Branigan & Pickering (B&P) argue that structural priming methods have “reached maturity” (target article, para. 2) to the point that they can inform not only language production and comprehension processes, but also the nature of grammar, as typically studied by linguists using different analytical tools and methods. This view appears overly optimistic; structural priming remains widely used but poorly understood, with little consensus about why the effect is observed or exactly what production and comprehension processes are promoted from prior exposure to a sentence. Moreover, the larger class of priming methods, to which B&P link structural priming, has been the target of extensive criticism and reassessment of what can be gleaned from the tasks (Cesario Reference Cesario2014). Here, we consider two perspectives on the nature of structural priming and their consequences for B&P's claims for grammar.
One perspective is that structural persistence is a strongly syntactic phenomenon: Encountering/producing a sentence somehow biases the language processing system to expect or produce a similar syntactic structure. B&P's logic is that, because the processing system draws on the grammar, patterns of priming must reveal the nature of the grammar. This thinking raises the classic issues of the competence-performance distinction. If language use is grammar + processing, there is a credit assignment problem for psycholinguistic data: Any linguistic behavior might reflect the grammar, processing mechanisms, or some combination. B&P make exactly this criticism of other psycholinguistic methods – for example, that Franck et al.'s (Reference Franck, Soare, Frauenfelder and Rizzi2010) studies of subject-verb agreement production might illuminate the nature of the grammar, or alternatively they might reflect production or comprehension processes and be uninformative about grammar. Crucially, this assignment problem applies equally to priming. Haskell et al. (2010) used priming to study agreement production and found that subject-verb agreement is sensitive to the statistical patterns in prior usage (the primes). These results could support a graded grammar in which statistical patterns shape grammatical representations (Bybee Reference Bybee2006). Researchers rejecting this approach, however, could instead attribute these priming data to processing, leaving the grammar untouched by the statistics of usage. Thus, given B&P's assumption of usage=grammar + processing, structural priming is just as much subject to interpretive uncertainty as any other measure.
Even more interpretive uncertainty arises from an alternative view of structural priming – that it is not strictly syntactic but rather a language example of a more general tendency to repeat prior actions. Cognitive models of motor planning suggest these reuse effects (termed hysteresis effects) arise because it is easier to recall a previously executed motor sequence than to generate alternative plans de novo (Rosenbaum et al. Reference Rosenbaum, Cohen, Meulenbroek, Vaughan, Latash and Lestienne2006). Our own research investigates the link between structural priming and domain-general plan reuse, and we have developed parallel tasks that yield reliable structural priming for dative sentence structures and priming of nonlinguistic manual actions in the same participants (Koranda et al. Reference Koranda, Bulgarelli, Weiss and MacDonald2016). We also observed a parallel effect of priming strength in both domains: Preferred sentences and movements are more easily primed than unpreferred ones, a phenomenon previously observed in structural priming (Bock Reference Bock1986). These findings raise the possibility that plan reuse may be a domain-general property of action planning. MacDonald (Reference MacDonald2013) suggested that a general plan reuse bias would ground patterns of language use in basic planning mechanisms, and the existence of a general plan reuse bias may also explain why some nonlinguistic motor sequences such as stacking blocks appear to prime sentence structure choices in language production or comprehension (Allen et al. Reference Allen, Ibara, Seymour, Cordova and Botvinick2010; Kaiser Reference Kaiser2012). On this domain-general view, sequences in one type of action may potentiate an analogous sequence in another domain under certain task demands (Van de Cavey & Hartsuiker Reference Van der Cavey and Hartsuiker2016). Clearly, the space of such domain-general priming effects is currently poorly understood, but if structural priming proves to be emergent from broader components of action planning, then there is little reason to expect that the phenomenon offers privileged insight into grammar.
Further investigation of the domain specificity versus generality of plan reuse will therefore be critical for gaining insight not only into priming as a tool, but also into the forces that shape implicit action choices, including, but not limited to, choices of syntactic structure. The mechanisms supporting plan reuse are likely to be highly conserved across domains and species, given that nonhuman primates exhibit homologous hysteresis effects (see Weiss & Wark Reference Weiss and Wark2009). A signature characteristic of hysteresis is asymmetry, such that a transition point between implicit action choices varies with prior history. For example, in studies in which reaching targets shift gradually clockwise or counterclockwise across trials, both human children and tamarin monkeys transition from left- to right-hand use at different points depending on past targets (Rostoft et al. Reference Rostoft, Sigmundsson, Whiting and Ingvaldsen2002; Weiss & Wark Reference Weiss and Wark2009). Our ongoing work investigates whether similar perseverative asymmetries are found in both motor and language production tasks with gradual changes in parameters that promote one versus another hand/syntactic choice. If so, these findings would suggest that structural priming is a subcategory of a broader cognitive heuristic. A related opportunity to study the domain-general versus specific nature of planning is the investigation of individual differences in plan reuse across domains. For example, working memory tasks are a classic locus of individual differences in cognitive performance, and both spatial and verbal working memory loads appear to interact with hysteresis effects in motor planning (Spiegel et al. Reference Spiegel, Koester and Schack2013). Such interactions are most consistent with a domain-general planning system, and individual differences in these interactions should further constrain theoretical accounts and also inform our understanding of priming. Indeed, individual differences in structural priming (Kaschak et al. Reference Kaschak, Kutta and Jones2011a; Kidd Reference Kidd2012) seem inconsistent with B&P's claims that priming reveals grammar, which is conceived as an abstract representation with only trivial variation across a language community. As we learn more about priming, we suspect that the lessons for grammar will not be the ones B&P promote but instead will suggest that the nature of domain-general action planning has an important role in patterns of syntactic structures in language use.
Branigan & Pickering (B&P) argue that structural priming methods have “reached maturity” (target article, para. 2) to the point that they can inform not only language production and comprehension processes, but also the nature of grammar, as typically studied by linguists using different analytical tools and methods. This view appears overly optimistic; structural priming remains widely used but poorly understood, with little consensus about why the effect is observed or exactly what production and comprehension processes are promoted from prior exposure to a sentence. Moreover, the larger class of priming methods, to which B&P link structural priming, has been the target of extensive criticism and reassessment of what can be gleaned from the tasks (Cesario Reference Cesario2014). Here, we consider two perspectives on the nature of structural priming and their consequences for B&P's claims for grammar.
One perspective is that structural persistence is a strongly syntactic phenomenon: Encountering/producing a sentence somehow biases the language processing system to expect or produce a similar syntactic structure. B&P's logic is that, because the processing system draws on the grammar, patterns of priming must reveal the nature of the grammar. This thinking raises the classic issues of the competence-performance distinction. If language use is grammar + processing, there is a credit assignment problem for psycholinguistic data: Any linguistic behavior might reflect the grammar, processing mechanisms, or some combination. B&P make exactly this criticism of other psycholinguistic methods – for example, that Franck et al.'s (Reference Franck, Soare, Frauenfelder and Rizzi2010) studies of subject-verb agreement production might illuminate the nature of the grammar, or alternatively they might reflect production or comprehension processes and be uninformative about grammar. Crucially, this assignment problem applies equally to priming. Haskell et al. (2010) used priming to study agreement production and found that subject-verb agreement is sensitive to the statistical patterns in prior usage (the primes). These results could support a graded grammar in which statistical patterns shape grammatical representations (Bybee Reference Bybee2006). Researchers rejecting this approach, however, could instead attribute these priming data to processing, leaving the grammar untouched by the statistics of usage. Thus, given B&P's assumption of usage=grammar + processing, structural priming is just as much subject to interpretive uncertainty as any other measure.
Even more interpretive uncertainty arises from an alternative view of structural priming – that it is not strictly syntactic but rather a language example of a more general tendency to repeat prior actions. Cognitive models of motor planning suggest these reuse effects (termed hysteresis effects) arise because it is easier to recall a previously executed motor sequence than to generate alternative plans de novo (Rosenbaum et al. Reference Rosenbaum, Cohen, Meulenbroek, Vaughan, Latash and Lestienne2006). Our own research investigates the link between structural priming and domain-general plan reuse, and we have developed parallel tasks that yield reliable structural priming for dative sentence structures and priming of nonlinguistic manual actions in the same participants (Koranda et al. Reference Koranda, Bulgarelli, Weiss and MacDonald2016). We also observed a parallel effect of priming strength in both domains: Preferred sentences and movements are more easily primed than unpreferred ones, a phenomenon previously observed in structural priming (Bock Reference Bock1986). These findings raise the possibility that plan reuse may be a domain-general property of action planning. MacDonald (Reference MacDonald2013) suggested that a general plan reuse bias would ground patterns of language use in basic planning mechanisms, and the existence of a general plan reuse bias may also explain why some nonlinguistic motor sequences such as stacking blocks appear to prime sentence structure choices in language production or comprehension (Allen et al. Reference Allen, Ibara, Seymour, Cordova and Botvinick2010; Kaiser Reference Kaiser2012). On this domain-general view, sequences in one type of action may potentiate an analogous sequence in another domain under certain task demands (Van de Cavey & Hartsuiker Reference Van der Cavey and Hartsuiker2016). Clearly, the space of such domain-general priming effects is currently poorly understood, but if structural priming proves to be emergent from broader components of action planning, then there is little reason to expect that the phenomenon offers privileged insight into grammar.
Further investigation of the domain specificity versus generality of plan reuse will therefore be critical for gaining insight not only into priming as a tool, but also into the forces that shape implicit action choices, including, but not limited to, choices of syntactic structure. The mechanisms supporting plan reuse are likely to be highly conserved across domains and species, given that nonhuman primates exhibit homologous hysteresis effects (see Weiss & Wark Reference Weiss and Wark2009). A signature characteristic of hysteresis is asymmetry, such that a transition point between implicit action choices varies with prior history. For example, in studies in which reaching targets shift gradually clockwise or counterclockwise across trials, both human children and tamarin monkeys transition from left- to right-hand use at different points depending on past targets (Rostoft et al. Reference Rostoft, Sigmundsson, Whiting and Ingvaldsen2002; Weiss & Wark Reference Weiss and Wark2009). Our ongoing work investigates whether similar perseverative asymmetries are found in both motor and language production tasks with gradual changes in parameters that promote one versus another hand/syntactic choice. If so, these findings would suggest that structural priming is a subcategory of a broader cognitive heuristic. A related opportunity to study the domain-general versus specific nature of planning is the investigation of individual differences in plan reuse across domains. For example, working memory tasks are a classic locus of individual differences in cognitive performance, and both spatial and verbal working memory loads appear to interact with hysteresis effects in motor planning (Spiegel et al. Reference Spiegel, Koester and Schack2013). Such interactions are most consistent with a domain-general planning system, and individual differences in these interactions should further constrain theoretical accounts and also inform our understanding of priming. Indeed, individual differences in structural priming (Kaschak et al. Reference Kaschak, Kutta and Jones2011a; Kidd Reference Kidd2012) seem inconsistent with B&P's claims that priming reveals grammar, which is conceived as an abstract representation with only trivial variation across a language community. As we learn more about priming, we suspect that the lessons for grammar will not be the ones B&P promote but instead will suggest that the nature of domain-general action planning has an important role in patterns of syntactic structures in language use.