Academic journal article Studia Psychologica

Lack of Publication Bias in Intelligence and Working Memory Research: Reanalysis of Ackerman, Beier, & Boyle, 2005

Academic journal article Studia Psychologica

Lack of Publication Bias in Intelligence and Working Memory Research: Reanalysis of Ackerman, Beier, & Boyle, 2005

Article excerpt

Introduction

Measures of fluid intelligence moderately correlate with wide repertoire of intellectual abilities. This well-known phenomenon is called positive manifold (Spearman, 1904). However, the relationship between measures of working memory and fluid intelligence is mostly known in the realm of cognitive psychology. Estimates of common variance of working memory capacity and fluid intelligence measures range from 50% (Kane, Hambrick, & Conway, 2005) to 92% (Colom, Rebollo, Palacios, Juan-Espinosa, & Kyllonen, 2004). As a result, working memory is asserted by some researchers as a base of fluid intelligence (Jensen, 1998; Colom, FloresMendoza, & Rebollo, 2003; Engle, 2002). These strong correlations affect the imagination of the researchers, who may think that working memory and fluid intelligence are highly related or even identical. Working memory capacity also strongly correlates with the following intellectual abilities considered to be components of intelligence: comprehension (Daneman & Carpenter, 1980), reasoning ability (Kyllonen & Christal, 1990), and test results, which reflect intellectual capacity - the SAT (e.g., Turner & Engle, 1989). It should be noted that there are no other candidates, besides working memory, so closely related to fluid intelligence (Kyllonen, 2002). Due to these facts, researchers might not be interested in reporting moderate correlation coefficients - since this resembles the positive manifold - and publish mostly results depicting strong correlations between working memory capacity and fluid intelligence measures.

The hypothesis that fluid intelligence and working memory are identical has no theoretical justification. Even the ideal correlation (R = 1) between two measures does not mean that the same mechanism is responsible for the variability of both measures. A good example of such a strong atheoretical correlation, from a slightly different field of science, is the relationship between chocolate consumption and the number of Nobel laureates in different populations (Messerli, 2012). The Pearson correlation coefficient in this case is equal to 0.79. It is unreasonable to interpret this relationship as causal, although there are studies showing a positive effect of the consumption of flavonoids contained in chocolate on cognitive functioning (see Nurk, Refsum, Drevon, Tell, Nygaard, Engedal, & Smith, 2002; Desideri, Kwik-Uribe, & Grassi, 2012; Corti, Flammer, Hollenberg, & Lüscher, 2009; Sorond, Lipsitz, Hollenberg, & Fisher, 2008; Bisson, Nejdi, Rozan, Hidalgo, Lalonde, & Messaoudi, 2008). The correlation is strong, but we do not have a theoretical model, which explains the linkage number of Nobel laureates to chocolate consumption. It can be presumed that this strong correlation is an effect of another variable, e.g. a socio-economic status. In countries characterized by wealth (people with high socio-economic status) there is a greater chance to conduct scientific research as well as to consume chocolate. Thus, even very high values of correlation coefficients between working memory capacity and fluid intelligence measures are not proof that we are dealing with the same phenomenon.

Leaving aside the issues of the mechanisms that are responsible for the observed strong correlations of working memory and fluid intelligence, let us consider this: Is there a phenomenon that can systematically overstate the value of correlation coefficients? The author suspects that correlation coefficients reported in the research on fluid intelligence and working memory are inflated due to these facts: researchers are interested in reporting strong relationships, since low correlations are explained as positive manifold. A higher value of correlation coefficient suggests that the factor is more important in the context of intelligence. Open Science Collaboration (2015) reported that replications usually end with the effects strength lower by half on average, compared to the original studies. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.