Gordon – Everyday Life as an Intelligence Test

Here’s the link (pdf). Some excerpts:

“this article examines the issue of general intelligence in everyday life, where many tasks tend to be performed widely if not universally. The aim is to elucidate both the practical importance of intelligence within that sphere and some major obstacles to the full recognition of that importance. The effects of such obstacles often exist side by side with a keen, if sometimes reticent, awareness by many individuals of the role of intelligence in its more standard applications, such as schooling and certain occupations. [...]

It is often not apparent to persons that cognitive tasks are embedded in many of life’s everyday activities, as those activities (say, parenting) often exhibit other, perhaps more salient, facets of content (warmth) that seem to outweigh any cognitive component (judgment) until the latter comes forcibly to attention (as when a warm parent leaves small children unattended with matches in the house). Empirically, however, such noncognitive facets may individually contribute much less than g does to variance in overall or long-term task performance (child safety) and, especially, to the task-as-item covariance (i.e., what is common to all such tasks) upon which any test depends for its reliability. [...]

Underrecognition of the cognitive component is abetted by numerous other features of the phenomenology and organization of intelligence in everyday life. Not least among these is the fact that there certainly are everyday activities in which a cognitive component is thought crucial, but such activities are set somewhat apart by our culture in special categories. The result is that the residual activities may unthinkingly be consigned by laypersons to the noncognitive realm by virtue of the contrast. Education and learning have, of course, been the chief repositories of tasks viewed as essentially cognitive, and it is only in recent times that this view has been broadened among specialists to include performance within jobs and participation in crime.
Although Singaporean mothers were able, when specifically asked, to perceive the role of intelligence in a variety of children’s behaviors, such as “Shows common-sense” and “Sizes up a situation badly,” it was “Learns quickly,” a typically academic marker, that had the highest g loading in a factor analysis of 55 ranked items (Nevo & Khader, 1995). Between first, third, and sixth grades, academic skills increasingly come to dominate children’s attention as examples of what it means to be smart or intelligent (Yussen & Kane, 1985, Tables 2-3). It would appear that social perception of the role of intelligence is drawn toward outcomes with the highest g loadings, which is not surprising, but it may sometimes be tacitly misconcluded as a result that other outcomes have no g loadings at all when their loadings are simply not as high. Just as individuals may often be assigned too hastily to only two categories on the g continuum, say, qualified and unqualified, so may the g-loadedness of outcomes be falsely dichotomized.
A crucial final point is that, as in aggregate data, repetition of a single task or response by multiple persons can produce regularities in percentages and averages that are as reliable in life (and as indicative of the operation of g) as the results of multiple tasks presented to a single person on tests. If two populations differ in the average g that they bring to a repeated single task in everyday life, reliable group differences in average performance will emerge, just as group differences emerge in rates of passing an individual test item. Support for the role of intelligence from aggregate data, when uncovered, thus makes it possible to work backward to the inference that intelligence was very likely an influential component of the individual behavior so aggregated if that was not already an accepted view [...]

What I am arguing against here, and hope to overcome with data, is a double standard in agnosticism among many test defenders concerning the potential g-loadedness of items, depending on whether the items appear on tests or in everyday life. [...]

Although again deceptively commonplace to test experts, Jensen’s (1986b,p. 109) third provision, that in order to measure individual differences in a group of people, “item difficulty (i.e., percent ‘failing’ the item) must be greater than 0 and less than loo%,” is of profound significance for understanding why the role of g in life tasks tends to be underestimated. Many everyday behaviors, such as operating a car, prove so easy for most persons that they seem not to depend on what the layperson thinks of as intelligence at all, and performing them produces no subjective sense of the effort known as “thinking.” Recall the estimate quoted above that some errors occur as seldom as one in 10,000 opportunities. Many such tasks, of course, were overlearned in childhood, when effort would have been more apparent. Adults who commit inexplicable errors on such tasks are greeted with special epithets, suggestive of no intelligence at all. [...]

Research on elementary cognitive tasks (ECTs), although conducted in the laboratory rather than on everyday tasks, provides especially informative examples of performances misperceived as making no demand on intelligence. ECTs are often so easy (pressing the button beside the light that goes on) that virtually no one gets them wrong, and participants cannot tell the difference between their own better and poorer performances (Jensen, 1980a, p. 691). Sensitive monitoring of reaction times (defined as the interval, in milliseconds, between the light signal and release of one’s finger from a home button) reveals, however, that speed of such performances does vary and is reliably correlated with g (Jensen,1993b). Jensen (1980b, p. 109) remarked that the cognitive demands of one particular ECT “are so extremely simple that it seems almost implausible that the procedure could yield any measurements that would be correlated with IQ.” The indefinite linearity of performance with IQ upwards (e.g., Hawk, 1970) appears to apply in the downward direction as well when appropriately measured, to include performance on tasks even as easy as these. [...] In test parlance, mundane life lacks sufficient “top” or “ceiling,” that is, lacks items at a sufficiently high level of difficulty to reveal clearly the advantages of high intelligence over average intelligence [...]

Almost all research on intelligence has been focused upon the individual level of analysis. For studied outcomes, research usually takes the form of correlating a measure of g with the outcome. For several reasons, some made understandable by the previous discussion of the test analogy, the theoretical value of such correlations is often underestimated. First, behaviors are rarely observed at the lowest level of performance, which would make their dependence on intelligence more apparent, and the correlations more convincing, because society is usually structured to prevent such poor performances from occurring. Second, performance failures, when witnessed, are often attributed to superficial causes, for example, not planning ahead, that are formulated in a manner that conceals the role of intelligence behind noncognitive, often motivational, terminology. Third, modest correlations that do get reported between IQ and outcomes are often dismissed as too inconsequential to motivate theory”

…

From the first 13 pages (of 118). I haven’t read much more than that yet, maybe I’ll post more on this later. Here’s a related paper written around the same time.