Understanding Advanced Level Academic Writing on Syntactic Complexity

Understanding Advanced Level Academic Writing on Syntactic Complexity Miao Wang Wander Lowie Shanghai International Studies University Rijksuniversiteit Groningen 550 Dalian Road (W) Oude Kijk in Het Jatstraat 26, 9712 EK Shanghai, China The Netherlands [email protected] [email protected] structures exhibited in language production (Lu, Abstract 2011; Ortega, 2015). Syntactic complexity is frequently used to assess writing quality. It can Recent publications have highlighted the combination of large-grained and fine-grained demonstrate learners’ language development and indices to tap into syntactic complexity, while syntactic maturity (Crossley et al., 2011; Lu, 2011; few studies have examined the writing Ortega, 2003; Wolfe-Quintero et al., 1998). produced by more advanced academic writers Several researchers have attempted to determine a from a cross‐linguistic perspective. This paper comprehensive and organic measuring system to reports on a corpus-based comparison of the assess L2 writing and L2 language proficiency native speaker (NS) and the Chinese advanced since 2009. The call for more comprehensive and academic writers’ written abstracts. Using 120 multi-dimensional constructs (Norris & Ortega, texts within the discipline of applied linguistics, 2009; Bulté & Housen, 2012) combined with the the study analyzed the extent to which the introduction of large-scale corpus-based Chinese and the native writers’ writing differed in 16 measures of syntactic complexity. The approaches (Biber & Gray, 2010; Biber, Gray, & article also compared across levels of English Poonpon, 2011), led to the attempted integration of language proficiency and language background. fine-grained measures of complexity at both the The results revealed significant differences in phrasal level and the clausal level and large- subordination, with the NS professional group grained measures of complexity at the clause or generated more subordination than the Chinese sentence level (the length-based indices, such as professional writer group. Within-group the length of clauses, T-units, and/or sentences). comparisons yielded statistically significant Many studies have shown that first language differences in fine-grained syntactic complexity (L1) background has a potential effect on the for the Chinese group. These results show quality of L2 written production using syntactic considerable differences between the Chinese academic writers and native speakers reside in complexity (Kuiken and Vedder, 2019; Lu and Ai, the distribution of subordination. The results 2015; Staples and Reppen, 2016). Some studies on have implications for incorporating both L1 and L2 syntactic complexity have focused subordination to writing instruction for L2 on primary school children and secondary school writers at more advanced level. students (Jiang et al., 2019). With the growing number of international students enrolled in higher education in English speaking countries in recent years, extensive research on L2 syntactic complexity has been dedicated to identifying 1 Introduction potentially different patterns of syntactic In the field of second language (L2) writing complexity between native speakers (NSs) and research, syntactic complexity refers to the range non-native speakers (NNSs) of English at and the degree of sophistication of syntactic university-level (Ai and Lu, 2013; Casal and Li, 2019; Lu and Ai, 2015; Staples and Reppen, 2016). 2 Literature Review However, a limited number of studies have examined the variation of L2 written syntactic 2.1 Writing quality and syntactic complexity complexity at more advanced levels of language development. It is still unclear to what extent the Syntactic complexity is an important construct for written syntactic complexity produced by highly second language acquisition (SLA). It refers to advanced non-native speaker (NNS) writers is “the range of forms that surface in language different from that of advanced NS writers. The production and the degree of sophistication of such investigation of the patterns of syntactic forms” in SLA (Ortega, 2003, p.492). Syntactic complexity at more advanced stages of language complexity has been increasingly taken as a proficiency would enrich the current literature on multidimensional construct in nature (Bulté and L2 syntactic complexity, and especially on “the Housen, 2012; Lu, 2011; Norris and Ortega, 2009) manifestation and development of complexity in a and the obtain a fair representation of this aspect of second language” (Housen et al., 2019). Moreover, writing SLA researchers are encouraged to it is estimated that 440-650 million Chinese measure overall complexity (mean length of T- citizens are English learners (He and Zhang, 2012). unit), subordination or coordination (clauses per T- Answers to the question about the extent of the unit), and subclausal or phrasal sophistication differences in written syntactic complexity (mean length of clause, complex nominal per between Chinese writers and NS writers at the clause) (Norris and Ortega, 2009). graduate level, and between professional Chinese In L2 writing research, the use of large-grained writers and professional NS writers would indices (the length-based indices) at the syntactic contribute to our understanding of the writing level has indicated that advanced L2 users would process, and is crucial to L2 writing pedagogy. produce longer sentences and more diverse or Narratives, expository, descriptive, and elaborate language (Housen and Kuiken, 2009; Lu, argumentative essays written by beginner-level L2 2011; Ortega, 2003; Wolfe-Quintero et al., 1998). writers and university-level L2 writers were Over the past 40 years, syntactic development and investigated in a large number of studies (Bulté syntactic complexity used to rely on clausal and Housen, 2018; Crossely & McNamara, 2014; subordination measures and T-unit based Lahuerta Martínez, 2018; Verpoor, Schimid, & Xu, measures, such as clauses per T-unit (C/T), mean 2012). However, very few of these studies have length of clause (MLC), mean length of T-unit focused on abstract writing (Ansarifar et al., 2018, (MLT), dependent clause per clause (DC/C) and is an exception), which is used for a variety of dependent clauses per independent clause purposes like degree projects, publications, or (Bardovi-Harlig and Bofman, 1989; Ellis and conferences (Lorés, 2004). Abstracts are expected Yuan, 2004; Larsen-Freeman, 2006; Wolfe- to represent the highest level of language Quintero et al., 1998). For instance, Ortega’s proficiency, yet little is known about the (2003) research synthesis found that mean length production of abstracts, produced by writers at of T-unit (MLT) and clause per T-unit (C/T) more advanced levels of language development widely used in college-level ESL and EFL writing, and how abstracts differ across writers from among the 27 studies from 1976 through 1998, 25 different L1 backgrounds. of them employed MLT and 11 of them used C/T. The present study aims to bridge this gap by Ortega (2003) indicated that to observe substantial investigating to what extent large-grained changes in MLT for college-level L2 and foreign complexity measures and fine-grained complexity language writers, a one-year observation period measures of syntactic complexity differ between needed to be taken into account. Several studies native English academic writers and Chinese revealed that MLT was significantly positively academic writers in abstract writing. To investigate correlated with L2 language proficiency (e.g. Yang this, we designed a cross-sectional study using six et al, 2015). The written texts produced by higher- written corpora. We hope to contribute to both the level L2 writers will exhibit an increase in both research of L2 academic writing and L2 writing MLT and C/T when L2 language proficiency is pedagogy in English for academic purposes for a defined by course levels in program, although there growing group of Chinese authors. may be no significant difference between immediately adjacent levels (Yang, 2013). In a suggested the sequence of development in L1 longitudinal study of university-level L2 syntactic complexity and; hypothesized that the complexity development, Bulté and Housen (2014) development of L1 academic writing relies heavily noted that MLT has shown an increase after four- on nominal structures; and, in particular, on the month long academic English instruction. usage of phrasal modification. Some evidence Correlations were found between MLT and from both NS and NNS studies supports the subjective writing quality ratings. developmental hypothesis that academic writing is However, the widely-spread use of large- “more syntactically complex than oral texts” and grained measures has been criticized by several phrasal level complexity such as noun phrases researchers (Biber et al., 2011; Bulté and Housen, (NPs) complexity is a later development in 2012; Larsen-Freeman, 2009; Norris and Ortega, academic writing (Ansarifar et al., 2018, Biber et 2009). As T-unit-based measures alone are not able al., 2011; Lahuerta Martínez, 2018; Lu, 2011; to “reflect accurately the knowledge of the learner” Parkinson and Musgrave, 2014; Yong, 2017). For (Bardovi-Harlig, 1982, p.391), besides, large- instance, Parkinson and Musgrave (2014) pointed grained indices are insufficiently sensitive to out that a wider variety of complex nominals, such capture the more subtle changes in writing as possessive nouns, participial adjectives, development (Larsen-Freeman, 2009). some prepositional phrases, and appositive noun phrases researchers have emphasized the use of a fine level were used significantly more frequently in a more of granularity in recent years (Bulté & Housen, proficient L2 English for Academic Purposes 2012; Larsen-Freeman, 2009; Norris & Ortega, group. Ansarifa et al. (2018) showed that L2 MA- 2009) to more accurately reflect the specific level writers and L2 expert writers differ in the use structures employed by the learners (Norris & of phrasal features, while L2 writers at the Ph.D. Ortega, 2009) and/or to get a comprehensive level only differed in the use of one phrasal feature profile of L2 language development (Bulté & (multiple prepositional phrases as post-modifiers) Housen, 2014). For instance, the same T-unit (the when compared with L2 expert writers. man takes the plane to work) in two different Different from Biber et al.’s developmental sentences have different grammatical features: hypothesis (2011), based on Halliday and (a) The American man takes the plane to work five Matthiessen's systemic functional grammar (1999), days a week; (b) The man takes the plane to Norris and Ortega (2009) hypothesized that the work, although it is very exhausting. In (a), an direction of the development of L2 syntactic adjective (American) is used to elaborate the noun complexity runs from coordination to (man). Moreover, the verb phrase is elaborated subordination, and then to greater use of phrasal with an adverb ( five days a week). In (b), the complexity at higher levels of proficiency. Several original T-unit is elaborated with a subordinate studies have confirmed this sequence of language clause (although it is very exhausting). Important development for English as a Foreign Language points can be extracted from the two examples (EFL) writers in secondary education and at the above. Even though the MLT score in two university level (Jiang et al., 2018; Lu, 2011). sentences is the same (12), the score alone only Moreover, studies have shown that subordination provides a general picture instead of a more or coordination does not necessarily decrease along detailed indication of the types of elaboration with the development of L2 language proficiency included in the text. MLT score does not provide (Lahuerta Martínez, 2018; Mancilla et al., 2017). enough interpretable information on syntactic The high-level NNS group produced more complexity (Norris and Ortega, 2009). Therefore, subordination (dependent clause per clause, the study of L2 development should include the dependent clause per T-unit) and coordination multiple dimensions of complexity (Bulté &and (coordinate phrases per clause, CP/C) than the low- Housen, 2020). level NNS group (Mancilla et al, 2017). This result Using large-scale corpus investigations, Biber is different from the previous claim that language et al., (1998) remarked that written language learners with advanced knowledge would exhibit "primarily relies on full noun phrases" and lower levels of subordination (Norris and Ortega, subordination structures are more common in 2009). A similar result was found in Lahuerta conversation. Furthermore, Biber et al. (2011) Martínez’s study (2018), which demonstrated a significant increase in sentence coordination and institution were not controlled. In a conceptual subordination in the higher grade of secondary replication study (e.g. Lu, 2011), Yoon and Polio education students. As for the development in (2017) indicated that significant genre effects were subordination for the NS writers, Pallotti (2009) detected on the length of production units (MLC, proposed that the decrease of subordination after a MLS, MLT) for both NNS and NS written certain point in the development of syntactic linguistic complexity. For instance, significant complexity 'might be interpreted as a sign of genre effects on the length of production units higher proficiency'. were found for argumentative essays than Some recent studies have explored both large- narratives on length of unit. The length of grained and fine-grained complexity measures to production units produced by both NS and NNS observe the relationship between L2 writing writers in argumentative essays was longer than quality and syntactic complexity (Casal and Lee, that of narrative essays. Different from NS writers, 2019; Crossly & and McNamara, 2014; Taguchi et NNS writers produced high phrasal-level al., 2013; Jiang et al., 2019). Taguchi et al. (2013) complexity in argumentative writing (coordinate reported that the L2 high-rated essays (essays score phrases per clause [CP/C], coordinate phrases per above 90) did not necessarily contain more T-unit [CP/T], complex nominals per clause complex language at the clausal complexity than [CN/C], complex nominals per T-unit [CN/T], verb the L2 low-rated essays (essays score below 80). phrases per T-unit [VP/T]). NS writing only However, the former group contains more phrasal exhibited differences on one measure (CN/C) in complexity (attributive adjectives and post-noun- both narrative and argumentative essays. modifying) than the latter. Similarly, Casal and Lee’s study (2019) revealed that clausal 3 Methodology complexity, such as T-units per sentence (T/S) and clauses per T-unit (C/T), is not a distinguishing 3.1 Description of corpora used factor in the assessed quality of first-year L2 undergraduate-level writing, but the high-rated The data consists of six corpora of 180 abstracts essays used more nominal modifiers (adjective pre- (dissertation, research articles) written by 30 modification, preposition post-modification, and Chinese Ph.D. writers, and 30 Chinese professional participle modification). This finding suggests that writers. 30 NS Ph.D. writers, and 30 NS as proficiency increases, L2 learners tend to use a professional writers from the applied linguistics wide range of complex nominals in academic program. We chose abstracts written from the same writing. discipline (applied linguistics) between 2005-2019. The Chinese corpora consist of random data 2.2 Language proficiency and syntactic selected from China National Knowledge complexity Infrastructure (CNKI), an online national database under the lead of Tsinghua University, including Variation can be observed when syntactic journals, doctoral dissertations, masters’ theses, complexity is assessed in different genres and proceedings, ebooks, and so on. Research article different language proficiency levels. Several L1 abstracts in the Chinese corpora were from the top- studies have investigated the effect of genre on rated 10 Chinese journals of applied linguistics and written syntactic complexity. The choice of foreign language teaching/learning issued in China. different measures of syntactic complexity, task All the Chinese writers are of Han nationality. types, and the level of participants has contributed Proquest database was used to obtain abstracts to the understanding of the broad range of L1 written by the NS writers. For the English data, writing development. research article abstracts were chosen from 10 Previous studies have shown that argumentative academic journals of applied linguistics and essays exhibited a higher degree of syntactic foreign language teaching/learning, such as Annual complexity than narratives. Lu (2011) reported Review of Applied Linguistics, Applied that, compared to narrative essays, argumentative Linguistics, Journal of Second Language Writing, essays produced by L1 Chinese university-level Language, Culture and Curriculum, Language writers were more complex on 13 of the total 14 learning, Language Teaching Research, Modern syntactic measures when timing condition and Language Journal, Second Language Research, English writing at the advanced level emphasizes Studies in Second Language Acquisition, TESOL the use of nominal modifiers. The eight fine- Quarterly. All the texts were contributed by single grained phrasal and clausal complexity measures authors. To control for the L1 background, we (see Table 3) in this study were processed using looked into all the NS writers’ education TAASSC 1.3.8 (Kyle, 2016). TAASSC includes backgrounds and personal websites to verify that the classic 14 indices of syntactic complexity they were native speakers of English. Section measures (L2SCA), seven phrase types, and ten headers, footers, and page numbers in the abstracts phrasal dependent types. were manually cleaned. Spelling errors in the raw Dimension Label Description data were corrected, such as “is build” and “softwares”. The words in the brackets were also Length of MLC Mean length of clause deleted. Since all abstracts used in the study were production unit MLS Mean length of either published or graded, we would expect MLT sentence Mean length T-unit corrections at the accuracy level but not at the Amount of C/T Number of clause per complexity level by instructors, reviewers, or subordination DC/C T-unit editors. Number of dependent The descriptive details of the text corpora are clauses per clause presented in Table 1. To balance the equal sizes of Amount of CP/C Number of coordinate the six corpora to compare grammatical features of coordination T/S phrases per clause interest, the first 150 words were sampled from all the abstracts to control the number of words in Number of T-units per each corpus (Ansarifar et al., 2018; Crawford and sentence Csomay, 2015). Degree of phrasal CN/C Number of complex sophistication nominal per clause Number of Number Mean length abstracts of words of abstract CHPhD 30 4,597 153.2 Table 2: The 8 large-grained syntactic complexity NSPhD 30 4,557 151.9 CHPW 30 4,592 153.0 measures (modified from Lu, 2017, p.503). NSPW 30 4,593 153.1 Total 180 27,615 153.4 Note: CHPhD = Chinese PhD-level; Grammatical structures Abbreviation NSPhD = native PhD-level Adjective modifiers Amod Table 1：Descriptive details of the six corpora. Noun as modifiers Nn 4.2 Syntactic complexity measures Prepositional Prep The analyses consisted of 8 large-grained and fine- Phrase grained measures of syntactic complexity based on previous research. The measurement of the possessives Poss traditional large-grained syntactic complexity in Subordinating conjunction Mark this study involved using L2SCA, a computational software developed by Lu (2010). This computational software yields 14 indices of Table 3: Fine-grained measures selected in this complexity targeting the length of production unit, study and analyzed by TAASSC amount of subordination, amount of coordination, and degree of phrasal sophistication. In the present study, eight measures were chosen (see Table 2). 4 Research Question T-units per clause (CT/C), complex T-units per T- RQ1. How is the large-grained and fine-grained unit (CT/T), clauses per sentence (C/S), and verb syntactic complexity associated with abstracts phrases per T-unit (VP/T) are not included in this produced by the Chinese Ph.D. group and the NS study, for they were poor candidates for Ph.D. group? developmental indices (Lu, 2011). Academic RQ2. Are there systematic patterns of large- 6.1.2 Fine-grained measures grained and fine-grained syntactic complexity on Similar to what we did in the previous section, the abstracts between the Chinese Ph.D. group and the alpha value for each comparison was at .006(.05/8) NS. Ph.D. group? to avoid Type I error. An independent samples t- test was conducted to determine whether there is 5 Results statistical evidence between groups of ( Ph.D., and professional writer) for ‘nouns as modifiers’(nn), t(58) = 3.246, p = .002, d = 0.83 between two 5.1 Differences between the Chinese and NS groups of experienced writers. Mann Whitney U group on syntactic complexity tests did not indicate significant differences between the two groups of Ph.D. writers and the 5.1.1 Large-grained measures two groups of professional writers for The first question in this study was to ask whether ‘possessives’ (poss), ‘prepositional phrases’ (prep), syntactic complexity measures differed ‘adjective modifiers’ (amod), ‘adjective modifiers’ systematically across the group of Chinese and NS (advcl), ‘clausal complement’ (ccomp), and writers. Table 4, Table 5, Table 6, Table 7 and ‘subjective conjunction’ (mark). Table 8 present the mean scores and standard Chinese professional writers produced more deviations of all the eight large-grained measures subordination structures (advcl, mark, ccomp) and for Chinese and NS writers. more phrasal level fine-grained complexity (amod, To determine which (if any) of the means were nn, prep, poss) compared with NS professional significantly different between each group of the writers on clausal level of fine-grained measures. writers (Ph.D., and professional), we ran both Chinese professional writers would produce more independent sample t-tests and Mann Whitney U nominal modifiers (except rcmod) than NS tests. To avoid type I error, the alpha value for professional writers. However, the Chinese Ph.D. each comparison set to .05/8 (.006), where .05 is group produced fewer fine-grained subordination the significance level for the complete set of tests, structures than the NS Ph.D. group. At the same and 8 is the number of tests performed. time, the phrasal level nominal modifiers they Independent samples t-tests showed that significant produced were lag behind the NS group. differences were found in DC/C between NS and MLS MLC MLT Chinese professional writer groups (t (53.602) = 3.688, p = .001; d = 0.952). The result indicates M SD M SD M SD Chinese professional writers produced less CHP 28.45 9.368 18.29 5.455 26.77 8.890 hD 17 99 52 89 44 52 subordination than NS professional writers. We expected that the Ph.D. group would demonstrate a CHP 25.25 5.648 16.09 3.100 22.53 5.158 similar pattern as the professional writer group on W 48 23 89 11 72 40 DC/C/. However, there were no significant differences in DC/C between the NS and Chinese NSP 23.45 4.732 16.28 4.058 22.71 5.461 Ph.D. groups. The effect size of DC/C between the hD 24 33 24 71 47 51 two professional writer groups exceeds 0.8. The Mann Whitney U tests revealed a significant NSP 25.42 6.128 14.40 3.799 24.80 8.227 difference in the C/T indices between the two W 55 54 21 49 02 91 groups of professional writers. The number of C/T produced by the Chinese professional writers is far Table 4: Descriptive statistics of large-grained less than the NS professional writers. The Mann measures concerning length of production. Whitney U tests revealed a significant difference in the C/T indices between the two groups of professional writers. The number of C/T produced by the Chinese professional writers is far less than the NS professional writers.  NSPhD .3386 .14251 C/T DC/C NSPW .3835 .14107 M SD M SD Nn CHPhD 1.0737 .24036 .2914 .13341 CHPhD .2366 .13709 CHPW 1.4023 .20352 .2615 .10553 CHPW .2690 .15651 NSPhD .2415 .10694 NSPhD 1.4602 .44458 .2959 .15747 NSPW .1700 .08891 NSPW 1.7650 .56507 .3804 .14170 Advcl CHPhD .0361 .06257 Table 5: Descriptive statistics of large-grained CHPW .0386 .04856 measures concerning amount of subordination. NSPhD .0307 .03970 NSPW .0330 .04050 T/S CP/C Rcmod CHPhD .0240 .02109 M SD M SD CHPW .0107 .01771 CHPhD 1.0737 .16322 .6625 .35315 NSPhD .0282 .03431 CHPW 1.1792 .24209 .6589 .37974 NSPW .0312 .03147 Ccomp NSPhD 1.0513 .16560 .5653 .30625 CHPhD .0403 .06410 NSPW 1.0460 .19398 .5560 .34769 CHPW .1260 .08184 NSPhD .0725 .07766 NSPW .0986 .07667 Table 6: Descriptive statistics of large-grained Mark SC measures concerning amount of coordination. CHPhD .0604 .05732 CHPW .1348 .08804 NSPhD .0827 .06793 CN/C NSPW .0961 .07732 M SD CHMA 2.5687 .60512 Table 8: Descriptive statistics for fine-grained CHPhD 2.8960 1.01858 syntactic complexity measures. CHPW 2.4818 .53478 NSMA 2.4178 .98027 5.2 Differences in syntactic complexity NSPhD 2.4807 .71658 between Chinese writer groups NSPW 2.2181 .69493 To investigate the differences in syntactic complexity among Chinese groups, we ran Table 7: Descriptive statistics for large-grained independent sample t-tests and Mann-Whitney U measures concerning degree of phrasal tests. A Mann-Whitney U test shows that the sophistication for Chinese group. Chinese professional writer group (Mdn= .1292) produced more ccomp (U= -4.119, p=0.000) than M SD the Chinese Ph.D. group (Mdn=.0000). Also, Poss Chinese professional writer group (Mdn= .1339) CHPhD .0502 .03380 produced more mark (U= -3.109, p=0.002) than CHPW .0536 .04818 the Chinese Ph.D. group (Mdn=.0000). NSPhD .0544 .05132 NSPW .0452 .05101 Prep 5.3 Differences in syntactic complexity CHPhD .3148 .8213 between NS writer groups CHPW .3279 .08213 NSPhD .3026 .0364 Having established the between-group differences NSPW .2969 .06758 in levels of syntactic complexity between the Amod Chinese groups, we continued to examine the CHPhD .4203 .10209 between-group differences in syntactic complexity CHPW .4152 .15956 between the NS groups. An independent samples t-test showed that significant differences were meet this prediction. The differences in C/T and found in nn between the NS Ph.D. group and DC/C align with Lu and Ai’s (2015) findings and the NS professional writer group (t (54.285) = signify that Chinese writers produced longer 2.991, p = .004; d = 0.77). clauses, more coordinate phrases, fewer subordinate structures, and more complex nominals, even though only the differences in 7. Discussion subordination reached significance. We suggest that the cause of this reduction in clausal This study investigated the use of grammatical complexity cannot be readily attributed to features representing different dimensions of proficiency but are likely to originate in language syntactic complexity associated with the Chinese typological differences. We will discuss it in the and English native speakers at advanced levels of following section. academic English across two L1s (Chinese, Our results on Chinese-NS differences in the English). We found differences in the syntactic mean values of phrasal sophistication (CN/C) did patterns used by the NS writers and the Chinese not yield statistical significance, which is in line writers in writing abstracts. In this section, we with Lu and Ai’s study (2015). Besides, the provide illustrations of these patterns， as well as Chinese group in our study did not significantly explanations for these differences related to differ from the NS group in terms of coordination. language proficiency and language typology In Lu and Ai’s (2015) study, the amount of between the NS group and the Chinese group, and coordination used by the upper-intermediate discuss differences observed within the NS group Chinese group was less than those of the NS group and within the Chinese group. and this feature was related to the use of 7.1 Writing quality and syntactic complexity punctuation to combine independent clauses in the Chinese language. One of the reasons for these Starting with the traditional large-grained measures conflicting results in coordination could be that the of syntactic complexity, we saw that there are Chinese participants in our study are advanced more similarities than differences. Significant academic English writers, so they have a written differences in C/T and DC/C were found between proficiency equivalent to that of NS writers in the NS professional writer group and the Chinese coordination, as opposed to those of the college- professional writer group. No substantial level writers used in Lu and Ai’s (2015) study. differences were detected between the NS Ph.D. Another reason could be the genre differences. As group and the Chinese Ph.D. across all the eight Graetz (1985) believed that 'the abstract avoids large-grained and fine-grained measures. There subordinate clauses, uses phrases instead of were no significant differences between the NS clauses'. group and the Chinese group on the eight fine- grained measures. 7.2 Cross-linguistic differences and syntactic According to the developmental prediction, complexity advanced proficiency L2 writers were expected to After establishing that there are statistically reduce the use of subordination at the clausal level, significant differences in subordination between but capitalize on phrasal level complexity (Norris the Chinese and the NS writers. We further & Ortega, 2009; Ortega, 2003; Wolfe-Quintero et examined the intra-group comparison to al. 1998). Corpus-based research (Biber et al., investigate whether within-group exists. As can be 2011) also has provided evidence that the use of seen in Table 9 and Table 10, the patterns of noun phrases is the main characteristic of academic syntactic complexity are varied across the two writing for advanced learners. Since both the languages in the study. Chinese group and the NS group can be The within-group statistical analysis revealed characterized as advanced-level writers of that the level of language proficiency did not differ academic English, we expected the two groups to significantly in the production of large-grained exhibit a similar pattern in the usage of syntactic complexity for the Chinese subgroups subordination and phrasal complexity. In our except for two fine-grained measures. Our results results, only the Chinese Ph.D. group appeared to are largely in line with the results Ai and Lu (2013) reported on the non-significant differences group produced significantly more clausal between the lower and high proficiency NNS complements and subordinate clauses than the groups in the dimension of MLC, DC/C, CP/C, Chinese Ph.D. group, which echoes what we found T/S, and CN/C. in large-grained clausal subordination measures. The NS writers in our study tend to generate Conversely, the NS subgroups did not differentiate shorter T-units and less subordination along with in all eight fine-grained complexity measures. improved language proficiency. There were no The consistent differences in both the large- statistically significant differences within the NS grained clausal level and the fine-grained clausal group on all the fine-grained measures, which is level subordination structures provide signs that different from what we found in the Chinese support the language typological differences group. between Chinese and English are likely to have an Systematic functional linguistics (Halliday and impact on English written syntactic complexity. Matthiessen, 1999, 2004) proposes that the Finally, previous studies have found sequence of language development runs from disciplinary variation for both phrasal features and coordination (‘to express ideas’) to subordination clausal features in academic writing. Biber and (‘to express the logical connection of ideas’), and Gray (2010) and Gray (2015) have shown that soft finally to nominalization with the employment of science is different from hard science writing in grammatical metaphors. Lower-level language phrasal complexity. Humanities writing relies less exhibits coordination, but move to subordination at on phrasal structures than science writing does. the intermediate level and more complex phrases at Our results in fine-grained measures are in line the advanced language level (Bardovi-Harlig, with the discipline norms. 1992; Norris and Ortega, 2009). If we look at the different patterns of syntactic complexity within 8. Conclusions the Chinese group and the NS group, the mean Using L2SCA and TAASSC, this study values of subordination for the NS group still examined the extent to which the multiple exceed those of the Chinese group. In the Chinese dimensions of syntactic complexity differ in language, zero connectives are very common in academic English writing across two languages subordination (Yu, 1993). Conjunctions in the background (Chinese, English) and language Chinese language can be left out as long as the proficiency through a corpus-based cross-sectional relationship is clear from the context. Logical approach. Both the traditional large-grained connectives are not always necessary to connect complexity measures and fine-grained complexity the semantic logical relationship between clauses. measures were employed. By examining abstract What we have found in this study is that L2 writing in this way, we have been able to answer proficiency does not always result in a similar the questions (a) whether the writings of Chinese variation in written syntactic complexity as the advanced-level writers better approximate the syntactic pattern of English native speakers, writing of NS advanced-level writers in syntactic especially when the target language is complexity as Chines writers advance to higher typologically different from the source language. levels of proficiency in English and (b) whether the Compared with the Chinese Ph.D. group, the L1 backgrounds differentiate syntactic complexity Chinese professional writer group can be expected at different levels of L2 proficiency. to have more experience in academic writing and Our results show that the Chinese advanced therefore produced more phrasal complexity, but level writers in the corpus data were syntactically less subordination. However, unlike our similar to the NS advanced level writers in large- expectations, the developmental trend in grained complexity measures, but the intra-group subordination and phrasal complexity is not statistical analysis results revealed different evident within the Chinese group in our results. patterns in syntactic complexity. Additionally, the within-group analysis has The findings from this study provide additional shown the significant differences in two insights into the relationship between patterns of subordinate structures (ccomp, mark) between the syntactic complexity, language proficiency, and L1 Chinese Ph.D. group and the Chinese professional backgrounds. Abstracts written by more advanced writer group. The Chinese professional writer academic English writers exhibited visible cross- Ansarifar, A., Shahriari, H., & Pishghadam, R.2018. linguistic influences in syntactic complexity. In Phrasal complexity in academic writing: A terms of syntactic complexity development, the comparison of abstracts written by graduate students differences in subordination produced by the and expert writers in applied linguistics. Journal of English for Academic Purposes, 31: 58-71. Chinese writers and the NS writers have exhibited the language typological differences between these Bardovi-Harlig, K. 1992. A second look at T-unit two languages. Moreover, the improved language analysis: Reconsidering the sentence. TESOL proficiency level did not link to the variation of quarterly, 26(2): 390-395. syntactic complexity within the Chinese groups. Bardovi-Harlig, K., & Bofman, T. 1989. Attainment of Further research would be needed to determine syntactic and morphological accuracy by advanced whether these patterns are similar for the same language learners. Studies in second language populations of writers from the same L1 acquisition, 17-34. backgrounds (Chinese, English) in performing Biber, D., Douglas, B., Conrad, S., & Reppen, R. multiple writing tasks. Future research should also 1998. Corpus linguistics: Investigating language address what L1 factors might be contributing to structure and use. Cambridge University Press. these differences across language proficiency. Biber, D., & Gray, B. 2010. Challenging stereotypes Academic writing plays an important role in about academic writing: Complexity, elaboration, university education and graduates school explicitness. Journal of English for Academic education for academic success. Our findings pose Purposes, 9(1): 2-20. useful implications for L2 writing pedagogy in Biber, D., Gray, B., & Poonpon, K. 2011. Should we course designs and assessment. When formulating use characteristics of conversation to measure writing evaluation criteria, L2 writing instructors grammatical complexity in L2 writing should take language proficiency and writing development?. Tesol Quarterly, 45(1): 5-35. purposes into account. The course design and assessment should strengthen L2 writers’ Bulté, B., & Housen, A. 2012. Defining and operationalising L2 complexity. Dimensions of L2 understanding of writing for different purposes and performance and proficiency: Complexity, accuracy employ appropriate language forms in academic and fluency in SLA, 23-46. writing. To improve the diversity and elaboration of syntactic structures in written productions, L2 Bulté, B., & Housen, A. 2014. Conceptualizing and writing instructors need to be able to deal with the measuring short-term changes in L2 writing linguistic differences of their students. Since we complexity. Journal of second language writing, 26, 42-65. found that typologically different languages, such as Chinese and English may differ in patterns of Bulté, B., & Housen, A. 2018. Syntactic complexity in subordinate structures in academic writing, writing L2 writing: Individual pathways and emerging group activities should involve L2 writers’ awareness trends. International Journal of Applied about how the different cross-linguistic factors Linguistics, 28(1): 147-164. affecting L2 academic writing. Moreover, attention Casal, J. E., & Lee, J. J. 2019. Syntactic complexity and and awareness need to be given to the instruction writing quality in assessed first-year L2 of syntactic complexity, especially on how to help writing. Journal of Second Language Writing, 44, 51- L2 advanced-level writers develop 62. multidimensional syntactic complexity in Crawford, W., & Csomay, E. 2015. Doing corpus performing academic writing tasks. linguistics. Routledge. Crossley, S. A., Weston, J. L., McLain Sullivan, S. T., & McNamara, D. S. The development of writing References proficiency as a function of grade level: A linguistic Ai, H., & Lu, X. 2013. A corpus-based comparison of analysis. Written Communication, 28(3): 282-311. syntactic complexity in NNS and NS university Ellis, R., & Yuan, F. 2004. The effects of planning on students’ writing. Automatic treatment and analysis fluency, complexity, and accuracy in second of learner corpus data, 249-264. language narrative writing. Studies in second Language acquisition, 26(1): 59-84. Gray, B. 2015. Linguistic variation in research articles. Martínez, A. C. L. 2018. Analysis of syntactic John Benjamins Publishing Company. complexity in secondary education EFL writers at different proficiency levels. Assessing Writing, 35, 1- Halliday, M. A. K., & Matthiessen, C. 1999. Construing 11. experience through meaning: A language-based approach to cognition. London: Cassell. Norris, J. M., & Ortega, L. 2009. Towards an organic approach to investigating CAF in instructed SLA: Housen, A., De Clercq, B., Kuiken, F., & Vedder, I. The case of complexity. Applied linguistics, 30(4): 2019. Multiple approaches to complexity in second 555-578. language research. Second language research, 35(1): 3-21. Ortega, L. 2003. Syntactic complexity measures and their relationship to L2 proficiency: A research Jiang, J., Bi, P., & Liu, H. 2019. Syntactic complexity synthesis of college-level L2 writing. Applied development in the writings of EFL learners: Insights linguistics, 24(4): 492-518. from a dependency syntactically-annotated corpus. Journal of Second Language Writing, 46, 100666. Ortega, L. 2015. Syntactic complexity in L2 writing: Kuiken, F., Vedder, I., Housen, A., & De Clercq, B. Progress and expansion. Journal of Second Language 2019. Variation in syntactic complexity: Writing, 29, 82-94. Introduction. International Journal of Applied Parkinson, J., & Musgrave, J. 2014. Development of Linguistics, 29(2): 161-170. noun phrase complexity in the writing of English for Kyle, K. 2016. Measuring syntactic development in L2 Academic Purposes students. Journal of English for writing: Fine grained indices of syntactic complexity Academic Purposes, 14, 48-59. and usage-based indices of syntactic sophistication. Staples, S., & Reppen, R. 2016. Understanding first- Larsen-Freeman, D. 2006. The emergence of year L2 writing: A lexico-grammatical analysis complexity, fluency, and accuracy in the oral and across L1s, genres, and language ratings. Journal of written production of five Chinese learners of Second Language Writing, 32, 17-35. English. Applied linguistics, 27(4)：590-619. Taguchi, N., Crawford, W., & Wetzel, D. Z. 2013. What Larsen-Freeman, D. 2009. Adjusting expectations: The linguistic features are indicative of writing quality? A study of complexity, accuracy, and fluency in second case of argumentative essays in a college language acquisition. Applied linguistics, 30(4) ： composition program. Tesol Quarterly, 47(2): 420- 579-589. 430. Li Charles, N., & Thompson Sandra, A. 1976. Subject Wang, L. 1945. Zhongguo yufa lilun (theories of and topic: a new typology of language. Subject and Chinese grammar). Beijing: Zhonghua Press. topic, 459-489. Wolfe-Quintero, K., Inagaki, S., & Kim, H. Y. Lorés, R. 2004. On RA abstracts: from rhetorical 1998. Second language development in writing: structure to thematic organisation. English for Measures of fluency, accuracy, & complexity (No. specific purposes, 23(3), 280-302. 17). University of Hawaii Press. Lu, X. 2011. A corpus-based evaluation of syntactic Yoon, H. J., & Polio, C. 2017. The linguistic complexity measures as indices of college-level ESL development of students of English as a second writers' language development. TESOL language in two written genres. Tesol Quarterly, 45(1), 36-62. Quarterly, 51(2): 275-301. Lu, X., & Ai, H. 2015. Syntactic complexity in college- level English writing: Differences among writers with diverse L1 backgrounds. Journal of Second Language Writing, 29, 16-27. Mancilla, R. L., Polat, N., & Akcay, A. O. 2017. An investigation of native and nonnative English speakers’ levels of written syntactic complexity in asynchronous online discussions. Applied Linguistics, 38(1):112-134.