonlinear effects on some but not the others?
For those still inclined to take this as some kind of confirmation about the sometimes importance of nonlinear and especially non-monotonic (curves that change direction, e.g. U shaped) effects of intelligence, consider that the typical finding in this field goes like this (2021 meta-analysis by Brown and colleagues):
Despite a long-standing expert consensus about the importance of cognitive ability for life outcomes, contrary views continue to proliferate in scholarly and popular literature. This divergence of beliefs presents an obstacle for evidence-based policymaking and decision-making in a variety of settings. One commonly held idea is that greater cognitive ability does not matter or is actually harmful beyond a certain point (sometimes stated as > 100 or 120 IQ points). We empirically tested these notions using data from four longitudinal, representative cohort studies comprising 48,558 participants in the United States and United Kingdom from 1957 to the present. We found that ability measured in youth has a positive association with most occupational, educational, health, and social outcomes later in life. Most effects were characterized by a moderate to strong linear trend or a practically null effect (mean R 2 range = .002–.256). Nearly all nonlinear effects were practically insignificant in magnitude (mean incremental R2 = .001) or were not replicated across cohorts or survey waves. We found no support for any downside to higher ability and no evidence for a threshold beyond which greater scores cease to be beneficial. Thus, greater cognitive ability is generally advantageous—and virtually never detrimental.
And no, there wasn’t a recent study showing this wasn’t the case for income either.
OK, so maybe this line of thinking doesn’t offer so much positive for this communication range idea. But then again, I imagine that the 150 IQ people now a days tend to find themselves in situations where they are around other people of similar intelligence, so they don’t suffer much from the “communication range” difficulties. Ironically, this did not apply to Grady Towers himself, as he worked as a security guard and seemingly died alone at age 55 (murdered on the job). This outcome is seemingly not so uncommon for people who get obsessed with their own hyper-intelligence.
Ending on a positive note, Towers was instrumental in showing that while mental health and intelligence go hand in hand in the right direction, there is seemingly a slight non-monotonic effect here, such that the very bright people feel more loneliness. This was found in the Terman sample, and also later in the SMPY (by my recollection of this talk, but the pattern wasn’t so clear). I imagine that this concern will be less of a worry in modern samples because it is so much easier with the internet and large connected populations to find someone as smart as yourself. If you are on the internet and cannot find someone smart enough to talk to, you are either a confirmed genius, or more likely, should to talk to a psychiatrist.
Conclusions
- There is an idea of a communication range or zone of tolerance, variously attributed to Arthur Jensen, Leta Hollingworth and probably others, but which really was promoted by Grady M. Towers, a Mensa-type with mental problems. The idea is that you cannot talk effectively or connect properly with other humans outside some IQ range (20-30).
- Evidence for this claim is scant and pretty weak. Best case is a study of business leaders showing a non-monotonic effect of intelligence on some aspects of perceived (subordinate-rated) leadership ability.
- But it is true that people cluster by intelligence. We knew that already. This is just a special case of social homophily, or assortative mating in terms of dating. Birds of a feather flock together. I’m sure a typical gifted person will have trouble relating to average intelligence people, but fortunately, it is pretty easy to find other smart people these days.
- In the Terman sample (very old!), the very bright people had somewhat more loneliness and social adjustment issues the smarter they were, probably related to their inability to find friends back then.
- If you think you are very, very smart, and can’t relate to others, and think everybody is too stupid to talk to, the problem doesn’t have much to do with intelligence, but with your other issues.
Extra material added after the initial post
Several readers wrote me to point out additional mentions of this idea.
Linda Gottfredson, wrote in a 2000 encyclopedia article about intelligence’s social aspects (thanks to Erwin Schmidt):
The basic issue is this: A difference in IQ of one standard deviation (about 15 points) is socially perceptible and meaningful. Interpersonal communication becomes fraught with increasing difficulty beyond this distance because of larger gaps in vocabulary, knowledge, and ability to draw inferences or “catch on,” as well as the emotional discomfort such gaps create. Figure 1 reveals how IQ ranges of about one standard deviation also mark off substantial differences in options for education, training, and career, and thus the likelihood of entering different social niches. As shown in the figure, the normal range of intelligence (IQ 70-130, which includes roughly 95 percent of the general white population) spans four standard deviations of IQ. Socially and cognitively, that is an enormous difference. How, then, do people com- municate and congregate across the IQ continuum in their daily lives? The average difference between siblings and spouses is about 12 IQ points, which means that most people in a biological family fall within the range of ready cognitive-communicability. Any two random people in the population, however, differ by 17 IQ points, which represents the borderline for communicating effectively and as social equals.
Communication, cooperation, and reciprocity. The ability to communicate as equals constitutes a social tie, as does the ability to trade information and assistance. Such reciprocity is the basis of longer-term cooperation. Lack of reciprocity creates not only social distance but also animosity where reciprocity had been expected. There are many bases for cooperation and reciprocity, but sharing information and helping to solve problems is crucial in many settings. Ethnographic studies of middle school children, for instance, show how patterns of mutual assistance and friend- ship, rather than resentment and unwillingness either to provide help to classmates or to seek it from them, evolve from similarities and differences in students’ competence in answering home- work and test items. Similar g-driven interpersonal relations can be expected in many workgroups and other settings in which teammates depend on one another for technical competence. People of markedly different ability levels also tend to have different interests, which further impedes their ability to develop rapport. Assortative mating studies show that individuals explicitly seek mates of similar IQ levels and that spouses’ IQs are, in fact, moderately correlated (about .4), perhaps more so than any other personal characteristic (except gender). Cognitive incompatibility is certainly responsible for the extreme social isolation often experienced by both the mentally retarded and the highly gifted. Extremely gifted children, who may be four standard deviations or more above. average (IQ 160 and above), often — feel, and are treated as, alien. These children are as different from the borderline gifted (IQ 130) as the latter are from the average child (IQ 100). With extraordinary vocabularies for their age, the highly gifted speak virtually a different language from their agemates. Although less extreme, the same type of alienation develops across much smaller gaps in IQ. In short, cognitive similarity seems to affect the formation of social bonds, which them- selves are the building blocks of “social structure.”
Social separation and segregation. Because rough similarity in g promotes interpersonal reciprocity and rapport, it should not be surprising that people segregate themselves somewhat by cognitive ability when free to do so, marriage being the most intimate example. Segregation occurs along IQ lines for other reasons as well, many related to the functional value of intelligence in obtaining higher education and better work.
But nothing here in terms of data.
In the book review by Christopher Langan with Arthur Jensen (Discussions on Genius and Intelligence Mega Foundation Interview with Arthur Jensen, 2002), this question was raised:
Chris Langan: The founders of Mensa, regarded by many as the original high IQ club, complained that the group had forsaken its original purpose…that instead of pooling its intellectual talent to solve the most urgent problems of society, it had fallen into aimless socializing and dilettantism. Since then, a small number of more rarified groups, known collectively as the UltraHIQ Community, have advocated a return to the original vision. What is your opinion regarding the concept of a pool of intellectual talent based strictly on high levels of g and dedicated to finding solutions for some of society’s more urgent problems?
Arthur Jensen: It’s hard to imagine how a group of high-IQ people with little else in common besides their IQ and probably differing in many other ways perhaps even more than a random sample of the population can do much to effect social change or carry out a large project with a unified aim. On the other hand, a group of persons with a wide range of IQs from average to very high who have come together as a group because they all have a similar philosophy and some realistic goal based on it could be a force for some concerted kind of achievement. If there were a subgroup of UltraHIQ individuals all with a similar vision, aim, and dedication to achieve their common purpose, that would be something!
But I wouldn’t apologize in the least for any High-IQ society that was intended as a purely social organization that qualified people could join simply because the find each others’ company more congenial than that of most of the people they would be apt to meet in other social groups. I suspect that the “zone of tolerance” for the intelligence levels of one’s friends and spouses is probably, at the outside, about one’s own IQ +/- 20. People in the upper-half of the IQ distribution are more closely assortative in this respect than are those in the lower half. In the general population, spouse similarity in IQ is about the same as full-sibling similarity. Assortative mating for a given trait has the effect of increasing the genetic variance in that trait in the offspring generation. It is estimated that some 15 to 20 percent of the population variance in IQ is attributable to the effect of assortative mating.
Again, not data. Jensen made the same connection to assortative mating I did. To be fair, I read this book years ago, but wasn’t able to locate it when I did my research this time around.
