There is abundant evidence from a wide range of studies that frequency of occurrence influences language use and language development (cf. Diessel, Reference Diessel2007). In light of this evidence, the hypotheses proposed by Ambridge, Kidd, Rowland, and Theakston (this issue) should not be controversial. As they point out, “all but the most hardened classist[s]” agree that frequency of occurrence has some effect on language acquisition. However, there is no consensus among linguists and psychologists as to how the various frequency effects can be explained. In particular, it is unclear how frequency affects the language user's grammatical (or syntactic) knowledge. As I see it, there are two general strategies to explain the influence of frequency on the development of syntactic structure.
In the generative (and structuralist) approach, syntactic categories are independent of the language users' linguistic experience. They are defined prior to syntactic analysis as primitive concepts that are universally applicable to all languages (Pinker & Jackendoff, Reference Pinker and Jackendoff2005). Building on this view of syntax, some researchers have argued that the generative model of grammar should be augmented by a probabilistic component in order to account for frequency effects in sentence processing and grammar learning. Specifically, these researchers suggest that syntactic categories (and syntactic rules) should be assigned ‘probability values' that reflect the language users' experience with particular categories (and rules) and that determine their ease of activation in comprehension and production and their development in L1 acquisition. There is a large body of research demonstrating that statistically enriched grammars of this sort, often referred to as ‘stochastic grammars', can predict the likelihood of garden-path effects in sentence processing and the occurrence of certain types of mistake in children's speech (cf. Jurafsky, Reference Jurafsky1996; Chater & Manning, Reference Chater and Manning2006; Bod, Reference Bod2009).
Challenging the generative view of syntax, usage-based linguists have argued that syntactic structure is emergent from the language users' experience with language. In the usage-based approach, grammar is a dynamic system of fluid categories and flexible constraints that are shaped by general cognitive mechanisms such as categorization, priming, and automatization (or entrenchment), which are crucially driven by frequency of occurrence (cf. Bybee, Reference Bybee2010; see also Diessel, Reference Diessel2011a).
The emergence of syntactic structures can be studied in two different time frames: in language acquisition and diachronic language change (see Diessel Reference Diessel, Heine and Narrog2011b, Reference Diessel, Bergs and Brinton2012a, for discussion). Research on grammaticalization indicates that bound morphemes, grammatical categories, and constructions are derived from frequent strings of linguistic elements. For instance, a large number of studies have shown that inflectional affixes such as tense, aspect, and mood are derived from free function morphemes that have fused with adjacent content words (cf. Bybee, Perkins, & Pagliuca, Reference Bybee, Perkins and Pagliuca1994), and that (free) grammatical markers such as prepositions, auxiliaries, and determiners are based on frequent lexical expressions, notably on nouns and verbs (cf. Hopper & Traugott, Reference Hopper and Closs Traugott2003), or spatial deictics (cf. Diessel, Reference Diessel, Breban, Brems, Davidse and Mortelmans2012b), that are routinely used in particular structural positions. Moreover, it has been argued that syntactic categories for grammatical relations such as subject and object are emergent from the language users' (unconscious) analysis of verb–argument constructions with overlapping structural and semantic features (cf. Croft, Reference Croft2001), and that phrase structure categories such as NP and PP are based on recurrent sequences of linguistic expressions that have developed into ‘processing units' through ‘repetition” or ‘chunking’ (Bybee, Reference Bybee2010). Generalizing across these findings, some researchers have argued that all aspects of syntactic structure are derivative. In the usage-based approach, syntax is an ‘emergent phenomenon’ (Hopper, Reference Hopper1987) grounded in the language user's experience with frequent strings of linguistic expressions (cf. Bybee & Hopper, Reference Bybee and Hopper2001).
These strings provide the ‘input’ for L1 acquisition. A fundamental problem young children face is to segment the linguistic sequences they encounter in the ambient language and to identify the units of adult speech (cf. Jusczyk, Reference Jusczyk1997). Frequency plays an important role in this process. There is good evidence from a number of studies that young children are very sensitive to distributional regularities. They recognize transitional probabilities in recurrent strings of speech sounds, which helps them to ‘break into linguistic structure’ and to ‘unpack’ the processing units they experience in adult language (e.g., Saffran, Newport, & Aslin, Reference Saffran, Newport and Aslin1996). Moreover, a number of studies have argued that distributional regularities play an important role in the acquisition of syntactic categories (and schemas). Evidence for this hypothesis comes from corpus-based research on children's errors in spontaneous speech (e.g., Rowland, Reference Rowland2007) and from experimental research on statistical grammar learning (Aslin and Newport, Reference Aslin and Newport2012, for a review). For instance, Gómez and Gerken (Reference Gómez and Gerken1999) conducted a series of learning experiments in which they taught twelve-month-old infants an artificial language consisting of monosyllabic nonce words (e.g., vot, pel, jic) that appeared in a set of particular sentence types (defined by word order and the number of words they include). After training, the nonce words were replaced by a novel set of nonce words and the infants were tested under two conditions. In condition 1, they were exposed to the same sentence types as during training, and in condition 2 they listened to a different set of sentence types. Although the infants had not heard the words of the test sentences before, they recognized the different structural patterns in the two conditions, indicating that children as young as one year of age are able to abstract beyond strings of concrete linguistic expressions and to form abstract categories (and constructional schemas). Similar results have been obtained in several other studies with both infants and adult speakers who were tested under various conditions and with a wide range of different stimuli (cf. Aslin & Newport, Reference Aslin and Newport2012).
Taken together, this research suggests that the acquisition of syntax does not need the support of pre-established (or innate) categories and rules. On this account, frequency does not only determine the activation value of syntactic categories (and rules), but also affects the way syntactic concepts are formed and organized in mental grammar.
There is abundant evidence from a wide range of studies that frequency of occurrence influences language use and language development (cf. Diessel, Reference Diessel2007). In light of this evidence, the hypotheses proposed by Ambridge, Kidd, Rowland, and Theakston (this issue) should not be controversial. As they point out, “all but the most hardened classist[s]” agree that frequency of occurrence has some effect on language acquisition. However, there is no consensus among linguists and psychologists as to how the various frequency effects can be explained. In particular, it is unclear how frequency affects the language user's grammatical (or syntactic) knowledge. As I see it, there are two general strategies to explain the influence of frequency on the development of syntactic structure.
In the generative (and structuralist) approach, syntactic categories are independent of the language users' linguistic experience. They are defined prior to syntactic analysis as primitive concepts that are universally applicable to all languages (Pinker & Jackendoff, Reference Pinker and Jackendoff2005). Building on this view of syntax, some researchers have argued that the generative model of grammar should be augmented by a probabilistic component in order to account for frequency effects in sentence processing and grammar learning. Specifically, these researchers suggest that syntactic categories (and syntactic rules) should be assigned ‘probability values' that reflect the language users' experience with particular categories (and rules) and that determine their ease of activation in comprehension and production and their development in L1 acquisition. There is a large body of research demonstrating that statistically enriched grammars of this sort, often referred to as ‘stochastic grammars', can predict the likelihood of garden-path effects in sentence processing and the occurrence of certain types of mistake in children's speech (cf. Jurafsky, Reference Jurafsky1996; Chater & Manning, Reference Chater and Manning2006; Bod, Reference Bod2009).
Challenging the generative view of syntax, usage-based linguists have argued that syntactic structure is emergent from the language users' experience with language. In the usage-based approach, grammar is a dynamic system of fluid categories and flexible constraints that are shaped by general cognitive mechanisms such as categorization, priming, and automatization (or entrenchment), which are crucially driven by frequency of occurrence (cf. Bybee, Reference Bybee2010; see also Diessel, Reference Diessel2011a).
The emergence of syntactic structures can be studied in two different time frames: in language acquisition and diachronic language change (see Diessel Reference Diessel, Heine and Narrog2011b, Reference Diessel, Bergs and Brinton2012a, for discussion). Research on grammaticalization indicates that bound morphemes, grammatical categories, and constructions are derived from frequent strings of linguistic elements. For instance, a large number of studies have shown that inflectional affixes such as tense, aspect, and mood are derived from free function morphemes that have fused with adjacent content words (cf. Bybee, Perkins, & Pagliuca, Reference Bybee, Perkins and Pagliuca1994), and that (free) grammatical markers such as prepositions, auxiliaries, and determiners are based on frequent lexical expressions, notably on nouns and verbs (cf. Hopper & Traugott, Reference Hopper and Closs Traugott2003), or spatial deictics (cf. Diessel, Reference Diessel, Breban, Brems, Davidse and Mortelmans2012b), that are routinely used in particular structural positions. Moreover, it has been argued that syntactic categories for grammatical relations such as subject and object are emergent from the language users' (unconscious) analysis of verb–argument constructions with overlapping structural and semantic features (cf. Croft, Reference Croft2001), and that phrase structure categories such as NP and PP are based on recurrent sequences of linguistic expressions that have developed into ‘processing units' through ‘repetition” or ‘chunking’ (Bybee, Reference Bybee2010). Generalizing across these findings, some researchers have argued that all aspects of syntactic structure are derivative. In the usage-based approach, syntax is an ‘emergent phenomenon’ (Hopper, Reference Hopper1987) grounded in the language user's experience with frequent strings of linguistic expressions (cf. Bybee & Hopper, Reference Bybee and Hopper2001).
These strings provide the ‘input’ for L1 acquisition. A fundamental problem young children face is to segment the linguistic sequences they encounter in the ambient language and to identify the units of adult speech (cf. Jusczyk, Reference Jusczyk1997). Frequency plays an important role in this process. There is good evidence from a number of studies that young children are very sensitive to distributional regularities. They recognize transitional probabilities in recurrent strings of speech sounds, which helps them to ‘break into linguistic structure’ and to ‘unpack’ the processing units they experience in adult language (e.g., Saffran, Newport, & Aslin, Reference Saffran, Newport and Aslin1996). Moreover, a number of studies have argued that distributional regularities play an important role in the acquisition of syntactic categories (and schemas). Evidence for this hypothesis comes from corpus-based research on children's errors in spontaneous speech (e.g., Rowland, Reference Rowland2007) and from experimental research on statistical grammar learning (Aslin and Newport, Reference Aslin and Newport2012, for a review). For instance, Gómez and Gerken (Reference Gómez and Gerken1999) conducted a series of learning experiments in which they taught twelve-month-old infants an artificial language consisting of monosyllabic nonce words (e.g., vot, pel, jic) that appeared in a set of particular sentence types (defined by word order and the number of words they include). After training, the nonce words were replaced by a novel set of nonce words and the infants were tested under two conditions. In condition 1, they were exposed to the same sentence types as during training, and in condition 2 they listened to a different set of sentence types. Although the infants had not heard the words of the test sentences before, they recognized the different structural patterns in the two conditions, indicating that children as young as one year of age are able to abstract beyond strings of concrete linguistic expressions and to form abstract categories (and constructional schemas). Similar results have been obtained in several other studies with both infants and adult speakers who were tested under various conditions and with a wide range of different stimuli (cf. Aslin & Newport, Reference Aslin and Newport2012).
Taken together, this research suggests that the acquisition of syntax does not need the support of pre-established (or innate) categories and rules. On this account, frequency does not only determine the activation value of syntactic categories (and rules), but also affects the way syntactic concepts are formed and organized in mental grammar.