We wholeheartedly agree with the authors about one thing: The demonstration of priming of a hypothesized representation is sufficient evidence for the existence of that representation. This is true even if this demonstration is possible only in controlled lab experiments. (In natural conversation, in fact, people do not have the “tendency to repeat their own and others' structural choices” (target article, abstract) as Healey et al. [Reference Healey, Purver and Howes2014] have demonstrated convincingly). Using structural priming as a necessary condition for the existence of representations “of any aspect of linguistic structure,” however, would be guaranteed to lead us astray.
First, here is an obvious methodological pitfall that we would be throwing ourselves into willingly. Not finding structural priming for a certain representation might have many different causes, the representation not existing being only one of them. It might also be that we haven't done the right experiment, or that the effect is too weak to detect. We might be unable to control for important confounding variables, be unable to find the proper stimulus materials, or lack the necessary statistical power. The key principle here is that absence of evidence is not evidence of absence.
Second, it is possible that the representation under investigation does, in fact, exist, but is for some reason not primeable in psycholinguistic experiments. There is no reason to expect that all linguistic representations posited in our theories are primeable. To be able to control for possible effects of meaning, most demonstrations of structural priming have involved a relatively small set of syntactic phenomena, such as the dative alternation or active vs. passive constructions (Mahowald et al. Reference Mahowald, James, Futrell and Gibson2016b). Does this mean that other syntactic constructions (e.g., small clauses) do not exist? As an example from a different linguistic domain, Tooley et al. (2014) found that they could not prime intonational phrases. But that should not (and did not) lead the authors to conclude that intonational phrases lack representation.
More generally, the astounding progress that the cognitive sciences in general, and psycholinguistics specifically, have made since the cognitive revolution is founded upon the post-behaviorist assumption that it is legitimate to posit internal representations and processes even if not every one of these components can be proved by direct demonstrations of their existence. Edward Tolman and his colleagues (Tolman Reference Tolman1948) provided convincing evidence of the existence of map-like mental representations in rats, but this never involved priming them.
As for abandoning acceptability judgments, the authors' proposal amounts to sawing off the branch we're sitting on. The success story of structural priming owes a large debt to half a century of previously developed linguistic theory about syntactic structures, massively informed by acceptability judgments. Without linguists identifying grammatical phenomena, determining their scope, and studying the relationships between different syntactic forms, we wouldn't even know where to start looking. How else would we know that The pirate gave the princess a parcel and The pirate gave a parcel to the princess (or an active and a passive version of a sentence) are semantically equivalent if not through using native-speakers' judgments to establish that they are both acceptable for describing a certain state of affairs? In fact, it would be impossible even to develop experimental stimuli without using our (implicit) acceptability judgments; there is a reason that structural priming experiments usually don't contain sentences such as Pirate a parcel the princess the gave.
Structural priming certainly deserves its place in the vast array of methods available to psycholinguists. But we see no reason to give it primacy over the many other paradigms that have been, and still are, essential pillars of cognitive science.
You have
Access
We wholeheartedly agree with the authors about one thing: The demonstration of priming of a hypothesized representation is sufficient evidence for the existence of that representation. This is true even if this demonstration is possible only in controlled lab experiments. (In natural conversation, in fact, people do not have the “tendency to repeat their own and others' structural choices” (target article, abstract) as Healey et al. [Reference Healey, Purver and Howes2014] have demonstrated convincingly). Using structural priming as a necessary condition for the existence of representations “of any aspect of linguistic structure,” however, would be guaranteed to lead us astray.
First, here is an obvious methodological pitfall that we would be throwing ourselves into willingly. Not finding structural priming for a certain representation might have many different causes, the representation not existing being only one of them. It might also be that we haven't done the right experiment, or that the effect is too weak to detect. We might be unable to control for important confounding variables, be unable to find the proper stimulus materials, or lack the necessary statistical power. The key principle here is that absence of evidence is not evidence of absence.
Second, it is possible that the representation under investigation does, in fact, exist, but is for some reason not primeable in psycholinguistic experiments. There is no reason to expect that all linguistic representations posited in our theories are primeable. To be able to control for possible effects of meaning, most demonstrations of structural priming have involved a relatively small set of syntactic phenomena, such as the dative alternation or active vs. passive constructions (Mahowald et al. Reference Mahowald, James, Futrell and Gibson2016b). Does this mean that other syntactic constructions (e.g., small clauses) do not exist? As an example from a different linguistic domain, Tooley et al. (2014) found that they could not prime intonational phrases. But that should not (and did not) lead the authors to conclude that intonational phrases lack representation.
More generally, the astounding progress that the cognitive sciences in general, and psycholinguistics specifically, have made since the cognitive revolution is founded upon the post-behaviorist assumption that it is legitimate to posit internal representations and processes even if not every one of these components can be proved by direct demonstrations of their existence. Edward Tolman and his colleagues (Tolman Reference Tolman1948) provided convincing evidence of the existence of map-like mental representations in rats, but this never involved priming them.
As for abandoning acceptability judgments, the authors' proposal amounts to sawing off the branch we're sitting on. The success story of structural priming owes a large debt to half a century of previously developed linguistic theory about syntactic structures, massively informed by acceptability judgments. Without linguists identifying grammatical phenomena, determining their scope, and studying the relationships between different syntactic forms, we wouldn't even know where to start looking. How else would we know that The pirate gave the princess a parcel and The pirate gave a parcel to the princess (or an active and a passive version of a sentence) are semantically equivalent if not through using native-speakers' judgments to establish that they are both acceptable for describing a certain state of affairs? In fact, it would be impossible even to develop experimental stimuli without using our (implicit) acceptability judgments; there is a reason that structural priming experiments usually don't contain sentences such as Pirate a parcel the princess the gave.
Structural priming certainly deserves its place in the vast array of methods available to psycholinguists. But we see no reason to give it primacy over the many other paradigms that have been, and still are, essential pillars of cognitive science.