Introduction
The construct of linguistic complexity has garnered much interest in second language (L2) writing research as well as in heritage language (HL) research more recently. Linguistic complexity is seen as an indicator of language development and mastery, as more proficient learners have been shown to use richer and more diverse vocabulary and grammatical elaboration when writing. Nonetheless, linguistic complexity is known to vary substantially in light of a text’s rhetorical purpose, audience, and form. Although strong lexico-syntactic sophistication and diversity are often considered markers of advanced writing in academic registers, that is not necessarily the case in more casual registers (e.g., Hyland, Reference Hyland2006; Norris & Ortega, Reference Norris and Ortega2009). Accordingly, research has begun to explore how L2 writers make distinctions in their linguistic complexity across communicative contexts involving different register configurations (e.g., Larsson & Kaatari, Reference Larsson and Kaatari2020; Qin & Uccelli, Reference Qin and Uccelli2020); however, HL writers have yet to occupy a focal point in this line of inquiry. Thus, the first goal of this study was to investigate whether Spanish HL and L2 writers show differential patterns of linguistic complexity variation across elicited register conditions.
Besides tasks’ situational characteristics, variability in writers’ linguistic complexity is also posited to be influenced by relevant cognitive-linguistic and affective individual differences, such as their language proficiency or perceived value attributed to writing (e.g., Kormos, Reference Kormos2012; Papi et al., Reference Papi, Vasylets, Ahmadian, Li, Hiver and Papi2022). Indeed, composing text is a complex endeavor that requires successful coordination of multiple resources and processes, and individuals are known to differ in their skills and motivation to do so effectively. Although L2 research has taken steps to investigate the contributions of learners’ proficiency and motivational beliefs to linguistic complexity in writing (e.g., Lee, Reference Lee2020; Qin & Uccelli, Reference Qin and Uccelli2020), these connections have not yet been explored in HL writing. Accordingly, the second twofold goal of this study was to examine whether proficiency and motivational beliefs differentially affect linguistic complexity in HL and L2 writing. To provide a more fine-grained analysis, the study considers different measures of proficiency (an elicited imitation task [EIT] and a cloze test) and task-value beliefs attributed to writing (intrinsic/interest and cognitive/linguistic value).
Background
Writing in a heritage and a second language
In the US, a HL learner is an individual who is “raised in a home where a non-English language is spoken, who speaks or at least understands the language, and who is to some degree bilingual in that language and in English” (Valdés, Reference Valdés, Peyton, Ranard and McGinnis2001, p. 38). HL learners are native bilinguals of a minority, nonhegemonic language who receive all or most of their formal education in the societal majority language, gradually becoming dominant in it. The early, naturalistic bilingual experiences of HL learners contrast with those associated with traditional adult L2 learners who first access the language in formal academic settings and are exposed to ample written input and metalinguistic information early in their developmental trajectories. Unlike L2 learners, HL learners acquire and continue to primarily use the language at home and in the community, and, although the population is highly heterogeneous, their linguistic knowledge is often described as implicit and automatized. Considering these differences in age and context of acquisition, it is not surprising to find that HL and L2 students report diverging self-perceptions about their linguistic strengths and needs. For instance, in a survey conducted among university students, Hedgcock and Leftkowitz (Reference Hedgcock, Leftkowitz and Manchón2011) found that HL students perceived “writing” to be the language skill they needed to develop the most, whereas L2 learners did so for “speaking.”
Different assumptions are also made about HL and L2 students’ literacy, which, in broad terms, refers to “reading and writing effectively in a variety of contexts” (Pilgrim & Martinez, Reference Pilgrim and Martinez2013, p. 60). Elola (Reference Elola2018) notes that HL writing tends to be characterized by “an unevenness [that] can be observed in overall proficiency across a variety of registers” (p. 120), which is attributed to limited opportunities for academic literacy development in the HL. In longitudinal research documenting academic writing skill development, Colombi (Reference Colombi, Schleppegrell and Colombi2002) observed that Spanish HL writers’ early essays show features that are typical of more “informal, oral-like registers” (p. 70) and later on incorporate features that are more common in academic texts, such as increased lexical density and grammatical complexity (e.g., Achugar & Colombi, Reference Achugar, Colombi, Ortega and Byrnes2008; Colombi, Reference Colombi, Schleppegrell and Colombi2002). A recent study by Bowles and Bello-Uriarte (Reference Bowles, Bello-Uriarte, Sato and Loewen2019) also revealed that Spanish HL learners’ argumentative writing demonstrates increased syntactic complexity and lexical sophistication over a semester of language study. Similar trends for enhanced linguistic complexity have been reported in longitudinal L2 research (e.g., Lu, Reference Lu2011; Yoon & Polio, Reference Yoon and Polio2017). Nonetheless, unlike the assumptions commonly put forth in HL writing research, L2 learners are often seen as having lower command of written registers in nonacademic contexts. Specifically, Qin and Uccelli (Reference Qin and Uccelli2020) suggest that, for L2 learners “who have studied [the L2] mostly in formal settings, academic [language] might be more accessible,” because “they may have been minimally exposed” (p. 4) to the L2 in more informal out-of-school contexts.
In sum, it is evident that, given the multifaceted differences that exist between HL and L2 learners’ bilingual experiences and knowledge, current L2 writing research findings cannot (and should not) be generalized to the HL context (see Elola, Reference Elola2018). Although there has been progress in understanding HL development in recent years, more studies are needed that investigate important questions in HL writing, along with research that also explores how relevant task and individual learner factors may differentially affect HL and L2 writing. Such research is theoretically and pedagogically significant because it offers important insights into students’ linguistic and literacy profiles, including factors that contribute to their variability, which are critical in guiding curricular and instructional decisions. With this in mind, the current study examines how Spanish HL and L2 learners’ writing performance—through the lens of linguistic complexity—differs across elicited register conditions and whether individual differences in Spanish proficiency and value beliefs attributed to Spanish writing differentially influence their writing performance.
Linguistic complexity in writing and the relevance of register
Linguistic complexity is a multidimensional construct referring to “the degree of elaboration, the size, breadth, width, or richness of the learners’ [linguistic] system or ‘repertoire’ (Bulté & Housen, Reference Bulté, Housen, Housen, Kuiken and Vedder2012, p. 25), with grammatical and lexical complexity viewed as its two major components. Ortega (Reference Ortega, Kortmann and Szmrecsanyi2012) notes that L2 research has employed linguistic complexity with three main goals—namely, to assess proficiency, explain performance, and identify L2 development benchmarks. Several measures of syntactic and lexical complexity have been identified as indicators of language proficiency and development (e.g., Bulté & Housen, Reference Bulté and Housen2014; Crossley & McNamara, Reference Crossley and McNamara2012; Lu, Reference Lu2011; Norris & Ortega, Reference Norris and Ortega2009; Ortega, Reference Ortega, Kortmann and Szmrecsanyi2012). For instance, lexical diversity—an index of how many unique words are present in a text—has been positively associated with L2 proficiency and text quality (e.g., Crossley & McNamara, Reference Crossley and McNamara2012; Crossley et al., Reference Crossley, Salsbury, McNamara and Jarvis2011). At a syntactic level, clausal subordination rate has also been connected with L2 proficiency, albeit less systematically (see Norris & Ortega, Reference Norris and Ortega2009). In a series of studies, Biber and colleagues have argued that phrasal complexity, especially noun phrase complexity, is a more relevant marker of writing proficiency in academic contexts, with clausal complexity being a stronger indicator in oral registers (e.g., Biber et al., Reference Biber, Gray and Poonpon2011, Reference Biber, Gray and Staples2016). Findings from L2 studies, conducted predominantly in English as a second or foreign language settings, have also provided support for this connection between phrasal complexity and academic writing proficiency (e.g., Bulté & Housen, Reference Bulté and Housen2014; Lu, Reference Lu2011; Qin & Uccelli, Reference Qin and Uccelli2020).
Although the links between linguistic complexity and proficiency are theoretically and empirically supported, it is well assumed that higher linguistic complexity is not always equated with higher command in a language (Qin & Uccelli, Reference Qin and Uccelli2020). Developing more sophisticated linguistic repertoires is an important goal in most language programs and an integral part of SLA (Ortega, Reference Ortega, Kortmann and Szmrecsanyi2012), but this goal is usually realized within the broader objective of increasing students’ communicative L2 competence. Understanding writing as “a dynamic social activity involving multiple situational factors” (Qin & Uccelli, Reference Qin and Uccelli2020, p. 3), researchers have been interested in investigating how L2 learners change the linguistic features of their texts in consideration of different audiences, communicative functions, and contextual variables. For instance, studies focusing on genre effects in L2 writing have documented that learners produce more syntactically and lexically complex language when writing argumentative or expository texts as opposed to narrative texts (e.g., Lu, Reference Lu2011; Qin & Uccelli, Reference Qin and Uccelli2020; Yoon & Polio, Reference Yoon and Polio2017).
Of central relevance to this study, recent work has focused on how L2 texts vary in linguistic complexity across register configurations. In broad terms, register refers to “a variety [of language] associated with a particular situation of use (including particular communicative purposes)” (Biber & Conrad, Reference Biber and Conrad2009, p. 6). Registers are usually described for their prevalent lexical and grammatical features and their situational contexts (e.g., whether they are communicated through oral or written channels) as well as the functional relationship between the two. The assumption is that linguistic attributes are commonly found within a register because they align well with the register’s goals and situational context (see Biber & Conrad, Reference Biber and Conrad2009). Of note, ample research has documented that academic texts tend to be characterized by greater lexical diversity and syntactic elaboration than nonacademic, more casual texts (see Hyland, Reference Hyland2006).
Considering their differing bilingual backgrounds, HL and L2 learners can be expected to produce texts that vary across key linguistic features, particularly in relation to different situational contexts. This assumption is grounded on theoretical proposals that conceptualize language development as stemming from learners’ socialization into different discourse communities. From such perspectives, proficiency in a language can depend strongly on learners’ opportunities to learn and engage in specific linguistic uses and literacy practices across a range of different sociopragmatic contexts (with diverse audiences, communicative purposes, etc.; Cazden, Reference Cazden2001; Heath, Reference Heath2012; Ninio & Snow, Reference Ninio and Snow1996). Because HL learners are exposed to and primarily employ the language at home and in the community, whereas traditional L2 learners are exposed to it first and mostly use it in instructed classroom environments, important differences can be assumed to exist in their cumulative opportunities to access and write in the language across diverse discourse contexts.
Although some research has begun investigating L2 and HL learners’ writing performance across registers, no contrastive studies to date have considered both populations. In the L2 writing domain, Qin and Uccelli (Reference Qin and Uccelli2020) examined how adult English as a foreign language (EFL) learners (from Chinese, French, and Spanish L1 backgrounds) shifted their lexico-syntactic and discourse complexity across two persuasive tasks that were designed to be similar in topic while eliciting different registers: a personal email to a close friend and an academic report to an educational authority. Students’ academic texts showed greater lexical diversity and noun phrase complexity, but no differences were found for clausal subordination.
Adopting a corpus-based approach, Larsson and Kaatari (Reference Larsson and Kaatari2020) explored the interaction between syntactic complexity and register using data from two L2 English learner corpora and one reference corpus. Texts from academic prose, popular science, news, and fiction writing were considered, with each register placed along a continuum of formality based on their situational characteristics. They found that phrasal complexity measures were correlated with formality and that L2 texts were similar to expert academic texts in some respects (e.g., reliance on longer T-units) but not others (e.g., lower use of complex nominals).
In the HL context, a notable study by Martinez (Reference Martinez2007) examined how adult Spanish HL learners’ use of subject pronoun expression varied when students completed a free, ungraded writing task during class compared with a formal, graded writing task as homework. There was a higher incidence of overt pronouns in the homework task such that pragmatic features influenced by English were more prevalent in HL students’ graded versus ungraded narratives. These findings were taken to suggest that HL writers’ linguistic choices are discursively situated and take into consideration their knowledge of textual expectations, based partly on factors pertaining to the communication channel or participants’ role structure.
Building on this emerging scholarship, the present study examines how Spanish HL and L2 writers’ linguistic complexity differentially varies across register contexts. Participants completed two writing tasks that differed in core characteristics of their contextual configuration, as implemented through the systemic-functional framework of field–tenor–mode (Halliday & Hasan, Reference Halliday and Hasan1985). The parameters of field (i.e., what is occurring; the topics or actions of the communicative event), tenor (i.e., who is participating; the nature of the relationship among participants, including their social status), and mode (i.e., what role language is playing; the mode and organization of the language, including the communication channel) are well theorized and have been used to describe textual variation in previous writing research (see, e.g., Martinez, Reference Martinez2007). Lastly, a methodological novelty of the present research was its analysis of linguistic complexity through principal component analysis. This statistical technique, as detailed subsequently, can provide an informative and reliable analysis by avoiding the redundancy problem derived from the fact that many complexity metrics measure highly similar things (see Norris & Ortega, Reference Norris and Ortega2009).
Individual differences in heritage and second language writing
Instructed SLA research has long acknowledged the role of individual differences in language learning; however, their contributions in L2 and (to a much lesser extent) HL writing specifically have begun to receive increased attention only recently (see Kormos, Reference Kormos2012; Papi et al., Reference Papi, Vasylets, Ahmadian, Li, Hiver and Papi2022). The following section presents relevant theoretical and empirical background for the focal cognitive-linguistic and affective factors of this study—namely, language proficiency and writing motivational beliefs.
Role of linguistic proficiency on writing performance
The construct of linguistic proficiency is recognized as essential in the understanding of language acquisition and, thus, has taken a central role in the study of L2 writing. In conceptualizing proficiency, Hulstijn (Reference Hulstijn2011, Reference Hulstijn2012) proposes a distinction between basic and higher language cognition (BLC and HLC). BLC is limited to the processing of oral language (listening and speaking) involving primarily high-frequency elements; in contrast, HLC entails the processing of written language (reading and writing) that may contain lower frequency lexicogrammatical features. Hulstijn (Reference Hulstijn2012) explains that BLC may be regarded as “the language knowledge shared by all adult native speakers (…), whereas HLC exhibits individual differences in language control, potentially affected by attributes such as literacy, age, level of education, profession, or leisure-time activities” (p. 429).
Based on current theoretical accounts, proficiency is expected to exert an important influence on HL and L2 writing processes and outcomes (e.g., Chenoweth & Hayes, Reference Chenoweth and Hayes2001; Kellogg, Reference Kellogg, Levy and Ransdell1996; McCutchen, Reference McCutchen2000), although such accounts do not make specific predictions about different components of language cognition or types of linguistic knowledge. In general terms, higher proficiency writers have access to broader and more automatized linguistic knowledge and tend to allocate processing time more evenly across composing stages (Roca de Larios et al., Reference Roca de Larios, Manchón, Murphy and Marín2008), promoting superior writing performance. It is assumed that proficiency underlies writers’ ability to effectively manage the competing attentional demands posed by textual planning, formulation, and monitoring processes. In all, higher proficiency writers are predicted to have advantages at all stages of writing, particularly when engaging in lexicogrammar and orthographic encoding, which depend strongly on linguistic means. Because high-level planning knowledge (e.g., regarding audience, tone), content knowledge (about the topic of writing), and discourse knowledge (e.g., regarding language forms, text format) all interact to influence writing processes and outcomes (McCutchen, Reference McCutchen2000), the predictive value of proficiency on writers’ performance may fluctuate across tasks with different functional/communicative demands.
Several studies have documented positive associations between proficiency and writing performance among L2 learners (e.g., Kim et al., Reference Kim, Tian and Crossley2021; Lee, Reference Lee2020; Qin & Uccelli, Reference Qin and Uccelli2020; Roca de Larios et al., Reference Roca de Larios, Manchón, Murphy and Marín2008; Sasaki & Hirose, Reference Sasaki and Hirose1996; Schoonen et al., Reference Schoonen, Gelderen, Glopper, Hulstijn, Simis, Snellings and Stevenson2003), although this connection remains largely unexplored in the HL context. Some of this work has also focused on the connection between proficiency and linguistic complexity specifically because, as noted by Ortega (Reference Ortega, Kortmann and Szmrecsanyi2012), “at the core of the construct is the claim that the ability to produce more linguistically complex oral or written texts reflects increasingly more developed and mature capacities to use the second language” (p. 127). In a particularly relevant study, Qin and Uccelli (Reference Qin and Uccelli2020) reported that higher L2 English proficiency (as measured by a standardized test) was accompanied by greater lexical diversity and syntactic complexity, mostly in terms of noun phrase complexity, in L1 Korean learners’ persuasive L2 writing. However, they found that more proficient learners’ texts also showed lower clausal subordination, especially in academic writing.
Echoing Hulstijn (Reference Hulstijn2012), it is useful to consider more closely how L2 proficiency has been operationalized in prior research to better understand any existing empirical gaps. Some L2 work has used researcher-designed tests (Schoonen et al., Reference Schoonen, Gelderen, Glopper, Hulstijn, Simis, Snellings and Stevenson2003) such as multiple-choice vocabulary tests and fill-in-the-blank grammar tests. Nonetheless, most L2 research has administered standalone standardized tests (e.g., Qin & Uccelli, Reference Qin and Uccelli2020; Roca de Larios et al., Reference Roca de Larios, Manchón, Murphy and Marín2008) such as the Oxford Placement Test, or specific sections of such tests, including vocabulary or reading comprehension portions (e.g., Kim et al., Reference Kim, Tian and Crossley2021; Lee, Reference Lee2020; Sasaki & Hirose, Reference Sasaki and Hirose1996). At times, researchers have provided an explicit rationale for the use of such subtests as proxy measures of overall linguistic knowledge (see Kim et al., Reference Kim, Tian and Crossley2021, regarding a vocabulary test). Of relevance to this study, an examination of current methodological practices points to a strong reliance on proficiency tests that primarily draw on learners’ ability to process language in the HLC domain (i.e., requiring literacy skills) as opposed to the BLC domain. Additionally, when multiple measures are employed, it is common for researchers to combine them for analysis without teasing out their independent effects. Thus, not much is known about how different types of linguistic knowledge affect HL or L2 writing performance. Considering Hulstijn’s (Reference Hulstijn2011, Reference Hulstijn2012) proposal, investigating how distinct proficiency measures—such as those involving only speech reception and production versus those involving literacy skills—are associated with writing outcomes would prove theoretically and empirically informative.
In light of these considerations, a primary aspect of this study concerns its assessment of proficiency using an oral/aural EIT and a written cloze test (adapted from the Diploma de Español como Lengua Extranjera [DELE]), to compare their differential roles in HL and L2 writing performance. These two measures were deemed especially pertinent in this study, given that HL and L2 learners’ varying bilingual trajectories are assumed to promote the development of different types of linguistic knowledge. The EIT requires listening and repeating spoken utterances, and its design aligns well with the BLC processing dimension, which “reflects the fact that speaking and understanding speech (…) involve parallel processing of phonetic-phonological, lexical, and grammatical information in high speed” (Hulstijn, Reference Hulstijn2011, p. 231). Unlike the EIT, where learners are expected to “rely on bottom-up information from the signal to form a representation of the sentence,” the cloze test requires them to “rely on top-down information about the language and about the text to predict missing words in the text” (Gaillard & Tremblay, Reference Gaillard and Tremblay2016, p. 427). Its design is likely to draw more heavily on declarative (rather than procedural) knowledge and explicit (rather than implicit) mental processes, given the speed at which linguistic information can be processed. In summary, although both tests are projected to share substantive variance, differences may be anticipated in their relative contributions to writing performance.
Influence of motivational beliefs on writing performance
Unlike proficiency, the role of motivational factors in L2 and HL writing is much less well understood. Research into such factors is nonetheless critical for advancing both theory and pedagogy, as students’ beliefs and perceptions about writing influence their cognitive, emotional, and behavioral engagement in writing activities (Papi et al., Reference Papi, Vasylets, Ahmadian, Li, Hiver and Papi2022). From this perspective, motivation has been widely acknowledged as a central determinant influencing learners’ intentional learning and writing achievement (e.g., Tahmouresi & Papi, Reference Tahmouresi and Papi2021; Tsao et al., Reference Tsao, Tseng and Wang2017). Its role has been empirically investigated through diverse conceptual frameworks, such as the L2 motivational self-system (see Dörnyei, Reference Dörnyei, Dörnyei and Ushioda2009), self-determination theory (see Ryan & Deci, Reference Ryan and Deci2020), or expectancy-value theory (see Wigfield & Eccles, Reference Wigfield and Eccles2000). Although these frameworks diverge in their primary foci (e.g., focus on self-concept alignment vs. subjective value attached to a task), they collectively recognize motivation as a multifaceted phenomenon that interacts with both internal psychological drivers and contextual influences.
Two recent L2 studies that have investigated motivational influences on writing through the L2 motivational self-system have highlighted the beneficial role of the ideal L2 self, in line with accumulated findings in the broader SLA literature. In the Korean EFL context, Jang and Lee (Reference Jang and Lee2019) examined how learners’ L2 ideal and ought-to selves affected students’ use of writing strategies and writing quality in a descriptive task that was scored using standardized descriptors. A stronger ideal L2 self was positively related to planning strategy use—and overall rate of strategy use—as well as writing quality, whereas a stronger ought-to L2 self was only associated with increased revising strategy use. In the Iranian EFL context, Tahmouresi and Papi (Reference Tahmouresi and Papi2021) explored connections among L2 writing future selves, intended L2 writing effort, L2 writing emotions, and L2 writing achievement. Students’ writing achievement was operationalized as their final English writing course grades. A stronger ideal L2 writing self was predictive of L2 writing joy, intended effort, and achievement, whereas a stronger ought-to L2 writing self was positively associated with L2 writing anxiety and achievement.
Additional research, guided by self-determination theory and expectancy-value theory, has further documented a beneficial role of intrinsic motivation for supporting writing outcomes and favorable affective dispositions (e.g., Tsao et al., Reference Tsao, Tseng and Wang2017; Yeşilyurt, Reference Yeşilyurt2008). Central to the present study are task value beliefs, which refer to individuals’ subjective judgments about the significance and worthiness of a particular task (Wigfield & Eccles, Reference Wigfield and Eccles2000). The concept is grounded in the assumption that individuals are more likely to expend time and be cognitively engaged in a task, such as writing, when they perceive it to be valuable. Specifically, students with strong intrinsic/interest value in a task—namely, those who attach heightened personal significance to it and derive more enjoyment from performing it—are likely to evidence greater attentional effort, persistence, and involvement with the task, as well as superior academic performance. Thus, positive effects of intrinsic/interest value are expected on both L2 and HL writing outcomes (Kormos, Reference Kormos2012; Papi et al., Reference Papi, Vasylets, Ahmadian, Li, Hiver and Papi2022).
Of relevance to this study, within the domain of L2 writing, Lee (Reference Lee2020) recently introduced the construct of cognitive/linguistic value belief, which refers to “individuals’ evaluative beliefs regarding the potential benefits of L2 writing on their cognitive and linguistic development” (p. 1241). To investigate this new construct’s predictive potential, Lee examined its influence alongside other individual difference variables on Korean EFL students’ writing performance, which was elicited using descriptive and argumentative tasks and rated using an analytic rubric. Notably, cognitive/linguistic value had a stronger positive influence on the quality of L2 writing than did intrinsic/interest value. The findings suggest that how students appraise the potential linguistic and cognitive advantages derived from L2 writing can be linked to their behavioral engagement, effort, and writing performance. More generally, they also underscore the importance of extending the exploration of task-value beliefs beyond the scope of intrinsic/interest value in this line of inquiry.
Building on this research, the current study investigates how HL and L2 learners’ intrinsic/interest and cognitive/linguistic value beliefs regarding Spanish writing are associated with their writing performance. Amid the various motivational factors relevant to writing (for a review, see Papi et al., Reference Papi, Vasylets, Ahmadian, Li, Hiver and Papi2022), this study chose to center on task-value beliefs due to their close connection with aspects linked to learners’ task-related effort, persistence, and overall performance, providing a targeted lens through which to explore connections with specific writing outcomes. A focus on intrinsic/interest and cognitive/linguistic value beliefs was also deemed pertinent considering that HL and L2 learners can evidence distinct affective dispositions toward writing on account on their different bilingual backgrounds, including varying formal classroom experiences. For instance, L2 learners may be more inclined to approach writing as a strategic way to improve their language abilities relative to HL learners; consequently, it is possible that individual differences in cognitive/linguistic value beliefs exhibit distinct predictive patterns across the two learner populations.
The present study
In light of the gaps identified in the current literature, the following research questions guided the study: (1) To what extent does Spanish HL and L2 writers’ linguistic complexity differ across elicited register conditions? (2) Does Spanish proficiency, as measured by elicited imitation and cloze tests, differentially affect Spanish HL and L2 writers’ linguistic complexity? and (3) Do intrinsic/interest and cognitive/linguistic value beliefs about Spanish writing differentially affect Spanish HL and L2 writers’ linguistic complexity?
Method
Design and protocol
Participants were HL and L2 Spanish learners recruited through various universities in the United States. Colleagues were contacted to share the information about the research opportunity with students, particularly those in upper-intermediate or more advanced courses, where greater emphasis on Spanish writing skills is typically observed. The study comprised two sessions, and all participants met individually online with the author or another trained researcher to complete all tasks. The average time interval between the two sessions spanned 10 days for both the HL and L2 groups. Total study participation lasted approximately 2 hours. Participants received extra course credit (n = 9 L2 writers) or a $30 Amazon gift card (n = 58 HL writers; n = 45 L2 writers).
Following a within-subjects design, all participants performed two persuasive Spanish writing tasks that were topically similar but differed in their contextual configuration. Task administration was randomized and counterbalanced. In the first session, participants performed the first writing task and responded to a task-perception questionnaire. Next, they completed two Spanish proficiency tests. Afterwards, they received questionnaires about their language history and writing motivational beliefs. In the next session, participants performed the second writing task and subsequent task perception questionnaire. Study materials are available on IRIS and https://osf.io/wf3x8/.
Participants
The final sample comprised 58 HL and 54 L2 Spanish learners.Footnote 1 All HL learners reported that they had been raised in a home where Spanish was spoken. Their mean age was 20.38 years (SD = 2.22; Min–Max = 18–30); 48 identified as female, 8 identified as male, 1 identified as gender variant/nonconforming, and 1 preferred not to say. Thirty-seven participants were born and had been raised in the United States, and 21 participants were born in a Spanish-speaking country and arrived in the USA before (n = 1) or during elementary school years (n = 17) or during middle school years (n = 3). Their families were from various Latin-American countries. Participants’ age of onset of Spanish-English bilingualism was 4.71 years old (SD = 3.88; Min–Max = 1–16). The average amount of prior Spanish instruction was 1.33 years (SD = 1.26; Min–Max = 0–5) at university, 1.93 years (SD = 1.66; Min–Max = 0–4) at high school, and 2.44 years at middle/elementary school (SD = 3.24; Min–Max = 0–11).
L2 learners’ mean age was 20.56 (SD = 1.84; Min–Max = 18–30); 43 identified as female, 8 identified as male, and 3 identified as gender variant/nonconforming. Fifty-six participants had been born and raised in the USA; two participants were born in Ukraine and China and moved to the USA as infants (both reported an age of exposure to English of ≤ 3 years old). Participants’ mean age of exposure to Spanish was 12.00 years old (SD = 4.02; Min–Max = 4–19), and their prior Spanish instruction amounted to an average of 2.35 years (SD = 1.19; Min–Max = 0–5.50) at university, 3.43 years (SD = 1.26; Min–Max = 0–6) at high school, and 3.19 years (SD = 2.85; Min–Max = 0–9) at middle/elementary school. Lastly, both HL and L2 learners described their frequency of language use in an average week (see Appendix S1 online).
Materials
Writing tasks
All participants completed two persuasive writing tasks that were similar in topic but differed with respect to their elicited register condition, following Qin and Uccelli (Reference Qin and Uccelli2020; see Appendix S2 online). In the Email to Friend task, participants wrote a personal email to a close friend to persuade them that either studying or not studying abroad for a year in Mexico was the best option. In the Letter to Dean task, they wrote a letter to their college dean to persuade them that either establishing or not establishing a study abroad requirement for language majors was the best option. In both tasks, participants were asked to write a minimum of 200 words within 30 minutes. No access to dictionaries or other resources was provided. Both tasks followed González-Lloret and Ortega’s (Reference González-Lloret and Ortega2014) conceptualization of tasks, as they (a) had a communicative goal with a primary focus on meaning, (b) required participants to use their own resources, and (c) had real-world relevance.
The tasks were strategically designed to elicit different registers by varying the main aspects of their contextual configuration, operationalized based on Halliday and Hasan’s (Reference Halliday and Hasan1985) field–tenor–mode framework. This well-established framework has been employed in prior research (e.g., Martinez, Reference Martinez2007; Qin & Uccelli, Reference Qin and Uccelli2020) and was also adopted here, as it provides a systematic lens for examining how learners adapt their writing to different subject matters, participants and social roles/statuses, and modes of communication. The social purpose of both tasks was similar in that participants were instructed to persuade the recipient and the text topic drew on vocabulary related to studying abroad. In the Email to Friend task, the personal relationship between participants was closely familiar and entailed equal social status (friend to friend). In contrast, in the Letter to Dean task, there was greater social distance and differences in power hierarchy (student to educational authority). Last, both tasks required composing text, but communication channels differed (email vs. letter).
Task perception questionnaire
The questionnaire, administered immediately after each task, was adapted from Révész et al. (Reference Révész, Kourtali and Mazgutova2017). It included statements on a 9-point Likert-type scale focusing on writers’ perceptions of (a) overall mental effort required to complete the task, (b) overall task difficulty, (c) difficulty in planning the essay, and (d) difficulty in linguistic encoding.
Spanish language proficiency measures
Elicited imitation task
The EIT asked participants to listen to and repeat 30 Spanish sentences of increasing length (from 7 to 17 syllables) and complexity. Utterances were minimally adapted from Ortega (Reference Ortega2000; see Faretta-Stutenberg & Morgan-Short, Reference Faretta-Stutenberg, Morgan-Short, Sanz and Morales-Front2018). After each sentence, there was a 2-s pause, followed by a 0.5-s tone that cued the participant to repeat the utterance out loud. Participants were told to repeat each utterance exactly as they heard it and to repeat as much of it as they could.
Cloze test
Participants were instructed to read a Spanish passage and choose which of three options best fit each of the blank spaces in it. The cloze test included 20 items and was part of a modified version of the Diploma de Español como Lengua Extranjera, widely used in previous HL research (see Montrul, Reference Montrul2016).
Motivational beliefs: Task value scales
Motivational value beliefs were operationalized as perceived intrinsic/interest and cognitive/linguistic value attributed to Spanish writing and were measured using seven Likert-type items each from Lee (Reference Lee2020), modified for Spanish (see Appendix S3 online). Participants were asked to indicate the extent to which they agreed with each statement using a 5-point scale anchored at 1 = Strongly disagree and 5 = Strongly agree. The intrinsic/interest value items tapped into participants’ enjoyment, interest, and involvement in Spanish writing (e.g., It is fun to write in Spanish), and the cognitive/linguistic value items measured beliefs regarding the cognitive and linguistic benefits derived from writing in Spanish (e.g., I can improve my Spanish vocabulary if I write in Spanish).
Coding and scoring
Writing performance
Multiple measures were selected to capture distinct aspects of linguistic complexity while ensuring limited conceptual overlap (see Norris & Ortega, Reference Norris and Ortega2009). The syntactic complexity measures were mean length of sentence (MLS), mean number of dependent clauses per sentence (DC), mean number of coordinate clauses per sentence (CC), and mean length of noun phrase (MLNP). MLS provides a broad metric by capturing complexity achieved by several means. DC and CC are more specific measures of clausal complexity via subordination and coordination devices, respectively, while MLNP measures complexification at the phrasal level.
The lexical complexity measures were mean word length (in characters; [MLW]), lexical frequency (LF), and Voc-D. MLW taps into sophisticated word use, as longer words tend to be less frequent (e.g., Yoon & Polio, Reference Yoon and Polio2017). LF was assessed using the MultiLingProfiler (Anthony et al., Reference Anthony, Finlayson, Marsden, Avery and Hawkes2020) and calculated as the percentage of words found among the top 2,000 most frequent Spanish words. Voc-D (Malvern et al., Reference Malvern, Richards, Chipere and Durán2004), calculated using D_Tools (Meara & Miralpeix, Reference Meara and Miralpeix2006), captures lexical diversity by assessing the degree of word repetition while accounting for text length. Misspelled words were corrected, and proper nouns and interjections were deleted for the analyses.
Following a multidimensional view of linguistic complexity, syntactic and lexical measures were submitted to separate principal component analyses. This technique takes advantage of the inherently coordinated nature of complexity metrics and avoids the need for multiple testing. Specifically, principal component analysis allows us to reduce multiple indices into one or more components, accounting for as much of the variance in the dataset as possible. It is also valuable for identifying any complexity indices that may pattern or cluster differently from the rest. Component scores were used in subsequent analyses.
Language proficiency
Elicited imitation task
Utterances were transcribed and rated following Ortega’s (Reference Ortega2000) guidelines.Footnote 2 Each utterance was scored for 0–4 points based on the accuracy and the idea units included. Each transcription was rated independently by the author and a trained researcher, and all scores were compared. Any discrepancies were discussed until 100% agreement was reached on all ratings. The mean score for HL and L2 writers was 103.75 (SD = 15.78; Min–Max = 52–120) and 68.94 (SD = 21.30; Min–Max = 30–112), respectively.
Cloze test
Participants received 1 point for each accurate response for a total possible score of 20 points. The mean score for HL and L2 writers was 13.84 (SD = 3.22; Min–Max = 5–19) and 9.54 (SD = 3.14; Min–Max = 4–16), respectively. Cronbach’s alpha was .75, indicating satisfactory reliability.
Writing motivational beliefs
For each scale (see Table 1 for descriptive statistics), its component structure was analyzed using principal component analysis with oblimin rotation and Kaiser’s criterion of eigenvalue > 1.0 (see Appendix S4 online).Footnote 3 For the intrinsic/interest value scale (Bartlett’s test: χ2(21) = 268.16, p <.001, KMO = .79), all but one of the items loaded onto the same component. This item was excluded from further analyses. For the cognitive/linguistic value scale (Bartlett’s test: χ2(21) = 455.21, p <.001, KMO = .86), all items loaded onto a single component. Cronbach’s alpha for the final intrinsic/interest and cognitive/linguistic value scales was .82 and .88, respectively. Component scores were used in subsequent analyses.
Descriptive statistics for task value beliefs
Task perception questionnaire
Both HL and L2 writers perceived that the Letter to Dean task required more mental effort and was more difficult overall (see Appendix S5 online for descriptive and inferential statistics). Mixed-effects models confirmed that task explained significant variance in all dimensions. LX status (HL or L2) explained variance only in mental effort and formulation difficulty, with L2 writers providing higher ratings for these categories relative to HL writers.
Analysis
Separate linear mixed-effects models were built using the GAMLj suite in Jamovi (v. 2.0.0.0). Models were computed with maximum likelihood as the estimation method, using the lmerTest R package with BOBYQA optimization. Best-fitting models were identified via chi-square test comparisons using model deviances to detect significant differences between full and reduced models (see Appendix S6 online for summary of model comparisons). Only parameters that significantly improved model fit were retained. The component scores for syntactic and lexical complexity were used as criterion variables. To address RQ1, task, LX status, and their interaction were entered as fixed effects. To address RQ2, task was employed as a covariate, and LX status, EIT z scores, cloze test z scores, and their interactions with LX status were entered as fixed effects. Last, to address RQ3, task was treated as a covariate, and LX status, intrinsic/interest value, cognitive/linguistic value, and their interactions with LX status were included as fixed effects. Random intercepts were included for participants in all models.Footnote 4
Results
RQ1: LX status and linguistic complexity across discourse contexts
Overall, both groups were similar in producing more grammatically and lexically complex language in the Letter to Dean than the Email to Friend task (see Tables 2 and 3). Nonetheless, HL writers’ essays evidenced greater linguistic complexity. Close observation of effect sizes reveals that changes in MLNP showed the most pronounced difference between HL and L2 groups when comparing registers, but DC evidenced the highest between-group differences overall.
Descriptive statistics for syntactic complexity
Descriptive statistics for lexical complexity
Linguistic complexity measures were submitted to principal component analyses using oblimin rotation (see Appendix S7 online for full details). Components with an eigenvalue greater than 1.0 were maintained. The initial analysis for syntactic complexity revealed two components (Bartlett’s test: χ2(6) = 268.73, p <.001, KMO = .59): MLS (.85), DC (.80), and MNLP (.83) loaded positively and strongly on the first component (eigenvalue = 2.18, % of variance = 54.51), whereas CC (.93) loaded strongly on the second component (eigenvalue = 1.05, % of variance = 26.15). Because CC loaded separately with no significant overlap, a follow-up component analysis was performed with MLS, DC, and MNLP only (Bartlett’s test: χ2(3) = 219.89, p <.001, KMO = .62). MLS (.91) showed the strongest loading, followed by DC (.85) and MLNP (.73). For lexical complexity, the analysis yielded a single component (Bartlett’s test: χ2(3) = 68.45, p <.001, KMO = .50). Both MLW (.85) and LF (−.87) showed similarly strong loadings, followed by Voc-D (.25). Component scores were employed in subsequent analyses.
The best-fitting model for syntactic complexity (Table 4) included LX status and task as significant predictors. This indicates that, overall, HL writers produced more grammatically complex essays than L2 writers and that the Letter to Dean task elicited greater syntactic complexity. The LX Status × Task interaction was also significant. Separate follow-up models revealed that, although both HL and L2 writers’ essays significantly varied in syntactic complexity across tasks, contextual distinctions were more pronounced among HL writers, B = −.92, SE = .13, t(58.00) = −7.21, p <.001, than L2 writers, B = −.48, SE = .09, t(54.00) = −5.39, p <.001, as illustrated in Figure 1. The marginal R 2 for the HL and L2 models was .19 and .11, respectively, suggesting that task explained more variance among HL writers.
RQ1: Optimal model on syntactic complexity
Note. R 2 marginal = .29, R 2 conditional = .65.
Syntactic complexity component scores by task and LX status.
Parameter estimates of the lexical complexity model (Table 5) revealed a different pattern. The optimal model included only a significant task coefficient, indicating that lexical complexity was generally greater in the Letter to Dean task. Inclusion of other fixed effects or interactions did not improve model fit.
RQ1: Optimal model on lexical complexity
Note. R 2 marginal = .20, R 2 conditional = .62.
RQ2: Proficiency and linguistic complexity
For RQ2, the best-fitting model for syntactic complexity (Table 6) included a significant EIT coefficient, indicating that higher EIT performance was associated with enhanced syntactic complexity. Inclusion of other fixed effects or interactions did not improve model fit.
RQ2: Optimal model on syntactic complexity
Note. R 2 marginal = .35, R 2 conditional = .63.
For lexical complexity, a more intricate picture emerged. The optimal model (Table 7) contained a significant cloze test coefficient, indicating that higher test accuracy was linked to increased lexical complexity. Both the LX Status × Cloze Test and the LX Status × EIT interactions were significant. Follow-up tests showed that, upon controlling for task effects, EIT performance was a significant predictor for L2 writers, B = .23, SE = .08, t(52.00) = 2.92, p = .005, but not HL writers, B = .15, SE = .12, t(56.00) = 1.29, p = .202. The cloze test was a significant predictor for both groups, but showed a greater effect among HL writers, B = .34, SE = .11, t(58.00) = 3.10, p = .003, than L2 writers, B = .17, SE = .08, t(54.00) = 2.12, p = .038.
RQ2: Optimal model on lexical complexity
Note. R 2 marginal = .33, R 2 conditional =.64.
RQ3: Motivational beliefs about LX writing and linguistic complexity
For RQ3, the best-fitting model for syntactic complexity (Table 8) included significant coefficients for intrinsic/interest value and cognitive/linguistic value, albeit with different predictive directions: Intrinsic/interest and cognitive/linguistic values were positively and negatively associated with syntactic complexity, respectively.
RQ3: Optimal model on syntactic complexity
Note. R 2 marginal = .34, R 2 conditional =.63.
Regarding lexical complexity, while a conditional model with intrinsic/interest value provided a slightly better fit to the data than the base model (-2LL Statistic = 6.49, p = .011), the intrinsic/interest value coefficient did not reach significance, B = .10, SE = .07, t(111.00) = 1.42, p = .158. The Bayesian information criterion difference between models was only 1.13, providing weak evidence for the conditional model’s strength against the base model (Raftery, Reference Raftery1995).
Discussion
LX status and linguistic complexity across registers
The first research question focused on the extent to which HL and L2 writers’ linguistic complexity production differs across elicited register conditions. Mixed-effects models showed that both HL and L2 writers evidenced greater lexico-syntactic complexity in the Letter to Dean task than in the Email to Friend task. Nonetheless, HL writers demonstrated more pronounced cross-register distinctions in syntactic complexity than did L2 writers. These findings provide novel comparative insights that critically expand prior HL and L2 writing research investigating intrawriter variability across discourse contexts (e.g., Martinez, Reference Martinez2007; Qin & Uccelli, Reference Qin and Uccelli2020).
The fact that both learner groups made consistent cross-task distinctions at the syntactic and lexical complexity levels indicates that university-level Spanish HL and L2 writers demonstrate a strong ability to adjust their linguistic styles across contextual configurations—encompassing primary variations in the field–tenor–mode parameters—that sought to elicit different registers. These results highlight learners’ ample sociopragmatic proficiency by revealing their ability to draw on their lexico-syntactic resources to construe different social meanings through writing. Although there has been continued discussion about limited register awareness in learner language (see, e.g., Gilquin & Paquot, Reference Gilquin and Paquot2008), this study suggests that adult Spanish HL and L2 learners with some formal classroom experience have developed a robust underlying register awareness that they draw from to structure their written discourse across communicative contexts.
Notably, although it is clear that both HL and L2 writers’ linguistic choices are influenced by situational characteristics—including the interpersonal and power relationships between the addresser and the addressee—cross-register variations in syntactic complexity were more robust among HL writers than L2 writers. That is, university Spanish HL writers adopted more grammatically complex language to meet the communicative functions of more formal registers relative to L2 learners, whereas the degree of cross-register differentiation for lexical complexity was comparable across groups. The results echo Ortega’s (Reference Ortega2020) observation that “the portrayal of HL speakers as uniformly weak in their HL literacy skills may be at least in part a methodological artefact” (p. 34), particularly when HL and L2 groups are compared. Although prior student self-report data has tended to emphasize an account of low HL literacy skills due, among other factors, to the linguistic insecurity commonly experienced by HL learners, direct measurement of writing performance reveals a different picture. The current findings present an alternative view that questions established assumptions about the prevailing cross-register asymmetries often associated with HL learners’ writing abilities.
Interestingly, cross-register differences between HL and L2 writers were most prominent for nominal complexity and not clausal complexity. Although HL and L2 learners’ essays showed virtually no differences in noun phrase length in the Email to Friend task, HL writers demonstrated considerably greater phrasal elaboration than L2 writers in the Letter to Dean task. By comparison, although the subordination measure showed the greatest overall discrepancy between groups, the magnitude of this difference remained stable across registers, with HL writers producing more subordinate clauses than L2 writers across the board. Thus, it appears that differences in Spanish HL and L2 writers’ cross-register patterns of grammatical complexity variation are marked for phrase-level complexity more so than for clause-level complexity. These observations are relevant in light of recent claims about the importance of nominal complexity as a predictor of formality (Larsson & Kaatari, Reference Larsson and Kaatari2020) and a superior indicator of academic writing proficiency relative to subordination (e.g., Biber et al., Reference Biber, Gray and Poonpon2011; Staples et al., Reference Staples, Egbert, Biber and Gray2016). Indeed, they suggest that HL writers exhibit a sophisticated command in the dense use of nominal complexity associated with academic registers, a pattern that is not (yet) observed among the L2 learner group of this study at the university level.
Roles of Spanish language proficiency measures in HL and L2 writing
The second research question focused on whether Spanish proficiency measures would differentially explain variability in HL and L2 writers’ linguistic complexity. The EIT and the cloze test scores emerged as positive predictors of syntactic and lexical complexity, respectively. Notably, the two proficiency measures were differentially linked to HL and L2 writing performance, as the EIT predicted lexical complexity only among L2 learners, whereas the cloze test did so among both LX groups, with a stronger effect for HL learners.
The findings lend broad support to previous studies documenting positive associations between proficiency and writing performance (e.g., Kim et al., Reference Kim, Tian and Crossley2021; Lee, Reference Lee2020; Roca de Larios et al., Reference Roca de Larios, Manchón, Murphy and Marín2008; Sasaki & Hirose, Reference Sasaki and Hirose1996) as well as linguistic complexity more specifically (e.g., Crossley et al., Reference Crossley, Salsbury, McNamara and Jarvis2011; Qin & Uccelli, Reference Qin and Uccelli2020). The results are in line with theoretical accounts predicting that individuals with higher linguistic knowledge can handle the competing demands imposed by writing processes on attentional resource limitations more effectively (e.g., Kellogg, Reference Kellogg, Levy and Ransdell1996; Kellogg et al., Reference Kellogg, Whiteford, Turner, Cahill and Mertens2013; McCutchen, Reference McCutchen2000) due to their broader, more automatized linguistic skills and more strongly developed strategic orientation to composing (e.g., Roca de Larios et al., Reference Roca de Larios, Manchón, Murphy and Marín2008; Schoonen et al., Reference Schoonen, Gelderen, Glopper, Hulstijn, Simis, Snellings and Stevenson2003). It follows that more proficient HL and L2 writers, who had greater access to a wider repertoire of complex linguistic features in long-term memory, were able to manipulate more information and elaborate the lexico-syntactic landscape of their essays on a larger scale, as evidenced by the greater syntactic and lexical complexity of their texts.
The associations observed between the EIT and syntactic complexity on one hand and the cloze test and lexical complexity on the other may be explained in relation to the linguistic abilities most strongly tapped by each proficiency test. As previously described, successful performance in the EIT requires learners to process and understand the utterances in order to repeat them, drawing on their BLC (Hulstijn, Reference Hulstijn2011, Reference Hulstijn2012). Mozgalina (Reference Mozgalina2015) further argues that the EIT measures “learners’ knowledge of core vocabulary, grammar, and phonology in combination with automated ability for their use in oral perception and production at the sentence level” (p. 40) (see Bowden, Reference Bowden2016, for further discussion). In line with this view, the results suggest that learners’ efficient access to such linguistic knowledge supports their ability to elaborate the syntactic landscape of written texts (e.g., through higher subordination rates, more phrasal complexity). That is, learners who evidence broader linguistic knowledge and more automatic processing of basic lexical and grammatical structures are better equipped to manage the competing demands of writing and, in particular, show advantages in the mapping of conceptual and syntactic encoding, leading to greater syntactic complexity.
In contrast, successful performance in the written cloze test stands as more likely to rely on the peripheral components of linguistic proficiency acquired through formal education, including explicit knowledge and access to lower frequency linguistic elements. The test design also allows for more controlled linguistic processing, tapping more strongly into declarative knowledge (Gaillard & Tremblay, Reference Gaillard and Tremblay2016). As lexical complexity is viewed as an indicator of the configuration and expansion of writers’ mental lexicon, it follows that learners with higher proficiency as measured by a cloze test will also have advantages in translation and revision activities during writing. Such predicted advantages will be most prominent for lexical encoding mechanisms and retrieval processes in consideration of a task’s communicative goals, promoting enhanced lexical complexity.
Of note, the results also showed that the EIT was linked to lexical complexity among L2 writers but not HL writers. A plausible explanation for this finding concerns differences in HL and L2 learners’ cumulative bilingual experiences and their measured linguistic knowledge. It is assumed that EIT performance requires automatic language processing for native speakers, but the same is not always true for L2 speakers. Particularly at lower levels of L2 experience, the cognitive operations underlying L2 learners’ EIT performance may at times be more effortful and involve more strategic control. This is not expected for HL learners, who are likely to have access to more automatized implicit knowledge of the language they acquired in childhood. Such differences in context and age of acquisition can account for why individual variability in EIT performance serves as a stronger discriminant of lexical complexity among L2 writers than HL writers.
Lastly, from a broader methodological standpoint, the findings highlight that writing research can benefit from investigating the relative contributions of different linguistic proficiency measures. Considering tests that involve only speech reception and production along with those that involve literacy skills, in line with Hulstijn’s (Reference Hulstijn2011, Reference Hulstijn2012) recommendation, stands as particularly informative to arrive at a finer-grained understanding of the roles of proficiency in writing processes and outcomes. As further discussed subsequently, one may argue that the need for such an approach is heightened in research considering HL learners, as the reliability and validity of using a written explicit language assessment—such as the DELE cloze test—as a single measure of HL proficiency has been questioned. Carreira and Potowski (Reference Carreira and Potowski2011), for instance, have called for the use of HL proficiency tests with an oral/aural component, particularly in research that compares L2 and HL learners. This study bolsters their argument and underscores its implications for HL and L2 writing research.
Influence of motivational value beliefs in HL and L2 writing
The third research question focused on whether task value beliefs would differentially explain variability in HL and L2 writers’ linguistic complexity. The results indicated comparable patterns for both learner groups, with intrinsic/interest value being a positive predictor and cognitive/linguistic value beliefs being a negative predictor of syntactic complexity. The finding concerning the supportive role of intrinsic/interest value to writing outcomes broadly echoes and expands previous evidence for the beneficial influence of intrinsic motivation in the area of L2 writing (e.g., Lin et al., Reference Lin, Cheng, Lin and Hsieh2015; Tsao et al., Reference Tsao, Tseng and Wang2017; Yeşilyurt, Reference Yeşilyurt2008). In line with the postulations of expectancy-value theory (e.g., Wigfield & Eccles, Reference Wigfield and Eccles2000), the results suggest that learners’ perceptions of value attributed to Spanish writing can influence their task behaviors in ways that meaningfully affect writing quality. It follows that among both HL and L2 students with greater interest or perceived enjoyment in Spanish writing, task performance is likely characterized by stronger attentional engagement, greater task interest, and more sustained involvement as they work toward meeting the communicative demands of tasks, leading to enhanced linguistic complexity.
The negative association between cognitive/linguistic value beliefs and syntactic complexity observed here departs from the results in Lee (Reference Lee2020), the only previous study to consider this motivational variable. Lee found a small but positive link between Korean EFL students’ cognitive/linguistic value belief and their writing performance in analytically scored descriptive and argumentative essays. Important methodological differences aside, one aspect that may have contributed to these seemingly divergent findings concerns students’ socioeducational context. Lee describes the Korean EFL context as “lacking a tradition of L1 writing pedagogy” (p. 1237) and mentions that L2 writing “is rarely taught in secondary schools” (p. 1244). Differences in learners’ age and context-specific writing experiences may have contributed to the observed disparities across Korea- and U.S.-based participants’ perceived cognitive/linguistic value of writing, which were considerably lower in Lee’s eighth-grade sample (M = 3.32, SD = .84) than in the present university-level sample (HL: M = 4.56, SD = .60; L2: M = 4.71, SD = .30).
A potential explanation for the negative role of cognitive/linguistic value beliefs found in this study is that participants who more strongly view writing as a tool to gain linguistic and cognitive benefits may also be more likely to adopt writing strategies that focus on lower level processes (e.g., accurate typing, word retrieval), leading to less fluent or efficient higher level processes (e.g., idea generation, development of text/argument structure). This may have had a negative indirect effect on how successfully writers were able to address the functional demands incumbent on persuasion, which are reliant of the effective use of grammatical structures to generate and elaborate arguments and their logical interconnections. The fact that persuasion places high demands on writers to deploy their syntactic resources (e.g., subordinating devices serve an important role in introducing reasons) may also help explain why the influence of both intrinsic/interest and cognitive/linguistic value beliefs was prominent for syntactic complexity but not lexical complexity in the current study. More research that considers both writing processes and outcomes is needed to elucidate the combined influences of these variables.
Last, given the paucity of previous data on cognitive/linguistic value beliefs, it is worth pointing out that results observed here for this variable cannot be explained based on students’ proficiency differences. Although it may be tempting to assume that learners who believe they can derive greater linguistic and cognitive benefits from practicing Spanish writing are also the ones with lower Spanish proficiency levels, follow-up analyses suggest that this assumption is not warranted. Specifically, correlations between cognitive/linguistic value beliefs and either EIT or cloze test z scores were nonsignificant (r = −.08, p = .270 and r = −.05, p = .461, respectively), suggesting that the behavioral outcomes linked to this motivational factor should not be interpreted as resulting from differences in linguistic ability.
Limitations and conclusion
Findings from this research should be interpreted in light of its limitations. While lexico-syntactic complexity was given a place of primacy here due to the study’s focus on cross-register variations, analysis of additional relevant linguistic and discoursal features could have provided a fuller picture of HL and L2 writers’ performance. Furthermore, only a narrow set of task-value beliefs were considered in the study given their potential relevance among HL and L2 samples. Subsequent research may wish to explore the contributions of other important motivational factors to students’ writing. Additionally, as this study involved participants from various universities, it was not feasible to collect detailed information about their individual institutional backgrounds, ongoing and previous coursework, or other pertinent aspects of their formal learning experiences. Incorporating this information in future research holds potential for understanding the effects of different instructional settings on both HL and L2 learners’ writing performance. Similarly, research with larger samples may also wish to examine HL and L2 writers’ performance based on proficiency levels to explore how proficiency interacts with other background factors to shape writing outcomes.
Despite these limitations, as the first study to compare cross-register variations among HL and L2 writers, the results provide new insights on how the linguistic dimensions of learners’ writing are differentially influenced by important task- and learner-related variables. The findings highlight the ability of both HL and L2 writers to adjust their linguistic complexity in response to shifting contextual demands. Notably, the results also prompt reconsideration of conventional narratives surrounding pervasive cross-register asymmetries in HL learners’ writing skills by demonstrating how these learners leverage their rich linguistic repertoire across different register configurations. Moreover, the study underscores the significance of adopting a multidimensional view of proficiency that considers aspects of both basic and higher language cognition to elucidate its contributions to HL and L2 writing performance more fully.
The results from this study also offer valuable pedagogical insights. As echoed by Qin and Uccelli (Reference Qin and Uccelli2020), teaching practices informed by the present findings challenge a monolithic view of linguistic complexity that portrays advanced writing as a mere aggregation of complex linguistic elements. Instead, educators should promote an understanding of writing as flexible construction of discourse serving distinct communicative purposes, introducing students to a variety of written registers while empowering them to navigate complexity dynamically based on contextual demands. From this perspective, the development of students’ critical communicative competence should ideally acknowledge and build on HL and L2 learners’ existing literacy experiences and sociopragmatic competence. Furthermore, assessments that account for students’ cross-register performance can provide a more precise and nuanced representation of both HL and L2 learners’ writing skills. Lastly, the results suggest that educators should strive to cultivate subjective interest and enjoyment toward Spanish writing among both HL and L2 students while avoiding a narrow view of writing as solely a tool for language learning. This entails going beyond portraying writing as a way to support linguistic development and instead highlighting its broader value and personal significance for individual students.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S027226312300058X.
Data availability statement
The experiment in this article earned the Open Materials badge for transparent practices. The materials are available at https://osf.io/wf3x8/.
Acknowledgments
This research was partly funded by a Provost’s Summer Research Award, a Provost’s Grant-in-Aid, and a College of Liberal Arts Research Award from Temple University. Additional support was received from Liberal Arts Undergraduate Research Awards. I would like to thank Felicia Trievel, Lytice Gordon, Gabriela Ingber, and Gillian Smyth for their help with various aspects of this project. I am also grateful to the SSLA editors and anonymous reviewers for their valuable feedback on an earlier version of this manuscript. All errors are my own.
