The GLM models presented in Figures 9.1-9.3 imply that highly specific learning occurs first, which then may generalize and transfer. Transfer can be defined as “the ability to use previously learned skills or knowledge in settings or on problems different from the original learning, including the capacity to distinguish when and where these learnings are appropriate” (Gentile, 2000, p. 13). Although transfer is valued by educators (and students), a century of research does not provide much evidence for learners’ ability to spontaneously transfer what they have learned. In perhaps the first systematic studies, Thorndike and Woodworth (1901; Thorndike, 1923) found that “formal disciplines” like math and Latin did not discipline the mind and improve thinking in general, thus refuting the dominant educational theory of that period. They proposed identical elements theory instead, which posited that the amount of transfer is a function of the number of common (identical or similar) elements between the original learning and target tasks or settings. Thus fluency in Latin might facilitate learning Spanish, but not Chinese, and learning to read Chinese characters may not facilitate speaking the language. Identical elements theory did not work in all situations, however. Critics such as Judd (1908) and the Gestalt psychologists showed that understanding principles (such as light refraction) could facilitate transfer (such as shooting at an underwater target).
By the 1960s, most of these ideas were incorporated into interference theory, which allowed a more complex study of transfer and memory processes in its proactive and retroactive interference (or facilitation) research designs. All learning occurs in the context of previous learning (a proactive effect where A influences B) and subsequent learning (a retroactive effect where B influences recall of A). The effects can be positive or negative; that is, they can facilitate or interfere with transfer or recall. Hundreds of studies were conducted in this paradigm (Deese & Hulse, 1967, and Ellis, 1965, provide excellent summaries) with concomitant recommendations for properly sequencing curricula to increase the probability of transfer.
By the 1980s, the cognitive revolution was in full force, emphasizing an active, socially mediated construction of meaning. From this conception came situated cognition theory (e.g., Brown, Collins, & Duguid, 1989; Brown & Duguid, 1993), which argued that what comes to be known is context-specific and largely determined by how it is learned. In this view, learning is so contextualized that “authentic” on-the-job training (as in apprenticeships) is necessary for positive transfer to occur. In addition, Singley and Anderson (1989) postulated identical productions theory, which predicts transfer based in part on whether the training-to-transfer tasks are both declarative (“knowing that”) or procedural (“knowing how”).
More recently, Barnett and Ceci (2002) provided a taxonomy for predicting transfer, with content and context as the major dimensions. Content includes the to-be-learned skill (procedure or principle), the required performance (speed, accuracy), and the memory demands. Context includes the domains of knowledge, as well as the physical, temporal, functional, social, and testing situations. Transfer can thus be considered to be “near” or “far” depending on how similar the transferred content and context are to the original.
Transfer is often cited as rare, although counterexamples are available (e.g., Pressley & Yokai’s 1994 review of Detterman & Sternberg, 1993). It is one thing, however, to find counterexamples and another to prove transfer. Nevertheless, finding counterexamples, discovering logical fallacies, and using the scientific method are themselves examples of thinking that are difficult to tie to specific contexts. Perkins and Salomon (1989) summarized the research and suggested two roads to transfer: the low road, which requires “much practice in a large variety of situations, leading to a high level of mastery and automaticity,” and the high road, which requires “deliberate mindful abstraction of a principle” (p. 22).
Given the paucity of strong evidence of transfer from classroom settings to other settings, it is surprising that a great deal of the research on video game effects can be interpreted as demonstrating “far” transfer. For example, meta-analyses demonstrate that violent video games predict (which could be reinterpreted as “train”) aggressive ways of thinking, aggressive feelings, and aggressive behaviors in the real world (e.g., Anderson et al., 2010; Ferguson, 2007). Similarly, repeated playing of prosocial video games seems to generalize and transfer to greater helpful and cooperative behaviors in the real world—that is, to situations that share almost no characteristics that are similar to those in the video games (e.g., Gentile et al., 2009). Games, therefore, deserve additional consideration as teaching tools.