I agree with Branigan & Pickering (B&P) that structural priming is an excellent method for tapping into speakers' linguistic representations. I agree with B&P that “researchers concerned with linguistic representations should not rely solely on [acceptability] judgments” (sect. 1.2, para. 10). I agree with B&P that, frequently, “evidence from structural priming will converge with evidence from acceptability judgments and hence provide strong support for specific representational claims” (sect. 1.5, para. 2).

I do not agree with B&P that “where acceptability judgment and priming evidence do not converge, evidence from priming should be favored” (sect. 1.5, para. 2). Why? The most suitable method for linguistic research depends on exactly what we want to know. Sometimes, structural priming is indeed the best choice: for example, when we want to know whether particular forms share some underlying representation. Sometimes, forced-choice comprehension is the best choice: for example, when we want to know if children understand the meaning of word order or case marking.

And sometimes, an acceptability judgment paradigm is the best choice: for example, when we want to know which of two similar forms is more consistent with adult speakers' underlying grammatical representations (e.g., *The funny clown giggled Bart vs *The funny clown laughed Bart). This really can be determined only using a Likert-scale-type judgment. Structural priming is all but useless here, because no adult native speaker of English is going to produce either sentence, no matter how much you prime her. The broader problem is that structural priming yields a binary outcome measure: You're primed, or you're not; you produce the sentence, or you don't. In contrast, acceptability judgments, if set up to do so, yield a continuous outcome measure. Crucially, the use of a relatively insensitive binary measure over a much more sensitive continuous measure can lead to erroneous conclusions regarding representation. Here's a case study.

B&P argue that “syntactic representations do not contain semantic information” (sect. 2.1, para. 2) and that “adults … appear to have abstract syntactic representations that are not specified for lexical or thematic content.” A crucial piece of evidence for this claim is a series of structural priming studies (Messenger et al. Reference Messenger, Branigan and McLean2012a; Reference Messenger, Branigan, McLean and Sorace2012b) in which passives with agent-patient, theme-experiencer and experiencer-theme verbs (e.g., The girl is being licked/scared/ignored by the cow) were equally effective at priming production of other passives. Hence – B&P's argument goes – adults' representation of the passive (a syntactic representation) does not care about the identity of the verb (semantic and/or lexical content) or, indeed, its thematic content (i.e., the way its syntactic arguments are mapped onto the semantic roles agent, patient, experiencer and theme).

This conclusion is incorrect, however. In one of my own studies (sorry!) based closely on Messenger et al. (Reference Messenger, Branigan, McLean and Sorace2012b), we asked adults to rate these types of sentences on a 5-point scale (Ambridge et al. Reference Ambridge, Bidgood, Pine, Rowland and Freudenthal2016). Passives with experiencer-theme verbs (e.g., ignore) were rated as less acceptable than passives with agent-patient (e.g., lick) and theme-experiencer verbs (e.g., scare), a pattern that did not hold for actives. This suggests that adults' representation of the passive construction is not purely syntactic but contains lexical/semantic/thematic-role information such that “[B] (mapped onto the surface subject [of a passive]) is in a state or circumstance characterized by [A] (mapped onto the by-object or an understood argument) having acted upon it” (Pinker et al. Reference Pinker, Lebeaux and Frost1987, p. 249). In other words, the more the first NP is affected by the action, the better the passive – which also explains why three hours can't be lasted by a film or five people slept by a tent (cf. The film lasted three hours; This tent sleeps five people).

Why did we find lexical/semantic/thematic-role differences between passive sentences, but Messenger et al. (Reference Messenger, Branigan, McLean and Sorace2012b) did not? Simple: We used a continuous DV (“How good is this sentence on a 5-point scale?”), while they used a less sensitive binary DV (“Does this sentence prime another passive?). This does not mean that acceptability judgments are always a better choice than structural priming; sometimes, the reverse is true. Horses for courses.