I do not know if this is strong enough evidence for you, but I offer the following text dump for consideration; from "The Predictive Value of IQ":

...Nuñes (formerly Carraher) has done a series of studies over the years investigating the mathematical skills of Brazilian street children (Carraher, Carraher, & Schliemann, 1985; Carraher, Carraher, & Schliemann, 1987; Nuñes, 1994; Nuñes, Schliemann, & Carraher, 1993). The example of Brazilian street children is an apt one for the illustration of the construct of practical intelligence, because as Nuñes points out, the survival of these children is threatened on a daily basis. If the children are unable successfully to run a street business, and lapse into crime, the chances of their being murdered are quite high. Nuñes has found that the same children who can do the mathematics to run a successful street business are often failing math in school or otherwise show only minimal competence in math in academic settings. Similar results have been obtained by Ceci and Roazzi (1994), suggesting the findings are generalizable across investigators. As pointed out by Anderson, Reder, and Simon (1996), one needs to be careful about the exact conclusions one draws from studies such as these. For example, the exact computations required in one situation may not be the same as the computations required in another. But in terms of [End Page 25] adaptive functioning, the point is that the people who are best able to adapt in one circumstance often are not those best able to adapt in another.

In a related study conducted near Kisumu, Kenya, we found that children's knowledge of the use of natural herbal medicines to combat illness is significantly negatively correlated with scores on tests of crystallized (Mill Hill Vocabulary in English and a comparable test in Dholuo, the home language) abilities (Sternberg, Nokes, et al., in press). In other words, practical intellectual skills were actually inversely associated with academic intellectual skills.

Lave (1988) also did related studies among Berkeley, California, housewives. She found that the same housewives who had no trouble doing comparative price calculations in the supermarket (before the introduction of unit pricing) were unable to complete most of the problems on a standard paper-and-pencil test of mathematical knowledge given in a classroom.

Investigating a different population, Ceci and Liker (1986) found that men's handicapping abilities for predicting outcomes of horse races were unrelated to their IQs. Moreover, successful handicappers had an average IQ of only about 100, despite the complexity of the handicapping task.

Ceci and Bronfenbrenner (1985) looked at quite a different task. They gave children a time-estimation task either in a classroom or at home. Strategies and quality of performance were very different in the two settings, suggesting that the context in which the judgments were made had a major impact both on how they were made and how well they were made.

In a very different context, Fiedler and Link (1994) reported that IQ positively predicted leadership performance under conditions of low stress but negatively predicted this same performance under conditions of high stress; in contrast, acquired knowledge of the kind that is essential for practical intelligence positively predicted leadership performance under conditions of high stress but negatively predicted under conditions of low stress.

Also somewhat relevant http://xa.yimg.com/kq/groups/190222/1077302733/name/Kwon is a brain imaging study of Korean Go players; the expert professional Go players averaged IQ of 93, the control group 101.