Transfer of Learning - Conceptual Foundation of Transfer Research - Mental Representations and Transfer: Common Element-based Vs. Schema-based Approaches

Mental Representations and Transfer: Common Element-based Vs. Schema-based Approaches

The majority of mental processes studied in research on human cognition have one thing in common: they all pertain in one way or another to the construction of mental representations. This is true, for instance, for perceiving, learning, problem-solving, reasoning and thinking, and recalling, as much as it is true for the phenomenon of transfer.

Although research on mental representation has been utterly manifold, two main traditions can be discerned. Some researchers have regarded mental representations in terms of abstract schemata, frames, patterns or mental models (Bartlett, 1932; Chase & Simon, 1973; Gentner & Stevens, 1984; Johnson-Laird, 1983; Johnson-Laird & Byrne, 1990; Minsky, 1975), while others have paid attention to semantic information and propositional nature of mental representations (Anderson, 1976, 1983, 1994; Collins & Quillian, 1968; Medin & Ribs, 2005; Medin & Smith, 1984; Minsky, 1968; Rosch, 1978). These differential conceptualizations have, in general, been driven by distinct psychological paradigms adopted, such as Associationism and Connectionism, Behaviorism, Gestaltism, and Cognitivism.

GOMS and ACT-based procedural transfer theses are a good example of modern explanations fitting the atomistic and mechanistic nature of the Connectionist paradigm, i.e., by seeing transfer as an effect of commonality in semantic conditions-action-goal structures, mainly instantiated as If-Then production rule associations overlap. This view on transfer clearly replaced Behaviorist explanatory concepts of stimuli and response with more sophisticated mental concepts that serve as units of transfer. The cognitive architecture background also added important processing capabilities and some degree of flexibility concerning the identicality constraint (e.g., declarative-to-procedural, and declarative-to-declarative transfer). It did not, however, essentially defy the common underlying common element-based thought model of transfer.

Both the original habitual response-based idea of common element transfer as well as the modern production rule compilation and knowledge encapsulation account are in their core assumptions refuted by Gestaltists' theories. Koffka's (1925) scrutiny of Thorndike's (1911, 1913) and Köhler's (1917) arguments and findings revealed that explanations of learning and transfer based on the notions of association and automation fall short of explicating the nature of mental activity even for simple problem-solving tasks. Novel explanatory concepts were needed to account for "learning by understanding" (Katona, 1940) and problem-solving transfer (Mayer & Wittrock, 1996). These were found with reference to the organization and structure of knowledge (Clement & Gentner, 1991; Gentner & Gentner, 1983; Gentner & Toupin, 1986), abstraction and general principle inferences (Bourne, Ekstrand, & Dominowski, 1971, p. 104ff.; Judd, 1908, 1939; Simon & Hayes, 1976), the goal- and meaning-directedness of thinking and its holistic nature (Bühler, 1907, 1908a; Holyoak, 1985; Humphrey, 1924; Selz, 1913, 1922), and functional relations (Duncker, 1935; Köhler, 1917). Because this tradition of investigating transfer is based on Gestaltist ideas, they could be summarized under the header of schema-based theories of transfer.

In accord with the traditions regarding research on mental representation, two mainstream explanatory models for transfer can be concluded to date. One is the model of common element-based transfer, rooting in Thorndikean ideas, which explains transfer as confined to elementary correspondences between a primary and a secondary learning situation, such as procedures and their automated effect (e.g., Allport, 1937; Singley & Anderson, 1985, 1989; Thorndike, 1924a, b). The other model emerging from the Gestalt tradition can be labeled schema-based or analogical transfer, emphasizing elementary loosened structural or principle/rule-based coherence between transfer source and target (e.g., Duncker, 1935; Gentner, 1983; Gentner & Gentner, 1983; Gick & Holyoak, 1980, 1983; Köhler, 1917/1957; Reed, 1993). They continued Judd's (1908) line of work, resulting in further accentuation of "insightful" transfer, using terms like knowledge structures and schemata, solution principles, and functionality (Katona, 1940; Wertheimer, 1945/1959).

The problem is that as far as transfer of learning in both traditions refers to one and the same phenomenon, there can not be a situation with two incompatible theoretical frameworks standing side-by-side. Conceptual resolution in some form is clearly imperative. Several efforts have been made in recent years to review and revive transfer research, and to resolve controversies (cf. the content- and apperception-based approach: Helfenstein, 2005), but empirical justification is still in early stages.