The representation of adjuncts: findings from structural priming

Structural priming, speakers’ tendency to reuse recently encountered syntactic structures, is enhanced when content words are repeated between primes and targets, a phenomenon termed the lexical boost (Mahowald et al., 2016). Studies suggest that the lexical boost is specific to the repetition of the syntactic head of the primed structure (e.g., the verb in a VP, Van Gompel et al., 2022). This is taken as evidence that structures are lexically represented with their subcategorising head but not with any other words in the sentence and is consistent with the residual activation account of priming (Pickering & Branigan, 1998), but not with models which maintain that a lexical boost occurs with the repetition of any content word (Chang et al., 2006; Reitter et al., 2011). We investigated how adjunct phrases, which are not subcategorized for by a verb, are represented. We asked whether adjuncts were associated with words in the sentence or whether their representation is lexically independent. Experiment 1 investigated this by testing if repeating a verb (e.g., shave) would enhance the priming of adverbial phrase (AdvP) position. We used primes with either a preverbal or postverbal AdvP (e.g., The driver carefully shaved vs. The driver shaved carefully). Participants read pre- or postverbal AdvP primes aloud and then described a target event using an adverb printed underneath (Fig.1). The verbs in the target events were either the same as in the primes or different. We observed a main effect of prime (p < .001) with more postverbal AdvP target responses after postverbal than after preverbal AdvP primes (78% vs. 60%); and a prime x repetition interaction (i.e., a lexical boost, p < .001) with stronger priming when the verb was repeated than when it was not (24% vs. 12%), although priming was significant in both the repetition and non-repetition conditions (Table.1). The findings suggest that the representation of AdvP position is associated with the verb even though the AdvP is an adjunct and therefore is not subcategorised for by the verb. However, some theoreticians do not treat all adverbs as adjuncts (Cinque, 1999); in addition, adverbs semantically modify the verb’s meaning (cf. compare the candle burnt partly vs. completely). Experiment 2 therefore investigated the lexical boost with temporal phrases (e.g., before breakfast), which are generally treated as adjuncts and modify the whole sentence rather than just the meaning of the verb. The temporal adjunct was placed either initially or finally in the prime sentence (e.g. Before breakfast the driver shaved vs The driver shaved before breakfast). The method used was the same as in Experiment 1 (Fig. 2). The analyses showed a main effect of prime (p < .001) with more adjunct-final responses after adjunct-final primes than after adjunct-initial primes (86% vs. 76%); and a prime x repetition interaction (p = .014) with a larger priming effect when the verb was repeated than when it was not (14% vs. 5%), although priming was observed both with and without verb repetition (Table 2). The results suggest that the location of temporal phrase adjuncts is also associated with the verb. We then asked if the repetition of adjunct phrases themselves could boost priming. Using the same methodology as before, we manipulated AdvP repetition (Experiment 3, Fig.1) and the repetition of the noun in the temporal phrase (Experiment 4, Fig. 2). Experiment 3 showed a main effect of prime (p < .001) with more postverbal AdvP responses after postverbal than after preverbal primes (77% vs. 34%). A significant prime x repetition interaction (p < .001) showed that although priming was evident when the AdvP was repeated (62% priming) and when it was not (25% priming), it was stronger in the former (Table 3). The results from Experiment 4 were similar: there was a main effect of prime (p < .001) with more adjunct-final responses after adjunct-final primes than after adjunct-initial primes (68% vs. 27%); and a prime x repetition interaction (p < .001) with stronger priming when the noun was repeated (47% priming) than when it was not (34% priming) (Table 4). The results suggest that the information about the adjunct location is also lexically represented with the adjunct itself. Together, our findings suggest that the syntactic position of adjuncts is associated with words in the sentence that do not subcategorise for them. This contrasts with evidence from argument phrases (e.g., in ditransitive structures The man gave the woman the book/the book to the woman, Carminati et al., 2019), which are represented with the head that subcategorises for them, but not with any other words.