Children’s acquisition of word order depends on syntactic/semantic role: Evidence from adjective-noun order

Abstract

Based on research on children’s verb production, Usage-Based theorists have argued that children learn grammatical abstractions in the preschool years. The fact that, in English verb clauses, word order determines semantic/syntactic roles leaves open the possibility that children are learning not just syntactic frames, but the relationship between order and semantic/syntactic roles. To clarify the nature of children’s abstract knowledge, we taught novel adjectives to English-speaking children (2 to 4 years), both prenominally and postnominally. Unlike verbs, adjective position in a sentence does not change the semantic/syntactic role of the adjective. Children showed sensitivity to the canonical order, but even four-year-olds frequently used novel adjectives postnominally. We argue that a strong motivation for ordering words grammatically is when order determines semantic/syntactic roles.

Keywords

adjective-noun strings language acquisition lexical categories semantic roles syntactic roles usage-based theory word order

In many languages, syntactic and semantic relations are expressed through word or morpheme order. Even in languages that allow multiple phrasal orders, different word orders often reflect different pragmatic motivations (Ocampo, 1991). Mastering word order is essential for children to process language for meaning. From the time children first start to put two words together in spontaneous speech (and perhaps earlier in comprehension – see Hirsh-Pasek & Golinkoff, 1996; Naigles, 2002), they almost always order the words according to grammatical rules of their input language (Brown, 1973; Wexler & Culicover, 1980). For example, English-speaking children often start to talk about possession with constructions like ‘Mommy sock’ that correspond to the typical English word order (Bloom, 1970). What kind of knowledge do children have that allows them to order their words according to grammatical rules of the input language? Until recently, the answer has often been that children must have access to some kind of abstract lexical categories before they start to produce word combinations. These abstract lexical categories could be grammatical, such as noun and verb (Pinker, 1984; Wexler & Culicover, 1980), or semantic, such as object and action (Dromi, 1987), or possessor and possessed (Bloom, 1970).

Young children seem to master syntactic frames, or orders of specific words, early (Abbot-Smith & Tomasello, 2010; Naigles, 2002; Naigles, Gleitman, & Gleitman, 1993; Tomasello, 2003; see also Mintz, 2003). Preschool children generally interpret ungrammatical sentences like ‘The zebra goes the lion’ with the meaning implied by the syntactic frame (i.e., the zebra makes the lion go) rather than the verb (i.e., the zebra goes to the lion; Naigles et al., 1993). From at least the age of two years, children can infer the meaning of a novel word on the basis of syntactic cues (Bélanger & Hall, 2007; Diesendruck, Hall, & Graham, 2006; Hall, Lee, & Bélanger, 2001; Hirsh-Pasek & Golinkoff, 1996; Klibanoff & Waxman, 1999; Taylor & Gelman, 1988). Some researchers have argued that children’s focus on order could be part of their pre-syntactic knowledge, either through representation (e.g., Pinker, 1984) or through cognitive processes (e.g., Slobin, 1973). Alternatively, children’s focus on word order could emerge as they reach a critical mass in vocabulary and need to structure their lexically based knowledge for multiple-word constructions (Bates & Goodman, 1999; Marchman & Bates, 1994; Nowak, Plotkin, & Jansen, 2000).

More recently, researchers working in the framework of a Usage-Based theory of grammar (Abbot-Smith, Lieven, & Tomasello, 2008; Goldberg, 2006; Tomasello, 2000a, 2000b, 2003) have argued that children may have little underlying abstract knowledge about grammar when they start producing linguistic constructions. Instead, children start by using words in set utterances, such as ‘Here you go’ and ‘Where’s Daddy’, eventually learning to make minor changes to those utterances by substituting single words in set phrases such as ‘Where’s X’ (Abbot-Smith & Tomasello, 2010; Lieven, Behrens, Speares, & Tomasello, 2003; Tomasello, 1992, 2003). With additional exposure to multiple exemplars, children can generalize according to abstract patterns (such as Subject-Verb-Object; see, for example, Savage, Lieven, Theakston, & Tomasello, 2003; see also Naigles, Fowler, & Helm, 1992). Under this framework, previous evidence showing that children are usually correct in ordering their words would be due to the fact that these studies have examined children’s spontaneous speech, rather than how children use novel words in combination (Naigles, 2002).

There is some compelling evidence to support a Usage-Based Theory. For example, Tomasello (1992) showed that his English-speaking daughter’s earliest uses of verbs were frequently very similar to previous uses she had made of the same verb (see also Goldberg, Casenhiser, & Sethuraman, 2004; Lieven et al., 2003). Other researchers have taught children verbs both in an order that corresponds to their input language and in an order that differs from the canonical order. Akhtar (1999) taught novel verbs in one of three word orders (SVO, SOV, or VSO) to English-speaking children aged two to four years. She showed that the two-year-olds used novel verbs in a non-SVO order 53% of the time (i.e., they corrected the novel verbs to SVO in 47% of their constructions), three-year-olds used a majority of SVO order (59% of their constructions) and four-year-olds almost never used an ungrammatical order. In contrast, when tested with familiar verbs, even the two-year-olds rarely used a non-SVO order (see also Abbot-Smith et al., 2008; Abbot-Smith, Lieven, & Tomasello, 2001; Naigles et al., 1992). Similar results were obtained from a study in which researchers chose existing verbs that varied in frequency. Matthews, Lieven, Theakston, and Tomasello (2005) tested preschool children on some high-frequency verbs and some low-frequency verbs, using some in a non-SVO order and some in the canonical order. The children were more willing to use the low-frequency verbs than the high-frequency verbs in a non-SVO order. Taken together, these results suggest that two-year-olds do not yet have abstract knowledge about English phrasal ordering, but they have some knowledge about where familiar (high-frequency) verbs go. Between two and four years of age, children gradually shift to using almost exclusively the canonical word order with novel words. By four years of age, children may start to abstract grammatical knowledge (rather than only surface knowledge) about word order and generalize to novel examples (Abbot-Smith & Tomasello, 2010; Naigles et al., 1993; Tomasello, 2003). Note that the claim here is not that children have no access to abstract grammatical knowledge before four years of age, but rather that grammatical knowledge becomes increasingly abstract over the preschool years.

Most of the research that has been done in a Usage-Based theoretical framework has focused on verb clauses. In verb clauses (in English, at least, the language most of the research has focused on), the word order corresponds to the semantic/syntactic role. That is, a noun preceding a verb in English is almost always the subject of the verb or the agent of the action. In fact, English constructions in which this basic order is not followed are often notoriously difficult for children, such as Object-Verb-er compounds like can opener (Clark, Hecht, & Mulford, 1986) and passive sentences (Koff, Kramer, & Fowles, 1980; Lempert, 1978). Thus, by focusing on verb clauses, it is not clear whether the observed changes in children’s usage between two and four years is due to increasing sensitivity to syntactic frames alone (Mintz, 2003), increasing sensitivity to semantic/syntactic roles (Braine & Brooks, 1995), or both (Tomasello, 2003).

There is extensive evidence showing that prelinguistic infants can abstract and predict the order of elements in novel strings (Gomez & Gerken, 1999; see review and discussion in Naigles, 2002). Mintz (2003) has shown that child-directed speech contains many frequent syntactic frames, which may allow children to predict which words could be slotted into those frames (see also Tomasello, 2003). These kinds of syntactic frames are a level of abstraction that is available to children very early in life (Perruchet & Pacton, 2006). So what kind of abstraction are children acquiring later, in the preschool years? Notably, Akhtar (1999) found that the rate of children’s correction to English sentential word order increased with age over the preschool years. One possible explanation of her finding is that preschool children gradually learn the semantic/syntactic roles of subject, verb, and object relative to order. A second possibility is that they simply learn that noun-verb-noun sentence frames are preferred in English.

The purpose of the present study was to try to elucidate the nature of children’s underlying grammatical knowledge, primarily between the ages of two and four years. We did this by using a weird word order paradigm (following Akhtar, 1999) with adjective-noun constructions. Unlike in verb clauses, the order of an adjective relative to a noun does not change the semantic/syntactic category of the adjective. For example, if someone said ‘A dog black walked on my lawn yesterday,’ a native English speaker would undoubtedly identify the sentence as ungrammatical but would be able to recover the intended meaning. Children who are in the process of learning word order and semantic/syntactic categories might similarly focus on their comprehension of the sentence as a whole rather than the order. In contrast, in a verb clause, the placement of a noun signifies its syntactic role (e.g., whether it is the subject or object of the verb) and therefore also affects the meaning of the sentence. If someone said, ‘The dog chased the cat’ when he or she meant, ‘The cat chased the dog’, there would be no way for a listener to recover the intended meaning on the basis of the sentence alone. If, between the ages of two and four, children’s production becomes increasingly sensitive to the underlying abstract syntactic frames in their language (Tomasello, 2003), then we would expect them to correct weird adjective order as often as they correct weird verb order (Akhtar, 1999). If the changes seen between two and four years of age are also due to children’s increasing sensitivity to how word order affects semantic/syntactic roles (Naigles, 2002), then children in this age range may still be willing to use adjective-noun combinations in noun-modifier constructions in a non-canonical order, since order is less linked to semantic role in adjective-noun constructions than it is in verb constructions.

English adjective placement and acquisition

Usage-Based theories rely on the surface ordering of words. Children are thought to acquire word order within constructions (Goldberg, 2006; Tomasello, 2003). A corollary relevant to the present studies is that children would learn the order of adjective-noun constructions (e.g., the black dog) independently of predicate adjectives (e.g., the dog is black) (Diesendruck et al., 2006; Nelson, 1976; Saylor, 2000). These studies have shown that children know that predicate adjectives, in contrast with attributive adjectives, tend to refer to temporary qualities that may or may not contrast with other individuals of the same kind. For example, the predicate phrase, ‘that dog is black’ does not imply the presence of other, different-colored dogs, whereas the attributive phrase, ‘look at the black dog’, does. The present studies focused exclusively on what children do with adjectives in noun-modifier (i.e., attributive adjective) constructions.

In English, the canonical surface order for simple adjectives in modifier-noun constructions is prenominal (as in big car). In some cases, modified adjectives can occur postnominally (as in hair whiter than snow). Also, some adjectives are used postnominally, usually adjectives borrowed from French (as in the wine extraordinaire) or with some quantifiers (as in something blue). Note that, like many other researchers, we assume that any change in order will have an effect on meaning (Diesendruck et al., 2006; Saylor, 2000). The key component of adjective-noun ordering for our study is that the use of an adjective before or after a noun changes neither the semantic nor syntactic function of either word. This point is true even of predicative adjectives. That is, using an adjective as a predicate clearly changes the nuance of the meaning, but it does not change the fact that it is an adjective and functions to describe the noun. In contrast, much previous research has focused on verb constructions, where a change in the order changes its semantic or syntactic function.

There is little research on children’s acquisition of the order of adjective-noun constructions. Most acquisitional work has been on children’s understanding of the semantics of adjectives and/or how the syntactic frame determines children’s understanding of the semantic category of an adjective (Akhtar, 2002; Akhtar & Montague, 1999; Diesendruck et al., 2006; Hall & Moore, 1997; Klibanoff & Waxman, 2000). However, the extant evidence on children’s spontaneous speech has shown that children are usually accurate in their ordering of adjectives and nouns. For example, Brown (1973) observed that English-speaking children used adjectives and nouns in the correct order, except when the copula was thought to be missing in a sentence (Nelson, 1976; Saylor, 2000; Tomasello, 1992). One elicitation study with children between three and five years of age showed that monolingual English-speaking children made less than 5% errors in adjective-noun order with familiar adjectives like long (Nicoladis, 2006). Even French-English bilingual children, who have to learn two different rules for adjective placement, order adjectives over 90% correctly in spontaneous speech in English and over 90% correctly in an elicitation task in English from at least the age of two and a half (Nicoladis, 2002, 2006). For French-English bilingual children, the correct ordering holds true even for code-mixed utterances;¹ that is, they follow adult norms in ordering adjectives from one language relative to nouns from the other language from the age of two years on (Paradis, Nicoladis, & Genesee, 2000).

On the basis of these studies, it would seem that the acquisition of either one or two adjective-noun orders is a trivial problem for children and that children are likely to order adjectives and nouns correctly. However, it should be noted that all of these studies concerned adjectives that children already knew. And, as we showed from the literature review above, children as young as two years are more likely to correctly order familiar words. In other words, data from familiar words may overestimate children’s abstract knowledge of canonical word order. For example, Akhtar (1999) showed that children were less likely to use the novel word order of a verb-noun construction with a familiar verb than with a novel verb (see also Matthews et al., 2005; Tomasello, 2000a). This is not to say that studies with novel words are more accurate than studies with familiar words. In fact, tasks such as Akhtar’s novel word order task may underestimate children’s knowledge because they rely on what children do on a small number of items in a short period of time. To determine children’s knowledge, then, it is necessary to have evidence from both familiar words and novel words.

Weird word order paradigm

The purpose of this experiment was to test English-speaking children’s knowledge of adjective-noun order. We used Akhtar’s (1999) weird word order methodology where children are taught novel words in both the correct (or default) order of their language and in at least one alternative order. Within adjective-noun constructions, there is only one other possible order (i.e., in English, postnominal). We presented children with play scenarios in which an experimenter taught them novel adjectives, two of which were presented in the canonical (prenominal) position and two of which were presented in the non-canonical (postnominal) position. Adjective-noun constructions usually occur in spontaneous speech to contrast one object with another from the same class. To elicit constructions with novel adjectives, we asked children to select an object when a number of contrasting objects from the same class were available (e.g., ‘which one would you like to play with?’). Children’s spontaneous use of adjective-noun constructions with the new words was recorded. Because the context of this experiment was a game, children may have been willing to go along with non-canonical word order for the purposes of the game. In other words, this paradigm probably underestimated children’s ability to produce the correct word order. For this reason, we taught children novel adjectives in both the canonical and novel orders, and compared their usage of both orders within the play scenario. In order to investigate the possible role of semantic/syntactic category in acquisition, we discuss how our results compare with those of Akhtar (1999), who used a similar method to research children’s use of novel verbs.

The present experiment was designed to explore three effects on children’s use of word order: (1) familiarity; (2) canonical order independent of semantic/syntactic role; and (3) age. We were able to compare our findings to previous studies to speculate about the effects of semantic/syntactic role in learning word order.

First, as noted earlier, there is evidence that the greater children’s experience with specific words, the less likely they are to mis-order those words. To explore this familiarity effect with adjective-noun constructions, children in the present study heard a familiar adjective (‘green’) in the postnominal position, in addition to novel adjectives. Based on previous findings with verbs, we predicted that children’s willingness to use an unconventional word order with a familiar word would decrease with age (e.g., Akhtar, 1999).

Second, to explore the effect of children’s use of canonical order independent of semantic/syntactic function, we compared their production of novel postnominal adjectives and prenominal adjectives. If children have an abstract representation of adjective-noun constructions, then we would expect them to reverse the order in the postnominal condition more often than in the prenominal condition. Finally, if children become increasingly sensitive to canonical order as they get older (Tomasello, 2003), then we may see an increase in children’s corrections to canonical order with age, particularly between two and four years. By four, children should almost never use the non-canonical order (cf. Akhtar, 1999). Alternatively, if three- to four-year-old children’s avoidance of non-SVO order is additionally due to their increasing sensitivity to the interaction between word order and syntactic/semantic role, then we may find little effect of age. In this case, even the four-year-olds might be willing to use the weird word order, as it does not violate the semantic function of the adjective.

Note that, as implied in the last paragraph, we are not directly testing the effect of syntactic/semantic function. To reach conclusions about its effect, we rely on comparisons with other studies on phrasal ordering using the same methodology (e.g., Akhtar, 1999).

Method

Participants

Thirty-five English-speaking children between two and four years of age participated in this study. The children were divided into three age groups, which we call for the sake of convenience with reference to the average age of the children ‘two-year-olds’, N = 12, range = 2;2–3;2, M = 2;7, 6 girls; ‘three-year-olds’, N = 12, range = 3;4–4;1, M = 3.8, 7 girls; and ‘four-year-olds’, N = 11, range = 4;2–4;9, M = 4;5, 7 girls. Children were tested individually in a quiet space in their daycare centre.

Materials

Children were presented with a farm play-mat which included a barn, a pond, some trees, and some chickens. Each of four novel adjectives and the control adjective green were taught using a different set of toys. Each of the novel adjectives had a clear meaning which is described in Table 1. We chose to create novel adjectives with the ending ‘-ish’ because of evidence that English-speaking children can understand this suffix as an adjectival marker before they are two years old (Waxman & Booth, 2001; Waxman & Klibanoff, 2000). For every object described by the novel adjective, there was at least one other object lacking the property to provide contrast. For example, for the adjective blickish, children were shown a duck that had the target property (a duck that was half pink and half yellow), as well as another duck that lacked the property (it was just yellow). Klibanoff and Waxman (2000) found that preschoolers more successfully map adjectives onto properties when the adjectives are taught in the context of contrasting properties.

Table 1.

Novel adjectives, their meaning and the objects that were paired with them.

Adjective	Meaning	Objects
groffish	legless	cow, horse, pig, chicken
strivvish	on skis	truck, car, motorcycle
blickish	two-colored	ball, star, duck, pencil
drackish	winged	turtle, frog, dolphin
green (control)	green	ball, dinosaur, pencil

Procedure

The adjectives were presented to children in a free play setting. The experimenter began by presenting objects that did not have the target property (e.g., ‘Here’s a black and white cow. Here’s a small cow’.). Then the experimenter presented the novel adjective paired with the object name and defined the adjective (e.g., ‘Here’s a groffish cow. It’s groffish because it has no legs’.). Experimenters used the adjective in several different sentence frames and contexts so that children would understand that the new words were adjectives (e.g. ‘This cow isn’t groffish; the groffish cow is sitting in the pool’.). To ensure that the children understood what was meant by each novel adjective, the experimenters posed a number of questions, such as, ‘Is this a groffish cow?’ and, ‘Which one of these is groffish?’ The experimenters had no set script to elicit adjective-noun constructions and were asked to concentrate on questions involving picking out a specific example of the objects. As a result, children were encouraged to produce adjectives in as natural a way as possible (e.g., in response to the question, ‘Which one do you want to play with now?’ or, ‘Which one should go in the pond now?’). Children sometimes produced the relevant constructions without any prompting at all. Note that experimenters and children were not limited to producing adjectives paired with the names of objects in Table 1. They could also refer to a superordinate category (e.g., groffish animals or groffish toys), replace the noun with one (e.g., groffish ones) or talk about non-present objects.

Each child heard four novel adjectives, the order of which was counterbalanced across children. Each child learned two adjectives in the postnominal position, and two adjectives in the prenominal position. Four different experimenters performed the task, each of whom learned only one set of adjective orders, so as not to mix up adjective ordering during the experiment. This was important, so that each adjective condition would have an equally confident experimenter. The control adjective, green, was always presented last, and always in the postnominal position (e.g., ‘a dinosaur green’). Each child was videotaped during the task.

Input

The data were collapsed across the two novel prenominal adjective trials and the two novel postnominal adjective trials for all analyses. Adjectives used in a predicate construction (e.g., ‘This cow is not groffish’) were excluded from the analyses. To check that input did not differ significantly across the two adjective orders, we examined type and token rates of use of pre- and postnominal adjective combinations in the experimenters’ input. To count types, plurals and diminutives were considered a single type (e.g., groffish pig, groffish piggy and groffish pigs were all considered a single type). There were no significant differences in the number of prenominal and postnominal orders in the input, either for tokens (prenominal mean = 30.5, postnominal mean = 30.8, t (34) = .26, p > .50) or types (prenominal mean = 7.1, postnominal mean = 7.1, t (34) = .09, p > .50).

Coding

We counted the number of adjective-noun constructions in both orders (e.g., a groffish cow, a cow groffish) that each child produced in the play session. The number of these constructions produced by children from each age group is summarized in Table 2. No child ever used any of the adjectives in Table 1 in a predicate construction.

Table 2.

Mean (SD) total number of AN or NA constructions by condition and age group.

	Prenominal	Postnominal
2-year-olds	6.1 (5.8)	3.4 (3.3)
3-year-olds	7.4 (7.4)	9.7 (8.5)
4-year-olds	10.2 (6.8)	13.2 (7.7)

Each prenominal and postnominal construction was coded. Our coding scheme deviated from Akhtar’s (1999), in that we included imitations in our counts of children’s usage of adjective-noun ordering. We included imitations for two reasons. First, from a Usage-Based perspective, imitations may indicate something about what children are willing to say, since they do not have to say anything at all (Tomasello, 2003). Second, particularly among the youngest children, a large proportion of their responses were imitations. Excluding them would entail excluding a great deal of data. In some analyses, we present imitations as a separate category.

The coding system included three mutually exclusive codes: imitation, extension and reversal. Imitation was the repetition of the novel adjective with the same noun in the same order as a researcher had used at least once, even if it was several minutes later. Extension referred to an extension of the modeled order to another noun. For example, if a researcher said ‘a groffish cow’, the child was counted as extending if he/she said ‘a groffish horse’. Reversal referred to a change of order from the one used by the researcher, for example, if a researcher said ‘a groffish cow’ and a child said either ‘a cow groffish’ or ‘a horse groffish’, this was considered a reversal. For some analyses, we present the data in terms of matches and mismatches to the modeled order. Matches include both imitations and extensions under our coding scheme and mismatches are reversals.

For each child we calculated the proportion of imitations, extensions and reversals out of the total number of responses for each adjective order (i.e., prenominal or postnominal). Some children did not ever use the novel adjectives in combination with nouns in a particular order. These children were excluded from analyses for that condition (for the prenominal adjectives, three two-year-olds and two three-year-olds were excluded; for the postnominal adjectives three two-year-olds and two three-year-olds were excluded; for the control adjective, four two-year-olds, three two-year-olds and one four-year-old were excluded). For that reason, different analyses are based on different numbers of children.

Results

The average proportion of matches in the prenominal condition for all the children was .99 (SD = .03) while the average proportion of matches in the postnominal condition was .72 (SD = .32). The average proportion of matches for the control adjective was .17 (SD = .33). A three by three between-within ANOVA with age group (between-subjects: two, three, four years) and condition (within-subjects: novel prenominal, novel postnominal, familiar postnominal) revealed a significant effect of condition, F (2, 40) = 97.32, p < .001, $η_{p}^{2}$ = .82. There was no interaction between condition and age, F (4, 40) = 1.69, p > .15, nor was there a main effect of age, F (2, 20) = 1.37, p > .25.

Pairwise contrasts confirmed that the children’s rate of matching utterances with novel prenominal adjectives was significantly higher than with novel postnominal adjectives, t (26) = 4.60, p < .001. The proportion of matches in both experimental conditions was significantly higher than in the control condition, (novel prenominal adjectives, t (24) = 6.84, p < .001, d = 3.64; novel postnominal adjectives, t (22) = 18.89, p < .001, d = 1.72).

Response type

Figure 1 summarizes the average proportion of response types (i.e., imitation, extension, reversal) by condition. Preliminary analyses revealed no main effects of age group or interactions with age for any response type, so age was removed as a factor from the following analyses. Three repeated-measures ANOVAs examined the differences between the relative proportion of each type of response across three adjective conditions (novel prenominal, novel postnominal, familiar postnominal). Each of these analyses found a strong effect of adjective condition on the proportion of each response type, imitation: F (2, 42) = 66.81, p < .001, $η_{p}^{2}$ = .761 extension: F (2, 42) = 7.08, p < .01, $η_{p}^{2}$ = .252; reversal: F (2, 42) = 95.62, p < .001, $η_{p}^{2}$ = .820. Pairwise comparisons revealed that children were less likely to imitate postnominal than prenominal novel adjectives, t (26) = 4.71, p < .001, d = .92, and even less likely to imitate the postnominal familiar control adjective green, t (24) = 5.03, p < .001, d = 1.56. In terms of extensions, the only significant difference was between the prenominal adjectives and the control adjective, t (22) = 3.85, p = .001, d = .54. Children were more likely to reverse postnominal than prenominal novel adjectives, t (26) = 4.60, p < .001, d = 1.21, They were also more likely to reverse the familiar control adjective than the novel postnominal adjective, t (24) = 6.84, p < .001, d = 1.69, and the novel prenominal adjective, t (22) = 18.89, p < .001, d = 2.92.

Figure 1.

Proportion of children’s responses that were categorized as imitations, extensions and reversals by condition in Experiment 1. Error bars depict one standard error.

Age effects

The preceding analyses revealed no significant effects of age group on any of our dependent variables. To investigate more closely whether age had any effect on children’s treatment of the novel and familiar adjectives, we performed correlations on the children’s age in months with their proportion of matching responses in each condition. We found a significant negative correlation between age and children’s proportion of matching utterances to the familiar postnominal (control) constructions, r (33) = -.41, p < .05. There was no significant correlation between age and the rate of matching novel prenominal adjectives, r (33) = -.18, p > .05, or novel postnominal adjectives, r (30) = .23, p > .05.

Discussion

The purpose of these experiments was to test three effects on children’s use of word order: (1) familiarity; (2) canonical order independent of semantic/syntactic role; and (3) age. We discuss the results of these experiments in light of the kind of abstract knowledge children might acquire over the course of the preschool years.

Familiarity with the word clearly played an important role in children’s willingness to use a word in a nonconventional order. Children were more likely to reverse the experimenter’s weird order of a familiar adjective (green) than to reverse the weird order of a novel adjective. Further, children’s willingness to use the familiar adjective in a non-canonical order decreased with age. These results replicate previous results with verb clauses, in which children were less likely to allow non-canonical order for a known verb than a novel verb (Akhtar, 1999; Matthews et al., 2005). These results lend further support to the argument that children’s familiarity with a word is an important determinant in their usage (Tomasello, 2003).

The canonical order of adjectives in English also played an important role. Children used more novel prenominal adjectives than postnominal adjectives. This result suggests that even very young children know the correct placement of adjectives in English (Tomasello, 2003). That is, children have some abstract knowledge about adjective-noun constructions. This level of abstraction could be due to having learned the conventional English syntactic frame for adjective placement (Mintz, 2003; Perruchet & Pacton, 2006). This result is in the same direction as findings with novel verbs showing that children as young as two years of age prefer the canonical order of their language (Abbot-Smith et al., 2001; Hirsh-Pasek & Golinkoff, 1996).

The present study revealed no effect of age on children’s willingness to use novel adjectives in a postnominal position. This result contrasts markedly with Akhtar’s (1999) finding that there is a clear negative relationship between age and children’s willingness to use a non-canonical verb phrase order. Even our four-year-olds averaged 59% matches with postnominal order.

We can think of two alternate accounts for the absence of age effects in children’s willingness to use non-canonical adjective orders. One possibility is that the high rate of matches to postnominal order was due to children’s willingness to switch adjectival order for the duration of the game. However, using similar methodology, Akhtar (1999) showed that four-year-olds almost never (less than 5% of their utterances) used unconventional phrasal ordering with verbs. It seems unlikely that four-year-olds would be willing to switch to an unconventional order for the purposes of a game with adjectives but not with verb phrases.

A second possibility has to do with the fact that, unlike Akhtar, we included imitations in our calculation of matches. This may have led to an inflated estimate of the children’s use of postnominal adjective constructions. However, even if we remove the imitations and just look at extensions and reversals, four-year-olds still averaged 41% matches with postnominal adjective order, still far above the rate of matches for the unconventional order observed with verbs.

Thus, the results of the present study differ in two important ways from previous results with verb clauses; we attribute these differences to the difference in semantic/syntactic function of adjective and verbs. First, we found no effect of age on children’s avoidance of non-canonical ordering. If children are increasingly likely to abstract patterns from an increased set of exemplars, children should be more likely to avoid non-canonical ordering as they get older (Tomasello, 2003). In fact, the results with verb clauses support the idea of increasing abstraction with age (e.g., Abbot-Smith et al., 2001; Akhtar, 1999). In contrast, the present results with children between two and four years of age showed no effect of age on their avoidance of non-canonical ordering. Second, where the results of this study depart from previous results with verb clauses is in the three- and four-year-olds’ degree of willingness to use the non-canonical order. For verb clauses, the majority of three- and four-year-olds did not use the non-canonical word order (Akhtar, 1999). In the current study, however, the majority of even four-year-olds’ productions of the novel adjectives presented in non-canonical order matched the order in which they were presented. This result suggests then that preschool children may simply be less invested in the correct ordering of novel adjective phrases than in that of novel verb clauses. We suggest that the reason for this difference is that the word order in verb clauses determines the semantic/syntactic role of the nouns. In contrast, the semantic/syntactic role of an adjective is not as dependent on its position relative to the noun. We recognize that the meaning of an adjective changes depending on order (e.g., attributive vs. predicative; Diesendruck et al., 2006), but the meaning change is subtle and does not affect the principal role of the adjective. In other words, these results suggest that children are more likely to attend to word order when a change in order makes a clear difference in semantic/syntactic role, as many researchers have suggested (e.g., Naigles, 2002; Slobin, 1973; see also Golinkoff & Markessini, 1980).

In arguing for the importance of semantic/syntactic role in children’s interest in canonical ordering, we have implicitly rejected some other possible interpretations. We next consider two alternative possibilities. One alternative explanation is that these results are due to methodological differences between our study and Akhtar’s (1999). While we retained the spirit of Akhtar’s method by teaching children novel adjectives in a naturalistic context, there were nonetheless methodological differences between the studies. For example, in this study, we used different elicitation questions than Akhtar did, in order to elicit adjective-noun constructions in the most naturalistic way possible (see Chang, Kobayashi, & Amano, 2009, for discussion of social effects in these kinds of experiments). We also used novel adjectives but did not use novel nouns. Children’s sensitivity to non-canonical word order should be tested with a variety of novel words. It will be important for future research to investigate which methodological parameters (e.g., the elicitation questions, the number of items, the number of exemplars in the learning phase, the lexical category of the novel words, the number of novel words, etc.) make a difference in children’s performance. While we cannot rule out the possibility that methodological differences between our study and Akhtar’s are responsible for the difference in results, it is not clear how these methodological differences could have resulted in the different pattern of results obtained.

Another possible interpretation of the present results is that children attend more to the order of verb clauses because verbs are required to make a grammatical sentence while adjectives are optional. We have collected data from three- and four-year-olds’ use of non-canonical ordering of another optional construction in English, that is, novel noun-noun compounds (Moroschan & Nicoladis, under review). As in adjective phrases, in English noun-noun compounds the first word modifies the second. We found that children were significantly less likely to use the non-canonical word order for novel nouns as modifiers than for novel adjectives as modifiers. In combination with the current data, these results support our conclusion that what is really at play is the effect of phrasal ordering on meaning. In noun-noun compounds, unlike adjective phrases, the order of the words makes a substantial difference to the meaning of the phrase (e.g., compare door factory to factory door). Unlike noun-noun compounds, and unlike verb phrases, adjective order does not determine the syntactic/semantic role of adjectives. These results support our interpretation that children are more likely to revert to the canonical order when the word order determines the syntactic/semantic role.

In sum, we have replicated previous studies’ results showing that from the age of two years on, children are sensitive to the canonical order in their language (e.g., Abbot-Smith et al., 2001). We have also replicated children’s insistence on using familiar words in the canonical order from the age of two years on (e.g., Akhtar, 1999), suggesting that they have access to abstract syntactic frames from at least two years of age. We have argued here that between three and four years of age, children become more sensitive to the semantic/syntactic roles of words relative to word order. These results underscore that children care about word order when the order makes a difference for meaning (Naigles, 2002). For this reason, they correct non-canonical verb clauses but are far less likely to correct non-canonical adjective phrases. Future research could focus on direct tests of this interpretation.

Footnotes

Acknowledgements

The directors of daycares, parents and children were all very helpful in encouraging participation. Alex Yiu, Elaine Greidanus, Devin Bruce and Monique Bunyan helped with the testing. Thanks to Johanne Paradis and Paula Marentette for their feedback on an earlier version of this paper. Samuel Navarro, Chris Sturdy, and Marcia Spetch were helpful in pointing us toward useful references.

Notes

References

Abbot-Smith

Lieven

Tomasello

(2001). What preschool children do and do not do with ungrammatical word orders. Cognitive Development, 16, 679–692.

Abbot-Smith

Lieven

Tomasello

(2008). Graded representations in the acquisition of English and German transitive constructions. Cognitive Development, 23, 48–66.

Abbot-Smith

Tomasello

(2010). The influence of frequency and semantic similarity on how children learn grammar. First Language, 30, 79–101.

Akhtar

(1999). Acquiring basic word order: Evidence for data-driving learning of syntactic structure. Journal of Child Language, 26, 339–356.

Akhtar

(2002). Relevance and early word learning. Journal of Child Language, 29, 677–686.

Akhtar

Montague

(1999). Early lexical acquisition: The role of cross-situational learning. First Language, 19, 347–358.

Bates

Goodman

J. C.

(1999). On the emergence of grammar from the lexicon. In MacWhinney

(Ed.), The emergence of language (pp. 29–79). Mahwah, NJ: Lawrence Erlbaum.

Bélanger

Hall

D. G.

(2007, March). Origins of lexical category to meaning links: The count noun/proper name distinction. Poster presented at the biennial meeting of the Society for Research in Child Development. Boston, MA.

Bloom

(1970). Language development: Form and function in emerging grammars. Cambridge, MA: MIT Press.

10.

Braine

M. D.

Brooks

P. J.

(1995). Verb argument structure and the problem of avoiding and overgeneral grammar. In Tomasello

Merriman

W. E.

(Eds.), Beyond names for things: Young children’s acquisition of verbs (pp. 353–376). Hillsdale, NJ: Lawrence Erlbaum.

11.

Brown

(1973). A first language: The early stages. Cambridge, MA: Harvard University Press.

12.

Chang

Kobayashi

Amano

(2009). Social factors in the acquisition of a new word order. First Language, 29, 427–445.

13.

Clark

E. V.

Hecht

B. F.

Mulford

R. C.

(1986). Coining complex compounds in English: Affixes and word order in acquisition. Linguistics, 24, 7–29.

14.

Diesendruck

Hall

D. G.

Graham

S. A.

(2006). Children’s use of syntactic and pragmatic knowledge in the interpretation of novel adjectives. Child Development, 77, 16–30.

15.

Dromi

(1987). Early lexical development. Cambridge: Cambridge University Press.

16.

Goldberg

A. E.

(2006). Constructions at work: The nature of generalization in language. Oxford: Oxford University Press.

17.

Goldberg

A. E.

Casenhiser

D. M.

Sethuraman

(2004). Learning argument structure generalizations. Cognitive Linguistics, 15, 289–316.

18.

Golinkoff

R. M.

Markessini

(1980). ‘Mommy sock’: The child’s understanding of possession as expressed in two-noun phrases. Journal of Child Language, 7, 119–135.

19.

Gomez

R. L.

Gerken

(1999). Artificial grammar learning by 1-year-olds leads to specific and abstract knowledge. Cognition, 70, 109–135.

20.

Hall

D. G.

Lee

Bélanger

(2001). Young children’s use of syntactic cues to learn proper names and count nouns. Developmental Psychology, 37, 298–307.

21.

Hall

D. G.

Moore

C. E.

(1997). Red bluebirds and black greenflies: Preschoolers’ understanding of the semantics of adjectives and count nouns. Journal of Experimental Child Psychology, 67, 236–267.

22.

Hirsh-Pasek

Golinkoff

R. M.

(1996). The origins of grammar: Evidence from early language comprehension. Cambridge, MA: MIT Press.

23.

Klibanoff

R. S.

Waxman

S. R.

(1999). Syntactic cues to word meaning: Initial expectations and the development of flexibility. In Greenhill

Littlefield

Tano

(Eds.), Proceedings of the 23rd Annual Boston University Conference on Language Development (pp. 361–372). Somerville, MA: Cascadilla Press.

24.

Klibanoff

R. S.

Waxman

S. R.

(2000). Basic level categories support the acquisition of novel adjectives: Evidence from preschool-aged children. Child Development, 71, 649–659.

25.

Koff

Kramer

P. E.

Fowles

(1980). Effects of event probability and animateness on children’s comprehension of active and passive sentences. The Journal of Psychology, 104, 157–163.

26.

Lempert

(1978). Extrasyntactic factors affecting passive sentence comprehension by young children. Child Development, 49, 694—699.

27.

Lieven

Behrens

Speares

Tomasello

(2003). Early syntactic creativity: A usage-based approach. Journal of Child Language, 30, 333–370.

28.

Marchman

V. A.

Bates

(1994). Continuity in lexical and morphological development: A test of the critical mass hypothesis. Journal of Child Language, 21, 339–366.

29.

Matthews

Lieven

Theakston

Tomasello

(2005). The role of frequency in the acquisition of English word order. Cognitive Development, 20, 121–136.

30.

Mintz

T. H.

(2003). Frequent frames as a cue for grammatical categories in child directed speech. Cognition, 90, 91–117.

31.

Naigles

L. G.

Fowler

Helm

(1992). Developmental shifts in the construction of verb meanings. Cognitive Development, 7, 403–427.

32.

Naigles

Gleitman

L. R.

Gleitman

(1993). Children acquire word meaning components from syntactic evidence. In Dromi

(Ed.), Language and cognition: A developmental perspective (pp. 104–140). Norwood, NJ: Albex.

33.

Naigles

L. R.

(2002). Form is easy, meaning is hard: Resolving a paradox in early child language. Cognition, 86, 157–199.

34.

Nelson

(1976). Some attributes of adjectives used by young children. Cognition, 4, 13–30.

35.

Nicoladis

(2002). The cues that children use in acquiring adjectival phrases and compound nouns: Evidence from bilingual children. Brain and Language, 81, 635–648.

36.

Nicoladis

(2006). Cross-linguistic transfer in adjective-noun strings by preschool bilingual children. Bilingualism: Language and Cognition, 9, 15–32.

37.

Nowak

M. A.

Plotkin

J. B.

Jansen

V. A. A.

(2000). The evolution of syntactic communication. Nature, 404, 495–498.

38.

Ocampo

(1991). Word order in constructions with a one-valency verb, a subject NP and a PP in spoken Rioplatense Spanish. Hispania, 74, 409–416.

39.

Paradis

Nicoladis

Genesee

(2000). Early emergence of structural constraints on code-mixing: Evidence from French-English bilingual children. Bilingualism: Language and Cognition, 3, 245–261.

40.

Perruchet

Pacton

(2006). Implicit learning and statistical learning: One phenomenon, two approaches. Trends in Cognitive Science, 10, 233–238.

41.

Pinker

(1984). Language learnability and language development. Cambridge, MA: Harvard University Press.

42.

Savage

Lieven

Theakston

Tomasello

(2003). Testing the abstractness of children’s linguistic representations: Lexical and structural priming of syntactic constructions in young children. Developmental Science, 6, 557–567.

43.

Saylor

M. M.

(2000). Time-stability and adjective use by child and adult English speakers. First Language, 20, 91–120.

44.

Slobin

D. I.

(1973). Cognitive prerequisites for the development of grammar. In Ferguson

C. A.

Slobin

D. I.

(Eds.), Studies of child language development (pp. 175–208). New York: Holt, Rinehart & Winston.

45.

Taylor

Gelman

S. A.

(1988). Adjectives and nouns: Children’s strategies for learning new words. Child Development, 59, 411–419.

46.

Tomasello

(1992). First verbs. A case study of early grammatical development. Cambridge: Cambridge University Press.

47.

Tomasello

(2000a). Do young children have adult syntactic competence? Cognition, 74, 209–253.

48.

Tomasello

(2000b). First steps toward a usage-based theory of language acquisition. Cognitive Linguistics, 11, 61–82.

49.

Tomasello

(2003). Constructing a language: A usage-based theory of language acquisition. Cambridge: Harvard University Press.

50.

Waxman

S. R.

Booth

A. E.

(2001). Seeing pink elephants: Fourteen-month-olds’ interpretations of novel nouns and adjectives. Cognitive Psychology, 43, 217–242.

51.

Waxman

S. R.

Klibanoff

R. S.

(2000). The role of comparison in the extension of novel adjectives. Developmental Psychology, 36, 571–581.

52.

Wexler

Culicover

P. W.

(1980). Formal principles of language acquisition. Cambridge, MA: MIT Press.