The authors present a wide-ranging theory of the evolution of intelligence. However, Burkart et al. seem to have confused the general intelligence factor (psychometric g) with fluid intelligence (Gf). Psychometric g is a statistical way of describing the positive manifold: the phenomenon that ability tests, each with different content, all correlate positively. As such, psychometric g is a necessary algebraic consequence of the positive manifold itself (Krijnen Reference Krijnen2004). The concept of general intelligence interprets psychometric g as a within-individual, domain-general cognitive ability that permeates all human mental activity so that different tests are functionally equivalent in the sense that they all measure this general ability to a varying extent. This is a sufficient, but not necessary, explanation of the positive manifold. Moreover, it is contradicted by evidence from cognitive neuroscience, neuropsychology, and the study of developmental disorders (e.g., Duncan et al. Reference Duncan, Burgess and Emslie1995; Vicari et al. Reference Vicari, Bellucci and Carlesimo2007; Wang & Bellugi Reference Wang and Bellugi1994).
Contrary to g, fluid intelligence can be meaningfully conceptualized as a domain-general ability involved in complex, novel problem solving – according to its definition, it is “an expression of the level of relationships which an individual can perceive and act upon when he does not have recourse to answers to such complex issues already stored in memory” (Cattell Reference Cattell1971, p. 115.) or “the use of deliberate and controlled mental operations to solve novel problems that cannot be performed automatically” (McGrew Reference McGrew2009, p. 5). In humans, fluid reasoning is usually measured with tests of nonverbal inductive reasoning. Gf shares nearly half of its variance with working memory (Kane et al. Reference Kane, Hambrick and Conway2005; Oberauer et al. Reference Oberauer, Schulze, Wilhelm and Süss2005), probably because they both tap executive/attentional processes to a large extent (Engle & Kane Reference Engle and Kane2004).
There are reasons that can lead one to think that Gf and g are the same: Gf is central to variation in cognitive abilities to the extent that g and Gf are statistically near-indistinguishable (Gustafsson Reference Gustafsson1984; Matzke et al. Reference Matzke, Dolan and Molenaar2010). Yet general intelligence and fluid reasoning are clearly different constructs (Blair Reference Blair2006) – and so are the psychometric factors g and Gf (Kovacs et al. Reference Kovacs, Plaisted and Mackintosh2006). Additionally, whereas the neural substrate of fluid intelligence is in the prefrontal and partly in the parietal cortex (Kane & Engle Reference Kane and Engle2002; Kane Reference Kane, Wilhelm and Engle2005), it is difficult to localize g, as results depend on the actual battery of tests used to extract g (Haier et al. Reference Haier, Colom, Schroeder, Condon, Tang, Eaves and Head2009). Also, different components of g are differently affected by aging or the Flynn effect (the secular increase in IQ), both of which manifest themselves more strongly on nonverbal than verbal tests (Flynn Reference Flynn2007; Horn & Cattell Reference Horn and Cattell1967; Trahan et al. Reference Trahan, Stuebing, Fletcher and Hiscock2014).
Verbal cognition itself is crucial from the target article's perspective when interpreting g. In humans, g is composed of crystalized intelligence (Gc), too: the ability to apply already acquired skills and knowledge, with an emphasis on language – vocabulary, reading comprehension, and verbal reasoning. This does not translate to nonhuman animals, making it very implausible that general factors reflect the same construct across species. The authors' approach to general intelligence, emphasizing problem solving in novel contexts, also in fact reflects fluid intelligence – the central component of g, but not the same as g. Finally, executive functions are more strongly related to Gf than to other components of g (Conway & Kovacs Reference Conway, Kovacs and Ross2013). In fact, given the authors' emphasis on problem solving in novel situations as well as on the role of cognitive flexibility and executive functions, we often had the impression when reading the target article that Burkart et al. in fact discussed fluid intelligence under the term general intelligence.
If g does not reflect a unitary domain-general cognitive ability and is not identical to Gf, then how can the general factor of intelligence be conceptualized? Or, more importantly, if mental tests do not all measure the same general intelligence, then why do tests with different content correlate so strongly?
There are two recent explanations of the positive manifold (with corresponding mathematical formulations) that do not propose a psychological equivalent of psychometric g: the mutualism model (van der Maas et al. Reference van der Maas, Dolan, Grasman, Wicherts, Huizenga and Raijmakers2006) and process overlap theory (Kovacs & Conway Reference Kovacs and Conway2016). Mutualism explains the positive manifold with mutually beneficial interactions between cognitive processes during development. Process overlap theory proposes a functional overlap of cognitive processes when people solve mental test items, such that executive/attentional processes are tapped by a large number of different items whereas domain-specific processes are tapped by specific types of tests only.
Both explanations conceptualize intelligence as a set of independent specific abilities and processes. According to the process overlap theory, g is an emergent rather than latent property of mental test scores. Technically, this means that g is conceptualized as a formative rather than reflective latent variable: the common consequence of the covariance among tests rather than its common cause. Another common example is socioeconomic status (SES), which clearly is the outcome, and not the cause of a number of indicators like family income, parental education, and so on (Fig. 1).
Figure 1. The structural model corresponding to process overlap theory on a simplified model.
Such a stance would contribute greatly to the authors' comparative approach, in which g would vary from species to species (depending on whether its exact composition includes social skills, language, etc.), whereas a reflective fluid intelligence could indeed be plausibly interpreted as an ability whose evolution was shaped by evolutionary pressures to solve novel problems. The evolution of fluid intelligence could probably be understood through disentangling the evolution of the prefrontal cortex and executive functions in a number of different species.
At the same time, applying a formative framework to g could contribute to a functionalist approach, because the primary role of formative constructs is predicting important real-life outcomes (Bagozzi Reference Bagozzi2007; Howell et al. Reference Howell, Breivik and Wilcox2007); in this case, evolutionary ones. Under such a formative/functionalist agenda, the focus would be on individually identifying the cognitive capabilities of each species, ranging from olfactory abilities to social cognition, and how they uniquely contribute to the given species chances of survival and reproduction.
The authors present a wide-ranging theory of the evolution of intelligence. However, Burkart et al. seem to have confused the general intelligence factor (psychometric g) with fluid intelligence (Gf). Psychometric g is a statistical way of describing the positive manifold: the phenomenon that ability tests, each with different content, all correlate positively. As such, psychometric g is a necessary algebraic consequence of the positive manifold itself (Krijnen Reference Krijnen2004). The concept of general intelligence interprets psychometric g as a within-individual, domain-general cognitive ability that permeates all human mental activity so that different tests are functionally equivalent in the sense that they all measure this general ability to a varying extent. This is a sufficient, but not necessary, explanation of the positive manifold. Moreover, it is contradicted by evidence from cognitive neuroscience, neuropsychology, and the study of developmental disorders (e.g., Duncan et al. Reference Duncan, Burgess and Emslie1995; Vicari et al. Reference Vicari, Bellucci and Carlesimo2007; Wang & Bellugi Reference Wang and Bellugi1994).
Contrary to g, fluid intelligence can be meaningfully conceptualized as a domain-general ability involved in complex, novel problem solving – according to its definition, it is “an expression of the level of relationships which an individual can perceive and act upon when he does not have recourse to answers to such complex issues already stored in memory” (Cattell Reference Cattell1971, p. 115.) or “the use of deliberate and controlled mental operations to solve novel problems that cannot be performed automatically” (McGrew Reference McGrew2009, p. 5). In humans, fluid reasoning is usually measured with tests of nonverbal inductive reasoning. Gf shares nearly half of its variance with working memory (Kane et al. Reference Kane, Hambrick and Conway2005; Oberauer et al. Reference Oberauer, Schulze, Wilhelm and Süss2005), probably because they both tap executive/attentional processes to a large extent (Engle & Kane Reference Engle and Kane2004).
There are reasons that can lead one to think that Gf and g are the same: Gf is central to variation in cognitive abilities to the extent that g and Gf are statistically near-indistinguishable (Gustafsson Reference Gustafsson1984; Matzke et al. Reference Matzke, Dolan and Molenaar2010). Yet general intelligence and fluid reasoning are clearly different constructs (Blair Reference Blair2006) – and so are the psychometric factors g and Gf (Kovacs et al. Reference Kovacs, Plaisted and Mackintosh2006). Additionally, whereas the neural substrate of fluid intelligence is in the prefrontal and partly in the parietal cortex (Kane & Engle Reference Kane and Engle2002; Kane Reference Kane, Wilhelm and Engle2005), it is difficult to localize g, as results depend on the actual battery of tests used to extract g (Haier et al. Reference Haier, Colom, Schroeder, Condon, Tang, Eaves and Head2009). Also, different components of g are differently affected by aging or the Flynn effect (the secular increase in IQ), both of which manifest themselves more strongly on nonverbal than verbal tests (Flynn Reference Flynn2007; Horn & Cattell Reference Horn and Cattell1967; Trahan et al. Reference Trahan, Stuebing, Fletcher and Hiscock2014).
Verbal cognition itself is crucial from the target article's perspective when interpreting g. In humans, g is composed of crystalized intelligence (Gc), too: the ability to apply already acquired skills and knowledge, with an emphasis on language – vocabulary, reading comprehension, and verbal reasoning. This does not translate to nonhuman animals, making it very implausible that general factors reflect the same construct across species. The authors' approach to general intelligence, emphasizing problem solving in novel contexts, also in fact reflects fluid intelligence – the central component of g, but not the same as g. Finally, executive functions are more strongly related to Gf than to other components of g (Conway & Kovacs Reference Conway, Kovacs and Ross2013). In fact, given the authors' emphasis on problem solving in novel situations as well as on the role of cognitive flexibility and executive functions, we often had the impression when reading the target article that Burkart et al. in fact discussed fluid intelligence under the term general intelligence.
If g does not reflect a unitary domain-general cognitive ability and is not identical to Gf, then how can the general factor of intelligence be conceptualized? Or, more importantly, if mental tests do not all measure the same general intelligence, then why do tests with different content correlate so strongly?
There are two recent explanations of the positive manifold (with corresponding mathematical formulations) that do not propose a psychological equivalent of psychometric g: the mutualism model (van der Maas et al. Reference van der Maas, Dolan, Grasman, Wicherts, Huizenga and Raijmakers2006) and process overlap theory (Kovacs & Conway Reference Kovacs and Conway2016). Mutualism explains the positive manifold with mutually beneficial interactions between cognitive processes during development. Process overlap theory proposes a functional overlap of cognitive processes when people solve mental test items, such that executive/attentional processes are tapped by a large number of different items whereas domain-specific processes are tapped by specific types of tests only.
Both explanations conceptualize intelligence as a set of independent specific abilities and processes. According to the process overlap theory, g is an emergent rather than latent property of mental test scores. Technically, this means that g is conceptualized as a formative rather than reflective latent variable: the common consequence of the covariance among tests rather than its common cause. Another common example is socioeconomic status (SES), which clearly is the outcome, and not the cause of a number of indicators like family income, parental education, and so on (Fig. 1).
Figure 1. The structural model corresponding to process overlap theory on a simplified model.
Such a stance would contribute greatly to the authors' comparative approach, in which g would vary from species to species (depending on whether its exact composition includes social skills, language, etc.), whereas a reflective fluid intelligence could indeed be plausibly interpreted as an ability whose evolution was shaped by evolutionary pressures to solve novel problems. The evolution of fluid intelligence could probably be understood through disentangling the evolution of the prefrontal cortex and executive functions in a number of different species.
At the same time, applying a formative framework to g could contribute to a functionalist approach, because the primary role of formative constructs is predicting important real-life outcomes (Bagozzi Reference Bagozzi2007; Howell et al. Reference Howell, Breivik and Wilcox2007); in this case, evolutionary ones. Under such a formative/functionalist agenda, the focus would be on individually identifying the cognitive capabilities of each species, ranging from olfactory abilities to social cognition, and how they uniquely contribute to the given species chances of survival and reproduction.
ACKNOWLEDGMENTS
The first author's research was supported by the grant EFOP-3.6.1-16-2016-00001 (“Complex improvement of research capacities and services at Eszterházy Károly University”).