Male hubris and female humility? A cross-cultural study of ratings of self, parental, and sibling multiple intelligence in America, Britain, and Japan.(Statistical Data Included)

1. Introduction

The topic of intelligence always generates controversy in psychology (Brand, 1996; Flynn, 1987; Halpern, 1997; Herrnstein & Murray, 1994; Howe, 1997, Lynn, 1982, 1994; Lynn & Mulhern, 1991; Lynn & Pagliari, 1994; Mackintosh, 1998; Sternberg, 1997). Popular debate over technical issues of IQ measurement led 50 of the world's experts on intelligence to write to the Wall Street Journal on 15th December 1994 (published Furnham, 1996). Debating the measurement of intelligence so frequently and in so much detail may well change public understanding over time. A number of studies published this decade have examined self-estimates of IQ, which may well have been informed by the public debate. This study focuses on cultural differences in self-estimated IQ, as well as sex and cultural differences in estimates of parents' and siblings' IQ.

There is a body of research literature on sex differences in estimated ability that demonstrates a consistent self-enhancing bias in men and a self-degrading bias in women (Beyer, 1990, 1998, 1999). A number of studies have specifically examined sex differences in the overall estimate (g) of one's own IQ (Beloff, 1992; Bennett, 1996, 1997, 2000; Byrd & Stacey, 1993; Furnham & Rawles, 1995; Hogan, 1978) (see Table 1). With few exceptions, the results confirmed early observations that there is a consistent significant sex difference, with males estimating their (general) intelligence higher than do females. Various studies also showed males and females rated their fathers as being more intelligent than their mother (Beloff, 1992; Byrd & Stacey 1993; Furnham & Rawles, 1995), and their grandfathers as more intelligent than their grandmothers (Furnham & Rawles, 1995). It is interesting to note that ratings of overall IQ (g) decline by about half a standard deviation for each generation back that was estimated. More recently, Furnham and Gasson (1998) found that parents estimated their male children as more intelligent than their female children (Table 1).

Other studies have looked at estimates of multiple intelligence rather than g (Bennett, 1996; 1997; 2000) based on the theory of Gardner (1983). It should be pointed out that none of these studies set out to test Gardner's (1983, 1998) original or updated theory for which there still remains little supportive evidence. Indeed, all the factor analytic work on ratings of the various intelligence types show that they do not reproduce the structure of the theory. Therefore, it could be argued that the theory has neither explicit nor implicit validity. Nevertheless, it provided a useful way to explore lay people's understanding of intelligence, which no doubt accounts for its popularity among nonscientists (Gardner, 1998).

Furnham, Clark, and Bailey (1999) asked male and female participants to rate each of the seven intelligence types set out by Gardner (1983). They found a sex difference only on the mathematical (logical) factor, but the seven intelligence types factored into three clear interpretable factors. Furnham, Fong, and Martin (1999) repeated this study on a bigger sample and found three significant differences: females rated themselves lower on mathematical, (logical) spatial and body-kinesthetic intelligence. More recently, Furnham (2000), in a study of parents' estimates of their own and their children's multiple intelligence types, found mothers gave lower estimates than fathers on their own mathematical (logical) and spatial intelligence. Both parents rated sons as having greater mathematical, spatial, and intrapersonal intelligence than daughters. Thus, the results seem to show consistent sex differences with respect primarily to mathematical and spatial IQ. Reviews of the intelligence literature suggest that these estimates may be accurate. Halpern (1997) wrote: "Males, on average, score higher on tasks that require transformations in visual-spatial working memory, motor skills involved in aiming, spatio-temporal responding, and fluid reasoning, especially in abstract mathematical and scientific domains" (p. 101). Indeed, it maybe that the more specific the ability the more accurate people are in the self-estimates (Table 2).

A few studies have examined the relationship between estimated and psychometrically measured IQ (De Nisi & Shaw, 1997; Furnham & Fong, 2000; Furnham & Rawles, 1999; Reilly & Mulhern, 1995), and each has reported positive and significant correlation in the range r = .13-.30, suggesting a weak but positive relationship. Some evidence suggests that outliers often repress these correlational coefficients, which may be as high as r = .40 (Furnham & Fong, 2000). Certainly, it does seem that overall IQ estimates may be a realistic measure of a person's psychometric IQ. Borkenau and Liebler (1993) found the correlation between self-rated and psychometrically measured intelligence to be r = .32 (N = 100), which dropped to r = .29 when sex and age were partialled out. Acquaintance ratings correlated very similarly to self-ratings with measured IQ (r = .29 or r = .31 when partialling took place).

More recently, Paulus, Lysy, and Yik (1998) noted correlations between single-item self-reports of intelligence and measured IQ scores in the range r = .20-.25 (among students). However, when data were aggregated, items weighted, and indirect questions addressed, the correlation rose to about r = .30. They thus argued that simple self-report measures are not particularly useful as proxies for IQ tests because they do not seem to reach the minimal standard of r = .55 and above. They conclude: "Both direct and indirect self-report measures of intelligence can reliably predict IQ scores. Because of the restricted range of abilities in competitive college samples, however, the validity limit appears to be 0.30" (p. 551).