Controlled Semantic Retrieval

Retrieving meaning in a context-relevant manner is critical to cognition. Doing so allows us to flexibly access information about concepts and objects in order to comprehend inputs and generate responses. Although on a continuum, there are two general routes to retrieving relevant semantic knowledge--automatic and controlled retrieval. Over the past seven years, our research has explored the neural mechanisms that support controlled semantic retrieval and that accompany the transition from controlled to automatic access. Initial results indicate that controlled semantic access partially depends on computations supported by the left ventrolateral PFC (VLPFC) [Wagner, Desmond et al., 1997; Wagner, Koutstaal et al., 2000; Wagner, Paré-Blagoev et al., 2001; Badre & Wagner, 2002; Wagner, 2002; Badre et al., 2005; Badre & Wagner, in press; Dobbins & Wagner, in press].

While left VLPFC is unambiguously engaged during semantic retrieval, a debate has emerged regarding how to understand VLPFC contributions to semantic processing. On the one hand, we hypothesized that left VLPFC guides top-down (controlled) retrieval of knowledge from long-term stores, with these computations being recruited when the strength of association between the retrieval cue and target knowledge is weak [Wagner, Paré-Blagoev et al., 2001]. From this perspective, VLPFC provides a top-down bias signal that guides the recovery of meaning under situations in which target knowledge is not retrieved through more automatic routes. By contrast, other theorists have proposed that left VLPFC does not support retrieval per se, but rather selects goal-relevant products of retrieval from amongst competitors.

To address this debate, we first demonstrated left VLPFC activation during controlled semantic retrieval in the absence of frank selection demands [Wagner, Paré-Blagoev et al., 2001; see also Bunge et al., 2005]. While potentially challenging the selection perspective, we noted that these controlled retrieval effects were observed in an anatomically distinct subregion of VLPFC from that reported by others to correlate with selection demands [Badre & Wagner, 2002]. Building on this potential anatomic segregation, we recently tested whether controlled retrieval and selection are distinct VLPFC processes or whether these models reduce to a common mechanism [Badre et al., 2005]. Using factor analytic techniques, we obtained evidence for a meta-factor that accounted for behavioral variance across tasks evoking semantic competition and for functional variance in left mid-VLPFC (inferior frontal gyrus pars triangularis), supporting a generalized control process that selects relevant knowledge from amongst competitors. By contrast, left anterior VLPFC (inferior frontal gyrus pars orbitalis) was sensitive to cue-target associative strength, but not competition, consistent with a control process that retrieves knowledge that does not automatically come to mind. These observations, together with other evidence that mid-VLPFC resolves competition within working memory [Badre & Wagner, in press] and during episodic recollection [Dobbins & Wagner, in press], motivate a two-process model of VLPFC function during semantic access, with distinct mechanisms mediating controlled retrieval and post-retrieval selection [for discussion, see Badre et al., 2005].