The lexical boost persists across intervening utterances with sufficient priming input

Speakers tend to reuse recently experienced syntactic structures, a phenomenon termed structural priming (Pickering & Ferreira, 2008). Priming is enhanced when prime and target sentences share lexical content (Pickering & Branigan, 1998). This effect is referred to as the lexical boost. Previous studies suggest that abstract structural priming (i.e., priming without word repetition) and the lexical boost differ in their temporal properties; abstract priming lasts across intervening sentences whereas the lexical boost decays fast (Hartsuiker et al., 2008), indicating that they may have different origins (Chang et al., 2016; Reitter et al., 2011). However, the lexical boost may differ from abstract priming, not in terms of whether it lasts across intervening sentences, but in the amount of priming input required to exert a long-lasting effect. We therefore tested whether increasing the number of primes would result in a lasting lexical boost (over two intervening fillers between a prime and a target). In Experiment 1, one prime preceded each target and in Experiment 2, two primes preceded it. In Experiment 1, 56 participants completed sentence fragment primes using the pictures provided. The prime fragments consisted of either an adjective-noun (AdjN) structure (e.g., the underlined ...), or a noun-relative clause (NRC) structure (e.g., the ... that is underlined), together with a picture (e.g., an underlined cat). Participants then read two unrelated filler constructions and completed a target fragment also using a picture (e.g., the... together with a picture of a highlighted strawberry, see Fig.1). The noun in the target was either the same as in the prime or different. In previous studies that examined the time course of the lexical boost (e.g., Hartsuiker et al., 2008; Branigan & McLean, 2016), the verbs in primes and targets were often similar in meaning even when they were not repeated (e.g., give and hand), potentially making the lexical boost less long lasting. The verbs were also often repeated across items, possibly interfering with the boost from a repeated verb in the prime. In addition, the primes were comprehended rather than produced, which might have decreased memory for the verbs and also made the boost last less long. In the present study, the nouns were clearly different in meaning in the conditions where they were not repeated, no nouns were reused across items and all nouns were produced aloud by referring to a picture. A mixed effects analysis showed a main effect of prime (p < .001) with more AdjN responses after AdjN primes than after NRC primes when the noun was repeated (87% vs 71%) and when it was not (88% vs 71%). No prime x repetition interaction (p = .662) was found (Table 1). In line with previous studies that used single primes, these results suggest that the lexical boost does not last across two intervening utterances while abstract structural priming remains. In Experiment 2, we used the same methodology as before except that now participants (n=56) had two primes on each trial: the first was a read-only prime (e.g., the circled cat/the cat that is circled) and the second was a fragment to be completed using a picture as before (Fig. 1). The results showed a main effect of prime (p < .001) with more AdjN responses after AdjN primes than after NRC primes when the noun was repeated (96% vs 69%) and when it was not (92% vs 75%), and a prime x repetition interaction (p < .001), indicating stronger priming when the noun was repeated (27%) than when it was not (17%) (Table 1). This suggests that the lexical boost does persist when two primes are used. Overall, these findings demonstrate that with sufficient input, the lexical boost can last longer than previously shown, indicating that the difference between abstract structural priming and the lexical boost is not in whether they last across intervening sentences, but in the amount of priming input that is required. This challenges current models of structural priming (Chang et al., 2016, Reitter et al., 2011), which have so far assumed that abstract structural priming and the lexical boost differ in their time course. We suggest that for longer-term priming, the lexical boost requires more priming input than abstract priming does, because there are many more competing word-structure associations than competing structural representations. As a result, over time (i.e. across intervening utterances), the interference from the many word-structure associations is greater than that from the fewer structural representations