Leech et al. make a compelling argument that analogical transfer may not require a deliberative or explicit mapping process. However, they provide no direct evidence showing successful implicit analogy. The analogy tasks reviewed in the target article do not provide evidence which rules out the use of deliberative strategies. In one such task the intent was to show that analogical transfer did not require a mapping of base to target (Ripoll et al. Reference Ripoll, Brude and Coulon2003). Although the finding tends to confirm that subjects perform analogical transfer without deliberative mapping, it does not rule out the possibility that analogical transfer was itself accomplished deliberately. Indeed, participants were all explicitly informed that their task was to complete an analogy. In another task, pre-exposure to a problem solution in the base task biased participants to employ a similar strategy in the target task (Schunn & Dunbar Reference Schunn and Dunbar1996). However, the only evidence that participants employed an implicit strategy was that they did not mention the base problem as a foundation for the solution to the target problem. Although these findings are both consistent with the use of implicit strategies, neither directly rules out the possibility that deliberative strategies were used.
Recently, direct evidence showed that analogical transfer occurs in the absence of deliberative strategies (Gross & Greene Reference Gross and Greene2007). Participants learned a transverse patterning sequence and transferred the learned relations to a novel stimulus set. For transverse patterning, given a choice between two faces, A and B, participants learn by trial and error to select A (A>B), and likewise they learn B>C and C>A. To test for analogical transfer, participants simultaneously learn a partial set of relations among different faces, X>Y and Y>Z, and are then tested on the novel pairing X?Z (the question mark indicated that no choice was ever reinforced). In the control condition, the training on the transverse patterning set is omitted, and participants uniformly infer that X>Z. However, with exposure to the transverse patterning set, a significant portion of participants apply the transverse patterning relations and select Z>X. This analogical transfer occurred without intent or awareness of the transfer. Two tests of awareness were used. First, a post-experimental questionnaire was employed with increasingly leading questions about the use of analogical transfer. Only those participants who asserted – in a forced-choice question – that no analogical reason existed for the Z>X choice were categorized as unaware. In addition, it was noted during the debriefing that many participants were surprised that they had not been trained on the Z?X choice. To test this, a recognition task revealed that most participants did not recognize Z?X as a novel configuration. Had a deliberative strategy been the foundation for analogical transfer, participants would have had to recognize the novel configuration to apply the mapping.
Analogical transfer is but one instance of relational priming. Evidence that complex relations can be learned implicitly is found in the contextual cueing task (Chun & Phelps Reference Chun and Phelps1999; Greene et al. Reference Greene, Gross, Elsinger and Rao2007) and the relational manipulation task (Ryan et al. Reference Ryan, Althoff, Whitlow and Cohen2000). Furthermore, evidence that implicit relational learning can be expressed under novelty is demonstrated by the transitive inference task (Ellenbogen et al. Reference Ellenbogen, Hu, Payne, Titone and Walker2007; Greene et al. Reference Greene, Spellman, Dusek, Eichenbaum and Levy2001; Reference Greene, Gross, Elsinger and Rao2006).
These additional relational priming tasks may also assess the semantic priming hypothesis put forth in the target article. The authors propose that relational priming is a facet of semantic priming (e.g., “chicken is to hen” may semantically prime the term “female” so that “horse is to mare” may be correctly selected), suggesting that implicit relational learning is verbally mediated. However, because several implicit relational learning tasks, including analogy, are nonverbal tasks, it is difficult to hypothesize that such verbal primes necessarily mediate performance, particularly in the spatial tasks (contextual cueing and relational manipulation). In fact, some evidence suggests that verbal strategies may bias the use of explicit strategies when implicit strategies could otherwise be employed (for a review, see Greene Reference Greene2007), suggesting that it may be more fruitful to explore the relational priming hypothesis using nonverbal stimuli.
The evidence that analogical transfer and other forms of relational priming can be accomplished implicitly forces a major change in the theory of memory. It has previously been asserted that only deliberative processes could support learning that is sufficiently flexible for abstract application under novelty (e.g., Cohen et al. Reference Cohen, Poldrack and Eichenbaum1997; Reber et al. Reference Reber, Knowlton and Squire1996). It is now evident that such relational flexibility is not solely the domain of declarative memory (Chun & Phelps Reference Chun and Phelps1999; Ellenbogen et al. Reference Ellenbogen, Hu, Payne, Titone and Walker2007; Greene Reference Greene2007; Greene et al. Reference Greene, Spellman, Dusek, Eichenbaum and Levy2001; Reference Greene, Gross, Elsinger and Rao2006; Reference Greene, Gross, Elsinger and Rao2007; Gross & Greene Reference Gross and Greene2007; Ryan et al. Reference Ryan, Althoff, Whitlow and Cohen2000). The consequence of the discovery of such implicit tasks is that the declarative memory model (e.g., Squire Reference Squire1992) must be considered inadequate in its current form.
