Session 2D

Syntactic complexity as an indicator of language performance and proficiency has received considerable attention from L2 writing researchers. Existing studies are mostly based on the written products of L2 learners, analyzing the sentence-level, clause-level, and phrase-level structures. Such performance data gathered from production tasks can provide direct information about learners' language use, but may not reflect the learner's non-observable aspects (Ionin & Zyzik, 2014). A receptive knowledge test can be used to gain information with greater reliability about how learners perceive syntactic complexity. 
In this presentation, we will report on how users of English from different groups perceive written syntactic complexity by making use of a syntactic complexity judgement test recently developed based on the stages for syntactic complexity features in L2 academic writing as proposed by Biber, Gray, and Poonpon (2011). According to Biber et al. (2011), L2 learners' development of syntactic complexity features progresses through certain stages "from simple clause structures to the more complex and elaborated clause structures that are supposedly typical of professional academic writing" (p. 9). We specifically aim to find out to what extent users of English from different groups (expert users of English, native speakers of English, and L2 learners of English) can distinguish the complexity differences between syntactic structures from different stages. 
Using a syntactic complexity judgment test (SCJT), we will collect data from three groups of users of English: 30 native speakers (NSs) of English who were undergraduate students taking Japanese classes at a university in the USA: 30 nonnative speakers (NNSs) of English who were undergraduate students enrolled in the English Language and Literature program at a university in Turkey: and 30 experts of English language (native and non-native) who completed a PhD program in (Applied) Linguistics. The SCJT includes 42 items with pairs of English sentences including syntactic structures from different developmental stages (both adjacent and non-adjacent). We will ask the participants to judge which of the structures in a pair they think is syntactically more complex than the other. The target structure types are adverbial clauses, complement clauses, adverbial phrases, prepositional phrases, and noun phrases. 
We expect to find differences in the judgments of users from the three groups. We expect that the judgments of the expert group will follow the proposed developmental stages more closely than the judgments of the native speakers and L2 learners of English. We hope that our findings will contribute to the field of L2 writing by providing teachers with an alternative means of assessing and understanding their students' syntactic complexity knowledge. Using this test can help in increasing instructional focus on complexity within L2 writing classrooms.
References
Biber, D., Gray, B., & Poonpon, K. (2011). Should we use characteristics of conversation to measure grammatical complexity in L2 writing development? TESOL Quarterly, 45, 5–35.
Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition, 27, 141–172. https://doi.org/10.1017/S0272263105050096

The aim of this study was to provide new insights on how the distribution of practice affects second language (L2) vocabulary and grammar learning. Previous research under lab conditions has consistently shown a lag effect for vocabulary learning, according to which more spacing between learning sessions leads to higher long-term gains (Cepeda et al., 2006). However, this effect has not been found in classroom studies with younger learners or for L2 skills practice in the lab (e.g., Rogers & Cheung, 2020; Suzuki & DeKeyser, 2017). The cause of this discrepancy is still unclear due to the variety of methodologies typically used in these studies and the lack of experimental control in classroom studies. 
In order to investigate where the lag effect applies in L2 learning, we conducted two experiments among the same population of secondary school learners (n = 140) in an international English school in Cambodia using cued-recall training with Quizlet, a methodology similar to previous lab studies but taking place in the classroom. One experiment used 16 grammar items, in which cues were scenarios written in English and target items were full sentences that exemplified a target grammar rule, and the other used 16 paired-associates including previously unknown vocabulary items. Participants learned two categories for grammar (future perfect and third conditional) and for vocabulary (animals and food), with two sessions per category spaced either 1 or 7 days apart, counterbalanced between participants, and with no additional instruction. They performed cued-recall tests at either 7 or 28 days after treatment. 
The results of the generalized linear models revealed that longer spacing was beneficial for vocabulary learning especially for long-term retention, contrary to previous findings reported for younger learners in classroom settings and suggesting that the lag effect does apply to younger learners but only for controlled retrieval-based vocabulary learning. However, there was no main effect of lag for grammar learning, in line with previous skills acquisition studies. These results will be interpreted according to the Desirable Difficulty Framework (Suzuki et al., 2019) and the Declarative/Procedural Model (Ullman, 2020).

References
Cepeda, N. J., Coburn, N., Rohrer, D., Wixted, J. T., Mozer, M. C., & Pashler, H. (2009). Optimizing distributed practice theoretical analysis and practical implications. Experimental Psychology, 56(4), 236–246. 
Rogers, J., & Cheung, A. (2020). Input spacing and the learning of L2 vocabulary in a classroom context. Language Teaching Research, 24(5), 616–641.
Suzuki, Y., & DeKeyser, R. (2017). Effects of distributed practice on the proceduralization of morphology. Language Teaching Research, 21(2), 166–188.
Suzuki, Y., Nakata, T., & DeKeyser, R. (2019). The Desirable Difficulty Framework as a Theoretical Foundation for Optimizing and Researching Second Language Practice. Modern Language Journal, 103(3), 713–720.
Ullman, M. T. (2020). The Declarative/Procedural Model. In B. VanPatten, G. D. Keating, & S. Wulff (Eds.), Theories in Second Language Acquisition (3rd ed., pp. 128–161). Third edition. | New York, NY: Routledge, 2020.

Previous experimental research on the non-native (L2) acquisition of word order has predominantly focused on learners' ability to distinguish grammatical from ungrammatical structures. Few studies have investigated the acquisition of permissible structural variation [e.g. 1,2]. Here we use an experimental approach to examine the word-order preferences of advanced L2 learners of German across different modalities, and the extent to which these are affected by participants' native language (L1) background (Dutch vs. Russian). 
German exhibits substantial variability regarding the linearisation of infinitival complements. They can be extraposed to the right of the matrix verb (1), intraposed to its left (2), or realised discontinuously as in (3) (the 'third construction' pattern) [3].
(1) EXTRAPOSITION: …dass Fred versucht [den Hund zu streicheln] 
(2) INTRAPOSITION: …dass Fred [den Hund zu streicheln] versucht 
(3) THIRD CONSTRUCTION: …dass Fred [den Hund] versucht [zu streicheln] 
'…that Fred tries to pet the dog'
Raising verbs (e.g. scheinen 'seem') normally require intraposed infinitives while control verbs (e.g. versuchen 'try') allow for more word-order variability, with extraposition being the preferred structural variant and third constructions rarely attested [4]. In Dutch, infinitival complements of control verbs may either be extraposed or linearised in a third construction pattern, whilst only the latter option is normally available for raising verbs. Russian differs from German and Dutch in that all infinitive-embedding verbs predominantly occur with extraposed infinitival complements.
Experiment 1 was a written production task examining the linearisation choices of L1 Russian (n=46) and Dutch-speaking (n=20) advanced (C1/C2) late learners of German and L1 German controls (n=46), and Experiment 2 was a scalar acceptability judgement task carried out with comparable groups of L2 speakers (Russian: n=34; Dutch: n=28). Our results revealed between-group differences indicative of L1 effects. In Experiment 1, although intraposition was the preferred choice for raising verbs, Russian speakers produced significantly more extraposed infinitives for raising verbs than L1 German speakers did, and Dutch speakers produced reliably more third constructions compared to the other participant groups. For control verbs, all participant groups preferred extraposition over other structural variants, with no statistical group differences. L1 effects were stronger in Experiment 2, where Russian speakers rated extraposition most favourably for raising verbs, whereas Dutch speakers favoured third constructions. For control verbs, Dutch speakers rated third constructions significantly more favourably relative to Russian speakers. 

Our results confirm and extend earlier findings of even advanced learners showing reduced sensitivity to distributional constraints on L2 syntactic variation [2] and of L1 effects on learners' word-order choices [1]. The observed task differences suggest that L1 distributional constraints affect L2 metalinguistic judgements more strongly than production. The implications of our findings for approaches to L1 transfer will be discussed. 

References
[1] Gries, S. Th., & Wulff, S. (2013). International Journal of Corpus Linguistics 18, 327–356. 
[2] Jäschke, K., & Plag, I. (2016). Studies in Second Language Acquisition 38, 485–521. 
[3] Haider, H. (2010). The syntax of German. CUP.  
[4] Bosch, S. et al. (to appear). Journal of Comparative Germanic Linguistics.

Our paper seeks to explore the relationship between formulaic sequences (FSs) and second language (L2) proficiency, building on earlier work by Myles (2012). We define FSs as frequently-occurring combinations of words which may allow for internal variation and which carry stable meanings (e.g., 'make a decision', 'make a ADJ decision'). Earlier work suggests that the use of FSs positively affects fluency and accuracy, and that L2 learners benefit from an implicit or explicit focus on FS during instruction (Boers et al., 2006, Wray, 2018). Yet, when a learner's language use contains a large number of FSs, this might also create a challenge for language teachers and testers: when trying to assess the proficiency level, the structures underlying FSs may or may not reflect a student's level of lexicogrammatical competence. For example, a learner who uses FSs such as 'how do you do' or 'why don't you V' may not yet have mastered wh-question formation with verb inversion. This inspired us to explore whether (and if so in what ways) FSs mask the proficiency of L2 learners.
To address this question, we extracted data on a selection of high-frequency FSs from a 24-million word subset of the Cleaned Subcorpus of the EF-Cambridge Open Language Database (EFCAMDAT; Shatz, 2020) of learner writing. We started from lists of frequent n-grams of various lengths and grouped files in EFCAMDAT by learner proficiency, ranging from A1 to C1 (according to the CEFR). Two of the FSs selected for our analysis, 'why don't you V' and 'I think you should V,' first appear in A1 learner responses to a prompt in unit 21 (of 128) on 'Giving suggestions about clothing'. Despite their frequent early occurrence in the corpus, these two FSs build on fairly complex morphosyntactic structures (e.g., subject-verb inversion, do-support, negation). We studied the use of these and other FSs across EFCAMDAT levels with particular attention paid to their variability and accuracy of use. For each slot in a FS we measured the lexical diversity (MATTR, MTLD) and predictability (normalized entropy) at each proficiency level.
Results point to an overall lack of productivity in the use of the focus FSs at the A1 and A2 levels, as well as low accuracy when Al/A2 learners move away from the initial fixed sequences, resulting in erroneous or unidiomatic uses such as 'why don't you joining' and 'I suggest you should V'. For variable slots in the FSs, we found an increase in lexical diversity and a decrease in predictability as learners move from beginner to intermediate and advanced levels. We also observed that 'I think you should V' shows productivity earlier than 'why don't you V,' suggesting that learners need longer to acquire the more complex morphosyntax of this FS. In line with earlier work, we argue that FSs play an important role in early L2 development and instruction. Our findings suggest that assessment of language containing many FSs benefits from taking productivity and variability of their underlying structures into account.