The basic continuity hypothesis of L1 to L2 production

Abstract

To what extent can second language (L2) speakers acquire a syntactic representation for an L2 structure absent in the first language (L1)? Findings from L2 structural priming studies are in conflict inasmuch as evidence for and against continuity between L1 and L2 sentence production has been shown. Furthermore, previous investigations have not adequately controlled for well-known animacy effects on choice of syntactic frames. I address the conflict of views in the field via three experiments of structural priming with native, Chinese, and Turkish speakers of English by means of an oral sentence-recall production task. The structure tested, the English genitive alternation, is subject to animacy effects as in the waiter’s photo / the photo of the waiter. Chinese and Turkish have no equivalent to English of genitives nor animacy effects in their genitive structure. Experiment 1 showed priming within-L1 English and found evidence of an animacy effect, albeit only numerical. Likewise, Experiment 2 showed priming within-L2 English and found the L1 Chinese were similarly susceptible to animacy effects. Experiment 3 also showed clear within-L2 English priming in Turkish speakers but the effects of animacy differed from the other groups. I argue the similarities between the native and L2 groups to constitute grounds for a basic continuity in L1 to L2 production.

Keywords

animacy genitives language production sentence production structural priming

I Introduction

A main hallmark of the field of second language (L2) acquisition has been to describe the abstract linguistic knowledge of L2 speakers and to understand the extent to which it is affected by previous knowledge of a first language (L1). Structural priming, the recycling of previously processed or produced structure in subsequent comprehension or production (Bock, 1986), has been productively used by L2 researchers to address such issues. In a recent meta-analysis, Mahowald et al. (2016) summarize the research designs adopted and the linguistic structures investigated in a large body of studies of priming in production. They point out that priming effects in production have been observed within-L1, within-L2, cross-linguistically from L1 to L2 and bi-directionally, or in a combination thereof. In contrast to an abundance of studies of within-L1 and bidirectional cross-linguistic structural priming, very little work has been done comparatively within-L2 (for this observation, see also Hartsuiker et al., 2008, 2015). This is surprising, since L2 priming is particularly informative to the field of L2 acquisition as it is widely believed to effectively capture abstract L2 representations. Furthermore, Mahowald et al. (2016) stress that while several studies have focused on priming structures at clause and sentence level such as passive/active, dative, transitive/intransitive, and relative clauses, far fewer have centered on phrase-level phenomena such as genitives.

A prominent hypothesis of L2 sentence production derived from priming experiments is the Shared Representation Hypothesis (SRH; Hartsuiker et al., 2004). The SRH posits that L2 speakers at advanced levels of proficiency resort to a single-level of syntactic representation which melds both the L1 and L2 grammatical architecture. Research in this domain has defined the scope of the SRH further, establishing, for example, that L1 and L2 word order for a given structure must be congruent for a shared representation to be accessed (Bernolet et al, 2007; Schoonbaert et al., 2007). Three studies of within-L2 priming have sculpted the limits of the SRH, Bernolet et al. (2013), Flett (2006), and Kim (2010). These studies asked whether L2 speakers are capable of building native-like representations for an L2 structure which is absent from the L1. The claims, however, have been contradictory and the research methods problematic. Flett (2006) claims that structures novel to the L2 speaker can indeed be acquired, whereas Kim (2010) believes it unlikely due to an inflexibility of the adult L2 production system. Lastly, Bernolet et al. (2013) maintain that L2 speakers can acquire a novel structure but that a new (perhaps non-L2-specific) node has to be added to the shared representation. One often neglected confound in studies of priming, however, is a word order effect tied to animacy found in native speakers, termed ‘conceptual accessibility’ (Branigan et al., 2008). Therefore, more recently studies of within-L2 priming have begun to factor in animacy effects (Hawkins et al., 2012; Romano, 2016).

In light of the above, this study asks, first, whether L2 speakers are capable of building representations for an L2 structure missing an equivalent in the L1. In this way, it seeks to build an informed argument as to whether a basic continuity between L1 and L2 production mechanisms exists or whether L2 representations are inflexible to novel structure, as claimed by Kim (2010). Furthermore, this study pays proper attention to the effect of conceptual accessibility by investigating a phrase-level structure which is theoretically surmised, but not yet empirically proven, to be subject to animacy effects: the English ’s versus of genitive alternation.

Prior to describing the key within-L2 priming studies, I introduce the basics of priming as a research paradigm. I then layout key theoretical notions relevant to the English genitive alternation, followed by predictions for the study and reports of three experiments completed to test them.

II The structural priming paradigm

Structural Priming is a well-established research paradigm in psycholinguistics, generally employed for the delineation of cognitive models in monolingual speakers. Structural or syntactic priming refers to the persistence in production or improved processing in comprehension of a previously experienced (“prime”) phrase, clause, or sentence (Bock, 1986). For instance, when mature speakers of English comprehend the double object sentence (1a) and are asked to use a prompt to produce a sentence either immediately after or with some delay to the comprehension of (1a), they are more likely to produce another double object structure rather than its structural alternative with identical meaning (1b), namely a prepositional object dative.

(1) a. The man gave the waiter a tip (double object)

NP V NP/DO NP/DO

b. The man gave a tip to the waiter (prepositional object dative)

NP V NP/DO PP/IO

Repetition of structure from a prime to a target sentence occurs even when the two share no lexical items (e.g. waiter or tip in (1)). Structural priming effects have been explored at large, in a variety of languages with a number of grammatical structures. These effects have been investigated via corpus linguistics and experiments with different populations, including L2 learners (Bernolet et al., 2007, 2009, 2012, 2013; Bock et al., 2007; Branigan et al., 2000; Ferreira et al., 2008; Gries, 2005; Hartsuiker and Kolk, 1998a; 1998b; Hartsuiker and Westenberg, 2000; Hartsuiker et al., 2004; Huttenlocher et al., 2004; Melinger and Dobel, 2005; Pickering and Branigan, 1998; Potter and Lombardi, 1998; Scheepers, 2003; Segaert et al., 2013 inter alia; for a critical review, see Pickering and Ferreira, 2008).

In a meta-analysis, Mahowald et al. (2016) lay out a number of key notions for the structural priming paradigm applied to production. They cite 10 constructions studied up until 2016: passive/active, dative, genitive, transitive/intransitive, locative inversion, modifier order, relative clauses, main verb-participle order, VP syntax, and complex NPs. As for populations, priming has been applied to study both native and non-native speech production in the form of L1 → L1 (within-L1), L2 → L2 (within-L2), L2 → L1 (cross-linguistic between L2-L1), and L1 → L2 (cross-linguistic between L1- L2) priming. The experiments reported here will bear on the first and second types only.

In contrast to the flourishing of syntactic priming applications to studies of monolinguals, interest in priming bilingual and L2 speakers at various levels of proficiency only began to grow in 2003 (Loebell and Bock, 2003). In a seminal study, Hartsuiker et al. (2004) employed a dialogue task to investigate whether Spanish active/passive sentences primed English active/passive targets and vice versa in late Spanish–English bilinguals. The task consists of the participant and a confederate implicitly priming each other by taking turns describing pictures to one another. Their results show participants produced more Spanish active targets after an active than passive English prime; similarly, they produced more Spanish passive targets after a passive than an active English prime. An identical pattern was observed from English to Spanish. These ‘fully-crossed’ results were hypothesized to indicate L2 speakers are likely to have a single shared L1–L2 representation for structures, which are sufficiently similar between the two languages. The findings in Hartsuiker et al. (2004) have since been replicated in several cross-linguistic priming studies and their hypothesis referred to as the Shared Representation Hypothesis (SRH hereafter; Bernolet et al., 2007; 2009; 2012; 2013; Cai et al., 2011; Desmet and Declercq, 2006; Kantola and Van Gompel, 2011; Meijer and Fox Tree, 2003; Salamoura and Williams, 2006; 2007; Schoonbaert et al., 2007; Shin and Christianson, 2009; 2012).

Subsequent investigations further defined the scope of the SRH, determining that for it to hold, a structure must additionally have the same word order in L1 and L2 (Bernolet et al., 2007; Schoonbaert et al., 2007). Bernolet et al. (2007) conducted a series of experiments priming NPs in Dutch–English bilinguals with a minimum 5 years and mean 10.6 years’ instruction in the L2. Their results, obtained under four designs, within-L1, between L1–L2, between L2-L1, and within-L2, found an absence of priming effects cross-linguistically, interpreted as evidence against the SRH. Lack of shared representations was attributed to differences in word order between English and Dutch: the order of the adjective red and the verb is in the baby that is red differs from the corresponding order in the Dutch de baby die rood is. In fact, an additional experiment testing L1 Dutch to L2 German where Dutch and German have identical word order for the given structure, found the expected priming effect.

The plausibility of word order equivalency as a prerequisite for the SRH was confirmed in Schoonbaert et al. (2007). In this study, conducted with adult bilingual Dutch–English speakers, the dative alternation was primed using the dialogue task within-L1, within-L2, and between languages in both directions. Findings again revealed a clear priming effect across the board, attributable to the equivalency of structures and word order between L1 and L2.

III Within-L2 priming studies

In contrast to between-language studies, studies of within-L2 priming remain sporadic. These studies have, in different ways, shaped the scope of the SHR. If L1 and L2 representations are shared so long as the order of constituents forming the target structure is the same, this raises the question of whether L2 speakers are capable of building native-like representations for a novel L2 structure that is altogether absent from the L1. Flett (2006) conducted a series of experiments combining English and Spanish as source and target languages, priming active/passive sentences and the dative alternation with participants. The L2 participants to the experiments were at two proficiency levels in English and Spanish, intermediate and advanced. Flett attempted to define the role played by the absence versus presence of an L2 structure from L1 representations: while active/passive voice are nearly identical in Spanish and English, the dative alternation is only possible in English. In relation to the former structure, Flett found a priming effect in Spanish to Spanish and English to Spanish language pairs but the effect was stronger in the latter (L2) case. She accounts for this as a frequency effect, as use of the passive in colloquial Spanish is less natural than English. The frequency of a structure in the input, thus, also appears to play a role in a bilingual’s construction of shared representations. In relation to the dative alternation, Flett found English to English and Spanish–English groups were primed to a similar degree, opening up the possibility that bilingual representations can mirror monolinguals’ even when an L2 structure is unavailable in the L1.

In contrast to Flett’s view, native and L2 representations have been argued to differ if the L2 structures primed are absent from the L1. Kim (2010) compared structural priming of English active/passive sentences and the dative alternation via the dialogue task in four groups, native English speakers and late L2 learners with L1 Japanese, Korean, and Mandarin Chinese. Her rationale for comparing these four groups was that while the structure, the word order, assignment of thematic roles, and grammatical functions of active/passive sentences are similar in the L2 and the three L1s, the dative alternation is only possible in English and Chinese.¹ Following the SRH then, Kim predicted priming effects to differ between the English, Japanese, and the Korean group with regards to the dative alternation. This is precisely what was found: while active/passive priming did not differ by group, all three groups’ results for the dative alternation differed from the native speakers’ (2010:107). In relation to the findings of similarity between active/passive sentence priming, Kim maintains her results indicate a basic continuity in sentence production mechanisms between native and bilingual speakers, despite broad typological differences between L1 and L2. In relation to the dative alternation findings, instead, Kim’s claims ‘the adult language production system is rather inflexible with regard to a novel L2 syntactic structure’ (2010: xviii).

One unexpected result, however, was a strong preference by the native speakers compared to the L1 Chinese for the production of double object constructions in both the double object and prepositional object condition of the experiment. According to Kim, this predominance was due to a ‘conceptual accessibility’ effect in the data. Conceptual accessibility is defined as retrieval bias in favor of entities which are more highly accessible (Branigan et al., 2008). Due to the incremental nature of language production, highly accessible entities such as animates are subject to earlier insertion into a syntactic frame. Native speakers of English, for instance, are known to prefer earlier placement of animate entities in a structure where an asymmetry in animacy exists even though an alternate structure with identical interpretation is available (Branigan et al., 2008; McDonald et al., 1993; Rosenbach, 2002, 2005, 2008; Tanaka et al., 2005; Tanaka et al., 2011, amongst others). To exemplify, native speakers are more likely to be primed for the doctor sent the patient to the hospital, where the first DO the patient is animate and the second IO the hospital is inanimate, than the alternative the doctor sent the hospital the patient with equivalent meaning, where the first DO/IO is inanimate and the second DO animate. Because the primes in Kim (2010) were constructed with animates preceding inanimates in both the double object and prepositional dative conditions IOs, the native speakers strongly disfavored inverting double object sentences (e.g. a librarian handed a boy a book) as this would have resulted in preposing inanimate to animate, against the animacy effect. This caveat suggests native and L2 representations may differ when syntactic structure interfaces with semantic concepts like animacy and calls for incorporating conceptual accessibility into experiments.

Conceptual accessibility was put to the test in a study conducted by Hawkins et al. (2012) with native, L1 Arabic, and Chinese speakers at three levels of L2 English proficiency: elementary, intermediate, and advanced/very advanced. The English dative alternation was used in a spoken sentence-recall test to measure the effects of alternating the animacy of the two objects on choice of structure. The L1s and L2 were fully matched, given that English, Arabic, and Chinese not only have the dative alternation, they also share the same word order, grammatical function, and semantic role assignment to the target structure (recall (1)). Although not openly addressed, Hawkins et al.’s locution suggests an assumption that no conceptual accessibility comparable to the one active in English dative alternation exists in the two L1s. The results fully replicated those in Flett (2006), finding a priming effect in all three groups. In addition to similar structural priming effects, the L2 speakers in Hawkins et al. (2012) also showed native-like sensitivity to the conceptual accessibility effect insofar as even after manipulating the symmetry of animacy for the two entities, differences in priming effect between native and advanced L2 speakers remained non-significant. By implication, an L2 representation is likely to mirror an L1’s even when syntactic structure interacts with other levels of representation such as the conceptual. The findings in Hawkins et al. (2012) must be interpreted with caution, however, due to the fairly limited size of their L2 sample (combined advanced/very advanced groups n = 8).

A further study relevant to the SRH and animacy effects is Bernolet et al. (2013). The genitive of and ’s alternation was primed from Dutch to English and within-L2 English, testing adult L1 Dutch speakers whose self-rated English skills proficiency ranged between 4.8 and 5.8 on a 7-point Likert scale. According to the authors, there is sufficient overlap between the linguistic features of Dutch and English genitives to consider the experiments a test of the SHR. Although a clear priming effect was found in both experiments, an unexpected strong bias favoring an of genitive target following both of and ’s primes obtained. The authors interpreted their results as the outcome of slight differences between the Dutch and English ’s genitives, in contrast to a full overlap between Dutch and English of genitives:

If the L2 structure is similar enough to a structure of the L1 that is already represented in the lexicon, the combinatorial node for the existing structure is used during syntactic processing in L1 and L2. If, on the other hand, the new L2 structure turns out to be structurally different from all syntactic structures that are represented in the lexicon, a new combinatorial node is formed. (Bernolet et al., 2013: 301)

In Bernolet et al.’s terms, a combinatorial node denotes a mapping between the lexicon (i.e. lemmas) and syntactic information relevant to the grammatical constructions in which lemmas may participate (e.g. combine with two NPs to form a passive; combine with an NP subject only to form an intransitive sentence) (2013: 288). Thus, if the L2 learner encounters a new structure not present in the L1, a new mapping between L2 lemmas and syntactic information relevant to that L2 structure will be created. The mechanism handling syntactic information, however, is not described in any detail by Bernolet et al.’s theory.

One possibility not discussed by the authors is intermediate transfer from French to English. The Dutch (Flemish) subjects, who were recruited in Belgium, had English as an L3 and French as an L2. That French only allows a prepositional genitive equivalent to the English of genitive, thus, may have engendered Bernolet et al.’s participant bias for these structures. Whether conceptual accessibility played a role in this study cannot be adequately determined insofar as the asymmetry in animacy of nouns in the genitive structures forming Bernolet et al.’s experimental items was investigated only in one direction. The design included only items with an animate possessor alongside an inanimate possessum (i.e. the item possessed) such as the egg of the witch/the witch’s egg. All in all then, experiments comparing priming across the full spectrum of mismatched (i.e. treatment) and matched (i.e. baseline) animacy combinations, as well as comparisons between priming in L2 and control groups of native speakers are likely to lead to clearer results.

In summary, there is good evidence from within-L2 priming experiments that bilingual speakers are likely to have shared L1/L2 representations that do not differ from native speakers’ provided the structure in question is represented in both L1 and L2. Conversely, there is less reliable evidence of the same outcome if a structure instantiated in the L2 is absent from the L1, when the word orders differ, or when conceptual accessibility effects are at stake. I tackle these questions head on by investigating conceptual accessibility effects on the English genitive alternation in native and L2 speakers whose L1s only allow one genitive construction.²

IV Genitive structures in English, Chinese, and Turkish

English allows two types of genitives, the ’s prenominal genitive exemplified in (2a) and (3a) and the of postnominal or prepositional genitive in (2b) and (3b). Lemmas retrieved from the lexicon can enter two potential syntactic frames which are morphologically marked by the possessive ’s or of. Each lemma is also assigned a semantic role, possessor (PR), the animate or inanimate entity that possesses something, or possessum (PM), the animate or inanimate entity that is possessed by the PR.

(2) a. The camera ’s batteries (s’ genitive, PR pronominal)

PR POSS PM

b. The batteries of the camera (of genitive, PR postnominal)

PM POSS PR

(3) a. The baker ’s mother (s’ genitive, PR pronominal)

PR POSS PM

b. The mother of the baker (of genitive, PR postnominal)

PM POSS PR

According to Rosenbach (2002:29), the PR–PM relation that determines the interpretation of a genitive structure can be of eight types: possessive, partitive, descriptive, appositive, idiomatic, collocations, subjective, and objective. Some meanings are possible in both syntactic frames, others only in one. For instance, a congressional medal of honor cannot alternate with *an honor’s congressional medal as the former identifies a collocation with unique meaning. Similarly, the driver’s seat and a stone’s throw lose their idiomaticity when inverted to the seat of the driver and the throw of a stone respectively. Thus, collocation and idiomatic relations are not experimentally-viable genitives for testing alternation.

The choice of syntactic frame is contingent upon other linguistic factors such as animacy, semantic roles, syntactic weight, definiteness, and phonology, along with non-linguistic factors such as topicality. On the basis of cross-linguistic analyses, Rosenbach (2008: 153) proposes the animacy scale in (4), where nouns are ranked from most to least animate from left to right:

(4) animate inanimate

human N > animal N > collective N > temporal N > locative N > common N

the boy’s > the dog’s > the company’s > Mondays’ > London’s > the building’s

bike collar director mail suburb door

Animacy of a noun, however, is also known to interact with semantic role in a variety of languages where the genitive alternation is possible (2008). The PR role in English, for instance, is more likely associated with a prenominal genitive as in the boy’s eyes preferred to the eyes of the boy and John’s wife preferred to the wife of John. The role PM, on the other hand, interacts with prosodic-syntactic factors such as syntactic weight whereby syntactically ‘heavier’ or larger NPs upon which the PM role is mapped are disfavored. For this reason, a museum’s reproduction of a Victorian drawing room is favored over a reproduction of a Victorian drawing room of a museum where the PM’s weight is too heavy for an of prepositional frame (Rosenbach, 2002:35). From a phonological perspective, nouns ending in a sibilant sound or bearing regular plural –s inflection may be preferred in the of genitive to avoid doubling or overlap between the noun’s coda sound and the sibilant phoneme of possessive ’s. Lastly, choice of genitive frame is contingent upon the PR and PM’s specifications for topicality. According to Sorhein (as cited in Rosenbach, 2002: 56), if the PM expresses new information, a PR expressing given information will favor the ’s genitive, while a PM expressing given information and a PR expressing new information will favor an of genitive. Of all the aforementioned factors, Rosenbach maintains animacy to be the strongest (2008: 152, 266). We take these linguistic observations into consideration in relation to the construction of the experimental items for this study.

From a purely syntactic standpoint, Adger proposes English ’s and of genitives to take the possessive DP structures in (5) and (6) respectively (2003: Chapter 7):

(5) The destruction of the enemy

(6) The enemy’s destruction

In (5), the objects destruction and the enemy are merged in the lower NP. The object the enemy bears a case feature which need be checked by a corresponding feature in the derivation. Case feature checking is achieved by merging an n with a weak [of] case feature whose ‘weakness’ results in the enemy remaining in situ. Subsequent movement of the PM destruction to the n head yields the surface prepositional genitive. In (6), the derivation takes place in the same manner, except n enters the derivation without a case feature which is instead realized as a strong (*) case feature valued [gen] on a null D merged with the nP. The strength of the [gen] feature on the null D triggers movement of the enemy to the Spec position of the DP possessive in order to satisfy a locality condition comparable to the EPP of main clauses, checking D’s case feature and valuing it [gen]. The output of the derivation is the intended prenominal genitive. The key difference in the two structures is the merging of n with a weak [of] case feature in prepositional genitives in contrast to merging of a null D with a strong [gen] case feature in prenominal genitives. Although Adger does not provide an account of how genitive alternation may obtain from the interaction of syntactic and semantic structure in the form of a conceptual accessibility effect, in previous work I proposed that this may be uniformly explained by postulating pre-syntactic if-then rules in the lexicon (Romano, 2016:119). What is relevant from an empirical perspective is that in order to produce a genitive of frame, speakers of a language which instantiates only a prenominal genitive comparable to English ’s will either have to acquire the of case feature realized by n or be able to access it from the universal inventory of features after exposure to positive evidence.

This is indeed the case in Mandarin Chinese and Turkish where only an English ’s-like genitive is possible:

(7) fúwùyuán-de zhàopiàn

waiter-PR/GEN photo

‘the waiter’s photo’ or ‘the photo of the waiter’

(8) garson-un fotoğraf-ın-ı

waiter-PR-GEN photo-POSS-PM/ACC

‘the waiter’s photo’ or ‘the photo of the waiter’

In (7), the Chinese genitive marker de encodes the genitive, signaling fúwùyuán ‘waiter’ as possessing zhàopiàn ‘photo’ (see Hsu, 2009:102 for parallel examples). Equally, in (8), the Turkish genitive case affix un/ün/in/ın, whose morphophonological form varies depending on vowel harmony and the coda vowel or consonant of the possessor, marks garson as the entity possessing the possessive accusative marked object fotoğraf. In Turkish, the PM always agrees with the genitive marker on the PR thanks to a morphological POSS marker whose form also varies depending on vowel harmony and coda vowel or consonant, determining the so-called Gen-Poss template (Badahır, 2012). Given that only one genitive form is possible in Turkish and Chinese, conceptual accessibility effects are not possible in these languages. For the purpose of the current study, I take the view that UG constrains interlanguage grammars (i.e. Full Access position).

V Present study

The present study was guided by the following research questions: (1) Do syntactic representations of English genitives in native speakers resemble those of L2 speakers’ whose L1 instantiates only a genitive comparable to English ’s? (2) Is L1 and L2 priming of English genitives affected by conceptual accessibility in similar ways? A first prediction is that both the Chinese (Experiment 2) and Turkish groups (Experiment 3) will show comparable structural priming effects to natives (Experiment 1), as found in Flett (2006), even after the effects of conceptual accessibility are factored in, as found in Hawkins et al. (2012), if the L1 does not have an impact on the L2 speaker’s capacity to represent L2 knowledge of genitive structures. Opposite to the first prediction, if the L1 has an effect on L2 representations, then a significant difference in structural priming between the groups is expected. A third prediction is that the native-L2 speaker patterns may differ, if L2 priming is unmoderated by conceptual accessibility. A final prediction implicit to a conceptual accessibility effect is that a stronger priming effect will be found for structures where animacy alternates and the first entity is animate, namely primes such as the criminal’s description (+AN ’s –AN) and the author of the novel (+AN of –AN), against baseline primes where animacy is matched (+AN ’s +AN, –AN ’s –AN, +AN of +AN, and –AN of –AN).

VI Experiment 1

1 Method

a Participants

Twenty native speakers, 10 female and 10 male, took part in the experiment. Participants were either native American or British English-speaking students, attending the University of Michigan and University College London respectively. Their age ranged between 21 and 35 with mean of 27 and standard deviation of 4.19. Three participants, one male and two females, performed below chance on the sentence-recall task, producing less than 77% usable data. Thus, the results of 17 participants were retained for analysis.

b Materials

Participants completed two tasks, a general English proficiency and a sentence-recall experiment. The purpose of the first task was to diagnose general proficiency in English via the grammar component of the Oxford Placement Test (2002) which includes 100 items testing knowledge of a variety of English grammar structures. Scores were as follows, reported in decimals: native controls (N = 17), range .92–1, mean .95, SD = .02; L1 Chinese (N = 26), range .54–.92, mean = .76, SD = .09; L1 Turkish (N = 18), range .53–.97, mean = .71, SD = .13. A t test assuming unequal variance between the L1 Chinese and Turkish groups finds the difference in mean proficiency scores was non-significant (two-tailed: t = 1.05, df = 28, p > .05). In light of the range of proficiency in the L2 groups, I will consider its impact on priming effects in the results section of Experiment 2 and 3.

The purpose of the second task, an oral sentence recall task with simultaneous bimodal (auditory and visual) presentation (Bock and Warren, 1985; Branigan and Feleki, 1999; Hawkins et al., 2012; McDonald et al., 1993; Potter and Lombardi, 1998; Tanaka et al., 2005, 2011), was to measure priming effects in the three groups. The task was presented as a ‘Memory for Words and Numbers’ task so as to induce participants to believe both numbers and words were being targeted. A PowerPoint presentation was used to project numbers and text on slides in black 24-point Arial font, centered on white background. Participants were then required to memorize words and numbers for two types of trials for a total of 56 trials. In the three experiments considered in this article, we report and discuss only Type I trials.

Type I trials, which include the 32 critical items of interest plus 32 fillers, tested priming in the baseline condition where animacy of the two entities in the genitive construction was matched (k = 16), and the experimental condition in which the animacy of the entities was manipulated to induce word order inversion during recall (e.g. from the photo of the waiter to the waiter’s photo) (k = 16).

Type I trials were made up of seven slides as shown in Figure 1. The purpose of using fillers was to disrupt unwanted priming effects of previous genitive structures on the subsequent. Thus, the structure of fillers was deliberately differentiated from that of targets (see Appendix 1). Interposing a numerical task between the filler and the prime on the second slide, as well as between the prime and the target recall on the fifth slide served the purpose of masking the filler and prime sentences as those to be recalled. In addition, the numerical tasks helped interrupt use of working memory capacity for recall.

Figure 1.

Type I trial.

Type II trials tested the effects of negative priming on choice of genitive construction. These trials were made up of 6 slides where slides 3–6 were identical in purpose and mechanism to the fourth through seventh slides of trial I as seen in Figure 1. However, instead of a filler followed by a numerical task as in trial type I, the first slide of type II trials contained a negative prime sentence (e.g. The girl with the iphone called the reporter’s assistant on Tuesday) followed by a memory-for-words question asking participants if a given word appeared in the previous slide (e.g. ‘Did you see the word girl?’). The question slide was then followed by the prime to be recalled and the sequence already seen in Figure 1.

The purpose of using negative primes was to examine whether priming is affected by an opposite structure, hence negative priming. For instance, the reporter’s assistant in the previous example negatively primed the recall of the father of the postman. In this case, the animacy of referents is held constant but the opposite genitive structure is used. In the case of an asymmetry in the animacy of referents in the negative prime, as in the concerto’s composer where the first noun is inanimate and the second animate, the sentence to be recalled would naturally present an inverted order of referents, in addition to genitive structure, as in the author of the novel. Thus, the structure of the negative primes was deliberately opposite that of the targets, termed ‘priming sentences’ in type 1 trials for the sake of avoiding confusion. The rationale for interposing a memory-for-words question, rather than a numerical task, between the negative prime and the target was to investigate whether recall could be affected by negative priming. There were in total 24 type II trials. For more details of these trials see Romano (2014).

The sentences to be recalled in type I and II trials were composed of countable, singular, definite and referential, non-pronominal, either human animate or common inanimate nouns that were always new information. In this way, the three factors, definiteness, phonology, and topicality, previously shown to interact with animacy in the genitive alternation, were controlled. The choice of human animate and common inanimate was dictated by a need to maximize differences in animacy between the two entities as per the animacy scale previously introduced.

There were in total 32 type I trials divided in 8 conditions with 4 items each. The 4 baseline conditions were (1) +AN ’s +AN, (2) +AN of +AN, (3) –AN ’s –AN, (4) –AN of –AN, while the 4 experimental conditions included (5) –AN ’s +AN, (6) –AN of +AN, (7) +AN ’s –AN, (8) +AN of –AN. An example for each condition is shown in (9a)–(9h).

(9) a. The secretary from the village contacted the postman’s father last night.

(+AN ’s +AN)

b. The dentist in town treated the cousin of the student during the holidays.

(+AN of +AN)

c. The musician from next door kept the hotel’s basement for a month.

(–AN ’s –AN)

d. The assistant in the shop replaced the handle of the saucepan in less than an hour.

(–AN of –AN)

e. The professor at the university thanked the concerto’s composer over dinner.

(–AN ’s +AN)

f. The cameraman at the studio hid the picture of the actor after tea.

(–AN of +AN)

g. The policeman in charge recognized the criminal’s description at once.

(+AN ’s –AN)

h. The publisher in the city avoided the author of the novel for several months.

(+AN of –AN)

The critical genitive constructions were part of sentences between 10 and 12 words in length (excluding the genitive marker) with the standard structure:

[NP PP] [V] [PossP] [AdvP]

Each sentence containing a critical item was composed of a nominal modified by a PP followed by a transitive verb, the genitive construction, and an adverbial phrase. During processing of the recall slide, a participant’s recall was aided by the three prompts in the middle of the slide (Figure 1). These always consisted of the noun subject, the uninflected form of the verb, and part of or the whole adverbial phrase. For example, after the target the musician from next door kept the hotel’s basement for a month, participants were shown musician keep month, in that order.³ Four versions of each critical item were generated to form four lists. Within the four lists, the order of critical items was pseudo-randomized such that critical items were always separated by a filler and no two consecutive critical items were from the same prime condition. The same fillers were used across the four lists. For illustrative purposes, I report one of the four parallel pseudo-randomized lists of the Type I trials used in Appendix 1.

a Procedure

Participants were tested individually in a lab. All three groups completed the proficiency task, which lasted on average 20 minutes, before the main experiment. Five practice trials were worked through and questions answered before starting the main experiment. The natives and L2 speakers alike completed the main experiment on average in 40 minutes. Two 60 second breaks were allowed. All participants were instructed not to read sentences or numbers aloud and only speak when the experiment required so. Responses were recorded on standard recording devices.

b Scoring

The dependent variable was coded as a binary response representing a correctly formed ’s or of genitive. For consistency with previous studies, responses in which one of the two entities was recalled with a near synonym (e.g. picture for drawing and vice versa, picture for photo and vice versa) also counted. All incomplete responses where information pertinent to the genitive construction was incorrect or missing were labeled ‘Other’ and eliminated from the analyses. These include cases where the PR, PM, or genitive marker is altogether missing or incorrect (e.g. student for teacher, use of noun modifiers instead of genitive) but not cases where the sentence-initial NP PP phrases were incorrect. Statistically, the data is best modeled by logistic regression for which the R package Generalized Linear Mixed Effect Regression (glmer) (Jaeger, 2008) was employed. Models built in glmer conveniently allow the simultaneous inclusion of by-participant and by-item analyses (i.e. random intercept analyses), as well as the determination of variation in the dependent variable due to the interaction of any of the fixed effect factors and the random intercepts (i.e. random slopes). We adopted a maximal random structure approach in which random intercepts and slopes were included when possible, in order to guard against inflated Type I error (Barr et al., 2013). When models with and without random slopes offered similar goodness-of-fit, visual diagnostics from the DHARMa package were used to select a final model after inspecting residuals and over/underdispersion. Main effects were detected via ANOVA type III analysis using the R package car (Fox and Weisberg, 2011) which runs log–likelihood ratio χ2 tests. In R, logistic regression model coefficients are estimated in terms of log odds of the outcome, either an ’s or an of genitive. The reference level of the outcome upon which all coefficients are calculated in experiments 1, 2, and 3 is specified to the log odds of an ’s genitive.

2 Results and discussion

For the oral sentence-recall task, a total of 528 usable observations were produced, of which 223 (42%) were ’s genitives, 226 (43%) were of genitives, and 79 (15%) were other responses. The latter were nearly equally distributed across the total possible ’s (11%) and of (11%) responses. Table 1 breaks down responses further according to the syntax and animacy baseline factors (+AN +AN, –AN –AN). Table 2 is identical to Table 1 except the animacy factor levels are animacy alternation (+AN –AN, –AN +AN).

Table 1.

Native ’s and of genitive responses by syntax and baseline animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN of +AN the cousin of the student	2	(5)	54	(95)	90
+AN ’s +AN the postman’s father	52	(98)	1	(2)	96
–AN of –AN the handle of the saucepan	1	(2)	54	(98)	96
–AN ’s –AN the hotel’s basement	58	(92)	5	(8)	84

Note. n = 17; +AN animate; –AN inanimate.

Table 2.

Native ’s and of genitive responses by syntax and alternate animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN ’s –AN the criminal’s description	52	(99)	1	(1)	98
+AN of –AN the author of the novel	1	(1)	54	(99)	98
–AN ’s +AN the concerto’s composer	51	(89)	6	(11)	78
–AN of +AN the picture of the actor	5	(10)	51	(90)	80

Notes. n = 17; +AN animate; –AN inanimate.

Table 1 suggests a very strong priming effect in the native speaker group across the two levels of the syntax factor, regardless of the animacy levels at baseline. In other words, there were far more ’s than of genitives following an ’s genitive prime and more of than ’s genitives following an of genitive prime. This is also reflected in the table by the values in the priming effect column. Likewise, Table 2 suggests a very strong priming effect across all levels of syntax and animacy factors: there were more ’s genitives following an ’s than an of prime, and far more of genitives following an of than an ’s prime, regardless of the way animacy alternated. Priming is stronger, however, when the prime is at the +AN ’s –AN (98%) and +AN of –AN (98%) levels of interest. These results are consistent with the expected preference for the first entity in the prime to be animate, as found in double object/dative and active/passive alternations. To further explore the differences in priming strength, an ANOVA Type III analysis in R using a model with syntax and animacy as fixed effects and participant and item random effects was run. The main effect of syntax was highly significant (χ2 (1) = 17.5, p < .001) confirming a strong priming effect, but neither an animacy effect (χ2 (3) = 3.38, p = .3) nor an interaction between syntax and animacy (χ2 (3) = 3.15, p = .3) were present.

To further illuminate on any possible interaction between syntax and animacy and retain statistical power, the two fixed-effects were reinserted into a new model transformed into a single syntax–semantics factor with eight levels matching the eight conditions shown in (9a)–(9h). An ANOVA type III analysis of a model comprising the syntax–semantics factor as fixed effect, random intercepts for participants and items, and a random slope for syntax–semantics by participants finds a significant main effect of syntax–semantics (χ2 (7) = 21.28, p < .01). Among the pairwise comparisons analyzed, I focus on those levels of the factor expected to show a conceptual accessibility effect (Table 3); for the full results, see Appendix 2.

Table 3.

Meaningful comparisons for conceptual accessibility effect.

Reference level	Comparison level	ß	SE	Wald Z	p
	+AN ’s +AN	1376	(3345)	411	***
	+AN of +AN	39	(1929)	20
+AN of –ANthe author of the novel	+AN ’s –AN	1198	(2998)	399	***
	–IN ’s +AN	1218	(3028)	402	***
	–IN ’s –IN	1212	(3077)	394	***
	–IN of +AN	282	(1789)	158
	–IN of –IN	14	(1736)	81
	+AN ’s +AN	178	(1660)	107
	+AN of +AN	–1159	(2838)	–408	***
+AN ’s –ANthe concerto’s composer	+AN of –AN	–1198	(2998)	–399	***
	–AN ’s +AN	20	(1355)	14
	–AN ’s –AN	14	(1301)	10
	–AN of +AN	–915	(2341)	–391	***
	–AN of –AN	–1057	(2663)	–396	***

Notes. n = 17; +AN animate; –AN inanimate; ***p < 001. Outcome is log odds of an ’s genitive response.

At the +AN of –AN reference level, the ß coefficients for all the comparison levels are positive, indicating that the log odds of an ’s genitive are higher than the reference level. Although this result suggests +AN of –AN primes are the least likely to be inverted, consistent with predictions, the comparisons to the baseline –AN of –AN and +AN of +AN levels and the experimental –AN of +AN levels fail to reach statistical significance. As for the second reference level, Table 3 also indicates the likelihood of inversion at the +AN ’s –AN level was rather low. As in the +AN of –AN level, however, differences fail to reach statistical significance compared to baseline +AN ’s +AN and –AN ’s –AN and experimental –AN ’s +AN levels. The positive ß coefficient for the +AN ’s +AN level suggests it to be the only level lower than the +AN of –AN for likelihood of inversion. Overall, the results from Experiment 1 are only partially consistent with conceptual accessibility and previous research documenting earlier serialization of animate entities in native speakers (Branigan et al., 2008; Hawkins et al., 2012; McDonald et al., 1993; Rosenbach, 2002, 2005, 2008; Tanaka et al., 2005; Tanaka et al., 2011 inter alia). To investigate whether L2 English representations differ from native speakers’, the experiment was replicated with Chinese speakers of English in Experiment 2.

VII Experiment 2

1 Method

a Participants

Twenty-seven L1 Chinese speakers, 15 female and 12 male, took part in the experiment. Participants were mostly undergraduate and graduate students recruited either from a university in the UK or in China. They declared having knowledge of no L2s other than English. Their age ranged between 16 and 31 with mean of 23.7 and standard deviation of 2.60. One male participant was excluded due to 66% of responses being unusable (i.e. 66% Other), leaving a total of 26 participants.

b Materials

These were the same as in Experiment 1.

2 Results and discussion

A total of 771 usable observations were produced in the oral sentence-recall task, broken down into 310 (40%) ’s genitives, 320 (42%) of genitives, and 141 (18%) Other responses. The latter were equally distributed across the ’s and of priming condition, amounting to 17% of possible responses. Table 4 breaks down the results according to the syntax and animacy factor (baseline). Table 5 is identical to Table 4 except the animacy factor levels are the experimental.

Table 4.

L1 Chinese ’s and of genitive responses by syntax and baseline animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN of +ANthe cousin of the student	20	(24)	64	(76)	52
+AN ’s +ANthe postman’s father	71	(96)	3	(4)	92
–AN of –ANthe handle of the saucepan	1	(1)	79	(99)	98
–AN ’s –ANthe hotel’s basement	65	(87)	10	(13)	74

Notes. n = 26; +AN animate; –AN inanimate.

Table 5.

L1 Chinese ’s and of genitive responses by syntax and alternate animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN ’s –ANthe criminal’s description	69	(89)	9	(11)	77
+AN of –ANthe author of the novel	3	(4)	72	(96)	92
–AN ’s +ANthe concerto’s composer	68	(81)	16	(19)	62
–AN of +ANthe picture of the actor	13	(18)	67	(82)	64

Notes. n = 26; +AN animate; –AN inanimate.

Table 4 shows variation in the magnitude of priming effect in the baseline condition, with effects ranging between 52% after +AN of +AN primes to 98% after –AN of –AN primes (last column). As found in Experiment 1, the strongest priming effects were registered at the +AN of +AN and –AN of –AN levels. Table 5 also shows a moderate range in priming effects for the experimental levels. The two highest effects were after the primes expected to show stronger priming, +AN ’s –AN (77%) and +AN of –AN (92%). These results mimic the native speakers’ as reported in Table 1 and 2. An ANOVA Type III analysis in R using a model with syntax and animacy as fixed effects, participant and item random effects, and a random slope with the interaction of syntax and animacy by participants, finds a highly significant main effect of syntax (χ2 (1) = 27.12, p < .001) indicating a clear priming effect, but no main effect of animacy (χ2 (3) = 2.08, p = .5) nor interaction effect between the two (χ2 (3) = 4.92, p = .17).

To look at the results in more detail and retain statistical power, pairwise comparisons were obtained from a new model where the two fixed-effects were recoded as a single syntax–semantics factor with eight levels matching; (9a)–(9h). As language proficiency was not homogenous, a continuous predictor for L2 proficiency level whose values were those obtained from the English proficiency task, was added. An ANOVA type III analysis of a model comprising the syntax–semantics factor, a proficiency covariate centered on its mean, and random intercepts for participants and items finds a significant main effect of syntax–semantics (χ2 (7) = 109.51, p < .001), no main effect of proficiency (χ2 (1) = .26, p = .6) and no significant interaction between the two (χ2 (7) = 2.79, p =.90). Figure 2 represents the model’s predicted probability of an ’s genitive across proficiency using the levels of syntax–semantics as a grouping variable.

Figure 2.

Probability of an ’s genitive response as a function of L2 proficiency and the 8 levels of syntax–semantics in Chinese speakers of English.

Each colored line indicates a trend for each level of syntax–semantics in the probability of an ’s genitive from lower to higher proficiency levels, while the shaded colored areas depict the confidence intervals for the associated colored trendline. The narrower the confidence interval, the more precise the model’s estimation of probability. The levels of interest are the blue (+AN ’s –AN) and purple (+AN of –AN) lines. The +AN ’s –AN trendline shows a steady increase with proficiency, plateauing at around 82% percent proficiency and 98% probability of an ’s genitive. In comparison to other relevant levels, the priming effect tops the levels in yellow and orange, once the 70% proficiency mark is reached, only remaining lower than the red line (+AN ’s +AN) which manifests the strongest priming effect across the board. As shown for the native speakers in Table 3, then, the probability of +AN ’s –AN priming is high, consistent with the conceptual accessibility effect. Pairwise comparisons, nevertheless, indicate the differences highlighted fail to reach significance (Appendix 2). The picture for the purple line resembles the blue line’s. The +AN of –AN primes were hardly ever inverted, as indicated by the steady flat trend starting at around the 2% probability of an ’s genitive on the y axis. Moreover, they were less likely to lead to inversion during recall than the levels represented by the green and brown lines, a result supported by inferential statistics (Appendix 2) (+AN of +AN: ß = 2.23, SE = .84, z = 2.655, p < .01; –AN of +AN: ß = 1.78, SE = .85, z = 2.080, p < .05). In summary, both the natives in experiment 1 and L1 Chinese speakers in Experiment 2 showed a structural priming effect, as well as conceptual accessibility effect when animacy alternates. Of the two levels expected to show this effect, however, only the +AN of –AN level responses differed significantly from other baseline and experimental levels in the L1 Chinese group. Crucially, the similarity of results between the two groups suggests L2 learners who lack an of genitive in the L1 are capable of building native-like representations for it. To confirm this finding, the experiment was replicated with a comparably proficient cohort of L1 Turkish L2 speakers, given Turkish, like Chinese, only instantiates a genitive similar to English ’s.

VIII Experiment 3

1 Method

a Participants

Twenty-four Turkish speakers, 12 female and 12 male, participated in the experiment. Participants were all undergraduate and graduate students or staff recruited from a university in Istanbul, whose age ranged between 19 and 31 with mean of 23 and standard deviation of 1.8. Results from six participants of Kazakh descent, 4 male and 2 female, were later discarded due to Russian being their principal L2 in addition to Turkish and English. Like English, Russian expresses possession by more than one structure, thus adding a confound to the data.

b Materials

Same as in Experiment 1 and 2.

2 Results and discussion

A total of 519 usable observations emerged from the oral sentence-recall task, of which 152 (29%) were ’s genitives, 263 (50%) of genitives, and 104 (20%) Other responses. The latter were equally distributed across the maximum possible ’s and of responses (19%). Table 6 reports responses by syntax and baseline animacy levels whilst Table 7 reports results by syntax and alternated animacy levels.

Table 6.

L1 Turkish ’s and of genitive responses by syntax and baseline animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN of +ANthe cousin of the student	4	(8)	45	(92)	84
+AN ’s +ANthe postman’s father	43	(88)	6	(12)	76
–AN of –ANthe handle of the saucepan	1	(1)	48	(99)	88
–AN ’s –ANthe hotel’s basement	29	(53)	25	(47)	6

Notes. n = 18; +AN animate; –AN inanimate.

Table 7.

L1 Turkish ’s and of genitive responses by syntax and alternate animacy primes.

Prime	Response				Priming effect %
	’s genitive		of genitive		Priming effect %
	Count	% total	Count	% total
+AN ’s –ANthe criminal’s description	32	(61)	20	(39)	22
+AN of –ANthe author of the novel	2	(3)	58	(97)	94
–AN ’s +ANthe concerto’s composer	39	(75)	13	(25)	50
–AN of +ANthe picture of the actor	2	(4)	48	(96)	92

Notes. n = 18; +AN animate; –AN inanimate.

In Table 6, the magnitude of priming in the baseline condition ranges from a low 6% with –AN ’s –AN primes to a high 88% with –AN of –AN primes. Likewise, Table 7 shows a large range in priming effects from as low as 22% in the +AN ’s –AN condition to a high of 94% in the +AN of –AN condition. On the surface, these results differ from those obtained for the L1 Chinese, namely 52–98% in Table 4 (baseline) and 62–92% in Table 5 (experimental), and natives, namely 84–96% in Table 1 (baseline) and 78–98% in Table 2 (experimental), where ranges were much more confined. The effect of priming at the levels of interest differ widely: at the +AN ’s –AN level priming is weak, implying a high degree of inversion, whereas at the +AN of –AN level priming is strong and inversions rare. An ANOVA Type III analysis of a model with syntax and animacy as fixed effects, participant and item random effects, and a random slope for syntax and animacy by participants, finds a highly significant main effect of syntax (χ2 (1) = 17.23, p < .001) indicating a clear priming effect, but no main effect of animacy (χ2 (3) = 1.97, p = .5) nor a main interaction between the two (χ2 (3) = .98, p = .80). These effects are comparable to the ones found in Experiment 1 and 2.

The syntax and animacy factors were recoded into a single syntax–semantics factor and a continuous covariate for proficiency centered on its mean was fit. An ANOVA type III analysis of a model with syntax–semantics and proficiency fixed effects, random intercepts for participants and items, and a random slope for syntax–semantics by participant, shows a significant main effect of syntax–semantics (χ2 (7) = 60.52, p < .001), no main effect of proficiency (χ2 (1) = .31, p=.57) and no significant interaction between the two (χ2 (7) = 2.29, p = .94). Figure 3 depicts the model’s predicted probability of an ’s genitive as a function of proficiency, grouped by the syntax–semantics factor.

Figure 3.

Probability of an ’s genitive response as a function of L2 proficiency and the 8 levels of syntax–semantics in Turkish speakers of English.

Like Figure 2, the colored lines in Figure 3 indicate a trend in the probability of an ’s genitive for each level of syntax–semantics from lower to higher proficiency levels. The shaded colored areas denote the upper and lower confidence interval limits for the color-matched trendlines. The blue and purple lines, respectively the +AN ’s –AN and +AN of –AN levels, are of interest. The blue trendline shows a steady decrease with proficiency, reflecting less than 50% probability of ’s priming when participant proficiency levels near 93%. In comparison to other levels, the priming effect of +AN ’s –AN is only higher than the yellow (–AN ’s –AN) and only once the 70% proficiency mark is reached. This contrasts with the trends already shown for Experiment 2 in Figure 2 where the blue line topped both the yellow and orange. As in Experiment 1 and 2, though, the difference to other levels is not confirmed by the inferential statistics (Appendix 2). The +AN ’s +AN red line is the level with the highest rate of priming compared to all levels apart from –AN ’s +AN where the difference remains numerical (+AN ’s –AN: ß = −1.89, SE = .746, z = −2.53, p = .011; –AN ’s –AN: ß = −2.11, SE = .737, z = −2.97, p < .01; –AN ’s +AN: ß = −1.08, SE = .750, z = −1.445, p = .14). The purple line trend shows the probability of inverting an +AN of –AN prime to an ’s genitive is consistently small, corresponding to a strong priming effect for an of genitive, in line with a conceptual accessibility. However, inferential statistics reveal that any differences between +AN of –AN and other levels are non-significant (Appendix 2). In summary, the L1 Turkish data overlaps with the native and L1 Chinese in that a priming effect for syntax but no main effects for animacy and no interaction between the two were found. This similarity suggests L2 learners who lack an of genitive in the L1 are nevertheless capable of building native-like representations for it. However, unlike the results in Experiment 1 and 2, the L1 Turkish produced an unexpected high rate of inversion in the +AN ’s –AN condition, particularly at higher proficiency levels, although the differences in rate of inversion to other relevant levels did not reach statistical significance.

To gauge whether there were differences at individual syntax–semantics levels between the three groups, a final analysis was computed. An ANOVA type III analysis on a model including syntax–semantics and group fixed effects alongside by-participant and item random intercepts finds a strong main effect of syntax–semantics (χ2 (7) = 153.71, p < .001), no effect of group (χ2 (2) = 4.76, p =.11), and a significant interaction between the two (χ2 (14) = 27.69, p = .015). The lack of a group effect indicates no significant difference between the three samples in the probability of a priming effect for an ’s genitive, the reference level of the model, after the syntax–semantics factor is controlled. Given the rate in production of postnominal of genitives is reciprocal to the priming effect of ’s, this result confirms L2 speakers have similar patterns to native speakers, despite their L1 only representing an English-like ’s genitive. To ascertain whether any differences in conceptual accessibility effects between the natives and L2 speakers exist, pairwise comparisons were computed (Appendix 2). I focus, in particular, on differences between groups in the effect of +AN ’s –AN and +AN of –AN primes. At the +AN ’s –AN level, there was no significant difference between the native and Chinese (ß = 2.05, SE = 1.195, z = 1.71, p = .08) but a highly significant one between the native and Turkish (ß = −4.07, SE = 1.182, z = −3.451, p < .001) where the large negative ß indicates much lower odds of a priming effect, and by implication much higher odds of inversion by the Turkish group. An analogous effect was found between the Chinese and Turkish groups (ß = −2.02, SE = .670, z = −3, p < .01). At the +AN of –AN level, there was no significant difference between the natives and Chinese (ß = −1.08, SE = 1.303, z = –.829, p = .40), the native and Turkish (ß = .94, SE = 1.382, z = .686, p = .49), and the Chinese and Turkish groups (ß = −.13, SE = 1.056, z = −1.25, p = .90). Thus, although the effects of conceptual accessibility where fairly similar between the natives and Chinese, the same cannot be said of the Turkish group who tended to invert +AN ’s –AN primes to a much higher extent.

IX General discussion

The purpose of this study was two-fold. First, it asked whether syntactic representations of English genitives in natives resemble those of L2 speakers’ whose L1 makes available only an English ’s-like genitive. A second purpose was to determine whether L1 and L2 structural priming are affected by conceptual accessibility, operationalized as an animacy effect, in a comparable manner. In the present study, the choice of an of versus ’s English genitive was predicted to be subject to an animacy effect, whereby structures where an asymmetry in the animacy of participants is present, as in the author of the novel or the criminal’s description, are more likely to be primed (or less likely to be inverted) than baseline structures where animacy is matched and structures where the animate entity follows the inanimate.

With regards to the first question, in all three experiments a significant structural priming effect was found, amounting to L1 English, L1 Chinese, and L1 Turkish speakers producing far more ’s than of genitives after an ’s genitive prime but far more of than ’s genitives after an of genitive prime. Moreover, the group analysis in Experiment 3 showed that there were no statistically significant differences in the structural priming effect across groups, after the effect of animacy effects is controlled. This finding corroborates previous within-L2 and between L1–L2 research whose design purposefully selected L1s that did not represent the L2 structure (Flett, 2006; Kim, 2010). If L2 learners transferred their L1 representation of the single prenominal genitive allowed in their L1, a majority of prenominal genitives would have obtained in both the ’s and of condition, contrary to fact. The similarity of structural priming effects across groups found in this study are consistent with the claim in Bernolet et al. (2013) that L2 speakers form a new combinatorial node which is added to a shared representation. One open question is how L2 speakers acquire the node tied to the of genitive structure which they cannot inherit from the L1. It is possible that after exposure to the novel genitive in the input, L2 speakers build a new node (i.e. mapping) that is L2 English-specific in Bernolet et al.’s terms, or access a feature which is made universally available by biological endowment in UG terms. The latter, which I prefer to Bernolet et al.’s (2013) combinatorial node approach, can be conceived as a matter of accessing the [of] case feature from the UG inventory and building the associated derivation already shown in (5) from Adger (2003) via the computational mechanisms intrinsic to the language faculty.

An issue tied to the activation of the [of] feature is how L2 learners achieve this. It is plausible that in the process of L2 acquisition, speakers exploit positive evidence in the input in the form of PPs without possessive meaning (e.g. time, location) in order to produce the novel genitive. These non-DP phrases are subsequently mapped onto those of genitives with meaning other than possessive, chiefly partitive, descriptive, appositive, idiomatic, collocations, subjective, and objective, introduced above. In practice, L2 speakers match PPs to possessive-like NPs such as a congressional medal of honor, only later activating the [of] characterizing ‘true’ possessives represented syntactically as postnominal genitives.

The target-like representations proper of the L2 speakers in this study should not be mistaken for a complete acquisition of the functional category DP. Instead, they should be taken to mean exactly what this research set out to examine: that the speech production mechanisms tied to the English genitive alternation are very similar in native and non-native speakers. Whether the L2 speakers in the present study have command of other features implicated by genitive DPs like definiteness and number (Adger, 2003: 279) is yet to be seen. Certainly, a flood of studies has already been conducted on L2 judgments and interpretation of various aspects of the English DP by speakers of languages which lack DP-related morphology such as articles or combine DP-related syntactic and semantic features such as definiteness and plural in very different ways than English (Snape et al., 2009, 2013 among others). These studies have shown that the acquisition and ultimate attainment of the L2 English DP is problematic due to L1 transfer effects (Robertson, 2000; White, 2003) but that in some cases L2 speakers can successfully recover from such effects (Ionin and Montrul, 2010).⁴

It has been shown that the representation of the genitive alternation in native and L2 speakers with L1 Chinese is fairly similar, reflected also by a strikingly analogous proportion of prenominal and postnominal genitives produced: 42% ’s versus 43% of genitives in Experiment 1 against 40% ’s versus 42% of genitives in Experiment 2. Although these proportions differ in Experiment 3 (29% versus 52%) for the L1 Turkish —a difference I turn to explain below— the rate of Other responses was always balanced across the two prime conditions in all three samples. These findings, on the one hand, improve our understanding of the nature of L1 versus L2 syntactic representation and the scope of the SRH. On the other, they crucially indicate an important continuity between L1 and L2 language production.

In relation to the question of conceptual accessibility effects, broad similarities between the native and Chinese speakers were found. The priming effect at the +AN of –AN and +AN ’s –AN levels was higher than the relevant baseline and experimental levels in both Experiment 1 and 2, consistent with the prediction that conceptual accessibility affects English genitives in ways similar to active/passive and double object constructions. Pairwise comparisons, however, did not reach significance except in the L1 Chinese group’s priming rates for an ’s genitive at the +AN of –AN level. It is plausible the lack of statistical significance was due to an unwanted effect of the +AN ’s –AN and its reciprocal –AN of +AN primes. To put it plainly, the criminal’s description inverts to the description of the criminal which is ambiguous between two possible PR–PM relations, the possessive and the descriptive. Participants may have misintended the PR function associated to of the criminal as modifier to the head NP, more simply qualifying the content of the description, ascribing an undesired interpretation. By the same token, primes such as the picture of the actor are subject to the same ambiguous interpretation whereby the picture may be misinterpreted as depicting rather than belonging to the actor. If this account is correct, it can also explain the unexpected low priming effect in the L1 Turkish group of Experiment 3 at the +AN ’s –AN level which resulted in significant differences to the native and L1 Chinese levels. One possibility is that the L1 Turkish speakers were more susceptible than the other groups to the ambiguity due to an L1 transfer effect. The potential confound is currently being addressed in a series of experiments with native and L1 Chinese speakers where the ambiguity has been remedied (Romano, in press).

Another significant pattern emerged from the data with regards to animacy effects. In particular, the experimental +AN ’s –AN and +AN of –AN conditions where an asymmetry was believed to result in earlier retrieval of the animate lemma from the lexicon were expected to show stronger priming effects than the four baseline conditions. On the contrary, one baseline condition, the +AN ’s +AN, had especially strong priming effects in all three experiments. To the best of my knowledge, this is a unique contribution to the theoretical and experimental field insofar as this study is the first to experiment with the interaction of the genitive alternation and conceptual accessibility. It confirms previous work (Romano, 2016), as well as analysis in theoretical linguistics (Rosenbach, 2008: 152), suggesting that in the lead up to L1 and L2 sentence production, the semantic role PR is privileged by assignment to first position.

In conclusion, the similarity of conceptual accessibility effects found in the present study between native and L2 speakers are reconcilable with views proposing a congruence between L1 and L2 sentence production mechanisms (Hawkins et al. 2012; Kim, 2010). In particular, I would like to hypothesize a Basic Continuity between L1 and L2 sentence production in the case of properties which require integration of syntactic and semantic structure. It has been shown that L2 groups integrate the semantic feature animacy and the syntactic [gen] case feature in native-like ways. It is hoped that our current experiments and future research will show that an improved version of the items utilized in the current study leads to a statistically warranted conceptual accessibility effect. It is worth pointing out that the Basic Continuity Hypothesis of L1 and L2 production that I propose is in line with earlier versions of the Interface Hypothesis (Sorace and Filiaci, 2006) claiming that L2 learners are more successful with grammar-internal than grammar-external interface properties, namely properties integrating grammatical information from formal domains such as syntax and semantics (Sorace, 2011; Tsimpli and Sorace, 2006). Moreover, the Basic Continuity Hypothesis has strong implications for future research in L1 and L2 structural priming and cognitive models of sentence production.

Supplementary Material

Supplementary Material, SLR_729423_Appendix_2 – The basic continuity hypothesis of L1 to L2 production

Supplementary Material, SLR_729423_Appendix_2 for The basic continuity hypothesis of L1 to L2 production by Francesco Romano in Second Language Research

Footnotes

Appendix 1 Acknowledgements

I would like to thank the audience of EUROSLA 26 (University of Jyväskylä, Finland, August 26th 2016) for comments received on a preliminary talk stemming from this paper. My gratitude also goes out to the School of Languages at Sabancı University where the L1 Turkish data collection took place and Ayşe Gürel for her comments on an earlier version of the manuscript. Finally, I am indebted to the editor of SLR and two anonymous reviewers for their dedication to improving my work and fast turnaround time for review. All remaining errors are mine.

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplementary Material

Supplementary material is available for this article online.

Notes

References

Adger

(2003) Core syntax: A minimalist approach. Oxford: Oxford University Press.

Badahır

(2012) Structural priming in Turkish genitive-possessive constructions. Unpublished doctoral dissertation, Middle East Technical University, Ankara, Turkey.

Barr

Levy

Scheepers

Tily

(2013) Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language 68: 255–78.

Bernolet

Hartsuiker

Pickering

(2007) Shared syntactic representations in bilinguals: Evidence for the role of word-order repetition. Journal of Experimental Psychology 33: 931–49.

Bernolet

Hartsuiker

Pickering

(2009) Persistence of emphasis in language production: A cross-linguistic approach. Cognition 112: 300–17.

Bernolet

Hartsuiker

Pickering

(2012) Effects of phonological feedback on the selection of syntax: Evidence from between-language syntactic priming. Bilingualism: Language and Cognition 15: 503–16.

Bernolet

Hartsuiker

Pickering

(2013) From language-specific to shared syntactic representations: The influence of second language proficiency on syntactic sharing in bilinguals. Cognition 127: 287–306.

Bock

(1986) Syntactic persistence in language production. Cognitive Psychology 18: 355–87.

Bock

Dell

Chang

Onishi

(2007) Persistent structural priming from language comprehension to language production. Cognition 104: 437–58.

10.

Bock

Warren

(1985) Conceptual accessibility and syntactic structure in sentence formulation. Cognition 21: 47–67.

11.

Branigan

Feleki

(1999) Conceptual accessibility and serial order in Greek language production. In: Hahn

Stones

(eds) Proceedings of the 21st Annual Conference of the Cognitive Science Society (pp. 96–101). Erlbaum, Mahwah, NJ.

12.

Branigan

Pickering

Mclean

(2000) Priming prepositional-phrase attachment during language comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition 31: 468–81.

13.

Branigan

Pickering

Tanaka

(2008) Contributions of animacy to grammatical function assignment and word order during production. Lingua 118: 172–89.

14.

Cai

Pickering

Yan

Branigan

(2011) Lexical and syntactic representations in closely related languages: Evidence from Cantonese–Mandarin bilinguals. Journal of Memory and Language 65: 431–45.

15.

Chen

Jia

Wang

Dunlap

Shin

(2013) Is word-order similarity necessary for cross-linguistic structural priming? Second Language Research 29: 375–89.

16.

Desmet

Declercq

(2006) Cross-linguistic priming of syntactic hierarchical configuration information. Journal of Memory and Language 54: 610–32.

17.

Ferreira

Bock

Wilson

Cohen

(2008) Memory for syntax despite amnesia. Psychological Science 19: 940–46.

18.

Flett

(2006) A comparison of syntactic representation and processing in first and second language production. Unpublished doctoral dissertation, University of Edinburgh, UK.

19.

Fox

Weisberg

(2011) An R companion to applied regression. 2nd edition. London: Sage.

20.

Gries

(2005) Syntactic priming: A corpus-based approach. Journal of Psycholinguistic Research 34: 365–99.

21.

Hartsuiker

Bernolet

(2015) The development of shared syntax in second language learning. Bilingualism: Language and Cognition 20: 219–34.

22.

Hartsuiker

Bernolet

Schoonbaert

Speybroeck

Vanderelst

(2008) Syntactic priming persists but the lexical boost decays: Evidence from written and spoken dialogue. Journal of Memory and Language 58: 214–38.

23.

Hartsuiker

Kolk

HHJ

(1998a) Syntactic facilitation in agrammatic sentence production. Brain and Language 62: 221–54.

24.

Hartsuiker

Kolk

HHJ

(1998b) Syntactic persistence in Dutch. Language and Speech 41: 143–84.

25.

Hartsuiker

Westenberg

(2000) Word order priming in written and spoken sentence production. Cognition 75: 27–39.

26.

Hartsuiker

Pickering

Veltkamp

(2004) Is syntax separate or shared between languages? Cross-linguistic syntactic priming in Spanish–English bilinguals. Psychological Science 15: 409–14.

27.

Hawkins

Althobaiti

(2012) Eliminating grammatical function assignment from hierarchical models of speech production: evidence from the conceptual accessibility of referents. Applied Psycholinguistics 35: 677–707.

28.

Hoji

(1985) Logical form constraints and configurational structures in Japanese. Unpublished PhD dissertation, University of Washington, USA.

29.

Hsu

(2009) Possessor Extraction in Mandarin Chinese. University of Pennsylvania Working Papers in Linguistics 15: issue 1, article 12. Proceedings of the 32nd Annual Penn Linguistics Colloquium. Available at: http://repository.upenn.edu/pwpl/vol15/iss1/12 (accessed September 2017).

30.

Huttenlocher

Vasilyeva

Shimpi

(2004) Syntactic priming in young Children. Journal of Memory and Language 50: 182–195.

31.

Ionin

Montrul

(2010) The role of L1-transfer in the interpretation of articles with definite plurals in L2-English. Language Learning 60: 877–925.

32.

Jackson

Ruf

(2017) The priming of word order in second language German. Applied Psycholinguistics 38: 315–45.

33.

Jaeger

(2008) Categorical data analysis: away from ANOVAs (transformation or not) and towards logit mixed models. Journal of Memory and Language 59: 434–446.

34.

Kantola

Van Gompel

RGP

(2011) Between- and within-language priming is the same: Evidence for shared bilingual syntactic representations. Memory and Cognition 39: 276–90.

35.

Kim

(2010) Syntactic representations of English in second language learners: An investigation of the process of English sentence production by bilingual speakers using a within-language (L2) structural priming paradigm. Unpublished Doctoral Dissertation, State University of New York Buffalo, New York, USA.

36.

Larson

(1988) On the double object construction. Linguistic Inquiry 19: 335–91.

37.

Loebell

Bock

(2003) Structural priming across languages. Linguistics 41: 791–824.

38.

McDonald

Bock

Kelly

(1993) Word and world order: semantic phonological and metrical determinants of serial position. Cognitive Psychology 25: 188–230.

39.

Mahowald

Futrell

AJR

Gibson

(2016) A meta-analysis of syntactic priming in language production. Journal of Memory and Language 91: 5–27.

40.

Mercan

(2016) Structural priming in L2 Turkish. In: Gürel

(ed) Second language acquisition of Turkish (pp. 313–32). Amsterdam: John Benjamins.

41.

Meijer

PJA

Fox Tree

(2003) Building syntactic structures in speaking: A bilingual exploration. Experimental Psychology 50: 184–95.

42.

Melinger

Dobel

(2005) Lexically-driven syntactic priming. Cognition 98: 11–20.

43.

Pickering

Branigan

(1998) The representation of verbs: Evidence from syntactic priming in language production. Journal of Memory and Language 39: 633–51.

44.

Pickering

Ferreira

(2008) Structural priming: A critical review. Psychological Bulletin 134: 427–49.

45.

Potter

Lombardi

(1998) Syntactic priming in immediate recall of sentences. Journal of Memory and Language 38: 265–82.

46.

Robertson

(2000) Variability in the use of the English article system by Chinese learners of English. Second Language Research 16: 135–72.

47.

Romano

(2016) Syntactic planning of English genitives in L1 and L2 production: The animacy rule model. Lingua 184: 104–21.

48.

Romano

(in press) Structural priming of English genitive alternation: more evidence for single-level syntactic planning models. Unpublished manuscript.

49.

Rosenbach

(2002) Genitive variation in English conceptual factors in synchronic and diachronic studies. New York: Mouton de Gruyter.

50.

Rosenbach

(2005) Animacy versus weight as determinants of grammatical variation in English. Language 81: 613–44.

51.

Rosenbach

(2008) Animacy and grammatical variation: Findings from English genitive variation. Lingua 118: 151–71.

52.

Salamoura

Williams

(2006) Lexical activation of cross-language syntactic priming. Bilingualism: Language and Cognition 9: 299–307.

53.

Salamoura

Williams

(2007) Processing verb argument structure across languages: Evidence for shared representations in the bilingual lexicon. Applied Psycholinguistics 28: 627–60.

54.

Scheepers

(2003) Syntactic priming of relative clause attachments: Persistence of structural configuration in sentence production. Cognition 89: 179–205.

55.

Schoonbaert

Hartsuiker

Pickering

(2007) The representation of lexical and syntactic information in bilinguals: Evidence from syntactic priming. Journal of Memory and Language 56: 153–71.

56.

Segaert

Kempen

Petersson

Hagoort

(2013) Syntactic priming and the lexical boost effect during sentence production and sentence comprehension: An fMRI study. Brain and Language 124: 174–83.

57.

Shin

Christianson

(2009) Syntactic processing in Korean–English bilingual production: Evidence from cross-linguistic structural priming. Cognition 112: 175–80.

58.

Shin

Christianson

(2012) Structural priming and second language acquisition. Language Learning 62: 931–64.

59.

Sorace

(2011) Pinning down the concept of ‘interface’ in bilingualism. Linguistic Approaches to Bilingualism 1: 1–33.

60.

Sorace

Filiaci

(2006) Anaphora resolution in near-native speakers of Italian. Second Language Research 22: 339–68.

61.

Snape

Garcia-Mayo

Gürel

(2013) L1 transfer in article selection for generic reference by Spanish, Turkish and Japanese L2 learners. International Journal of English Studies 13: 1–28.

62.

Snape

Garcia-Mayo

Gürel

(2009) Spanish, Turkish, Japanese and Chinese L2 learners’ acquisition of generic reference. In: Bowles

et al . (eds) Proceedings of the 10th Generative Approaches to Second Language Acquisition Conference. Somerville, MA: Cascadilla Proceedings Project.

63.

Tanaka

Branigan

Mclean

Pickering

(2011) Conceptual influences on word order and voice in sentence production: Evidence from Japanese. Journal Memory Language 65: 318–30.

64.

Tanaka

Branigan

Pickering

(2005) The role of animacy in Japanese sentence production. Unpublished paper presented at the CUNY Sentence Processing Conference Tucson.

65.

Tsimpli

Sorace

(2006) Differentiating Interfaces: L2 performance in syntax–semantics and syntax–discourse phenomena. In: Bamman

Magnitskaia

Zaller

(eds) Proceedings of the 30th annual Boston University Conference on Language Development [BUCLD 30] (pp. 653–64). Somerville MA: Cascadilla Press.

66.

White

(2003) Fossilization in steady state L2 grammars: Implications of persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6: 129–41.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.66 MB