Journal of Experimental Child Psychology 138 (2015) 15–30

Contents lists available at ScienceDirect

Journal of Experimental Child Psychology journal homepage: www.elsevier.com/locate/jecp

Exploring interactions between semantic and syntactic processes: The role of animacy in syntactic priming Perla B. Gámez a, Marina Vasilyeva b,⇑ a b

Department of Psychology, Loyola University Chicago, Chicago, IL 60660, USA Lynch School of Education, Boston College, Chestnut Hill, MA 02467, USA

a r t i c l e

i n f o

Article history: Received 29 July 2014 Revised 29 April 2015 Available online 27 May 2015 Keywords: Language production Semantics Animacy Structural priming Passives Syntax

a b s t r a c t The current study addressed the relation between syntactic and semantic processes during language production in 5- and 6-year-old children. A priming paradigm was used to examine children’s production of passives in describing transitive scenes (target pictures) following exposure to the experimenter’s sentences (primes). The key question was whether the tendency to repeat the syntactic form of the prime was affected by the animacy features in the prime and the target picture. In Experiment 1, children heard either passive or active primes with varied animacy configurations (e.g., animate patient/inanimate agent vs. inanimate patient/animate agent). The animacy features of the prime matched those of the target. Similar to prior studies, results showed a greater use of passives following passive, as opposed to active, primes. Critically, the difference between the two priming conditions varied as a function of animacy; it was larger when the prime and the target included an animate patient/inanimate agent than with the reversed animacy. In Experiment 2, the animacy configuration of the prime either matched or did not match that of the target. Results showed a greater likelihood of producing a passive when the target picture contained an animate patient versus an inanimate patient, and this effect was stronger when the prime had the same animacy features. The findings indicating that syntactic priming is moderated by animacy are discussed in the broader context of understanding the role of semantics in guiding the choice of syntactic structure. Ó 2015 Elsevier Inc. All rights reserved.

⇑ Corresponding author. E-mail address: [email protected] (M. Vasilyeva). http://dx.doi.org/10.1016/j.jecp.2015.04.009 0022-0965/Ó 2015 Elsevier Inc. All rights reserved.

16

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

Introduction When producing language, speakers must choose words to express an intended meaning, assign these words their syntactic roles, and assemble them within a particular sentence structure. The fundamental conceptual issue examined in the current study concerns the nature of the relation between semantic and syntactic processes involved in children’s language production. We ask to what extent semantic properties, specifically animacy, modify the processes guiding the choice of syntactic form of a sentence. We address this question by leveraging the priming paradigm to test whether children’s tendency to repeat the syntactic structure of the prime is affected by the animacy configuration in the prime and in the child’s intended message. Processes involved in language production In an influential theoretical model of language production, Levelt (1989) provided a framework for exploring the processes underlying language production and the potential interplay between them. The model postulates distinct stages involved in preparation for articulating speech: conceptualization, in which speakers conceptualize a message, and formulation, in which conceptual representations undergo grammatical encoding. At the stage of message conceptualization, speakers nonlinguistically represent the aspects of meaning that are to be conveyed in a sentence. Such conceptual representations may include thematic roles (e.g., the agent or patient of an action) and the animacy of the entities involved in the event (Pickering & Ferreira, 2008). The preverbal information represented in the conceptual message must be mapped onto specific language devices, which occurs at the formulation stage. It is at this stage that speakers choose words that express their intended meanings, assign these words their respective syntactic roles, and combine them to form sentences. These processes comprise grammatical encoding, which precedes the final production stage where sounds and intonation are used to create phonological representations that are ultimately articulated by speakers. Although this model provides a general outline of the processes involved in language production, the specifics concerning the structure of information at different stages and possible mechanisms of interrelations between stages remain unclear. One methodological tool that has been used to explore the nature of syntactic processing and its relation to semantic levels of processing is the structural priming technique (see Pickering & Ferreira, 2008, for a review). This experimental methodology involves exposing participants to sentences that have a predetermined structure (i.e., primes) and exploring whether any aspect of that structure is repeated in the participants’ subsequent sentence production. The priming paradigm has been used to examine speakers’ tendency to repeat both the syntactic form and semantic features of the prime. Thus, this paradigm can provide information relevant to the key conceptual issue addressed in the current study, namely the relation between semantic and syntactic processes involved in language production. To the extent to which the repetition of the syntactic form (i.e., syntactic priming effect) is independent of the semantic features of the stimuli, one could argue that the syntactic level of processing is encapsulated from semantic information (Pickering & Ferreira, 2008). On the other hand, if repetition of the syntactic form is sensitive to manipulations of semantic features, this would suggest an interaction between the levels of processing. Using priming methodology to investigate processes involved in language production Bock and colleagues used a priming procedure to explore the nature of syntactic processes, in particular, their lexical independence (Bock, 1986; Bock & Loebell, 1990). The results showed a clear syntactic priming effect: after exposure to a prime sentence with a particular syntactic structure (e.g., passive), speakers were more likely to produce their own sentence with the same structure (e.g., passive) rather than the alternate structure (e.g., active). Furthermore, because speakers’ own sentences contained a different set of words than the prime, the findings were interpreted as showing that the

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

17

process of syntactic priming is lexically independent. Yet, although there is strong evidence that syntactic priming operates independent of specific words in both adults and children (Bencini & Valian, 2008; Huttenlocher, Vasilyeva, & Shimpi, 2004), it is possible that this syntactic process may draw on nonsyntactic information from general semantic categories that capture conceptual aspects of speakers’ message. Two such categories have attracted particular attention: thematic roles and animacy. Several studies used the priming paradigm to reveal the effect of thematic roles on sentence production (e.g., Cai, Pickering, & Branigan, 2012; Chang, Bock, & Goldberg, 2003; Vernice, Pickering, & Hartsuiker, 2012). The term thematic role refers to the semantic role that the verb arguments play in the event described in the sentence (e.g., agent, goal, theme, location). In one study, Chang and colleagues (2003) demonstrated speakers’ tendency to reproduce the order of thematic roles from prime to target; participants were more likely to produce sentences with a particular thematic order (e.g., agent–theme–location) after primes that had the same order (e.g., ‘‘The maid rubbed polish onto the table’’) versus an alternate order (e.g., ‘‘The maid rubbed the table with polish’’). In another study, Vernice and colleagues (2012) captured speakers’ tendency to repeat the thematic emphasis from prime to target; participants were more likely to produce a passive after a complex sentence that emphasized the patient rather than the agent of the action. The effects observed in these studies can be described as nonsyntactic structural priming. Note that Chang and colleagues’ (2003) study manipulated thematic role assignments but not the syntactic form of the prime, and in the study by Vernice and colleagues (2012) the prime and target sentences were syntactically unrelated. Thus, although it is clear from these studies that thematic roles affect sentence production, their methodology does not allow for an examination of a potential interaction between semantic and syntactic processes, that is, whether semantic processes modify the effect of syntactic primes on sentence production. When manipulating the syntactic structure, as well as thematic roles, of the prime, Bock and colleagues (Bock & Loebell, 1990; Bock, Loebell, & Morey, 1992) found that a prime with a particular syntactic form increased the production of participants’ responses that had the same form, and this was regardless of whether or not the prime and the response had a matching order of thematic roles. In other words, the effect of syntactic primes on participants’ sentences did not appear to be influenced by thematic roles. Similarly, in a study that involved both children and adults, Messenger, Branigan, and McLean (2012) showed that the syntactic priming effect was not modified by the manipulation of thematic role order. Chang and colleagues (2003) argued that a possible reason for the discrepancy in findings is that syntactic distinctions (e.g., between active and passive primes) may override thematic distinctions. We would add that some thematic distinctions are less obvious than others. For example, Messenger and colleagues (2012) reported that children showed a similar increase in the production of patient–agent passives (e.g., ‘‘A girl is being hit by a sheep’’) following passive primes that involved different thematic structures—patient–agent (matching the target), theme–experiencer (e.g., ‘‘A girl is being loved by a sheep’’), and experiencer–theme (e.g., ‘‘A girl is being shocked by a sheep’’). Yet, despite the nuanced differences in the thematic roles of arguments in these examples, children might not be sufficiently sensitive to such distinctions. They may interpret the thematic relationship captured in each sentence as that between the character that initiates the event (e.g., hitting, loving, shocking) and the character that is on the receiving end of that event. The lack of sensitivity to thematic role distinctions is a potential explanation for the null effect of thematic roles reported by Messenger and colleagues. Effects of animacy on sentence production Other kinds of semantic information may present more salient distinctions than thematic roles, for example, the difference between animate and inanimate objects, which is recognized even by young children (Mahajan & Woodward, 2009). Whereas thematic roles capture relational meanings that are specified with respect to the whole event, animacy is a basic characteristic of individual objects. Objects’ animacy has been shown to affect the ease of retrieval of corresponding lexical items (Bock & Warren, 1985; McDonald, Bock, & Kelly, 1993), and the lexical items that are recalled more easily

18

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

are more likely to occupy the initial position in the sentence (Prat-Sala, Shillcock, & Sorace, 2000). Thus, animacy may have a more direct effect on the ordering of sentence constituents than thematic roles. As reviewed below, there is indeed substantial evidence indicating that the animacy of the entities involved in an event affects the choice of syntactic form in a sentence that describes that event. Evidence from nonpriming studies The effects of animacy on the choice of syntactic form during spontaneous sentence production in picture description tasks have been documented with both adults and children (Corrigan, 1986; Harris, 1978; Lempert, 1990). Animate characters are used more frequently as sentence subjects than are inanimate characters. Related to this tendency, speakers are more likely to produce a passive sentence (placing a patient in the subject position and an agent in the object position) if the action captured by the sentence involves an animate patient and inanimate agent. In fact, there is an accumulating body of evidence that the strong preference for active sentences, over passives, is reduced when the patient of the action is animate in a variety of languages, including English (McDonald et al., 1993; Prat-Sala & Branigan, 2000), German (Van Nice & Dietrich, 2003), Japanese (Branigan, Pickering, & Tanaka, 2008), and Spanish (Prat-Sala, 1997). One example of the link between animacy and the choice of syntactic form was demonstrated in a picture description task with Catalan-speaking children (Prat-Sala et al., 2000). In that study, children were shown pictures in which the depicted patient was either animate or inanimate. On trials where the patient was animate, children were more likely to choose a syntactic construction that highlighted the patient by placing it at an initial position in a sentence in comparison with trials where the patient was inanimate. The authors explained these results in terms of the ease with which material can be retrieved from memory and entered into the processes of grammatical encoding. They argued that because animate entities are retrieved faster than inanimate entities, due to their conceptual accessibility, information about the animate patient becomes available early during the syntactic stage of processing, facilitating the choice of a sentence form with the patient in the initial position. Evidence from priming studies Although language researchers generally acknowledge the role of animacy in sentence production, there has been limited research examining this issue within the priming paradigm. Yet, this paradigm allows researchers to go even further in exploring the link between semantics and syntax than a nonpriming picture description task. In particular, the priming methodology allows us to vary the animacy features of the stimuli while simultaneously varying the syntactic form of the prime. Doing so can provide evidence concerning a potential interaction between syntactic and semantic processes that have been separately shown to have an effect on the choice of the sentence form. One priming study in which the role of animacy was systematically investigated involved the priming of active and passive structures in adults (Bock et al., 1992). In that study, researchers manipulated the syntactic form of the prime sentence (active vs. passive) as well as the animacy of its arguments (animate vs. inanimate subject). The animacy features of the target pictures that were to be described by participants were kept constant; these pictures contained animate patients and inanimate agents. Participants’ responses showed (a) a main effect of syntactic priming, with active primes (regardless of animacy) eliciting more active responses than passive primes and vice versa, and (b) a main effect of animacy, with primes containing animate subjects (both passive and active) eliciting more responses with animate subjects compared with primes inanimate subjects. Critically, there was no interaction between animacy and syntactic priming; the increase in the production of passive sentences following a passive prime was the same regardless of the animacy of the prime’s subject. The lack of interaction between animacy and syntactic priming led researchers to conclude that syntactic priming is independent of semantic-level processing. Even though the animacy of the subject in the prime increased the choice of subject animacy in the participant’s own sentence, it did not have a corresponding effect on the choice of the syntactic structure. These results have been identified as a critical piece of evidence for the independence of syntactic and semantic processes in language production: ‘‘This [finding] implies that semantic level processing that underlies the animacy effect is encapsulated from the syntactic level processing that underlies the syntactic priming effect. The alternative—that the animacy priming ‘boosts’ the syntactic priming, suggesting interactivity between the

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

19

levels—is disconfirmed’’ (Pickering & Ferreira, 2008, p. 432). Yet, our analysis of Bock and colleagues’ (1992) study suggested that certain aspects of its methodology may have reduced the likelihood of finding an interaction, as explained below. Thus, we decided to further pursue the issue of a potential interaction between syntactic and semantic levels of processing, specifically, by examining animacy effects in the context of priming. The current study One difference between the current study and the study by Bock and colleagues (1992) concerns the animacy of characters depicted in the target pictures. Recall that Bock and colleagues varied the animacy of characters in the primes, but all of the target pictures had the same animacy characteristics. Sentence production on a priming task can be influenced not only by priming stimuli but also by the features of target pictures. It is possible that having the same animacy across targets reduced a potential effect of the prime’s animacy on participants’ choice of syntactic form. Thus, in the current study, we varied the animacy of characters not only in the prime sentences but also in the targets. Second, we investigated the effects of manipulating both animacy and syntactic prime in children. Several studies using a priming paradigm with children raised the possibility that they may be particularly attuned to nonsyntactic information captured in the prime. For example, one study with 4- and 5-year-olds provided evidence for the role of topical similarity in syntactic priming (Goldwater, Tomlinson, Echols, & Love, 2011). The findings revealed a boost of syntactic priming in the high-similarity condition (e.g., when a sports scene in the prime was paired with a different sports scene in the target picture) compared with low similarity (e.g., when a sports scene in the prime was paired with the food scene in the target). A study with Russian speakers showed that a passive prime increased children’s production of alternative syntactic structures that preserved the emphasis on the patient, whereas adults preserved not just the semantic emphasis but also the syntactic form of the prime (Vasilyeva & Waterfall, 2012). Based on these findings, we expected a high likelihood of obtaining semantic effects in the priming study with children. As argued by Pickering and Ferreira (2008), a key piece of evidence for the interaction between semantic and syntactic levels of processing would be provided by finding that syntactic priming—the process that affects language production at the grammatical encoding stage—can be boosted by semantic factors such as animacy. If animacy and syntactic priming affect the choice of sentence structure independent of each other, we should see additive effects of these factors on language production. That is, (a) based on what is known about the effects of animacy on sentence production, one would expect participants to be more likely to produce a passive sentence in describing pictures with animate patients; (b) based on what is known about syntactic priming, one would expect participants to be more likely to produce a passive sentence following a passive prime; and (c) the magnitude of the increase in the production of passives will be the same regardless of the animacy features. If, however, there is an interaction between animacy and syntactic priming, we would expect syntactic priming to be moderated by animacy. Specifically, the increase in the production of passive responses after passive primes, compared with actives, would be largest on the trials involving animate patients and inanimate agents and would be smallest on the trials involving inanimate patients and animate agents. To distinguish between these possibilities, we conducted two experiments examining the joint contribution of animacy and syntactic form of the prime to children’s production of passives. Participants’ ages varied from 4.5 to 6.5 years. By this age, children demonstrate lexically independent syntactic priming (e.g., Huttenlocher et al., 2004) and at the same time show sensitivity to general semantic features of the prime (Goldwater et al., 2011; Vasilyeva & Waterfall, 2012). In both experiments, children took turns with the experimenter to describe pictures of transitive events (the descriptions produced by the experimenter are referred to as primes, and the pictures that were to be described by children are referred to as targets). In Experiment 1, children received either passive or active primes, depending on the priming condition. Similar to prior research that employed a syntactic priming paradigm with children, the syntactic form of the prime varied between subjects (Goldwater et al., 2011; Huttenlocher et al., 2004; Thothathiri & Snedeker, 2008). On each trial, the animacy of characters in the experimenter’s picture paralleled the animacy of characters in the target picture. In Experiment 2, the animacy features of the prime either matched or did not match the

20

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

animacy features of the target picture, which allowed us to examine separately the contribution of the animacy relations captured in the prime and in the target. Experiment 1 Method Participants The study involved 38 children (21 girls and 17 boys) ranging in age from 4;5 (years;months) to 6;2 (mean age = 5;2). They were recruited from private suburban schools in the greater Boston area in the northeastern United States. Based on the information provided by parents, all participants were native speakers of English. Materials The materials included 24 black-and-white drawings that depicted transitive actions (e.g., building, carrying, pouring, catching). Half (12) of these drawings were designated as the experimenter’s set (to be described by the experimenter with priming sentences), and the other 12 were designated as the child’s set (to be described by the child). The two sets had the same distribution of animate and inanimate characters. Specifically, each set included pictures with four combinations of characters: animate patient/animate agent, animate patient/inanimate agent, inanimate patient/animate agent, and inanimate patient/inanimate agent, with 3 pictures per combination. Lexical items referring to objects and actions depicted in both sets were high-frequency words (based on developmental norms of the MacArthur Communicative Development Inventories (CDI, http://www.sci.sdsu.edu/lexical). Appendix A contains descriptions of characters in the two sets. The experimenter’s drawings differed from the child’s drawings in terms of objects and actions depicted so as to minimize a possible lexical overlap. For each drawing in the experimenter’s set, two sentences that were to be used as primes were created: an active and a passive. Both types of primes described a given drawing using the same object names and the same verb; the only difference was in the syntactic form. Procedure Children were randomly assigned to either the active (n = 18) or passive (n = 20) priming condition. In both conditions, the experimenter told the child that they were going to take turns describing pictures. On each trial, the experimenter showed the child a picture and described it with a prime sentence (either active or passive, depending on the priming condition) and then showed the child a new picture and asked him or her to describe it. There were 12 trials; the experimenter’s pictures were presented in random order across trials. The animacy of characters in the child’s picture on any given trial matched that of the experimenter’s picture. The procedure was audiotaped, and children’s responses were later transcribed and coded for syntactic form. To establish reliability, responses from 8 participants (20% of the sample) were transcribed and coded by two independent researchers. They showed 96% agreement between their transcripts (computed as the percentage of identically transcribed utterances out of all responses provided by the 8 children). The few cases of disagreement were resolved in consultation with a third researcher. The agreement rate calculated for the coding of children’s responses into three categories—active, passive, and other—was 100%. Responses coded as active contained an agent in the subject position, followed by a transitive verb and a patient in the object position (e.g., The bunny was eating the flower). Responses coded as passive contained a patient in the subject position, followed by an auxiliary and then a transitive verb with a by-phrase (e.g., The girl was tickled by the monkey). Of note, some responses categorized as passives contained an incorrect verb form (e.g., aten or eated instead of eaten). Finally, responses that contained none of the forms described above were coded as other. Results First, we examined the types of utterances produced by children. Table 1 presents the frequency of active, passive, and other responses in each priming condition. As shown in the table, the majority of

21

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30 Table 1 Types of utterances produced in two priming conditions: Experiment 1. Priming condition

Active prime (n = 18) Passive prime (n = 20)

Transitive sentences

Other utterances

Active

Passive

169 (78%) 134 (56%)

15 (7%) 58 (24%)

32 (15%) 48 (20%)

Note: Percentages were computed out of all responses produced in a given priming condition. Children in the active condition produced a total of 18  12 = 216 responses. Children in the passive condition produced a total of 20  12 = 240 responses.

responses across the two priming conditions were transitive sentences (actives or passives). Consistent with prior experimental work, as well as observations of everyday speech, children showed a preference for the active form of the transitive. Yet, within this overall tendency, the frequency of using a particular transitive form varied with the experimental condition; the average percentage of passives was higher following passive primes compared with active primes, whereas the percentage of actives showed the opposite pattern. In terms of individual participants, 18 children in the passive priming condition (90%) and 9 in the active condition (50%) produced at least one passive response. All participants in both conditions produced active responses. In addition to full passives and actives, children’s responses in both priming conditions contained a variety of sentence forms that were categorized as other. This category included partial sentences (e.g., The girl) as well as complete sentences with intransitive verbs (e.g., The boy cried), linking (e.g., The dog was on the guy’s leg), and truncated/adjectival passives (e.g., The bear is hurt). It should be noted that some responses in the other category reflected children’s attempts to emphasize the patient of the transitive action without using a syntactic passive structure. These included ergatives, in which the patient was placed in the subject position and a verb was used intransitively (e.g., The glass broke). There were also some utterances, in which children incorporated elements of the passive while keeping the verb in the active form (e.g., The tree fell down by the lightning bolt). Importantly, most responses emphasizing the patient without using a passive structure were produced in the passive condition (10 vs. 2 in the active condition). Turning to the main question of the study, we examined the production of passives as a function of both the syntactic structure of the prime and the animacy of characters. The average percentages of passive responses across all experimental conditions are shown in Fig. 1. In the active priming condition, we observed little variation in the use of passives as a function of animacy. In contrast, in the passive priming condition, the use of passives varied substantially as a function of animacy. The highest frequency of passives (50%) was observed when the prime and the target contained an animate patient and an inanimate agent (i.e., when both animacy components could be expected to increase the use of passives by emphasizing the patient and deemphasizing the agent of the transitive action). Conversely, the lowest frequency of passives (7%) was observed with an inanimate patient and an animate agent (i.e., with both components expected to decrease the use of passives). The other two animacy configurations produced intermediate levels of passive use following passive primes. In our statistical analysis, we tested the hypothesis that children’s tendency to repeat the syntactic structure of the prime may be facilitated or impeded by animacy features of the stimuli. As noted earlier, we hypothesized that the production of passive responses after passive primes would be largest on the trials involving animate patients/inanimate agents and smallest on the trials involving inanimate patients/animate agents. Thus, our hypothesis focused on the two animacy configurations where the contrast in animacy between the patient and the agent may create a powerful force to either promote or demote the role of the patient in a sentence. Consequently, our statistical analysis focused on the key comparison between these two animacy conditions in relation to syntactic priming. To compare the likelihood of passive responses as a function of syntactic prime (passive vs. active) and animacy (animate patient/inanimate agent vs. inanimate patient/animate agent), we conducted a mixed-effects logistic regression analysis (Jaeger, 2008). During recent years, priming researchers have adopted this statistical methodology over the previously used analysis of variance as a more statistically appropriate tool for the analysis of data with binomial dependent variables. The logit

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

Percentage of passive responses

22

Animate paent/ Inanimate agent

Animate paent/ Inanimate paent/ Inanimate paent/ Animate agent Inanimate agent Animate agent

Fig. 1. Production of passives as a function of syntactic priming and animacy: Experiment 1. Percentages were calculated out of all responses produced in a given condition. In each of the four animacy conditions (3 trials per condition), there were 18  3 = 54 responses produced after active primes and 20  3 = 60 responses produced after passive primes.

mixed-effects models predict the variability of a certain outcome (in the current study, production of a full passive response) as a function of manipulated experimental conditions. The fixed effects included syntactic prime (passive = 1, active = 0), which was a between-participant factor, and animacy relation (animate patient/inanimate agent = 1, inanimate patient/animate agent = 0), which was a within-participant factor; participants and items were included as random effects. For simplicity of presentation, the first animacy condition is henceforth referred to as ‘‘animate patient’’ and the second as ‘‘inanimate patient.’’ First, we tested a model that included only main effects of prime and animacy. Part A of Table 2 presents the results for this model. As shown in the table, the analysis indicated significant effects of both syntactic prime and animacy. Next, we added to the model an interaction term, Prime*Animacy. Part B of Table 2 presents the results for the model that includes both main effects and the interaction. This analysis showed a significant interaction effect, whereas the main effects became nonsignificant. A comparison of the goodness of fit for the interaction model (Akaike information criterion [AIC] = 138.49, Bayesian information criterion [BIC] = 58.51) against the main-effects-only model (AIC = 141.90, BIC = 52.131) suggested that the interaction model had a better fit with the data and, in fact, adding the interaction term significantly improved the model (v2 = 40.83, p < .001). Finally, following up on the interaction finding, we examined simple main effects. This analysis showed that there was a significant difference in the production of passives between the passive Table 2 Mixed-effects logistic regression model results: Experiment 1. Fixed effects

Estimate

SE

A. Main effects only Intercept Prime Animacy

5.55 2.53 2.96

1.02 0.69 0.62

B. Interaction added Intercept Prime Animacy Prime*Animacy

3.97 0.61 1.14 2.23

1.00 1.22 0.84 1.10

z-Value

p-Value

Confidence interval

5.46 3.65 4.79

.000 .000 .000

[ 7.54, 3.56] [1.17, 3.89] [1.75, 4.18]

3.96 0.49 1.35 2.03

.000 .620 .180 .030

[ 5.94, 2.00] [ 1.80, 3.00] [ 0.52, 2.79] [0.07, 4.39]

Note: Prime (Passive = 1, Active = 0), Animacy (Animate Patient = 1, Inanimate Patient = 0).

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

23

and active priming conditions when the pictures displayed an animate patient/inanimate agent (log odds coefficient = 3.37, SE = 0.72, z = 4.70, p < .001). In contrast, the difference between the two priming conditions was not significant when the stimuli displayed an inanimate patient/animate agent (log odds coefficient = 1.14, SE = 0.86, z = 1.33, p > .05). Experiment 2 In the previous experiment, the animacy of agent/patient in the experimenter’s picture (and, consequently, in the priming sentence) matched the animacy configuration depicted in the target picture on any given trial. The findings indicated that, following the passive prime, children were most likely to produce a passive with the animate patient/inanimate agent and were least likely to produce a passive with the inanimate patient/animate agent. However, because the prime and the target were matched in terms of animacy, it remained to be determined whether the animacy characteristics of the priming sentence or the target picture, or a combination of a particular animacy configuration and a similarity between the prime and the target, was responsible for the observed effect. Experiment 2 was designed to tease apart these possibilities. Given the lack of animacy effects on the production of passives in the active priming condition, Experiment 2 tested children’s performance following a passive prime while manipulating the animacy relations within the prime and the target. Based on prior literature, as well as the findings of Experiment 1, we expected that exposure to a target picture with an animate patient/inanimate agent would increase the production of passives. Furthermore, we hypothesized that the combination of a target picture with an animate patient, paired with a prime picture that similarly features an animate patient, would have a greater effect on the production of passives in comparison with trials on which the animacy configurations of the prime and the target were reversed. That is, we conjectured that a matching prime may activate the notion of animacy and the binding between patient and animacy, thereby further increasing the effect of the animate patient in the target. Method Participants The study involved 15 children (8 girls and 7 boys) ranging in age from 5;3 to 6;3 (mean age = 5;6). Children were recruited from the same schools as the Experiment 1 participants (although none of them took part in the previous experiment). Based on the information provided by parents, all participants were native speakers of English. Materials As in Experiment 1, the materials included 24 black-and-white drawings depicting transitive actions, 12 of which were designated as the experimenter’s set and the other 12 of which were target pictures to be described by the child. Appendix B contains descriptions of characters used in the two sets. Half of the pictures in Experiment 2 were identical to those in Experiment 1 (the overlapping pictures are highlighted in bold in the Appendixes A and B), whereas the other half were new pictures. Procedure Children were tested using the same procedure as in the previous experiment (i.e., the experimenter and child took turns describing pictures) with the exception that all participants in Experiment 2 received passive primes. The animacy was manipulated to create four conditions (3 trials each). Two of these conditions (animate patient/inanimate agent in both the prime and the target and inanimate patient/animate agent in both the prime and the target) were parallel to Experiment 1. In the other two conditions, the animacy features of the prime were different from those of the target (animate patient/inanimate agent in the prime but inanimate patient/animate agent in the target and vice versa). As in Experiment 1, responses were coded into three categories—active, passive, and other— according to the criteria outlined above.

24

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

Results

Percentage of passive responses

Even though participants in this experiment were all presented with a passive prime, the majority of their responses were active sentences, reflecting a general preference for the syntactic form that is prevalent in everyday speech. Of a total of 180 responses (15 children  12 trials), 104 (58%) were full active sentences, 41 (23%) were full passive sentences, and 35 (19%) were coded as other. Most responses in this category (26 of 35, 74%) were incomplete sentences or sentences with intransitive and linking verbs. Of note, the remaining 26% of other responses represented syntactic forms that emphasized the patient, namely truncated/adjectival passives and ergatives. Fig. 2 shows the distribution of passive responses across the four animacy conditions. The findings from the two conditions with matching animacy configurations in the prime and the target replicated the results from the two corresponding conditions in Experiment 1. The highest frequency of passive responses (44%) was observed when both the prime and the target had an animate patient/inanimate agent, and the lowest frequency was observed when the animacy was reversed (9%). When the animacy relation captured in the target did not match that of the prime, it appeared that the target’s animacy configuration played a more critical role than the prime’s animacy. In particular, 24% of passives were produced when the target contained an animate patient/inanimate agent (with the opposite animacy configuration in the prime) and 13% were produced when the target contained an inanimate patient/animate agent (with the opposite animacy configuration in the prime). To compare the likelihood of passive responses as a function of animacy relation in the prime and the target, we conducted a mixed-effects logistic regression analysis. The fixed effects included target animacy (animate patient/inanimate agent = 1, inanimate patient/animate agent = 0) and prime animacy (animate patient/inanimate agent = 1, inanimate patient/animate agent = 0), both of which were within-participant factors; participants and items were included as random effects. As in Experiment 1, we first tested a model that included only main effects and then added an interaction term. The model was not improved by the addition of interaction (p > .05). The best-fit model is presented in Table 3. Guided by our hypothesis that the effect of having an animate patient in the target may be strengthened by exposure to a prime with matching animacy, we compared the production of passives in describing pictures with animate patients following matching versus mismatching primes. In other words, even though the interaction between the prime and target animacy was not significant, we conducted further analyses because the lack of the interaction may reflect smaller effect sizes

Animate paent/Inanimate agent Inanimate paent/Animate agent

Animate paent/Inanimate agent

Inanimate paent/Animate agent

Fig. 2. Production of passives as a function of animacy of the target and the prime: Experiment 2. Percentages were calculated out of all responses produced in a given condition. In each of the four animacy conditions (3 trials per condition), there were 15  3 = 45 responses produced.

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

25

Table 3 Mixed-effects logistic regression model results: Experiment 2. Fixed effects Intercept Target animacy Prime animacy

Estimate 2.52 1.47 0.77

SE 0.41 0.38 0.40

z-Value 6.11 3.89 1.89

p-Value

Confidence interval

.000 .000 .059

[ 3.33, 1.71] [0.73, 2.21] [ 0.03, 1.57]

Note: For both target animacy and prime animacy, animate patient/inanimate agent = 1 and inanimate patient/animate agent = 0.

compared with the main effects (Leon & Moonseong, 2009). Thus, it may be useful to examine the effects of one of the variables at different levels of another variable that are based on theoretical predictions. Indeed, these analyses revealed that in describing target pictures with animate patients, children produced more passives when the prime’s patient had the same animacy as the target compared with when it had a different animacy (log odds coefficient = 0.91, SE = 0.30, z = 3.02, p < .001). In contrast, when that target had an inanimate patient, there was no significant difference in performance as a function of prime’s animacy (log odds coefficient = 0.46, SE = 0.89, z = 0.51, p > .05). General discussion The current study examined the nature of the relation between syntactic and semantic processes during children’s language production. General questions concerning the interplay between the form and content of thought have been debated for centuries by philosophers. More recently, linguists and psychologists have started addressing specific questions about the links between syntactic processes (e.g., the choice of a sentence form) and semantic factors reflecting sentence meaning. The current study contributes to this line of investigation by examining the relation between syntactic priming and animacy, which constitutes one of the basic semantic properties. In Experiment 1, we exposed participants to either active or passive prime sentences that varied in their animacy configurations; the animacy of characters in the target pictures matched that of the prime. In Experiment 2, the animacy configuration of the target pictures either matched or did not match that of the prime. Establishing the interaction between syntactic form and animacy The results of the first experiment replicated the findings of earlier studies demonstrating syntactic priming effects (e.g., Huttenlocher et al., 2004); children produced more passive responses following passive primes as opposed to active primes. Extending the findings of previous research, our study showed that this priming effect was particularly strong when the prime and the target included an animate patient/inanimate agent compared with the condition in which the animacy was reversed. In other words, the process of syntactic priming (captured by the difference in the percentage of passives produced after passive primes compared with active primes) was moderated by animacy. Note that 50% of all responses that were produced in Experiment 1 after passive primes with an animate patient/inanimate agent were passives. This is higher than the typical percentages reported in other priming studies with children (e.g., Hupp & Jungers, 2009; Huttenlocher et al., 2004; Savage, Lieven, Theakston, & Tomasello, 2003). The difference is likely due to the fact that these prior studies either included inanimate patients in all of their stimuli or averaged the percentage of passives across trials that involved different animacy configurations. As shown in the current study, the highest proportion of passives can be observed when the syntactic and semantic factors converge to facilitate the production of sentences describing a transitive action with a focus on the patient. Although the first experiment provided evidence of the interaction between syntax and animacy, it did not allow us to tease apart the contribution of animacy of the prime versus the target to children’s sentence production. To address this issue, we followed up with the second experiment, in which we crossed the animacy of agent/patient in the prime with that of the target. The results showed that children’s production of passives following passive primes was strongly affected by the animacy relation

26

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

depicted in the target picture; children were more likely to produce a passive response when the target had an animate patient compared with an inanimate patient. In addition, this effect was boosted by having an animate patient in the prime. That is, the combination of the prime and the target with an animate patient was more likely to elicit a passive response than was any other animacy combination. Together, the findings of the two experiments provide a picture of how syntactic and semantic properties of the stimuli interact to determine the likelihood of producing a passive sentence. Interpreting the interaction findings As noted earlier, research using picture description tasks provided initial evidence of the link between semantic features and the choice of a sentence form, showing, for example, that speakers are more likely to produce a passive sentence when describing a transitive action with an animate patient (Lempert, 1990; Prat-Sala et al., 2000). The role of animacy observed in these studies can be explained by an information accessibility approach (Bock et al., 1992). When a child conceptualizes a message depicted in the target picture, the message-level representation captures the semantic features of the scene such as the animacy of the characters. Given that animate entities are more accessible, they enter the next stage of grammatical encoding earlier than nonanimate entities. Thus, the choice of a passive sentence form, in which the patient appears in the initial position, may reflect the order in which semantic information is accessed during the syntactic stage of processing. The priming task used in the current study provided a means of further investigating the link between semantic features and syntactic processes. In particular, it allowed us to examine whether semantic factors known to affect the choice of sentence structure (e.g., animacy) interact with syntactic factors known to affect the choice of sentence structure (e.g., syntactic priming) during the process of language production. Indeed, we found that animacy moderated the effects of syntactic priming—a finding that is relevant to a theoretical understanding of language production. In the view of language production as outlined in Levelt’s model (Bock & Levelt, 1994; Levelt, 1989), there is a strong presupposition of modularity in the system. At the level of conceptualization, the content of the preverbal message is determined, including the entities that are involved in the event and their features. Semantic factors that are processed at this stage, such as animacy, determine the conceptual content of the message and may affect the order in which lexical items are accessed at the next stage, namely formulation. Beyond this influence, however, semantic factors have been viewed as not interfering with the processes that occur at the stage of grammatical formulation. A key goal of this stage is to choose a sentence form within which the message can be expressed. One process that influences the choice of a constituent structure at the stage of grammatical formulation is syntactic priming. Exposure to a prime sentence that has a particular form (e.g., a passive) activates a corresponding syntactic structure, thereby increasing the likelihood that the speaker will choose this structure to express the message. Our findings show that the effect of the syntactic form of the prime (passive vs. active), which takes place at the grammatical formulation level, varies as a function of the animacy features that are encoded in the conceptual message level, suggesting an interaction between these levels. Alternative models of language production, beyond that of Levelt (1989), assume interactivity between the different stages of production. For example, there are models that assume ‘‘cascading activation between levels’’ in which different levels may operate simultaneously (see Vigliocco & Hartsuiker, 2002). Other theoretical accounts, such as the functionalist account (Goldberg, 1995), posit not just an interaction between semantic and syntactic representations but rather an existence of linguistic representations (referred to as constructions) that are defined by both syntactic and semantic properties. In this view, a passive prime not only activates the corresponding syntactic structure but also brings the focus of the transitive action to a patient because the patient appears in the prominent position of a sentence subject. Thus, syntactic and semantic features of the prime together activate the emphasis on the patient. When a child who has been exposed to this type of prime is presented with a picture that has a prominent patient (due to its animacy), the likelihood of producing a picture description with an emphasis on the patient increases even further. Although the functionalist account is consistent with the current findings, our results do not lead to a definitive conclusion that semantic information is represented as part of the syntactic structure

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

27

because alternative explanations are also plausible. However, although the current study does not identify the specific mechanism underlying animacy effects on syntactic priming, it does call for a revision of theoretical accounts that posit the encapsulation of semantic and syntactic processes in language production. In accordance with Pickering and Ferreira (2008), if both semantic and syntactic factors influenced sentence production but did so independent of each other, we would expect main effects of syntactic prime and animacy but not the interaction effects observed in the current study. In contrast, our interaction findings suggest that the syntactic and semantic processes involved in sentence production are not encapsulated in that the effect of syntactic factors on the choice of a sentence form depends on semantic properties of the sentence. Animacy effects on syntactic priming: A broader view To our knowledge, the current study is the first investigation involving children that systematically manipulated both animacy and syntactic form so as to examine their interaction in the course of priming. Our findings differ from those obtained with adults, specifically by Bock and colleagues (1992). That study revealed main effects of animacy and syntactic priming, but no interaction between these factors, in the production of passives, suggesting that processes of syntactic and animacy priming were independent of each other. One possible explanation for this discrepancy relates to the methodological differences between the two studies. Recall that in Bock and colleagues’ study, animacy features varied across primes but not across targets. In contrast, in Experiment 1 of the current study, the animacy features of the prime were manipulated so that in one experimental condition both the prime and the target had an animate patient, whereas in another condition both the prime and the target had an inanimate patient. As noted earlier, having an animate patient in the prime may have increased the effect of patient animacy in the target, making this condition particularly distinct from the one with a reversed animacy configuration. Thus, in the passive priming condition, animacy effects on the speaker’s syntactic choice should be more pronounced when the target’s animacy matches the prime as opposed to being fixed. Another possible explanation for the discrepancy between the current results and previous findings by Bock and colleagues (1992) is that it reflects developmental differences between children and adults. There is no agreement at the moment as to how children’s sensitivity to semantic versus syntactic features of sentences compares with that of adults. Some evidence suggests that when children are exposed to passive primes, it leads to a particularly strong activation of the focus on the patient of the transitive action. The evidence comes from work with speakers of languages that allow for a greater flexibility of word order than English (e.g., Spanish: Gámez, Shimpi, Waterfall, & Huttenlocher, 2009; Gámez & Shimpi, 2015; Russian: Vasilyeva & Waterfall, 2012; see also Gámez & Vasilyeva, 2015). As a result, the emphasis on the patient can be achieved not only with a participial passive but also with alternative syntactic structures. Studies with Spanish- and Russian-speaking children showed that presenting them with a prime that had a structure of a participial passive resulted in a significant increase of responses that had alternative syntactic forms but shared with the prime an emphasis on the patient. Russian-speaking adults presented with the same task were more likely to preserve not just the focus on the patient but also the syntactic structure of the prime. Thus, the prominence of specific syntactic forms and semantic features, as well as a possibility of an interaction between them, may differ as a function of prior language experience—a possibility that should be addressed in future research. Role of other semantic factors in syntactic structure choice Research on the relation between animacy and syntactic priming represents just one of several directions exploring the contribution of semantic processes to syntactic structure choice. A growing body of research has addressed another semantic property—thematic roles of sentence constituents—in relation to syntax. Bock and Loebell (1990) showed that thematic role differences had no effect on syntactic priming. Several other studies, on the other hand, revealed that thematic roles did influence sentence production (Chang, Bock, & Goldberg, 2003; Hare & Goldberg, 1999; Salamoura & Williams, 2007). For example, Hare and Goldberg (1999) showed the effects of thematic role order

28

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

on the production of dative sentences. Datives can take a form of a prepositional order (PP) sentence (e.g., ‘‘The girl gave the book to the teacher’’ = agent, theme, recipient) or a double object (DO) sentence (e.g., ‘‘The girl gave the teacher the book’’ = agent, recipient, theme). The authors found that primes such as ‘‘The officers provided soldiers with guns,’’ which had a syntactic form of a PP dative but a thematic role order of a DO dative, increased the frequency of DO responses. Thus, the choice of a particular dative structure was determined by the thematic role order of the prime. When addressing possible reasons for the inconsistency among findings regarding thematic roles, Chang and colleagues (2003) highlighted the importance of considering the relative sensitivity of speakers to particular syntactic forms as well as semantic distinctions. Specifically, Chang and colleagues argued that thematic differences in the Bock and Loebell (1990) study, in which a thematic role manipulation was crossed with a syntactic form manipulation, may not have been detectable because they were overshadowed by syntactic differences. The relative power of specific syntactic and semantic manipulations may play a role in determining the outcome of investigations exploring the interaction between semantics and syntax. A further problem with exploring semantic processes is that they are not as easily formalized as syntactic processes. Sentence types, such as transitives and datives, have a delimited number of alternate forms; thus, manipulating these structures is a relatively straightforward process. In contrast, the distinction between semantic properties may be less clear. For example, some studies have manipulated animacy by contrasting animate and inanimate entities, whereas others have done so by contrasting humans and animals. Speakers in general, and children in particular, may be more sensitive to some of these distinctions than others. In a similar vein, certain differences in thematic roles may be more obvious than others, again leading to discrepancies across studies. This somewhat ambiguous nature of semantic properties makes their investigation challenging. The current study contributes to this area of inquiry by providing evidence of an interaction between semantic and syntactic processes in children’s production of passives on a priming task. Accumulating this type of evidence, as well as identifying the cases where semantic and syntactic processes behave autonomously, will lead to a more nuanced understanding of which semantic features matter most for the choice of syntactic structure.

Appendix A. Patient/Agent animacy in the target picture and the prime: Experiment 1. Target

Prime

Animate patient (butterfly)/Animate agent (frog); verb: grab Animate patient (boy)/Animate agent (bee); verb: sting Animate patient (postman)/Animate agent (dog); verb: bite Animate patient (man)/Inanimate agent (water); verb: splash Animate patient (bear)/Inanimate agent (trap); verb: catch Animate patient (girl)/Inanimate agent (hose); verb: sprinkle Inanimate patient (window)/Inanimate agent (ball); verb: break Inanimate patient (fence)/Inanimate agent (car); verb: smash

Animate patient (student)/Animate agent (teacher); verb: help Animate patient (hippo)/Animate agent (giraffe); verb: lick Animate patient (horse)/Animate agent (farmer); verb: feed Animate patient (cat)/Inanimate agent (curtain); verb: pull Animate patient (snowman)/Inanimate agent (sun); verb: melt Animate patient (monkey)/Inanimate agent (rock); verb: hurt Inanimate patient (city)/Inanimate agent (river); verb: flood Inanimate patient (leaves)/Inanimate agent (wind); verb: blow

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

29

Appendix A. (continued) Target

Prime

Inanimate patient (tree)/Inanimate agent (lightning); verb: strike Inanimate patient (trunk)/Animate agent (beaver); verb: eat Inanimate patient (flowers)/Animate agent-person (woman); verb: water Inanimate patient (shoe)/animate agent (dog); verb: rip

Inanimate patient (boat)/Inanimate agent (waves); verb: rock Inanimate patient (book)/Animate agent (doctor); verb: pick up Inanimate patient (nest)/Animate agent (bird); verb: build Inanimate patient (milk)/Animate agent (woman); verb: serve

Note: The lexical items listed in the ‘‘Target’’ column reflect the scenes depicted in the target pictures, but they do not necessarily reflect the actual words used by children in describing these pictures (e.g., a child may have described a picture of a postman being bitten by a dog as ‘‘the dog bit the man’’).

Appendix B. Patient/Agent animacy in the target picture and the prime: Experiment 2. Target

Prime

Animate patient (man)/Inanimate agent (water); verb: splash Animate patient (bear)/Inanimate agent (trap); verb: catch Animate patient (girl)/Inanimate agent (hose); verb: sprinkle Animate patient (clown)/Inanimate agent (cake): verb: hit Animate patient (squirrel)/Inanimate agent (snow); verb: cover Animate patient (boy)/Inanimate agent (glass), verb: cut Inanimate patient (trunk)/Animate agent (beaver): verb: eat Inanimate patient (flowers)/Animate agent-person (girl); verb: water Inanimate patient (shoe)/animate agent (dog); verb: rip Inanimate patient (board)/Animate agent (student); verb: split Inanimate patient (leaves)/Animate agent-person (wolf); verb: blow Inanimate patient (hole)/animate agent (builder); verb: dig

Animate patient (cat)/Inanimate agent (curtain); verb: pull Animate patient (snowman)/Inanimate agent (sun); verb: melt Animate patient (monkey)/Inanimate agent (rock); verb: hurt Inanimate patient (dish)/Animate agent (rabbit); verb: wash Inanimate patient (tower)/Animate agent (baby); verb: destroy Inanimate patient (salad)/Animate agent (cook); verb: make Inanimate patient (book)/Animate agent (doctor); verb: pick up Inanimate patient (nest)/Animate agent (bird); verb: build Inanimate patient (milk)/Animate agent (woman); verb: serve Animate patient (surfer)/Inanimate agent (wave); verb: carry Animate patient (tiger)/Inanimate agent (fire): verb: start Animate patient (donkey)/Inanimate agent (rope); verb: trip

References Bencini, G. M. L., & Valian, V. (2008). Abstract sentence representation in 3-year-olds: Evidence from comprehension and production. Journal of Memory and Language, 59, 97–113. Bock, K. (1986). Syntactic persistence in language production. Cognitive Psychology, 18, 355–387. Bock, K., & Levelt, W. (1994). Language production: Grammatical encoding. In M. A. Gernsbacher (Ed.), Handbook of psycholinguistics (pp. 945–983). San Diego: Academic Press.

30

P.B. Gámez, M. Vasilyeva / Journal of Experimental Child Psychology 138 (2015) 15–30

Bock, K., & Loebell, H. (1990). Framing sentences. Cognition, 35, 1–39. Bock, K., Loebell, H., & Morey, R. (1992). From conceptual roles to structural relations: Bridging the syntactic cleft. Psychological Review, 99, 150–171. Bock, K., & Warren, R. (1985). Conceptual accessibility and syntactic structure in sentence formulation. Cognition, 21, 47–67. Branigan, H. P., Pickering, M. J., & Tanaka, M. (2008). Contributions of animacy to grammatical function assignment and word order during production. Lingua, 188, 172–189. Cai, Z. G., Pickering, M. J., & Branigan, H. P. (2012). Mapping concepts to syntax: Evidence from structural priming in Mandarin Chinese. Journal of Memory and Language, 66, 833–849. Chang, F., Bock, K., & Goldberg, A. E. (2003). Can thematic roles leave traces of their places? Cognition, 90, 29–49. Corrigan, R. (1986). The internal structure of English transitive sentences. Memory & Cognition, 14, 420–431. Gámez, P. B., & Shimpi, P. (2015). Structural priming in Spanish as evidence of implicit learning. Journal of Child Language. http:// dx.doi.org/10.1017/S0305000915000161 (in press). Gámez, P. B., Shimpi, P. M., Waterfall, H. R., & Huttenlocher, J. (2009). Priming a perspective in Spanish monolingual children: The use of syntactic alternatives. Journal of Child Language, 36, 269–290. Gámez, P. B., & Vasilyeva, M. (2015). Increasing second language learners’ production and comprehension of developmentallyadvanced syntactic forms. Language Learning & Development, 11(2), 128–151. Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press. Goldwater, M. B., Tomlinson, M., Echols, C. H., & Love, B. C. (2011). Structural priming as structure-mapping: Young children use analogies from previous utterances to guide sentence production. Cognitive Science, 35, 156–170. Hare, M. L., & Goldberg, A. E. (1999). Structural priming: Purely syntactic? In M. Hahn & S. C. Stones (Eds.), Proceedings of the 21st annual meeting of the Cognitive Science Society (pp. 208–211). Mahwah, NJ: Lawrence Erlbaum. Harris, M. (1978). Noun animacy and the passive voice: A developmental approach. Quarterly Journal of Experimental Psychology, 30, 495–501. Hupp, J. M., & Jungers, M. K. (2009). Speech priming: An examination of rate and syntactic persistence in preschoolers. British Journal of Developmental Psychology, 27, 495–504. Huttenlocher, J., Vasilyeva, M., & Shimpi, P. (2004). Syntactic priming in young children. Journal of Memory and Language, 50, 182–195. Jaeger, T. F. (2008). Categorical data analysis: Away from ANOVAs (transformation or not) and towards logit mixed models. Journal of Memory and Language, 59, 434–446. Lempert, H. (1990). Acquisition of passives: The role of animacy, salience, and lexical accessibility. Journal of Child Language, 17, 677–696. Leon, A. C., & Moonseong, H. (2009). Sample sizes required to detect interactions between two binary fixed-effects in a mixedeffects linear regression model. Computational Statistics and Data Analysis, 53, 603–608. Levelt, W. G. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press. Mahajan, N., & Woodward, A. (2009). Seven-month-old infants selectively reproduce the goals of animate but not inanimate agents. Infancy, 14, 667–679. McDonald, J. L., Bock, K., & Kelly, M. H. (1993). Word and world order: Semantic, phonological, and metrical determinants of serial position. Cognitive Psychology, 25, 188–230. Messenger, K., Branigan, H. P., & McLean, J. F. (2012a). Is children’s acquisition of the passive a staged process? Evidence from six- and nine-year-olds’ production of passives. Journal of Child Language, 39, 991–1016. Messenger, K., Branigan, H. P., McLean, J. F., & Sorace, A. (2012b). Is young children’s passive syntax semantically constrained? Evidence from syntactic priming. Journal of Memory and Language, 66, 568–587. Pickering, M. J., & Ferreira, V. S. (2008). Structural priming: A critical review. Psychological Bulletin, 134, 427–459. Prat-Sala, M. (1997). The production of different word orders: A psycholinguistic and developmental approach. (Unpublished doctoral thesis). University of Edinburgh. Prat-Sala, M., & Branigan, H. P. (2000). Discourse constraints on syntactic processing in language production: A cross-linguistic study in English and Spanish. Journal of Memory and Language, 42, 168–182. Prat-Sala, M., Shillcock, R., & Sorace, A. (2000). Animacy effects on the production of object-dislocated descriptions by Catalanspeaking children. Journal of Child Language, 27, 97–117. Salamoura, A., & Williams, J. N. (2007). Processing verb argument structure across languages: Evidence for shared representations in the bilingual mental lexicon. Applied Psycholinguistics, 28, 627–660. Savage, C., Lieven, E., Theakston, A., & Tomasello, M. (2003). Testing the abstractness of children’s linguistic representations: Lexical and structural priming of syntactic constructions in young children. Developmental Science, 6, 557–567. Thothathiri, M., & Snedeker, J. (2008). Syntactic priming during language comprehension in three- and four-year-old children. Journal of Memory and Language, 58, 188–213. Van Nice, K., & Dietrich, R. (2003). Task sensitivity of animacy effects: Evidence from German picture descriptions. Linguistics, 41, 826–849. Vasilyeva, M., & Waterfall, H. (2012). Beyond syntactic priming: Evidence for activation of alternative syntactic structures. Journal of Child Language, 39, 258–283. Vernice, M., Pickering, M. J., & Hartsuiker, R. J. (2012). Thematic emphasis in language production. Language and Cognitive Processes, 27, 631–664. Vigliocco, G., & Hartsuiker, R. J. (2002). The interplay of meaning, sound, and syntax in language production. Psychological Bulletin, 128, 442–472.