Relational reasoning, or the fundamental human cognitive ability to discern patterns within any stream of information (Alexander and the DRLRL, 2012; Bassok, Dunbar, and Holyoak, 2012; Goswami, 2013), is a critical mental process that has been empirically linked to the learning and problem solving of students across the gamut of academic contexts and domains from early reading (Ehri, Satlow, and Gaskins, 2009) to clinical reasoning in medical school (Dumas, Alexander, Baker, Jablansky, and Dunbar, 2014). This current conceptualisation of relational reasoning that exists today within the educational psychology literature – as well as the empirical research that supports that theorising – has been evolving for more than a century (James, 1890; Sternberg, 1977; Alexander et al., 2012). For example, in his seminal text Principles of Psychology, William James described the human mental ability to discern relations of ‘difference and similarity’ among cognitive representations as a highly foundational cognitive process on which other functions, such as problem solving, may depend (James, 1890: 346).

Without this ability, we would necessarily inhabit a world of entirely isolated information. Given the fact that people are incessantly exposed to large volumes of information in daily life (e.g. students regularly access a variety of information sources for learning including conventional curricular materials, online documents, magazines, etc., Braasch et al., 2013), this ability to discern patterns within all these stimuli is therefore especially important in today’s age of information-inundation.

Later, in formulating his theory of intelligence, Charles Spearman defined his overall g factor as the ‘cognition of relations’ that supported all other more specific intellectual functions (Spearman, 1927: 165). As the theoretical conceptualisation of intelligence evolved throughout the mid-twentieth century, Raymond Cattell (1940) and John Raven (1941) both created intelligence tests (i.e. Cattell’s Culture-Free IQ test and the Raven Progressive Matrices) in which participants were required to reason with complex relations and that were designed to be influenced as little as possible by previously habituated cognitive processes (e.g. skills learned in school). These measures were soon (e.g. Cattell 1943) used to assess a specifically theorised form of intelligence – called fluid intelligence – in which participants were required to reason with complex relations that they could not have been previously exposed to. In contrast to fluid intelligence, crystallised intelligence was more specifically dependent on previously learned skills and knowledge and was measured with test items that had a strong reliance on prior knowledge (usually knowledge that would be taught in school). In theory, both fluid and crystallised tests would have required some ability on the part of participants to reason with relations (Alexander et al., 2016), but today relational reasoning is much more commonly associated with fluid intelligence, and Raven’s Progressive Matrices (Raven, 1941), which was originally a fluid intelligence measure, is commonly given as a measure of relational reasoning (e.g. Crone et al., 2009)

Later in the twentieth century, Sternberg (e.g. 1977) and Alexander and colleagues (e.g. White and Alexander, 1986) demonstrated that students’ ability to reason relationally was highly malleable when exposed to direct instruction of reasoning strategies, opening the door to a body of research concerning the specific processes that students must engage in to successfully map a relation among distal concepts or ideas (e.g. Gentner and Markman, 1997; Holyoak and Thagard, 1997). This body of work utilised a number of cognitive science methodologies, including neuroscientific methods (e.g. Krawczyk, McClelland, Donovan, Tillman, and Maguire, 2010), to elucidate the processes by which patterns are inferred and relations are mapped. By the beginning of this decade (i.e. the 2010s), this psychological literature had proliferated into a massive collection of investigations that, taken together, held many critical insights not only about the cognitive processes associated with relational reasoning, but also the diverse academic contexts in which relational reasoning plays a role.

In 2012, a review of relational reasoning literature focused on underlying neural mechanisms (i.e. Krawczyk, 2012) was published, effectively organising much of the cognitive science literature on relational reasoning. However, this review was not specifically concerned with the educational contexts in which relational reasoning may manifest and, therefore, Dumas, Alexander, and Grossnickle (2013) followed Krawczyk’s review with their own systematic synthesis of the extant literature, specifically in regard to the educational relevance of relational reasoning. By combining the findings of their synthesis of the empirical literature, as well as previously forwarded theoretical perspectives, these researchers posited a novel theoretical conceptualisation of relational reasoning: one that specifically described the construct as multidimensional – consisting of multiple forms or types of relational reasoning – as opposed to the previously held perspective of relational reasoning as a unitary cognitive ability. These researchers (Alexander et al., 2012; Dumas et al., 2013) were certainly not alone in attempting to broaden the conceptualisation of relational reasoning (cf. Holyoak, 2012; Sagi, Gentner, and Lovett, 2012), but they were the first to formally propose multiple forms of the construct that may be differentiated by the way they manifest within education and the academic context. In this chapter, the forms of relational reasoning put forward through this synthesis of the extant literature are presented and explained. Then, the development and initial reliability and validity evidence for one multidimensional measure of relational reasoning (i.e. the Test of Relational Reasoning; TORR; Dumas and Alexander, 2016) is presented and briefly reviewed.