Selection Demands and Working Memory Mediate Interference during Naming

• ¹ Rice University, Psychology, United States

Naming pictures in semantically related (DOG, CAT, PIG) vs. unrelated groups (DOG, BUS, EAR) across numerous cycles (blocked-cyclic naming; BCN) results in increasingly slower response times (RTs) (e.g., Belke, 2008). This semantic interference (SI) is thought to arise because selecting responses is more demanding in the context of similar vs. dissimilar items (e.g., Schnur, Schwartz, Brecher, & Hodgson, 2006). We may attenuate SI depending on our capacity to reduce interference from competitors during lexical selection (selection capacity; cf. Schnur et al., 2009). Alternatively, we reduce SI by strategically holding names in working memory (WM) to anticipate pictures (Belke, 2008; Crowther & Martin, 2014), which attenuates interference after the first cycle such that interference does not accumulate (Belke & Stielow, 2013). If WM is the primary contributor to performance in BCN, this suggests BCN is not an appropriate measure of interference during lexical selection (cf. Navarette, Del Prato, Peressotti, & Mahon, 2014). Here we investigated the degree to which SI in BCN reflects selection demands, WM processes, or both. Method Using two variations of BCN, we tested 12 patients (from Harvey & Schnur, 2015) with variable selection capacity and WM impairments following stroke (not all patients could complete all tasks, see Figure for details) and 122 healthy subjects. We measured patients’ selection capacities using a high vs. low selection demand sentence completion task (cf. Snyder & Munakata, 2008), where subjects provided a final word to complete a sentence. High demand sentences had many possible responses (e.g., “He couldn’t think of anyone less ______.”), while low demand sentences had generally one response (e.g., “He mailed a letter without a _______.”). Selection capacity was the percent correct difference between these two conditions (low – high). We tested patients’ WM capacities using rhyme and category probe tasks (e.g., Martin, Shelton, & Yaffee, 1994). Subjects were given lists of words to remember (CAR, PEAR, DOG) and asked if a probe word (BAR) either rhymed with a word in the list (rhyme probe) or was from the same semantic category as a word in the list (category probe). Results and Conclusions Analyzing RTs, we replicated previous SI effects in BCN in healthy subjects and patients (reported effects’ p’s < .05). Subjects responded more slowly to items in semantically related vs. unrelated groups, and this interference increased across cycles (evidence of increasing selection demands; Figures 1a/c). Indicative of WM, healthy subjects were faster to name items as the number of unrelated items to be named (i.e., remembered) decreased. However, this was attenuated by selection demand, as there was no decrease in RTs across repeated related items (Figure 1b). Converging evidence from patients demonstrated more impaired selection capacity produced increasingly slower RTs across item positions within the related condition (high selection demand) (Figure 1d). Larger WM capacity (rhyme probe but not category probe, p = .54) produced increasingly faster RTs across items positions in the unrelated condition (low selection demand) (Figure 1e). Together, these findings suggest that both selection demand and phonological WM processes mediate interference during naming.