Frequency at the syntax–discourse interface: A bidirectional study on fronting options in L1/L2 German and L1/L2 English

Abstract

The present study investigates whether second language (L2) speakers are sensitive to the information-structural constraints and frequency distributions at the syntax–discourse interface in the L2. L1-German–L2-English and L1-English–L2-German speakers completed a speeded naturalness judgment task. For sentences presented in broad or narrow-focus contexts, they judged the naturalness of fronted locative (LP) and temporal (TP) adverbial phrases and fronted objects in both English and German. English and German differ in the frequency with which they employ these constructions. With high-frequency fronted-LP and TP sentences, both the L2 English and L2 German speakers exhibited equivalent judgments as their L1 counterparts, in spite of differences in the perceived naturalness and relative frequency of these constructions in English vs. German. Like L1 speakers, L2 English and L2 German speakers also judged the less-frequent fronted objects as more natural in narrow-focus than broad-focus contexts, showing successful acquisition at the syntax–discourse interface. However, they judged fronted object sentences as more natural overall than their respective L1 counterparts in both English and German. Together, these findings suggest that convergence at the L2 syntax–discourse interface is possible per se, but that lower construction frequency in the input entails persistent overgeneralization of non-canonical options in the L2.

Keywords

frequency information structure non-canonical word syntax–discourse interface word order

I Introduction

Mastering a second language (L2) comprises acquiring the grammatical options of the target language and identifying the appropriate contexts of their use. While adult L2 speakers acquire the word order options in the target language (TL) at advanced proficiency levels, they often continue to struggle to restrict them to the discourse contexts in which they are pragmatically appropriate (e.g., Carroll et al., 2000; Sorace, 2011). In other words, adult L2 speakers fail to converge on the discourse-appropriate use of alternating L2 word orders, instead exhibiting residual optionality and first language (L1) transfer at the syntax–discourse interface, i.e., the mapping of discourse constraints on word order (e.g., Belletti et al., 2007; Hopp, 2009; Roberts et al., 2008; Sorace and Filiaci, 2006). Sorace formalized this observation in the Interface Hypothesis (e.g., Sorace, 2005, 2011), arguing that syntax–discourse mappings are vulnerable in advanced L2 acquisition. More recently, Slabakova (2015) proposed that these difficulties are aggravated if discourse-to-syntax mappings differ in the frequency with which they occur across L1 and L2.

However, it remains an open question whether L2 speakers can successfully configure discourse–syntax mappings and approximate the frequency with which different word orders are employed in the L2 across different discourse contexts, especially if analogous word order options in the L1 are realized at different frequencies (Slabakova, 2015). Only if they have mastered both the mappings and the frequency distributions in the L2 can they be said to have fully acquired L2 word order options. To address this question, the present study explores the syntax–discourse interface in a bidirectional study of fronting tendencies in German and English declarative main clauses by L1 and L2 speakers. The left periphery of the sentence acts as the link between discourse and syntax where topic and focus are marked to construct the sentential information structure (IS; e.g., Birner and Ward, 1998; Chafe, 1974; Gundel, 1988; Lambrecht, 1994; Ward, 1988). In investigating the use of the left periphery in German and English, this article serves a dual purpose. First, it establishes a baseline of fronting tendencies in German and English by testing the acceptability of non-subject-initial sentences in different contexts among L1 speakers of each language. In doing so, the article charts microtypological variation in argument and adjunct fronting between German and English, i.e., fine-grained differences across comparable constructions within a single language family in the use of the sentential left periphery and its links to discourse information. Second, the article explores how adult L2 speakers acquire non-canonical word orders in the L2. Specifically, we address whether L2 speakers are sensitive (1) to information-structural constraints on non-canonical fronting in the L2 at the syntax–discourse interface and (2) to frequency differences between fronting options in the target language. In this respect, German and English provide an ideal test bed for current models of L2 acquisition, because options for fronting adjuncts and arguments differ in the frequencies with which they are employed in the two languages.

The article is structured as follows: First, we provide an overview of previous research on the L2 syntax–discourse interface in general and fronting tendencies in an L2 in particular. We then review the word order options in the left periphery of German and English main clauses. Subsequently, we present the results of our study and, finally, discuss these findings from a cross-linguistic perspective and with respect to the Interface Hypothesis.

II The Interface Hypothesis

A large body of research on adult L2 speakers reports that late-learning L2 speakers have persistent difficulties with word order optionality, in particular when discourse constrains the syntactic options available for expressing a particular interpretation. For instance, Belletti et al. (2007) show that near-native L2 English speakers have difficulty coordinating the expression of subjects with topic and focus (see also Roberts et al., 2008; Sorace and Filiaci, 2006; Wilson, 2009; but see Kras, 2008; Slabakova et al., 2012). Further, intermediate or advanced L2 speakers of Italian and Spanish fail to converge on the discourse constraints of left-dislocated clitics, although more advanced L2 speakers approximate native performance (e.g., Donaldson, 2011, 2012; Ivanov, 2012; Leal et al., 2017; but see also Valenzuela, 2006). Within a generative framework, Sorace (2005, 2011) captures these difficulties in her Interface Hypothesis, according to which external interfaces of syntax, i.e., interfaces with the non-linguistic cognitive system, are particularly vulnerable in bilingual and adult L2 acquisition. Most research framed within the Interface Hypothesis has focused on mappings between syntax and discourse. According to the Interface Hypothesis, coordinating grammatical and discourse information exceeds the processing capacity of bilingual and L2 speakers, leading to residual optionality. In other words, since L2 speakers cannot compute the discourse restrictions on syntactic variation, they allow for more variation in word order than L1 speakers and extend variation to contexts that do not license optionality in the target language. Several authors have argued that the syntax–discourse interface presents difficulty irrespective of whether the L1 and L2 exhibit analogous or different interface mappings (Serratrice et al., 2009; Sorace, 2011). At the same time, many studies report L1 effects for interface properties since adult L2 speakers rely on L1 discourse-to-syntax mappings in order to ease the processing burden of computing the TL mappings (Sorace and Filiaci, 2006). L1 effects have been widely been reported in the overproduction of overt subjects, e.g., by English L2 speakers of Italian or Spanish (e.g., Belletti et al., 2007), differences in the interpretation of pronouns (Roberts et al., 2008; Wilson, 2009) or the transfer of Dutch IS-strategies to express focus in German scrambling (Hopp, 2009).

Similar effects also surface in the L2 acquisition of fronting options in German and English. In a dynamic picture description task that manipulated the location and given/new status of different objects, O’Brien and Féry (2015) found that English L2 speakers of German exhibited a strong preference for producing subject-initial sentences over sentences with fronted locative phrases (LPs), producing similar rates of subject-initial sentences in both their L1 English and L2 German, and they found only limited evidence of fronted LPs among their most proficient L2 German speakers (for L1-Swedish–L2-German, see also Bohnacker and Rosén, 2008). In contrast, German L2 speakers of English produced significantly more subject-initial sentences in their L2 English than their L1 German, even though the overall frequency of subject-initial sentences in L2 English was still significantly lower than the frequency of subject-initial sentences produced by L1 English speakers. In line with the Interface Hypothesis (Sorace, 2011), O’Brien and Féry attribute their L2 findings to continued influence from L1 preferences in expressing information structure, combined with an incomplete understanding of the ways in which discourse-pragmatic principles guide the frequency of producing non-canonical word order in the TL.

Using a picture description task, Carroll et al. (2000) found that near-native L1 English L2 speakers of German could produce sentences containing locative information in sentence-initial position at rates similar to L1 German speakers, even though placing such information in sentence-initial position was less preferred when they completed the same task in their L1, i.e., English. However, subtle differences remained in the types of locative adverbials that appeared sentence initially in their L2 German vs. in the productions of L1 German speakers. Similar to O’Brien and Féry (2015), Carroll et al. attribute these differences to residual influence from preferences for the conceptualization of space and place in the speakers’ L1 English.

At the same time, relying solely on production tasks to measure L2 speakers’ knowledge of L2 syntax–discourse and IS-constraints may underestimate L2 knowledge, since production requires the selection of a single structure from two or more options. This demand may compel learners to overproduce a simpler, unmarked canonical structure, even though their interlanguage grammar may license non-canonical structures. Thus, it is critical to also investigate L2 speakers’ preferences via comprehension measures, as such measures allow speakers to accept multiple possible structures in any given context (for discussion of production vs. comprehension processes, see, for example, Pickering and Garrod, 2013; Tooley and Bock, 2014).

For comprehension, Hopp (2009) compared the acceptability and on-line reading of fronted objects in embedded clauses, so-called scrambling, among advanced to near-native L2 German speakers with either English, Dutch or Russian as their L1. In German, scrambling is only licit for defocused objects. Russian displays comparable IS-constraints on scrambling, while Dutch differs in that scrambling is limited to contrastive topics. Finally, English does not exhibit any scrambling. For sentences embedded in discourse and question contexts that manipulated focus structure, an acceptability judgment task found that both L1 English near-native and L1 Russian L2 speakers of German were sensitive to the IS-constraints on scrambling, while the Dutch speakers were not. During on-line processing, a similar pattern held, with the English and Russian speakers showing target differences in reading times according to whether the context designated the fronted object as topic or focus, while the Dutch learners did not differ in reading times across contexts. These findings suggest, first, that target-like syntax–discourse mappings are acquirable, since English L2 speakers of German came to be sensitive to IS-constraints on scrambling despite the L1 offering no analogue. Second, L1 mappings delimit convergence on target IS-syntax mappings, since the different discourse constraints on scrambling in Dutch interfered in speakers’ L2 German (see also Smeets, 2017).

For the acquisition of object fronting in L2 English by L1 Spanish speakers, Slabakova (2015) compared topicalization and (contrastive) focus-fronting, as in (6) (capitalization indicates stress).

(1) a. A JACKET he bought, not a shirt. (Focus fronting)

b. The salmon I haven’t tried yet. (Topicalization)

A contextualized acceptability judgment task showed that the Spanish L2 speakers of English clearly distinguished between context-appropriate and context-inappropriate uses of focus fronting, but they did not recognize the context-appropriateness of English topicalization. In contrast, a comparable group of English L2 speakers of Spanish had no trouble judging the context-appropriateness of Spanish clitic left dislocation, which works analogously to English topicalization. As clitic left dislocation is more frequent in Spanish than topicalization is in English, Slabakova interprets this asymmetry between languages as evidence of frequency effects in the acquisition of interface properties that are similarly realized in L1 and L2. More specifically, she argues that L2 speakers need to encounter such properties with sufficient frequency in the L2 input to acquire them, since oftentimes the relevant syntactic structure itself is comparatively rare in the input, and the discourse-level properties guiding the appropriate use of such structures may therefore remain unclear.

The study by Slabakova (2015) highlights the importance of construction frequency in two respects. First, low frequency of a construction in the L2 input may restrict the learnability of interface properties due to their restricted availability in the input. Second, becoming sensitive to the frequency, or distributional properties, of optional forms in the input is itself part of the learning task. This second question of whether L2 speakers are sensitive to distributional properties of forms in the input has featured prominently in usage-based approaches to L2 acquisition (e.g., Collins et al., 2009), whereas it has rarely been investigated in the context of the Interface Hypothesis. Critically, mastery of an L2 includes both knowledge of the discourse constraints on optional word orders and sensitivity to the frequency of use of non-canonical orders in different discourse contexts. As Jackson and Ruf (2017) suggest, convergence on these L2 properties may also be affected by the transfer of properties of the analogous L1 construction along with its L1 usage frequency. For L1 English learners of German, Specifically, they showed that learners adopt L1 frequency preferences of discourse-related fronting of adverbials, suggesting that L1 transfer may include information on how frequently a non-canonical syntactic option is employed.

To address the combined impact of construction frequency on the appropriateness and usage frequency of phenomena at the syntax–discourse interface, the present study explores discourse-driven fronting tendencies in a bidirectional study of German and English speakers. Specifically, we establish a directly comparable L1 baseline of the options used in the sentential left periphery as a starting point to then investigate whether adult L2 speakers are sensitive (1) to the discourse conditions that license various types of fronting in English and German and (2) to the frequency of their realization in these contexts. We address the question of whether L2 speakers converge on the distributional properties of different structures in the L2, as measured by responses in a speeded naturalness judgment task. We go beyond previous research by considering different constructions at the syntax–discourse interface that diverge in their frequency of use within a single language as well as across languages. Beyond looking at L2 frequency, this set-up allows us to investigate whether frequency differences between structures within the L1 affect naturalness judgment rates in the L2 (see also Jackson and Ruf, 2017).

III Fronting tendencies in German and English

In both German and English, subject-initial word order is the unmarked option in declarative clauses. However, German and English allow for other constituents to be fronted into sentence-initial position, mainly for discourse-related reasons (Bader and Häussler, 2010; Verhoeven, 2015). Both English and German allow for adverbials to be preposed, as in (2) and (3), and for objects to be topicalized (4) or to appear in it-cleft constructions (5). In the present study, we examine different types of fronting in the left periphery of main clauses: (2) preposing of temporal phrases (TPs), (3) preposing of locative phrases (LPs) and (4 and 5) sentence-initial objects derived through topicalization or object clefts (object NPs).

(2) Am Vormittag hat der Bäcker einen Kuchen gebacken. (fronted TP)

in.the morning has the baker a cake baked

‘In the morning the baker baked a cake.’

(3) Im Garten hat der Mann einen Kaffee getrunken. (fronted LP)

in.the garden has the man a coffee drunk

‘In the garden the man drank a cup of coffee.’

(4) Einen Ring hat die Frau unter dem Sofa gefunden. (topicalization)

a ring has the woman under the sofa found

‘A ring the woman found under the sofa.’

(5) Es war ein Ring, den die Frau unter dem Sofa gefunden hat. (object cleft)

it was a ring that the woman under the sofa found has

‘It was a ring that the woman found under the sofa.’

As can be seen in examples (2–5), German and English differ in word order of non-subject-initial clauses due to German being a verb-second language, which entails that the finite verb is always the second sentential constituent in main clauses. In contrast, English, being an SVO language, has verb-third order in non-subject-initial sentences.

In terms of frequency, subject-initial declaratives are the most common sentence types in German and English (e.g., Bader and Häussler, 2010; Bohnacker and Rosén, 2008). In both languages, clause-initial adjuncts, i.e., adverbials, are more common than clause-initial arguments, i.e., objects (Doherty, 2005). While these relative differences in frequency are similar in German and English, the two languages differ in the absolute frequency with which the options in (2–5) are employed. As for adverbials, German allows more liberally for initial adverbials than English, with the largest cross-linguistic differences occurring with LPs compared to TPs (e.g., Jackson and Ruf, 2017; see also Doherty, 2005). As for fronted arguments, German strongly prefers topicalization over clefting. Object-initial main clauses account for approximately 10% of all declaratives in German (Bader et al., 2017), while clefts are much rarer (Bohnacker, 2010). English, however, prefers object clefts over topicalization. Topicalization in English is exceedingly rare: in the 1.1 million Wall Street Journal corpus, Postolache (2005) finds 10 instances of topicalization (for spoken data, see also Gregory and Michaelis, 2001). In contrast, clefts occur more often, with reported occurrences around 61 out of 1 million sentences in written corpora (Roland et al., 2007). Still, cleft sentences in English are overall comparatively rare compared to topicalization in German. In the present study, we test the preposing of temporal and locative adverbials in both German and English to investigate frequency differences within constructions, and we examine topicalization in German vs. object clefts in English to examine frequency effects between constructions.¹

The non-canonical orders in (2–5) are subject to information structural constraints, expressible, e.g., in terms of topic and focus. In both German and English, constituents can be fronted, irrespective of whether they are topics or in focus. In this article, we limit our discussion of the orders in (2–5) to focus fronting, since we test effects of (narrow) focus. In IS terms, focus is associated with new or non-presupposed information in a sentence (for discussion, see Krifka, 2008). Many approaches make a further distinction between presentational or information focus, on the one hand, and contrastive focus, on the other (Gundel, 1988; Jackendoff, 1972; É Kiss, 1998). In the present study, we concentrate on information focus by virtue of employing wh-questions that either induce wide (all sentence) focus (6a) or narrow focus on the object (6b) for the sentences in (2–5).

(6) a. What happened?

b. What did the woman find?

Although narrow focus does not necessitate fronting, the left periphery preferably hosts focused constituents, marked prosodically by stress or pitch contour in both German and English. In consequence, the orders in (2–5) are felicitous in narrow focus contexts (6b), yet less felicitous in wide focus contexts (6a). Despite these broad similarities in IS terms, there are also differences between topicalization and it-clefts. Unlike topicalization in German, cleft sentences give rise to an exhaustiveness implicature (e.g., É Kiss, 1998), and cleft sentences are prima facie biclausal (for monoclausal analyses, see Frascarelli and Ramaglia, 2013; Meinunger, 1998).²

In many syntactic approaches to fronting, the focus-marked constituent is moved to a (focus) projection in the left periphery with an unpronounced copy of the fronted constituent remaining in its base position (e.g., Chomsky, 1977; for clefts, e.g., Haegeman et al., 2015; É Kiss, 1999). For the purposes of this article, nothing hinges on the particulars of a syntactic analysis, since both English and German share all options in (2–5). A learner will likely thus transfer the options from the L1.³ The learning task for L2 learners is thus not to acquire or reassemble features or restrict the inventory of features or operations, but to calibrate the acceptability of non-canonical orders to the appropriate discourse contexts and target levels of use in the L2. Specifically, English learners of German need to acquire that fronting is employed more liberally in German than in English and that topicalization constitutes the preferred means for fronting arguments in German. Conversely, German learners of English need to converge on a lower use of preposed adverbials, and they need to come to express argument preposing by virtue of clefting rather than topicalization.

IV Research questions and predictions

Given the overall similarities in the IS-constraints governing the use of fronting in English and German (e.g., Bader and Häussler, 2010; Verhoeven, 2015), we predict discourse constraints that license fronting in the present study, operationalized as broad vs. narrow focus, to affect the ratings of non-subject-initial sentences in both languages in similar ways. Based on previous corpus and experimental research (e.g., Bader and Häussler, 2010; Bohnacker and Rosén, 2008; Roland et al., 2007; Jackson and Ruf, 2017; Weskott et al., 2011), we expect German and English native speakers to differ in the extent to which they license fronting, with German allowing for more fronting overall than English, leading to higher acceptability rates for non-subject-initial sentences in German than English, independent of discourse constraints. For L2 learners, we pose the following research questions.

Research question 1: Do late L2 learners converge on the discourse–syntax mapping of fronting options in the TL?

For L2 acquisition, the Interface Hypothesis (IH; Sorace, 2011) predicts L2 learners to experience difficulty in restricting fronting options to appropriate discourse contexts, in particular when constructions differ between the L1 and TL. In this study, which tests the preposing of locative and temporal adverbial phrases and of object NPs, the IH thus predicts that L2 speakers will show lower sensitivity to the difference between broad and narrow focus contexts for object fronting, given that the preferred means for fronting object NPs are (information-)structurally more dissimilar between English and German than the preposing of locative and temporal adverbial phrases.

Research question 2: Do late L2 learners converge on the frequencies with which discourse-related frontings are used in the licensing contexts in the TL?

Previous research has identified construction frequency as an obstacle in the L2 acquisition of interface properties (Slabakova, 2015), and the frequency of a construction in the L1 has been found to correlate with accuracy during L2 production (Jackson and Ruf, 2017).

Against this backdrop, we predict late-learning L2 speakers to diverge from L1 speakers in their absolute levels of acceptance of L2 fronting options, with the greatest divergence occurring with those constructions that are least frequent in the L2 input and differ most from the frequency of use in the L1. In our study, we thus expect to find a larger divergence in naturalness judgments between L2 speakers and L1 speakers for object fronting than for sentence-initial adverbials.

V Methods

1 Participants

Twenty-five L1-English–L2-German speakers in the United States and 24 L1-German–L2-English speakers in Germany participated in the study. One L1 English speaker and two L1 German speakers were excluded because they grew up in bilingual households, and one L1 English speaker and one L1 German speaker were excluded because they did not complete the proficiency task. For proficiency assessment, participants completed the LexTALE proficiency task (Lemhöfer and Broersma, 2012) in both English and German. LexTALE is a lexical decision task that has been validated as a general measure of L2 proficiency and produces comparable scores across both languages. Based on the LexTALE scores, we excluded an additional five L1 English speakers because they scored below 50% accuracy on the German LexTALE task, indicating low L2 German proficiency.⁴ The final pool of participants included 18 (13 female; 5 male) L1-English–L2-German speakers and 21 L1-German–L2-English speakers (18 female; 3 male). All participants began learning their respective L2 (English or German) at age six or later through classroom-based instruction at school (i.e., not in an immersion-based environment). A t-test comparing the LexTALE results from each group’s respective L2 revealed no significant differences in L2 proficiency (t(37) = 1.74, p = .091).⁵ Complete biographical information is provided in Table 1.

Table 1.

Biographical information for L1 and L2 speakers.

	L1-English–L2-German speakers (n = 18)		L1 German-L2 English speakers (n = 21)
	M (SD)	Range	M (SD)	Range
Current age (years)	25.9 (6.4)	19–40	21.4 (3.4)	18–35
Age of first L2 exposure (years)	15.4 (5.6)	11–36	8.0 (1.2)	6–11
Length of residence in L2 country (months)	18.7 (27.3)	0–83	5.8 (7.0)	0–24
LexTALE score (in %): English	97.2 (6.2)	73.8–100	71.3 (11.7)	55.5–100
LexTALE score (in %): German	64.8 (11.5)	50.0–92.5	86.3 (7.8)	72.1–100

2 Materials

In a speeded naturalness judgment task, participants judged whether a target sentence was a natural answer to a question in relation to pictures depicting the events expressed in the target sentence. Target sentences varied according to the critical constituent (LP vs. TP vs. Object NP) and the placement of this critical constituent (nonfronted vs. fronted), as in examples (7)–(12).

(7) Der Mann hat einen Kaffee im Garten getrunken. (LP; nonfronted)

the man has a coffee in.the garden drunk

‘The man drank a cup of coffee in the garden.’

(8) Im Garten hat der Mann einen Kaffee getrunken. (LP; fronted)

in.the garden has the man a coffee drunk

‘In the garden the man drank a cup of coffee.’

(9) Der Bäcker hat am Vormittag einen Kuchen gebacken. (TP; nonfronted)

the baker has in.the morning a cake baked

‘The baker baked a cake in the morning.’

(10) Am Vormittag hat der Bäcker einen Kuchen gebacken. (TP; fronted)

in.the morning has the baker a cake baked

‘In the morning the baker baked a cake.’

(11) Die Frau hat einen Ring unter dem Sofa gefunden. (object NP; nonfronted)

the woman has a ring under the sofa found

‘The woman found a ring under the sofa.’

(12) Einen Ring hat die Frau unter dem Sofa gefunden. (object NP; fronted)

a ring has the woman under the sofa found

‘It was a ring that the woman found under the sofa.’

To make sure that case and gender information would not be necessary for a target interpretation of the sentence, we chose animate subject nouns and inanimate object nouns for all sentences. Hence, changes in word order could not possibly lead to different interpretations of the sentences.

Each target sentence was embedded in a question context about the event expressed in the sentences (see examples 13–16), which varied the information structure of the answer, i.e., the target sentence. In broad-focus conditions, the question did not focus on any particular element in the target sentence, as in (13). In narrow-focus conditions, the question induced focus on the critical constituent, i.e., either the location where an event took place (LP items), the time at which an event took place (TP items), or the direct object of the event, as in (14)–(16), thereby establishing a context in which it would be felicitous to front the critical constituent.

(13) Was ist passiert? (all constituents; broad focus)

what is happened

‘What happened?’

(14) Wo hat der Mann einen Kaffee getrunken? (LP; narrow focus)

where has the man a coffee drunk

‘Where did the man drink a cup of coffee?’

(15) Wann hat der Bäcker einen Kuchen gebacken? (TP; narrow focus)

when has the baker a cake baked

‘When did the baker bake a cake?’

(16) Was hat die Frau unter dem Sofa gefunden? (Object NP; narrow focus)

what has the woman under the sofa found

‘What did the woman find under the sofa?’

The task included 28 target sentences for each of the three critical constituents. These target sentences were distributed across four lists, so that participants saw seven items in each condition (nonfronted-broad focus; fronted-broad focus; nonfronted-narrow focus; fronted-narrow focus), but only one version of any given item. These 74 target sentences were presented along with 46 filler items. Twenty filler sentences were grammatical in both German and English (e.g., Die Schülerin stellte dem Lehrer eine Frage ‘The student asked the teacher a question’); and 26 filler sentences contained a word order or verb tense violation in both languages, rendering them ungrammatical (e.g., *Der Patient muss gehen zum Krankenhaus ‘*The patient must to the hospital go’). These filler sentences were preceded by questions that induced either broad or narrow focus. The target and filler items were presented in a different pseudorandomized order for each participant, with the provision that no more than two items from the same condition appeared consecutively.

3 Procedure

All participants were tested individually in a quiet room. Participants first filled out a language background questionnaire. Then they completed the speeded naturalness judgment task in either German or English, with the order of languages counterbalanced between participants. This task was implemented in E-Prime 2.0 (Psychology Software Tools, 2012). All trials began with a fixation cross for 500 ms. Then the target event picture was displayed for 2,000 ms. Prior to each target sentence, participants saw a colored picture on the screen, as in Figure 1 (used for target sentences 7–8). This was then replaced with the target question, which appeared for 2,000 ms. This was followed by a blank screen for 500 ms, after which the target sentence was presented in a Rapid Visual Serial Presentation (RSVP), i.e., word-by-word manner, with each word appearing in the center of the screen for 400 ms. At the conclusion of the sentence, a screen containing three question marks appeared, and participants were instructed to judge whether the sentence was a natural answer to the question by giving a binary yes-no response. This prompt remained on the screen until participants made their judgment, or for a maximum of 2,000 ms (see Hopp, 2010; for similar procedures, see Bader and Meng, 1999). A blank screen then appeared for 500 ms, after which the next trial began. Participants received six practice trials to familiarize themselves with this procedure. After completing the speeded naturalness judgment task, participants took the LexTALE proficiency task in whichever language they had just completed the speeded naturalness judgment task. Then they completed the LexTALE proficiency task in the other language, followed by the speeded naturalness judgment task in the other language.

Figure 1.

Sample target event picture.

VI Results

Analyses of the filler items revealed no significant difference in participants’ judgments for the filler items in their respective L2, i.e., L2 German for the L1-English–L2-German speakers (M = 83.8, SD = 15.5) and L2 English for the L1-German–L2-English speakers (M = 89.4, SD = 5.7; β = −0.49, SE = 0.41, z = −1.19, p = .236),⁶ providing additional evidence that these two groups were of comparable L2 proficiency. Similarly, there was no significant difference in participants’ judgment for filler items in their respective L1, i.e., L1 English for the L1-English–L2-German speakers (M = 91.3, SD = 13.1) and L1 German for the L1-German–L2-English speakers (M = 94.7, SD = 5.0; β = −0.77, SE = 0.54, z = −1.44, p = .151).

Analyses for all experimental items were conducted using mixed-effect logistic regression models with the lme4 package in R version 3.2.5 (R Development Core Team, 2016). Participants’ binary judgments of whether the target sentence was a natural response to the question was the dependent variable. Fronting (nonfronted vs. fronted) and focus (broad focus vs. narrow focus) were entered as fixed effects in each analysis. First, to establish whether there were differences between L1 English and L1 German in fronting tendencies and whether IS affects fronting, we compared the judgments of the L1 speakers across the languages (subsection: L1 English vs. L1 German). For these analyses, language (L1 German vs. L1 English) was entered as an additional fixed effect. Second, to test whether there were differences between L1 and L2 speakers in fronting tendencies within a language, we compared the L1 and L2 speakers in English and in German, respectively (subsections: L1 English vs. L2 English; L1 German vs. L2 German). For these two analyses, group (L1 speaker vs. L2 speaker) was entered as an additional fixed effect. All fixed effects were sum-coded as –.5 and .5. To capture the repeated measures design, the random effects structure included random intercepts for participants and items, as well as all random slopes justified by the design (all main effects plus interactions between group/language and the sentence-level predictors of fronting and focus). Where this maximal model did not converge, we first removed the by-item, and then the by-participant random correlation parameters (for discussion, see Barr et al., 2013: 276). Finally, to avoid over-fitting, any remaining random correlations greater than .95 were removed, leaving in the decorrelated random slopes (Baayen et al., 2008). The initial model for each comparison included all higher order interactions. The fixed effect structure was then fitted using a backwards stepwise procedure, removing non-significant higher order interactions (ps > .1) from the final models reported here. Descriptive results by sentence type are presented in Figures 2 –4.

Figure 2.

Judgment rates, by participant group and stimulus type: Locative phrases.

Figure 3.

Judgment rates, by participant group and stimulus type: Temporal phrases.

Figure 4.

Judgment rates, by participant group and stimulus type: Object NPs.

Initially, we also compared participants’ reaction times for making their judgments (for complete results of these analyses, see Appendix 1). Across all analyses, there was a significant effect of focus, because both L1 and L2 speakers in both languages exhibited longer response latencies for broad-focus than narrow-focus conditions. In the analysis of LPs between L1 German and L2 German speakers, this was qualified by a significant 3-way interaction between group, fronting and focus because the L2 German speakers’ reaction times were longer for broad focus than narrow focus sentences overall (broad focus: M = 659 ms, SD = 310; narrow focus: M = 515 ms; SD = 222), whereas the L1 German speakers’ reaction times were longer only for fronted-LPs in the broad focus condition (M = 685 ms; SD = 317). Critically, with the exception of the analysis of TPs between L1 German and L2 German speakers, there were no significant main effects of group, indicating that judgment times were similar for L1 and L2 speakers across conditions.⁷

1 L1 English vs. L1 German

The first analysis compared judgments from the L1 English speakers in English and the L1 German speakers in German to establish whether there were any significant differences between the two languages. In a preliminary analysis including sentence type (LP vs. TP vs. Object NP) as an additional variable, there was a significant 3-way interaction between sentence type, fronting and focus (object NPs vs. adverbial phrases: β = −2.51, SE = 0.80, z = −3.14, p = .002; TPs vs. LPs: β = −0.27, SE = 0.68, z = −0.40, p = .688).⁸ Therefore, in the analyses presented here, we treat each sentence type separately.

With sentences containing LPs, the 3-way interaction between language, fronting and focus did not reach statistical significance and was removed from the final model (β = −1.55, SE = 1.35, z = −1.15, p = .249). As seen in Table 2, there was a significant effect of fronting, which was qualified by a marginal 2-way interaction between language and fronting. While participants in both languages judged nonfronted-LPs as natural more often than fronted-LPs, this difference was greater in English (nonfronted-LPs: M = 95.6, SD = 11.0; fronted-LPs: M = 74.6, SD = 29.9) than in German (nonfronted-LPs: M = 91.8, SD = 10.9; fronted-LPs: M = 83.3, SD = 18.3). No other effects or interactions were significant.

Table 2.

Summary of mixed logit model for English vs. German LPs: L1 English vs. L1 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	3.05	0.34	8.96	< .001
Language	0.25	0.54	0.45	.650
Fronting	−1.55	0.62	−2.50	.013
Focus	−0.42	0.46	−0.91	.363
Language × Fronting	−1.80	1.02	−1.76	.078
Language × Focus	0.17	0.61	0.28	.783
Fronting × Focus	0.21	0.99	0.21	.832
Random effects*	Name		Variance	SD
Participant	Intercept		1.65	1.28
	Fronting		5.30	2.30
	Focus		0.50	0.71
	Fronting × Focus		4.58	2.14
Item	Intercept		0.10	0.32
	Language		0.00	0.00
	Fronting		0.00	0.00
	Focus		0.00	0.00
	Language × Fronting		0.00	0.00
	Language × Focus		0.00	0.00
	Fronting × Focus		1.00	1.00

Note. * by-item correlation parameters removed.

With sentences containing TPs, the 3-way interaction between language, fronting and focus did not reach statistical significance and was removed from the final model (β = −0.77, SE = 1.79, z = −0.43, p = .668). As seen in Table 3, there was a significant effect of fronting, which was qualified by a significant 2-way interaction between language and fronting. While participants judged nonfronted-TPs as natural more often than fronted-TPs in English (nonfronted-TP: M = 93.3, SD = 18.6; fronted-TP: M = 82.9, SD = 26.4), there was no difference in German (nonfronted-TP: M = 91.8, SD = 11.4; fronted-TP: M = 90.1, SD = 8.6). No other effects or interactions were significant.

Table 3.

Summary of mixed logit model for English vs. German TPs: L1 English vs. L1 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	4.00	0.58	6.86	< .001
Language	0.98	0.90	1.09	.277
Fronting	−1.98	0.82	−2.43	.015
Focus	−0.14	0.69	−0.21	.835
Language × Fronting	−2.37	1.01	−2.36	.018
Language × Focus	−0.49	0.88	−0.55	.581
Fronting × Focus	−1.17	1.17	−1.00	.316
Random effects*	Name		Variance	SD
Participant	Intercept		4.71	2.17
	Fronting		1.38	1.18
	Focus		1.72	1.31
	Fronting × Focus		1.83	1.35
Item	Intercept		0.07	0.27
	Language		0.00	0.00
	Fronting		0.01	0.09
	Focus		0.00	0.05
	Language × Fronting		0.00	0.00
	Language × Focus		1.09	1.04
	Fronting × Focus		0.54	0.73

Note. * by-item correlation parameters removed.

With sentences containing object NPs, the 3-way interaction between language, fronting and focus did not reach statistical significance and was removed from the final model (β = 2.00, SE = 1.49, z = 1.35, p = .178). As seen in Table 4, there was a significant effect of fronting and a significant effect of focus, which were qualified by a significant interaction between fronting and focus. As seen in Figure 4, participants were equally likely to judge nonfronted-object NPs as natural regardless of focus, but they judged fronted-object NPs as natural more often in narrow focus contexts than in broad focus contexts. No other main effects or interactions were significant. In sum, German allows for more fronting than English, as shown by the interactions between language and fronting, with fronting tendencies being susceptible to information structure in similar ways in both languages.

Table 4.

Summary of mixed logit model for English vs. German object NPs: L1 English vs. L1 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	2.30	0.42	5.45	< .001
Language	0.74	0.77	0.96	.339
Fronting	−3.24	0.72	−4.53	< .001
Focus	1.33	0.52	2.58	.010
Language × Fronting	−0.51	1.22	−0.42	.674
Language × Focus	0.11	0.75	0.14	.888
Fronting × Focus	2.46	1.01	2.43	.015
Random effects*	Name		Variance	SD
Participant	Intercept		3.98	1.99
	Fronting		8.22	2.87
	Focus		0.97	0.98
	Fronting × Focus		4.40	2.10
Item	Intercept		0.01	0.10
	Language		0.00	0.00
	Fronting		0.26	0.51
	Focus		0.00	0.00
	Language × Fronting		0.00	0.00
	Language × Focus		0.97	0.98
	Fronting × Focus		0.00	0.00

Note. * by-item correlation parameters removed.

2 L1 English vs. L2 English

We turn next to the comparison of L1 and L2 speakers’ judgments in English to test how L2 speakers who allow for more fronting in their L1, i.e., German, judge sentences in an L2 that is more conservative in its fronting tendencies. In a preliminary analysis including sentence type (LP vs. TP vs. Object NP) as an additional variable, there was a significant 3-way interaction between sentence type, fronting and focus (object NPs vs. adverbial phrases: β = −2.38, SE = 0.76, z = −3.13, p = .002; TPs vs. LPs: β = −0.34, SE = 0.64, z = −0.53, p = .598). Therefore, in the analyses presented here we treat each sentence type separately.

With sentences containing LPs, no 2- and 3-way interactions reached statistical significance and were removed from the final model (3-way interaction: β = −1.81, SE = 1.24, z = −1.46, p = .145; Group × Fronting: β = −0.72, SE = 1.00, z = −0.73, p = .469; Group × Focus: β = −0.02, SE = 0.70, z = −0.03, p = .975; Fronting × Focus: β = −0.54, SE = 0.99, z = −0.55, p = .586). As seen in Table 5, there was a significant effect of fronting. Both L1 and L2 German speakers judge nonfronted-LPs as natural more often (L1 speakers: M = 95.6, SD = 11.0; L2 speakers: M = 94.6, SD = 5.5) than fronted-LPs (L1 speakers: M = 74.6, SD = 29.9; L2 speakers: M = 73.5, SD = 28.5). No other effects were significant.

Table 5.

Summary of mixed logit model for English LPs: L1 English vs. L2 English data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	2.94	0.34	8.53	< .001
Group	0.37	0.56	0.66	.511
Fronting	−2.08	0.56	−3.72	< .001
Focus	−0.08	0.42	−0.19	.851
Random effects*	Name		Variance	SD
Participant	Intercept		2.01	1.42
	Fronting		5.03	2.24
	Focus		1.82	1.35
Item	Intercept		0.09	0.29
	Group		0.09	0.31
	Fronting		0.04	0.20
	Focus		0.49	0.70

Note. * by-item correlation parameters removed.

With sentences containing TPs, the 3-way interaction between group, fronting and focus did not reach statistical significance and was removed from the final model (β = −0.06, SE = 1.54, z = −0.04, p = .967). As seen in Table 6, there was a significant effect of group and a significant effect of fronting, which were qualified by a significant interaction between group and fronting. While the L1 and L2 speakers judged fronted-TPs as natural at equal rates (L1 speakers: M = 82.9, SD = 26.4; L2 speaker: M = 81.6, SD = 14.0), the L1 speakers judged nonfronted-TPs as natural more often (M = 93.3, SD = 18.6) than the L2 speakers (M = 86.1, SD = 16.4). No other effects or interactions were significant.

Table 6.

Summary of mixed logit model for English TPs: L1 English vs. L2 English data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	3.32	0.47	7.01	< .001
Group	1.80	0.84	2.13	.033
Fronting	−1.76	0.62	−2.85	.004
Focus	−0.25	0.46	−0.55	.585
Group × Fronting	−2.10	0.91	−2.31	.021
Group × Focus	−0.42	0.63	−0.66	.509
Fronting × Focus	−0.09	0.78	−0.12	.905
Random effects*	Name		Variance	SD
Participant	Intercept		3.95	1.99
	Fronting		1.10	1.05
	Focus		0.14	0.38
	Fronting × Focus		2.43	1.56
Item	Intercept		0.13	0.36
	Group		0.71	0.84
	Fronting		0.00	0.00
	Focus		0.00	0.00
	Group × Fronting		0.00	0.00
	Group × Focus		0.00	0.00
	Fronting × Focus		0.00	0.00

Note. * by-item correlation parameters removed.

With sentences containing object NPs, the 3-way interaction between group, fronting and focus did not reach statistical significance and was removed from the final model (β = 1.93, SE = 1.85, z = 1.04, p = .297). As seen in Table 7, there was a significant effect of fronting and a significant effect of focus, which were qualified by a significant interaction between fronting and focus. As seen in Figure 4, while participants judged nonfronted-object NP sentences as equally natural regardless of focus, fronted-object NPs were judged as more natural in narrow focus contexts than in broad focus contexts. There was also a significant interaction between fronting and group. The L2 English speakers were more likely to judge fronted-object NPs as natural (M = 76.5, SD = 21.8) than L1 English speakers (M = 60.3, SD = 36.3) overall. No other main effects or interactions were significant.

Table 7.

Summary of mixed logit model for English object NPs: L1 English vs. L2 English data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	3.29	0.44	7.45	< .001
Group	0.04	0.70	0.06	.950
Fronting	−3.49	0.71	−4.92	< .001
Focus	1.02	0.46	2.19	.028
Group × Fronting	−2.51	1.27	−1.97	.049
Group × Focus	0.55	0.80	0.69	.492
Fronting × Focus	2.60	0.97	2.69	.007
Random effects*	Name		Variance	SD
Participant	Intercept		3.32	1.82
	Fronting		9.84	3.14
	Focus		1.95	1.40
	Fronting × Focus		13.01	3.61
Item	Intercept		0.00	0.00
	Group		0.00	0.00
	Fronting		0.16	0.40
	Focus		0.35	0.59
	Group × Fronting		0.05	0.23
	Group × Focus		0.00	0.00
	Fronting × Focus		0.40	0.63

Note. * by-item and by-participant correlation parameters removed.

In sum, the English results show similar judgment patterns for L1 and L2 English speakers; yet, the L2 English speakers accepted the fronting of objects in English more than the L1 speakers.

3 L1 German vs. L2 German

Finally, we turn to the comparison of L1 and L2 speakers’ judgments in German to test how L2 speakers whose L1, English, is more conservative in its fronting tendencies judge sentences in an L2 that allows for more fronting. In a preliminary analysis including sentence type (LP vs. TP vs. Object NP) as an additional variable, there was a significant 3-way interaction between sentence type, fronting and group (object NPs vs. adverbial phrases: β = 1.02, SE = 0.84, z = 1.21, p = .226; TPs vs. LPs: β = 1.97, SE = 0.67, z = 2.93, p = .003). Therefore, in the analyses presented here we treat each sentence type separately.

With sentences containing LPs, no 2- and 3-way interactions reached statistical significance and were removed from the final model (3-way interaction: β = 0.20, SE = 1.22, z = 0.17, p = .865; Group × Fronting: β = 0.43, SE = 0.74, z = 0.58, p = .560; Group × Focus: β = 0.00, SE = 0.51, z = −0.01, p = .994; Fronting × Focus: β = 0.66, SE = 0.59, z = 1.11, p = .265). As seen in Table 8, there was a significant effect of fronting. Both the L1 and L2 German speakers judge nonfronted-LPs as natural more often (L1 speakers: M = 91.8, SD = 10.9; L2 speakers: M = 94.8, SD = 5.9) than fronted-LPs (L1 speakers: M = 83.3, SD = 18.7; L2 speakers: M = 90.9, SD = 19.6). There was a marginal effect of group. The L2 German speakers judged sentences as natural more often overall (M = 92.9, SD = 10.6) than the L1 German speakers (M = 87.6, SD = 12.6). No other effects were significant.

Table 8.

Summary of mixed logit model for German LPs: L1 German vs. L2 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	2.97	0.26	11.22	< .001
Group	0.84	0.46	1.83	.068
Fronting	−0.75	0.35	−2.13	.033
Focus	−0.33	0.23	−1.42	.156
Random effects*	Name		Variance	SD
Participant	Intercept		1.22	1.10
	Fronting		1.90	1.38
	Focus		0.00	0.00
Item	Intercept		0.00	0.00
	Group		0.00	0.00
	Fronting		0.00	0.00
	Focus		0.06	0.24

Note. * by-item and by-participant correlation parameters removed.

With sentences containing TPs, the 3-way interaction between group, fronting and focus did not reach statistical significance and was removed from the final model (β = 1.56, SE = 1.09, z = 1.43, p = .152). As seen in Table 9, there was a significant effect of fronting, which was qualified by a significant interaction between group and fronting. While the L1 and L2 speakers judged fronted-TPs as natural at equal rates (L1 speakers: M = 90.1, SD = 8.6; L2 speakers: M = 92.9, SD = 14.1), the L1 speakers judged nonfronted-TPs as natural more often (M = 91.8, SD = 11.4) than the L2 speakers (M = 79.4, SD = 32.5).

Table 9.

Summary of mixed logit model for German TPs: L1 German vs. L2 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	3.39	0.39	8.67	< .001
Group	0.03	0.67	0.05	.960
Fronting	0.74	0.35	2.11	.035
Focus	0.20	0.38	0.52	.602
Group × Fronting	2.25	0.69	3.28	.001
Group × Focus	0.34	0.85	0.40	.687
Fronting × Focus	0.49	0.52	0.94	.348
Random effects*	Name		Variance	SD
Participant	Intercept		2.94	1.71
	Fronting		1.14	1.07
	Focus		2.11	1.45
	Fronting × Focus		0.00	0.00
Item	Intercept		0.00	0.00
	Group		0.52	0.72
	Fronting		0.00	0.00
	Focus		0.00	0.00
	Group × Fronting		0.00	0.00
	Group × Focus		3.54	1.88
	Fronting × Focus		0.00	0.00

Note. * by-item and by-participant correlation parameters removed.

With sentences containing object NPs, the 3-way interaction between group, fronting and focus did not reach statistical significance and was removed from the final model (β = −0.08, SE = 1.33, z = −0.06, p = .951). As seen in Table 10, there was a significant effect of group. The L2 German speakers judged sentences as natural more often overall (M = 86.5, SD = 21.8) than the L1 German speakers (M = 75.5, SD = 18.7), although one sees that, descriptively, this effect is driven largely by the L2 German speakers’ judgments for fronted-object NPs, especially in narrow focus contexts. There was a significant effect of fronting, which was qualified by a significant interaction between fronting and focus. As seen in Figure 4, participants were equally likely to judge nonfronted-object NPs as natural regardless of focus, but they judged fronted-object NPs as natural more often in narrow focus contexts than in broad focus contexts. No other interactions were significant.

Table 10.

Summary of mixed logit model for German Object NPs: L1 German vs. L2 German data.

Predictor	Parameter estimates		Wald’s test
Fixed effects	Estimate	Standard error	z-value	Pr (>\|z\|)
(Intercept)	2.70	0.36	7.44	< .001
Group	1.75	0.64	2.75	.006
Fronting	−2.19	0.53	−4.10	< .001
Focus	0.62	0.46	1.34	.179
Group × Fronting	1.03	0.88	1.17	.244
Group × Focus	−0.72	0.64	−1.11	.265
Fronting × Focus	1.82	0.80	2.28	.023
Random effects*	Name		Variance	SD
Participant	Intercept		2.63	1.62
	Fronting		2.78	1.67
	Focus		0.41	0.64
	Fronting × Focus		1.31	1.14
Item	Intercept		0.07	0.27
	Group		0.04	0.20
	Fronting		0.00	0.00
	Focus		0.45	0.67
	Group × Fronting		0.59	0.77
	Group × Focus		0.00	0.00
	Fronting × Focus		0.00	0.00

Note. * by-item correlation parameters removed.

In sum, the German results show similar judgment patterns for L1 and L2 English speakers; yet, the L2 German speakers accepted the fronting of objects in German more than the L1 speakers.

4 Within-group comparisons

In a final set of comparisons, we directly compared participants’ ratings in their L1 vs. their L2 to confirm that participants judged the naturalness of sentences differently across their two languages. For these analyses, we conducted separate analyses for each participant group (L1 German; L1 English), in which L1/L2 (L1 vs. L2) was entered as an additional fixed effect, alongside fronting and focus. These analyses revealed a significant effect of fronting or a fronting × L1/L2 interaction among both the L1 German speakers and L1 English speakers for both fronted-LP and TP sentences. Neither group rated fronted sentences as similarly natural across both of their languages (L1 German speakers LPs: fronting × language β = −1.29, SE = 0.46, z = −2.80, p = .005; L1 German speakers TPs: language β = −0.79, SE = 0.31, z = −2.58, p = .010; L1 English speakers LPs: fronting × language β = −3.00, SE = 1.10, z = −2.74, p = .006; L1 English speakers TPs: fronting × language β = −4.69, SE = 1.30, z = −3.62, p < .001). As seen in Figures 2 and 3, participants rated fronted LP and TP sentences in German as natural more often than fronted LP and TP sentences in English, regardless of which language was their L1 or L2. The analysis of object NP sentences similarly revealed a significant fronting × L1/L2 interaction among the L1 German speakers (β = 1.53, SE = 0.47, z = 3.25, p = .001) and a significant effect of L1/L2 among the L1 English speakers (β = −1.43, SE = 0.70, z = −2.03, p = .043). As seen in Figure 4, similar to the primary analyses, these two effects are driven by the fact that participants rated fronted object NP sentences as natural more often in their L2 than their L1, regardless of which language was their L1 or L2.

VII Discussion

The present study probed how L1-English–L2-German and L1-German–L2-English speakers judge the naturalness of different types of non-canonical fronted constructions across different discourse contexts in both German and English. It systematically investigated whether L2 speakers are sensitive (1) to the information-structural (IS) constraints that license fronting and (2) to the frequency with which such constructions are used in the L2 — especially when the frequency of these constructions differ between languages. The major findings are as follows:

Comparisons between the L1 English and L1 German results confirmed (1) similar cross-linguistic IS-constraints in that fronted-object sentences were judged as more natural in narrow-focus contexts than broad focus contexts in both German and English and (2) cross-linguistic frequency differences in that fronted-LP and TP sentences were judged as more natural in German than in English, paralleling differences in the production frequency of these constructions between German and English (Carroll et al., 2000; Doherty, 2005; Jackson and Ruf, 2017; O’Brien and Féry, 2015).

For fronted-LP and TP sentences, both the L2 English and L2 German speakers exhibited equivalent judgments as their L1 counterparts in spite of cross-linguistic differences in the perceived naturalness and overall frequency of these constructions — as shown by the significant differences in participants’ judgments between their respective L1 vs. L2.

Both L2 English and L2 German speakers overgeneralized the perceived naturalness of fronted-object sentences, as they were more likely to judge fronted-object sentences as natural than their L1 counterparts in both English and German.

To address research question 1, we first explore whether the L2 groups are sensitive to interface conditions licensing the sentences in (7) to (12), as would be reflected in differences in judgments between broad and narrow focus contexts. Turning first to fronted-object sentences, both participant groups show the appropriate L2 syntax-to-discourse mappings for the use of object-first constructions in German and object-cleft constructions in English, highlighting that advanced L2 speakers can map IS-functions onto syntactic structures in the L2.⁹ In the present study, we cannot tell whether they used the IS-constraints from their respective L1 to judge sentences in the L2, or if learners extracted the appropriate IS-conditions from the L2 input. In any case, the present study shows that L2 speakers converge on TL mappings at the syntax–discourse interface. Our findings are in line with other studies documenting convergence at the L2 syntax–discourse interface (e.g., Kras, 2008; Donaldson, 2012; Leal et al., 2017, Slabakova et al., 2012), and they underscore that IS-mappings at the syntax–discourse interface do not constitute persistent difficulty in L2 acquisition.

Second, we turn to L2 sensitivity to frequency differences between German and English of various fronting options across discourse contexts (research question 2). For fronted-LP and TPs, we found non-significant differences between L1 and L2 speakers within each language and significant differences between participants’ respective L1 and L2 within each group. These findings suggest that both the German and English L2 speakers successfully came to reflect the L2 frequency of non-canonical fronted sentences with adverbial phrases in their judgments rather than continuing to make judgments based on the perceived naturalness and frequency of such sentences in their respective L1. Together, these results highlight that convergence on target-like distributional patterns is possible, even when L2 speakers are faced with identifying subtle differences in the perceived naturalness of non-canonical word orders with adverbial phrases between their L1 and L2. This contrasts with previous research on the L2 production of fronted adverbials (Carroll et al., 2000; O’Brien and Féry, 2015; see also Bohnacker and Rosén, 2008) and suggests that production research may underestimate L2 speakers’ sensitivity to L2 syntax-to-discourse mappings, because outcome measures in production are limited to the selection of a single structure from multiple structural alternatives.

At the same time, while acceptability ratings for fronted-TP sentences did not differ between L1 and L2 speakers in either language, both the L2 English speakers and L2 German speakers were less likely to rate nonfronted-TP sentences as natural compared to their respective L1 counterparts (see Figure 3). We hypothesize that this difference stems from underlying cross-linguistic differences in word order constraints for TPs vs. direct objects (DOs) in the midfield. German allows for both TP-DO and DO-TP word order, with a preference for TP-DO word order when the DO is a mass noun or preceded by an indefinite article, as in the present study (Helbig and Buscha, 2001). In contrast, the DO-TP order is strongly preferred in English, since TP-DO orders can only be realized for heavy or contrastively focused objects by virtue of extraposition (heavy NP-shift). As nonfronted-TP sentences in the present study contained the less marked word order in both languages, i.e., TP-DO order in German and DO-TP order in English, we speculate that when making judgments about nonfronted-TP sentences in their L2, both participant groups may have been influenced by word order constraints from the L1. A more systematic investigation of these potential differences for non-fronted sentences warrants future research.

In contrast to fronted-LPs and TPs, where L2 speakers converged on the L1 judgment levels, we found differences for fronted-object sentences, as revealed by both the significant differences between L1 and L2 speakers within each language, and by the significant differences between participants’ respective L1 and L2 within each participant group. Both participant groups were more likely to judge fronted-object sentences as natural in their respective L2 compared to their L1 counterparts. Hence, the L2 speakers exhibited continued difficulty in adjusting their perceived naturalness of fronted-object constructions to reflect the distributional properties of such constructions in the L2. In the following, we consider different potential causes to account for such difficulty.

The overacceptance of fronted objects in the L2 could reflect L1 frequency patterns. Object-first constructions in German are more frequent than object-cleft constructions in English (Bader et al., 2017; Roland et al., 2007; Weber and Müller, 2004). If the L1 frequency of a given or related structure drove the acceptability of such non-canonical word orders in the L2, then, the L1-German–L2-English speakers should have shown higher ratings for fronted objects in their L2 English than the L1 English speakers. While this is borne out in the data, an L1-frequency based account would predict the L1-English–L2-German speakers to rate fronted-object sentences in their L2 German as less natural than their L1 German counterparts. The fact that both groups overgeneralized the naturalness of fronted-object sentences in the L2, irrespective of whether the L1 realizes fronted objects more or less frequently, suggests that the frequency of a given structure in the L1 is not the primary driving force behind the acquisition of native-like sensitivity to the distributional properties of non-canonical word orders in the L2. Rather, based on the present findings, we hypothesize that the locus of continued optionality among even advanced L2 speakers lies in the frequency with which they encounter canonical vs. non-canonical word orders in the L2 input, irrespective of the frequency of these orders in the L1. In both German and English, the fronting of sentential arguments, like object NPs, is relatively rare, especially compared to the fronting of sentential adjuncts, like adverbial phrases (Doherty, 2005; Jackson and Ruf, 2017). Finding an asymmetry in convergence in judgments on the frequency distribution of fronted adverbials vs. objects in the L2 supports the idea proposed by Slabakova (2015) that a certain threshold of how frequent a particular non-canonical structure is in the L2 input must be achieved before convergence is possible — a threshold that is met for fronted adverbial phrases but not for fronted-object NPs in the input available to German and English learners.

Alternatively, the different degrees of discourse relatedness of fronted-object NPs vs. fronted adverbials could affect the degree of target-likeness among the L2 speakers. Fronting of sentential adjuncts, like temporal and locative adverbials, is more liberal than argument fronting and can also be used as a scene-setting device (Birner and Ward, 1998). These differences in the discourse functions between fronted adverbials and fronted objects become clear in the present judgment data: only the perceived naturalness of fronted objects was modulated by broad vs. narrow focus contexts; for fronted TPs and LPs there was never a main effect or interaction involving the factor focus in either English or German. Accordingly, L2 speakers might continue to struggle to calibrate the more specific discourse conditions for object fronting compared to adverbial fronting. However, the more specific discourse functions of fronted objects are conflated with their frequency in the input, since more discourse-specific orders necessarily occur less frequently than less discourse-specific orders, all else being equal. Hence, it is near impossible to disentangle the overall frequency of a given structure from the number and type of IS-functions it serves in a language. In any case, together, the present findings highlight the complex ways in which IS-constraints and L2 frequency interact to influence the L2 acquisition of non-canonical word orders and the syntax–discourse mappings that govern their use.

Importantly, the present study suggests advanced L2 speakers do not have residual difficulty with the discourse-to-syntax mappings at the external interfaces but rather with calibrating the levels of appropriate use to native levels in some cases. We have suggested that determining the relative frequency of a given structure in the L2 is contingent on sufficient evidence in the input and that frequency differences of discourse-related phenomena in the input lead to differential acquisition at the interfaces. These effects hold over and above any L1 frequency effects. In the absence of actual frequency data in the input available to L2 speakers, this conclusion must necessarily remain somewhat speculative, but it highlights the need to consider frequency differences in the input as an important factor in future research on convergence at the syntax–discourse interface. At the same time, it may lead to a reappraisal of apparently incompatible findings in previous studies on difficulties at the syntax–discourse interface in L2 acquisition. Different findings across phenomena or languages may reflect different degrees of frequency of these phenomena in the L2 input, similar to how fronted adverbials and fronted objects showed asymmetries for L2 convergence in the present study.

VIII Conclusions

The present study aimed to identify the specific difficulties faced by L2 speakers in the context of L2 syntax-to-discourse mappings. In a bidirectional study on German and English adult L2 speakers, we investigated not only whether L2 speakers can converge on discourse-appropriate uses of non-canonical TL word orders, but also whether they are sensitive in their judgments to the frequency with which such non-canonical word orders are employed across different discourse contexts in the TL. The results show that L2 speakers are sensitive to IS-constraints on the fronting of objects, and they provide novel evidence to suggest that L2 speakers become sensitive to the frequency with which non-canonical fronted adverbials are used in the L2, even when this frequency distribution differs from that of the L1. At the same time, the over-acceptance of fronted objects among both L2 English and L2 German speakers suggests that L2 input must contain sufficient evidence of the contexts in which more marked word orders are used before convergence on TL distributional patterns is possible. As such, the present study highlights the importance of L2 frequency in explaining residual optionality at the syntax–discourse interface.

Supplemental Material

Oct2017_Materials – Supplemental material for Frequency at the syntax–discourse interface: A bidirectional study on fronting options in L1/L2 German and L1/L2 English

Supplemental material, Oct2017_Materials for Frequency at the syntax–discourse interface: A bidirectional study on fronting options in L1/L2 German and L1/L2 English by Holger Hopp, Joseph Bail and Carrie N. Jackson in Second Language Research

Footnotes

Appendix 1 Acknowledgements

We thank three anonymous reviewers and the editor for helpful comments on earlier versions of this manuscript.

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) disclosed receipt of the following financial support for the research, authorship, and/ or publication of this article: This research was funded in part by the National Science Foundation under grant OISE-0968369 (PI: JF Kroll; co-PIs: PE Dussias and JG van Hell).

Supplemental material

Supplemental material for this article is available online.

Notes

References

Baayen

Davidson

Bates

(2008) Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language 59: 390–412.

Bader

Häussler

(2010) Word order in German: A corpus study. Lingua 120: 717–62.

Bader

Meng

(1999) Subject–object ambiguities in German embedded clauses: An across-the-board comparison. Journal of Psycholinguistic Research 28: 121–43.

Bader

Ellsiepen

Koukoulioti

Portele

(2017) Filling the prefield: Findings and challenges. In: Freitag

Bot

Schloterbeck

(eds) Two perspectives on V2: The invited talks of the DGfS 2016 workshop “V2 in grammar and processing: Its causes and its consequences”. Konstanz : University of Konstanz, pp. 27–49.

Barr

Levy

Scheepers

Tily

(2013) Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language 68: 255–278.

Belletti

Bennati

Sorace

(2007) Theoretical and developmental issues in the syntax of subjects: Evidence from near-native Italian. Natural Language and Linguistic Theory 25: 657–89.

Birner

Ward

(1998) Information status and noncanonical word order in English. Amsterdam/Philadelphia, PA: John Benjamins.

Bohnacker

(2010) The clause-initial position in L2 Swedish declaratives: Word order variation and discourse pragmatics. Nordic Journal of Linguistics 33: 105–143.

Bohnacker

Rosén

(2008) The clause-initial position in L2 German declaratives: Transfer of information structure. Studies in Second Language Acquisition 30: 511–38.

10.

Carroll

Murcia-Serra

Watorek

Bendiscioli

(2000) The relevance of information organization to second language acquisition studies: The descriptive discourse of advanced adult learners of German. Studies in Second Language Acquisition 22: 441–66.

11.

Chafe

(1974) Language and consciousness. Language 50: 111–33.

12.

Chomsky

(1977) On wh-movement. In: Culicover

Wasow

Akmajian

(eds) Formal syntax. New York: Academic Press, pp. 71–132.

13.

Collins

Trofimovich

White

Cardoso

Horst

(2009) Some input on the easy/difficult grammar question: An empirical study. Modern Language Journal 93: 336–53.

14.

Doherty

(2005) Topic-worthiness in German and English. Linguistics 43: 181–206.

15.

Donaldson

(2011) Left-dislocation in near-native French. Studies in Second Language Acquisition 33: 399–432.

16.

Donaldson

(2012) Syntax and discourse in near-native French: Clefts and focus. Language Learning 62: 902–30.

17.

É Kiss

(1998) Identificational focus vs. information focus. Language 74: 245–73.

18.

É Kiss

(1999) The English cleft construction as a focus phrase. In: Mereu

(ed.) Boundaries of morphology and syntax. Amsterdam: John Benjamins, pp. 217–229.

19.

Frascarelli

Ramaglia

(2013) (Pseudo) clefts at the syntax–prosody–discourse interface. In: Hartmann

Veenstra

(eds) The structure of clefts. Amsterdam: Benjamins, pp. 97–140.

20.

Gregory

Michaelis

(2001) Topicalization and left dislocation: A functional opposition revisited. Journal of Pragmatics 33: 1665–1706.

21.

Gundel

(1988) The role of topic and comment in linguistic theory: Outstanding dissertations in linguistics. New York: Garland.

22.

Haegeman

Meinunger

Vercauten

(2015) The syntax of it-clefts and the left periphery of the clause. In: Shlonsky

(ed.) Beyond functional sequence: The cartography of syntactic Structures: Volume 10. Oxford: Oxford University Press, pp. 73–90.

23.

Helbig

Buscha

(2001) Deutsche Grammatik: Ein Handbuch fur den Ausländerunterricht [German grammar: A handbook for foreigner education]. Munich: Langenscheidt.

24.

Hopp

(2009) The syntax–discourse interface in near-native L2 acquisition: Off-line and on-line performance. Bilingualism: Language and Cognition 12: 463–83.

25.

Hopp

(2010) Ultimate attainment in L2 inflectional morphology: Performance similarities between non-native and native speakers. Lingua 120: 901–31.

26.

Ivanov

(2012) L2 acquisition of Bulgarian clitic-doubling: A test case for the interface hypothesis. Second Language Research 28: 345–68.

27.

Jackendoff

(1972) Semantic interpretation in generative grammar. Cambridge, MA: MIT Press.

28.

Jackson

Ruf

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

29.

Kam

Franzese

(2007) Modeling and interpreting interactive hypotheses in regression analysis. Ann Arbor, MI: University of Michigan Press.

30.

Kras

(2008) Anaphora resolution in near-native Italian grammars: Evidence from native speakers of Croatian. EUROSLA Yearbook 8: 107–34.

31.

Krifka

(2008) Basic notions of information structure. Acta Linguistica Hungarica 55: 243–76.

32.

Lambrecht

(1994) Information structure and sentence form: Topic, focus and the mental representation of discourse referents. Cambridge: Cambridge University Press.

33.

Leal

Slabakova

Farmer

(2017) The fine-tuning of linguistic expectations over the course of L2 learning. Studies in Second Language Acquisition 39: 493–525.

34.

Lemhöfer

Broersma

(2012) Introducing LexTALE: A quick and valid lexical test for advanced learners of English. Behavior Research Methods 44: 325–43.

35.

Meinunger

(1998) A monoclausal structure for (pseudo-)cleft sentences. In: Tamanji

Kusumoto

(eds) Proceedings of NELS (North East Linguistic Society) 28. Amherst, MA: Graduate Linguistic Student Association, pp. 283–98.

36.

O’Brien

Féry

(2015) Dynamic localization in second language English and German. Bilingualism: Language and Cognition 18: 400–18.

37.

Pickering

Garrod

(2013) An integrated theory of language production and comprehension. Behavioral and Brain Sciences 36: 329–47.

38.

Postolache

(2005) Information structure in written English: A corpus study. Unpublished paper presented at the IGK colloquium, Universität des Saarlandes, Saarbrücken. Available at: http://www.coli.uni-saarland.de/projects/irtg/contents/Colloquium/WS-05/oana.pdf (accessed September 2018).

39.

Psychology Software Tools, Inc. (2012). E-Prime 2.0.

40.

Roberts

Gullberg

Indefrey

(2008) Online pronoun resolution in L2 discourse: L1 influence and general learner effects. Studies in Second Language Acquisition 30: 333–57.

41.

Roland

Dick

Elman

(2007) Frequency of basic English grammatical structures: A corpus analysis. Journal of Memory and Language 57: 348–79.

42.

Serratrice

Sorace

Filiaci

Baldo

(2009) Bilingual children’s sensitivity to specificity and genericity: Evidence from metalinguistic awareness. Bilingualism: Language and Cognition 12: 239–57.

43.

Slabakova

(2015) The effect of construction frequency and native transfer on second language knowledge of the syntax–discourse interface. Applied Psycholinguistics 36: 671–99.

44.

Slabakova

Kempchinsky

Rothman

(2012) Clitic-doubled left dislocation and focus fronting in L2 Spanish: A case of successful acquisition at the syntax–discourse interface. Second Language Research 28: 319–43.

45.

Smeets

(2017) The acquisition of object movement in Dutch: L1 transfer and ultimate attainment. In: LaMendola

Scott

(eds) Proceedings of the 41th Boston University Conference on Language Development. Somerville, MA: Cascadilla Press, pp. 588–601.

46.

Sorace

(2005) Selective optionality in language development. In: Cornips

Corrigan

(eds) Syntax and variation: Reconciling the biological and the social. Amsterdam: John Benjamins, pp. 55–80.

47.

Sorace

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

48.

Sorace

Filiaci

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

49.

Tooley

Bock

(2014) On the parity of structural persistence in language production and comprehension. Cognition 132: 102–36.

50.

Valenzuela

(2006) L2 end state grammars and incomplete acquisition of the Spanish CLLD constructions. In: Slabakova

Montrul

Prévost

(eds) Inquiries in linguistic development: In honor of Lydia White. Amsterdam: John Benjamins, pp. 283–304.

51.

Verhoeven

(2015) Thematic asymmetries do matter!: A corpus study of German word order. Journal of Germanic Linguistics 27: 45–104.

52.

Ward

(1988) The semantics and pragmatics of preposing. New York/London: Garland Publishing.

53.

Weber

Müller

(2004) Word order variation in German main clauses: A corpus analysis. In: Proceedings of the 20th International Conference on Computational Linguistics. Geneva, pp. 71–77.

54.

Weskott

Hörnig

Fanselow

Kliegl

(2011) Contextual licensing of marked OVS word order in German. Linguistische Berichte 225: 3–18.

55.

Wilson

(2009) Processing at the syntax–discourse interface in second language acquisition. Unpublished PhD dissertation, University of Edinburgh, Edinburgh, UK.

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

0.08 MB