Language provides a way to map meaning onto form and vice versa. We know, from direct observation, that the form side of the mapping consists of a string of words, inflected in some languages and bare in others, arranged in a particular order. And we know, from inference, that the meaning side includes at its core a representation of a predicate and its argument(s), along with ways to express information-related contrasts and scopal relations.
We do not know how form and meaning are related to each other. It is generally assumed that the relationship is mediated by a syntactic representation that organizes words into ever-larger hierarchically organized constituents. Drawing on the hypothesis that priming is sensitive to constituent structure, Branigan & Pickering (B&P) argue for a comparatively flat syntactic structure, as illustrated in Figure 1.
Figure 1.
Flat syntactic structure.
Another idea, common in work on cognitive science, proposes a direct mapping between strings of words and semantic representations, mediated by processing operations that make no reference to traditional syntactic structure (e.g., O'Grady Reference O'Grady, MacWhinney and O'Grady2015). On this view, a sentence such as Andy reads books to children is mapped onto a semantic representation by the processing operations paraphrased in (1).
-
(1)
Andy reads books to children.
-
a.
Andy is identified as the first argument of the predicate read: (a = Andy)
-
b.
books is identified as the second argument of read (b = books)
-
c.
children is identified as the third argument of read (c = children)
In this approach, priming is the product of processing: The activation of the three operations in (1) favors their reactivation, in the same order, at the next opportunity – the classic priming effect.
One way to tease apart the two views of priming is to focus on VP constituents, whose existence is posited by the flat-structure theory (see Fig. 1) but not by the direct-mapping theory, which eschews syntactic structure. A constituent-based account of priming such as B&P's makes two predictions: (1) constituents (such as VP) should trigger priming effects; and (2) non-constituents should not. I will focus here on the latter prediction, with special attention to the subject–verb relation.
Because subjects and verbs do not form a constituent, B&P's theory predicts an absence of priming effects for this relation. In contrast, the direct mapping view predicts that there could and should be priming effects of this type, because there is a processing operation that associates a clause's first argument with the verb (see [1a]). To my knowledge, there have been no specific tests of this claim, but several candidate phenomena come to mind, of which I will mention four here.
First, it is well known that certain items manifest variation in the type of agreement they trigger in the verb. Neither is one such word (for a discussion of a similar effect with certain collectives, such as couple, see Bock et al. 2007).
-
(2)
-
a. Neither is satisfactory.
-
b. Neither are satisfactory.
If the choice of number agreement (is versus are) in these patterns could be primed by the choice made in a previous pattern with a similar type of subject, we would have prima facie evidence of priming by a subject-verb combination, contra the constituent-based theory.
A second candidate phenomenon occurs in languages that allow partial agreement with a coordinate NP subject. As illustrated in (3), verbs in Tsez can agree either with a full coordinate NP, yielding a plural form, or with just the nearest conjunct, giving a singular form. (Tsez is a Nakh-Dagestanian language spoken in the NE Caucasus; the data below are from Polinsky [Reference Polinsky2009]; abs = absolutive, pl = plural, pst = past, m = masculine,
sg
= singular.)
-
(3)
-
a. Full agreement with the subject:
-
b. Partial agreement with the subject's second conjunct:
Here again, evidence that either pattern of subject-verb agreement triggers a priming effect would create a challenge for a constituent-based theory but not for a theory based on processing operations and direct mapping.
A third phenomenon of interest involves optional subject marking, which is especially common in ergative languages. The following example is from the Tibeto-Burman language Mongsen Ao (McGregor Reference McGregor2009, p. 496). (erg = ergative, det = determiner, prs = present, decl = declarative)
-
(4) a-hən (nə) a-tʃak tʃàʔ-ə`ɹ-ùʔ
Because ergative case marks a relationship between a subject and a transitive verb, its primability would further challenge either the existence of VPs, or the viability of the view that priming faithfully reflects constituent structure, or both.
Finally, given possible independent restrictions on certain types of inflectionally triggered structural priming (Santesteban et al. Reference Santesteban, Pickering, Laka and Branigan2015), it is important to consider subject-verb relationships that do not involve this phenomenon. The scopal ambiguity illustrated in (5) is a case in point.
-
(5)
Everyone didn't finish the project.
This sentence permits two interpretations: Not everyone finished, with wide scope for the negative, and No one finished, with wide scope for the universal quantifier. Viau et al. (Reference Viau, Lidz and Musolino2010) showed that children's access to the first interpretation increases significantly when primed by exposure to similar sentences in contexts that support the not every reading. Although B&P suggest that scopal relations are represented at a special level of semantic structure, the fact they can be primed by a subject–verb combination creates a potential problem for the constituent-based account.
Priming is important in its own right and as a window into the workings of language. In the best case, it may even provide insights into the nature of syntactic representations. Crucially, however, those insights may well point to a type of sentential architecture quite different both from the widely held view and from B&P's proposal.
Language provides a way to map meaning onto form and vice versa. We know, from direct observation, that the form side of the mapping consists of a string of words, inflected in some languages and bare in others, arranged in a particular order. And we know, from inference, that the meaning side includes at its core a representation of a predicate and its argument(s), along with ways to express information-related contrasts and scopal relations.
We do not know how form and meaning are related to each other. It is generally assumed that the relationship is mediated by a syntactic representation that organizes words into ever-larger hierarchically organized constituents. Drawing on the hypothesis that priming is sensitive to constituent structure, Branigan & Pickering (B&P) argue for a comparatively flat syntactic structure, as illustrated in Figure 1.
Flat syntactic structure.
Another idea, common in work on cognitive science, proposes a direct mapping between strings of words and semantic representations, mediated by processing operations that make no reference to traditional syntactic structure (e.g., O'Grady Reference O'Grady, MacWhinney and O'Grady2015). On this view, a sentence such as Andy reads books to children is mapped onto a semantic representation by the processing operations paraphrased in (1).
(1) Andy reads books to children.
a. Andy is identified as the first argument of the predicate read: (a = Andy)
READ
<a…>
b. books is identified as the second argument of read (b = books)
READ
<a b…>
c. children is identified as the third argument of read (c = children)
READ
<a b c>
In this approach, priming is the product of processing: The activation of the three operations in (1) favors their reactivation, in the same order, at the next opportunity – the classic priming effect.
One way to tease apart the two views of priming is to focus on VP constituents, whose existence is posited by the flat-structure theory (see Fig. 1) but not by the direct-mapping theory, which eschews syntactic structure. A constituent-based account of priming such as B&P's makes two predictions: (1) constituents (such as VP) should trigger priming effects; and (2) non-constituents should not. I will focus here on the latter prediction, with special attention to the subject–verb relation.
Because subjects and verbs do not form a constituent, B&P's theory predicts an absence of priming effects for this relation. In contrast, the direct mapping view predicts that there could and should be priming effects of this type, because there is a processing operation that associates a clause's first argument with the verb (see [1a]). To my knowledge, there have been no specific tests of this claim, but several candidate phenomena come to mind, of which I will mention four here.
First, it is well known that certain items manifest variation in the type of agreement they trigger in the verb. Neither is one such word (for a discussion of a similar effect with certain collectives, such as couple, see Bock et al. 2007).
(2)
a. Neither is satisfactory.
b. Neither are satisfactory.
If the choice of number agreement (is versus are) in these patterns could be primed by the choice made in a previous pattern with a similar type of subject, we would have prima facie evidence of priming by a subject-verb combination, contra the constituent-based theory.
A second candidate phenomenon occurs in languages that allow partial agreement with a coordinate NP subject. As illustrated in (3), verbs in Tsez can agree either with a full coordinate NP, yielding a plural form, or with just the nearest conjunct, giving a singular form. (Tsez is a Nakh-Dagestanian language spoken in the NE Caucasus; the data below are from Polinsky [Reference Polinsky2009]; abs = absolutive, pl = plural, pst = past, m = masculine, sg = singular.)
(3)
a. Full agreement with the subject:
kid-no uži-n b-ik'-s.
girl-abs boy-and pl -go-pst
The girl and the boy went.
b. Partial agreement with the subject's second conjunct:
kid-no uži-n Ø-ik'-s.
girl-abs boy-and m,sg -go-pst
The girl and the boy went.
Here again, evidence that either pattern of subject-verb agreement triggers a priming effect would create a challenge for a constituent-based theory but not for a theory based on processing operations and direct mapping.
A third phenomenon of interest involves optional subject marking, which is especially common in ergative languages. The following example is from the Tibeto-Burman language Mongsen Ao (McGregor Reference McGregor2009, p. 496). (erg = ergative, det = determiner, prs = present, decl = declarative)
(4) a-hən (nə) a-tʃak tʃàʔ-ə`ɹ-ùʔ
det-chicken erg det-paddy eat-prs-decl
The chickens are eating paddy.
Because ergative case marks a relationship between a subject and a transitive verb, its primability would further challenge either the existence of VPs, or the viability of the view that priming faithfully reflects constituent structure, or both.
Finally, given possible independent restrictions on certain types of inflectionally triggered structural priming (Santesteban et al. Reference Santesteban, Pickering, Laka and Branigan2015), it is important to consider subject-verb relationships that do not involve this phenomenon. The scopal ambiguity illustrated in (5) is a case in point.
(5) Everyone didn't finish the project.
This sentence permits two interpretations: Not everyone finished, with wide scope for the negative, and No one finished, with wide scope for the universal quantifier. Viau et al. (Reference Viau, Lidz and Musolino2010) showed that children's access to the first interpretation increases significantly when primed by exposure to similar sentences in contexts that support the not every reading. Although B&P suggest that scopal relations are represented at a special level of semantic structure, the fact they can be primed by a subject–verb combination creates a potential problem for the constituent-based account.
Priming is important in its own right and as a window into the workings of language. In the best case, it may even provide insights into the nature of syntactic representations. Crucially, however, those insights may well point to a type of sentential architecture quite different both from the widely held view and from B&P's proposal.