Revisiting the Cognition Hypothesis: Bridging a ... - Wiley Online Library

3 downloads 0 Views 254KB Size Report
Eun Young Kang. Teachers College, Columbia University, New. York, USA. Correspondence. ZhaoHong Han, Teachers College, Columbia. University, Box 66 ...
Received: 16 March 2017

Revised: 22 November 2017

Accepted: 2 March 2018

DOI: 10.1111/ijal.12209

ORIGINAL ARTICLE

Revisiting the Cognition Hypothesis: Bridging a gap between the conceptual and the empirical ZhaoHong Han

|

Eun Young Kang

Teachers College, Columbia University, New York, USA Correspondence ZhaoHong Han, Teachers College, Columbia University, Box 66, 525 W. 120 Street, New York, NY 10027, USA. Email: [email protected]

Task‐based language learning has commanded ample attention in the field of instructed second language acquisition (ISLA) over the last two decades. Much of the research has centered on the Cognition Hypothesis, particularly testing out its key tenet that there is a correlation between task complexity and the syntactic complexity of task‐doers' or L2 learners' speech production. Extant studies, however, have led to inconclusive findings, and methodological inadequacy has generally been deemed the source of divergence. In this article, we take issue with this line of reasoning, arguing that the fundamental problem is theoretical more than methodological. We suggest that thought complexity rather than cognitive complexity should be investigated as a core construct mediating the task and language relationship in L2 learners. KEYWORDS

language complexity, second language learners, task complexity, thought complexity

과업중심 언어학습은 지난 이 십년 동안 교실상황에서의 제 2언 어 학습 분야 (ISLA)에서 많은 관심을 받아왔다. 연구의 대부분 은 인지 가설을 중심으로 이뤄져왔고, 특별히 과제의 복잡성과 과제 수행자의 언어 복잡성의 관계에 관한 가설을 평가하는데 초점을 두어왔다. 기존의 연구들은, 그러나, 다양한 결과를 보 여주고 있고 이에 대해서는 일반적으로 연구방법의 부적합성이 주된 이유로 손꼽히고 있다. 본 연구에서는 이러한 결론에 문제 를 제시하고 근본적인 문제는 연구방법이 아닌 이론적인 문제 라는 것을 제시하고자한다. 본 저자는 인지의 복잡성이 아닌 생

Int J Appl Linguist. 2018;1–15.

wileyonlinelibrary.com/journal/ijal

© 2018 John Wiley & Sons Ltd

1

2

HAN AND KANG

각의 복잡성이 과제와 제2 언어 학습자의 언어와의 상관성을 변 화시키는데 주된 개념으로 조사되어져야한다고 본다. 키 워드

과제복잡성, 언어복잡성, 생각복잡성, 제2 학습자

1

|

I N T RO DU CT I O N

Over the last three decades, with the steady rise of task‐based language teaching (TBLT) (Ellis, 2003; Long, 1985, 2015; Long & Crookes, 1992; Van den Branden, Norris, & Bygate, 2009) as a theory and research‐driven pedagogical paradigm has come a heightened interest in the effects of task characteristics – especially task complexity – on L2 production and ultimately on L2 learning. The impetus for the now rich body of research stems in large part from Robinson's (1995, 2001) Cognition Hypothesis (CH), which, inter alia, stipulates a positive relationship between the cognitive demand of a task and the task‐doer's, in this case the L2 learner's, use of language. However, recent research syntheses and meta‐analyses (Jackson & Suethanapornkul, 2013; Sasayama, Malicka, & Norris, 2015) have cast doubts on the validity of the hypothesized linkage between task complexity and language complexity. Specifically, extant research has shown a lack of correlation between the syntactic complexity of L2 learners' output and the purported task complexity. Methodological factors – such as varied operationalizations, treatments, and measures, and the neglect to validate the independent variables (e.g., task complexity) – are considered underlying the deviant findings (Jackson & Suethanapornkul, 2013; Révész, 2014), feeding the perception that once the methodological weaknesses are overcome, findings from future research could be expected to fall in line with the CH. In this article, we point out that the issues may not entirely be methodological; rather, there are issues inherent in the hypothesis that have yet to be addressed, not the least concerning the alleged relationship between task complexity and language complexity. We argue that the relationship is convoluted in L2 learners, and, accordingly, a more nuanced approach must be sought when theorizing the task and language relationship, one that should prioritize thought complexity over cognitive complexity. The CH has evolved noticeably since its postulation (Robinson, 1995, 2001). Robinson (2011a), for example, expanded its scope to include five predictions, all revolving around the effects of task complexity on the task doer in terms of output (a), uptake and interaction (b), memory and retention (c), automaticity (d), and individual difference (e), with (a) being the original prediction and (b), (c), (d) and (e) additional spin‐offs. In this article, we seek to engage with the first and original prediction, namely that there is a correlation between task complexity and task doer's output complexity. Our intent is to provide, primarily, a conceptual pathway to resolving the empirical and theoretical variance, and, secondarily, a perspective – complementary to existing thinking (e.g., Révész, Michel, & Gilabert, 2016; Sasayama, 2016) – on how task complexity can be established rather than assumed in CH research. To those ends, in what follows we first briefly review the CH, with a view to tracing its epistemic source and exposing conceptual issues. After that, we posit thought complexity as a latent variable of the CH, discussing its linguistic correlates and how these may play out differentially in native speakers' versus L2 learners' language output. In the final section, we summarize our arguments and contemplate the implications for future CH research.

2

|

T H E CO G N I T I O N H Y P O T H E S I S

The CH (Robinson, 1995, 2001, 2005, 2009, 2011a, 2011b) was spurred by concerns over the syllabus design of TBLT, in particular, the grading and sequencing of pedagogic tasks, a long‐standing issue harkening back to the early

HAN AND KANG

3

days of the TBLT proposals (see, e.g., Candlin, 1984; Long, 1985). At its inception, Robinson drew on insights from first language acquisition research (Brown & Bellugi, 1964), L2 research (e.g., Long, 1985; Meisel, 1987), and functional linguistic theory (Givón, 1985). The main thrust of the hypothesis was, then, that there exists a correlation between task complexity, a task‐inherent attribute, and linguistic complexity, an attribute of the task‐doer's language use. Thus, if a task is complex, then the task doer's language will also be complex, or as Givón (1985: 1021) put it, “greater structural complexity tends to accompany greater functional complexity in syntax.” For instance, a simpler narrative task, requiring reference to the here‐and‐now, would correspondingly elicit simpler output from a task‐doer than would a more complex task – such as one requiring reference to the there‐and‐then in, which would elicit linguistically more complex output of language. Robinson (1995: 101) claimed: Tasks requiring accurate or precise use of language, and requiring the expression of multiple propositions drawn from memory, are more likely to “stretch” the interlanguage resources of second language users (Long, 1989) than are tasks not requiring them, and so will lead to greater communicative resource expansion. On this conception, tasks, by virtue of their complexity, can fashion learners or task‐doers' output in one way or another. It follows that for L2 instruction, task complexity can be manipulated to exert desired influences on the task‐doer's output (cf. Skehan, 1998), an implication most of the empirical work on task complexity to date has picked up on. What exactly is task complexity? Defining the construct has proven a daunting task. In the applied linguistics literature, there exists a myriad of definitions (see, e.g., Candlin, 1987; Long, 1985; Nunan, 1989; Skehan, 1996; Widdowson, 1990), which, for space limitation, will not be delved into here. Instead, we will confine our discussion to Robinson's conception of task complexity. According to Robinson (1995, 2001), task complexity derives from the information processing or cognitive demands of a task that emanate, in turn, from task design characteristics that are relatively fixed and invariant. The processing demands of tasks, these, in essence, the attentional, memory, and reasoning demands, comprise and vary along two dimensions: the resource‐directing and the resource‐depleting or dispersing. The resource‐directing dimension putatively consumes attentional resources, and the resource‐dispersing dimension moderates – at times on a temporary basis rather than permanently – the intensity of resource consumption. The resource‐directing dimension revolves around such variables as +/− here‐and‐now, +/− few elements, and +/− reasoning. The resource‐dispersing dimension, on the other hand, subsumes variables like +/− planning time, +/− prior knowledge, and +/− single task. Importantly, it is argued that this differentiation “[identifies] an important difference in the way these dimensions affect resource allocation during L2 task performance” (Robinson, 2001: 30), suggesting that task design features are a source of leverage for channeling learners' attention to different aspects of language use – accuracy, fluency, and complexity – during task performance. The list of variables associated with either the resource‐directing or the resource‐dispersing dimension has noticeably grown over the years, as seen in Robinson (2011a), summarized in Table 1. As shown in Table 1, the resource‐directing dimension of task complexity is connected to a task's cognitive or conceptual demands, and the resource‐dispersing dimension to a task's performative demands. These links, apart from showing the two dimensions of task complexity as distinct from each other, should, by Robinson's conjecture, trigger two other sets of associations between task complexity and learning affordances. The resource‐directing dimension of task complexity feeds developmental complexity, manifested behaviorally as changes to the accuracy and (syntactic and lexical) complexity of learners' or task‐doers' output, while the resource‐dispersing dimension regulates performative complexity, evident behaviorally in changes to the automaticity or fluency of learners' output. Still, the resource‐directing and resource‐dispersing dimensions can interact to modulate task performance and learning. Consider a task that is simple on the resource‐dispersing dimension but complex on the resource‐directing dimension. Such a task may allow “optimum resource allocation to satisfy the linguistic demands of the task”

4

HAN AND KANG

TABLE 1

Task complexity

Task complexity (Cognitive factors) (Classification criteria: cognitive demands) (Classification procedure: information‐theoretic analyses)

Subcategories Resource‐directing variables making cognitive/conceptual demands

Resource‐dispersing variables making performative/ procedural demands

+/− +/− +/− +/− +/− +/−

+/− +/− +/− +/− +/− +/−

here and now few elements spatial reasoning causal reasoning intentional reasoning perspective taking

planning time prior knowledge single task task structure few steps independency of steps

(Robinson, 2001: 31), and simultaneously afford the learner an opportunity to develop accuracy and complexity (cf. however, Skehan, 1998). Displayed also in Table 1 are the seeming dichotomous values of the sub‐variables of the resource‐directing and resource‐dispersing dimensions of task complexity; these, too, are likely to form continua (Robinson, 2011a). Undergirding Robinson's CH is a distinction he makes between task complexity and task difficulty. “It is important to distinguish between differentials in the processing demands of tasks, which are a consequence of task structure and design, and differentials in the resources learners bring to tasks which are attributable to a range of individual difference variables” (Robinson, 2001: 29). Put another way, while task complexity exclusively concerns task design features, task difficulty relates to the resources that individual learners bring to a task. Task complexity should, therefore, result in within‐learner variance, while task difficulty would find its expression in between‐learner variance. Thus, if the same learner were subject to two tasks of differential attentional demands, his/her performance would vary across these two tasks; conversely, two learners would react to the same task differently, since they might each bring a different set of resources to bear on their task performance. But even here, Robinson was quick to speculate that task complexity and task difficulty are likely to interact, forming stable relationships between increases in task complexity (the cognitive demands of tasks) and learner perceptions of difficulty, and this has been empirically proven (see, e.g., Ishikawa, 2011; Préfontaine & Kormos, 2015; Robinson, 2011a; Tavakoli, 2009). A central tenet of the CH is that “increasing the cognitive demands of tasks contributing to their relative complexity along resource‐directing dimensions should push learners to greater accuracy and complexity of L2 production in order to meet the consequently greater functional/communicative demands they place on the learner, while negatively affecting fluency, compared to simple task performance” (Robinson, 2011a: 18). Simply put, depending on the relative degree of task complexity, that is, how the resource‐directing and the resource‐depleting variables play out, often evaluated within the same task type, learner output differentials can be expected in terms of complexity, accuracy, and fluency, and so can different aspects of linguistic development, broadly conceived to entail structuring of new L2 knowledge and restructuring of existing knowledge representations (Robinson, 2001). Robinson has consistently underscored attention as the lynchpin between task complexity and learner language, stating: • Task demands can focus attention on specific concepts required for expression in the second language (L2) and prompt effort to grammaticize them in ways that the L2 formally encodes them, with consequences for improvements in accuracy of production. • Simple task demands can promote access to and automatization of currently emerged interlanguage means for meeting these demands, with consequences for improved fluency of production. • Task demands can also promote effort at reconceptualizing and rethinking about events, in ways that match the formal means for encoding conceptualization that L2s make available.

5

HAN AND KANG

• Complex tasks can prompt learners to attempt more ambitious, complex language to resolve the demands they make on communication success, thereby stretching interlanguage and promoting syntacticization, with consequences for improved complexity of production. (Robinson, 2011b: 2; emphasis original) These assertions further elucidate Robinson's reasoning about the link between task demands, task‐doer's output, and developmental consequences. For him, what mediates task complexity and output complexity boils down to “action control” – “Consequently, increasing complexity along various dimensions of tasks, such as increasing the amount of reasoning a task requires, promotes greater effort at controlling production and more vigilant monitoring of output” (Robinson, 2011b: 12), hence resulting in greater accuracy and greater complexity of learner output. This line of reasoning appears somewhat oversimplistic and reductionist to us. For one, it fails to recognize the task doer, the L2 learner, as the mediator of task complexity and learner language. The task doer is the agent of attention and action control, and for that reason, the alleged ‘conversion’ of task complexity into complex language may or may not happen. But more important, the CH ignores the role ‘thought’ plays in mediating the task‐to‐output process. Attention does not equal thought; attention is arguably a process but thought is a product. Thus, while the CH is concerned with the relationship between attention (regulated by cognitive complexity of a task) and language, we believe that the relationship between thought and language is more crucial to understanding the relationship between task and language, and it is far from straightforward or ipso facto in L2 learners, as would seem with native speakers. With L2 learners, for the most part, there is neither a one‐to‐one mapping between thought and language nor between attention and thought. Therefore, their eventual output may very well be neither a reliable reflection of attention nor thought. Thought may not go hand in hand with attention, either, even with native speakers. While attention may be task‐induced, the thought that results can transcend the scope of attention. All told, the alleged correlation between attention (induced by task complexity) and output (task‐induced production) may not be tenable, at least for L2 learners. Before we further delve into these arguments, a selective review of the empirical research into the task and language relationship is in order.

3 | E M P I R I C A L R E S E A RC H ON TA S K C O M P L E X I T Y A N D L E A RN E R LANGUAGE The earliest empirical attempt at understanding task complexity and its influence on learner output was Robinson (1995), which sought to assess whether differences in task complexity would affect learner language in measurable ways. Six hypotheses in line with the CH were formulated, among them that there‐and‐then narrative tasks will elicit greater propositional complexity of production than here‐and‐now narrative tasks, as reflected in higher numbers of multipropositional utterances (a); that there‐and‐then narrative tasks will elicit greater syntactic complexity of production than here‐and‐now tasks, as reflected in higher numbers of S‐nodes per T‐unit (b); and that there‐and‐then narratives will elicit more lexical content than Here‐and‐Now narratives, as reflected in higher ratios of lexical to grammatical words (c). Two monologic narrative tasks, one operationalizing the here‐and‐now (H/N) condition and the other the there‐and‐ then (T/T) condition, were administered to a group of L2 learners. The T/T condition, being non‐context‐supported, was considered more complex than the H/N condition, being context‐supported. The more complex task, evoking greater attentional and memory resources as demanded by the construction and expression of more complex meanings, was expected to yield greater propositional complexity, more lexical content, greater syntactic complexity, greater accuracy, and higher ratios of lexical to grammatical words, but lesser fluency. Participants were 12 intermediate‐level students from a variety of linguistic backgrounds, ages between 19 to 25 years. Randomly assigned to either of two groups, each

6

HAN AND KANG

participant produced three narratives according to three different wordless cartoon strips, resulting in an equal distribution of the two conditions, H/N and T/T, across the two groups and in a total of 36 narratives. In the H/N condition, participants looked at the picture strip while telling the story; in the T/T condition, they turned away the picture before beginning their narration. Three sets of measures – grammatically defined, intonationally and pausally defined, and lexcically and pausally defined – were deployed of the narrative accuracy (i.e., targetlike use of articles), complexity (multi‐propositional utterances, S‐nodes per T‐unit, and ratio of lexical to grammatical words), and fluency (the number of pauses and words per utterance). Robinson (1995: 111) reasoned: [T]he functional complexity of maintaining displaced reference would be matched by the learner's effort to produce greater structural complexity. This could be quantified as a greater proportion of S‐nodes to T‐ units, indicating a greater degree of subordination and embedding. It would also lead the learners to attempt to maintain greater accuracy. Statistical results pertinent to the three hypotheses on task complexity and linguistic complexity showed that the more complex T/T condition did not elicit greater propositional complexity – it was not clear how in the study multipropositional utterances were coded – nor greater syntactic complexity. However, there was a significant difference between the more complex T/T condition and the less complex H/N condition vis‐à‐vis use of lexical words. The more complex condition elicited greater use of lexical content; the lexical‐to‐grammatical word ratio was higher in the complex than in the simple condition. Another significant finding concerns the targetlike use of articles; the more complex condition promoted greater accuracy. The fluency measure, on the other hand, only yielded “a trend approaching significance.” Taken together, the findings from the study gave “some reason to believe that the There‐and‐Then/Here‐and‐Now distinction (as operationalized here) represents a valid difference in the difficulty level of oral narrative production tasks” (Robinson, 1995: 121). Importantly, Robinson acknowledged that the hypothesis that the simpler task would induce a pragmatic mode of production in the task doer and the complex task a syntactic mode was not supported by the data. In a subsequent study, Robinson (2001) continued to investigate the effects of task complexity on learner language, employing, this time around, one‐way interactive tasks rather than monologic tasks as in the earlier study. Robinson suspected that the predictions of the CH could be modulated by the interactive nature of a task. Forty‐four Japanese university undergraduates participated in the study, which controlled for sequence (simple‐complex or complex‐simple) and role (speaker or hearer) as between‐subject variables and task complexity (simple or complex) as a within‐subject variable. A simple version and a complex version of a map reading task were created, with the task differentials being +/− few elements and +/− prior knowledge, operationalized as giving directions about a known location in a small area versus an unknown location in a larger area. Learner language was assessed for accuracy via error‐free C‐units, for fluency via the number of words per C‐unit, and for complexity via clauses per C‐unit for syntactic complexity and type‐token ratio for lexical complexity or variety. The results confirmed, albeit only to a moderate extent, a priori hypotheses, in that task complexity significantly affected the speaker's production, engendering greater lexical variety and lesser fluency. The results on accuracy, on the other hand, were not significant. Equally noteworthy, there was a lack of correlation between task complexity and syntactic complexity. Nevertheless, the findings as a whole were interpreted as largely in keeping with the predictions of the CH. The two studies by Robinson, both focusing on determining the relationship between task complexity and learner language, produced incongruent as well as congruent findings. First off, they both reported a positive correlation between task complexity and lexical load, but a lack thereof between task complexity and syntactic complexity. Second, both studies found a positive relationship – though not to the same extent – between task complexity and fluency. The two studies, however, were at straight odds over accuracy: the 1995 study illustrated a positive relationship between task complexity and accuracy, but such was not observed in the 2001 study. While an explanation was attempted for the positive finding, little was given for the lack thereof. These two early studies portended well for the ensuing empirical research on the CH writ large, where mixed findings were a staple. As a matter of fact, in spite of two decades of research resulting in a sizable number of studies

HAN AND KANG

7

(129 studies over the last two and a half decades, by Sasayama, Malicka, & Norris's 2015 count), the general understanding still remains blurred, particularly concerning what task complexity does to learner language, as a recent meta‐analysis by Jackson and Suethanapornkul has revealed. Jackson and Suethanapornkul (2013: 356) attempted to synthesize and meta‐analyze quantitative, primary studies (published between 1995 and 2009) seeking to validate the CH, especially its overarching prediction that “increasing task demands benefits L2 learners' accuracy and complexity, but hinders their fluency.” Their analysis showed, inter alia, the following: that the studies have limited comparability, due to their deployment of an “assortment of treatments and measures” (Jackson & Suethanapornkul, 2013: 330) (a); that the studies collectively yield small positive effects for accuracy, and small negative effects for fluency (b); and in contrast, that there is much stronger evidence disconfirming the relation between task complexity and syntactic complexity (c). The researchers, therefore, claimed to have found only partial support for the CH. Two observations among many from this synthesis and meta‐analysis are noteworthy for our purposes. First is the striking dissimilarity in operationalizing the resource‐directing dimension of task complexity across the empirical studies. Although studies all seemed driven by an intense interest in investigating the +/− here‐and‐now, +/− few elements, and/or +/− reasoning conditions, there were substantial variations in how these resource‐directing dimensions were operationalized; “current understanding of these variables is dynamic and evolving” (Jackson & Suethanapornkul, 2013: 353). It is conceivable that the across‐the‐board inconsistencies may have compromised the results of the meta‐analysis. The authors were certainly wise to pose the question: How do such manipulations impact the quality versus quantity of learner production? Another observation is that in spite of outcome measures aplenty – as many as 84 across the studies and 23 in a single study– applied to evaluate the complexity, accuracy, and fluency of learner production, there was very little consistency in their deployment across the board. In sum, the empirical support for the CH is as yet selective, lending credence to its predictions on accuracy and fluency but casting doubt on syntactic complexity as a correlate of task complexity. It is easy, and natural, to fault this on methodological procedures and to call for improvements. Suggestions tendered by Jackson and Suethanapornkul (2013) were many: broadening the scope of research on the resource‐directing dimension of task complexity, adding, for example, +/− perspective taking as a variable (a); investigating multiple variables, and their interaction, within and across the resource‐directing and resource‐dispersing dimensions (b); exploring the validity and reliability of the many measures of complexity, accuracy, and fluency of learner production (see, e.g., Lu, 2010) (c); continuing to differentiate general and specific measures (d); exploring developmentally more sensitive measures of syntactic complexity (e); balancing and going beyond monologic and dialogic task types (f); and examining written tasks as well as oral tasks (g). Specific to improving the observed relation between task complexity and syntactic complexity, Jackson and Suethanapornkul (2013: 357) recommended a regimen of “a more balanced selection of complexity measures, in order to arrive at interpretations based on increases within specific subconstructs of complexity at specific proficiency levels.” Current research on the CH, therefore, seems to hold out the hope that additional empirical studies coupled with an improved methodology will contribute to “a fuller picture of task‐based learning framed by the Cognition Hypothesis” (Jackson & Suethanapornkul, 2013: 358). This, however, is not the tack we take in this article. Instead, as noted earlier, we believe that there are lingering conceptual concerns about the CH that appear to have largely evaded researchers in investigating the task and language relationship.

4 | I S S U E S A T T H E CO RE OF T H E T H E O R E T I C A L A N D E M P I RI C A L VARIANCE As noted earlier, the CH presupposes a chain and linear relationship between task, cognition, and language. Task complexity is deemed isomorphic with cognitive complexity (i.e., attentional demands), and, in turn, correlating with linguistic complexity. Such conception may sound plausible for native speakers, but not for L2 learners, for the simple

8

HAN AND KANG

fact that L2 proficiency may affect the mapping between task complexity and learner language (see Sasayama, 2016). It is not hard to imagine what could happen in the task and language process for L2 learners. A complex task, by virtue of its conceptual and communicative demands, may prompt greater learner attention to speech content or thought but its ultimate expression is foiled owing to a lack of linguistic resources to encode it. Alternatively, the complex task does not induce cognitive complexity (i.e., garnering due attention) in the L2 learner for any number of reasons, not the least the learner's lack of experience with it, in which case, there is little attention to the speech content resulting in little thought and subsequently simple(r) language. Put another way, the relationship between task complexity and linguistic complexity is typically non‐linear, if not short circuited, in L2 learners. The ability to produce syntactically complex output on a complex task is subject to at least three variables, not one: attention to the demands of a task (a); but additionally, actual thought induced (b); and language ability (c). Over the last couple of years, work has surfaced that sought to validate – rather than assuming as had been all along – task complexity and cognitive complexity (in line with (a)), with “cognitive task complexity” either as a dependent or independent variable, as in Sasayama (2016). In contrast, work on (b) and (c) either to gauge actual thought resulting from cognitive complexity or to understand how available linguistic resources constrain articulation of thought has yet to emerge. This, to us, represents a major gap in current CH research, hampering progress in understanding the task and language relationship in L2 learners. Interestingly, few people have realized that attention to thought was actually somewhat present in the earliest research on the CH. To his credit, in Robinson (1995), there was a conscious attempt to measure “propositional complexity,” a construct akin to thought complexity, operationalized in the study as the number of major arguments and predications in utterances. But as it turned out, data analyses of propositional complexity and syntactic complexity did not find any significant differences for the simple versus complex task conditions. This might have been why propositional complexity fell by the wayside in subsequent CH research (see, e.g., Robinson, 2001). In hindsight, Robinson (1995) made what appears to us a critical speculation, namely that the prediction on a positive link between task complexity and syntactic complexity might be appropriate for adult native speakers, the focus of Givón (1985, 2009), not for adult second language learners. Proficiency was, therefore, implicated as a variable leading to the null finding: “perhaps students' proficiency in English was not sufficient to allow them to demonstrate the expected differences” (Robinson, 1995: 125). However, these emerging attempts and insights did not catch on in ensuing studies. Instead, the preponderance of CH researchers has since opted for a simpler albeit conceptually weak approach, by superficially investigating the relationship between task characteristics (e.g., monologic versus dialogic, open versus closed) and linguistic (i.e., syntactic and lexical) complexity.

5

|

T HO U GH T C O M P L E X I T Y

As is clear from our discussion so far, we believe that where the CH has failed to show validity has to do with its lack of theorizing on the cognitive consequences of task demands for L2 learners, who, unlike native speakers, have a weaker ability to map meaning onto form. Including thought complexity in its calibration may, therefore, create a new opening in CH research. We contend that while a cognitively more complex task may call for greater attentional resources from the task doer than a cognitively simpler task, what differentiates the complex and simple tasks at the level of cognition should culminate in differential thought complexity. Only with complex thought is the language likely to be complex. In the terms of Levelt's (1989) speech production model, thought resides in the Conceptualizer and manifests via the Articulator. Importantly, for native speakers, the intermediary Formulator may occur ipso facto, seamlessly weaving meaning (thought) and form (language) together, but for L2 learners such ‘automatic’ operation is much less likely. Proficiency, among other possible factors, may hamper meaning‐form mapping in the Formulator, resulting in the Articulator producing language that under‐represents thought (see, e.g., Préfontaine & Kormos, 2016).

HAN AND KANG

9

We further entertain the notion that thought complexity is not monolithic. It has at least two dimensions: breadth and depth. Linguistically, these aspects should manifest differently as well. We base our argument on an important finding from extant CH research, including Robinson's early studies. These studies, setting out to measure syntactic and lexical complexity vis‐à‐vis task complexity, have yielded an all but categorical finding of an asymmetry on lexical versus syntactic complexity – that tasks that are more complex tend to correlate with greater lexical complexity but not greater syntactic complexity – of learner output. This gives some validity to using “lexical load” (Robinson, 1995) as a measure of cognitive complexity, what we would call thought complexity , to which we now turn.

6 | L E X I C A L A N D F U N C T I O N A L W O R D S A S P O S S I B LE I N DE X E S O F THOUGHT COMPLEXITY Lexical load, be it measured in terms of the ratio between content and grammatical words or the ratio between lexical words and the total number of words produced, known as lexical density (LD), addresses the amount of content produced (see, e.g., Michel, Kuiken, & Vedder, 2007). The higher the load or density, the more content expressed. Granted, we observe, however, that even if a further analysis of the lexical profile in terms of the frequency band of the content words used might throw some light on whether the meanings expressed are concrete or abstract, such measures are, after all, about content words only. Function words, on the other hand, have largely been sidelined in extant CH research (see Table 3 of Jackson & Suethanapornkul, 2013), due to the pervasive conception that function words are mostly meaningless, if not flat‐out semantically empty (VanPatten, 1996). This may have been a grave mistake (Han, 2008, 2011). Work by the psychologist James Pennebaker at the University of Texas at Austin is particularly enlightening when it comes to the role of function words in communication. Countering the general bias against function words as a vehicle of meaning, Pennebaker, amply and convincingly, demonstrates the semantic and pragmatic value of function words. Content and function words serve different purposes as vehicles of meaning: in any given sentence, content words – known also as lexical or open‐class words – such as nouns and verbs provide basic content and meaning, whereas function words – known also as style or closed‐class words‐ such as pronouns, articles, prepositions, and conjunctions – serve support functions, connecting, shaping, and organizing content words. Pennebaker (2013) noted: • The content of speech can be distinguished from the style of speech. • Style or function words can say more about a person than the more meaningful ones ‐ style words can reveal aspects of people's personality, social connections, psychological states, and cultural backgrounds. • Each style word may indicate the speaker's mental model as to what is said. • Style words, meaningless and useless on their own and often misunderstood, convey deep, implicit, and pragmatic meanings. By way of illustration, consider the following pair of sentences: 1a. I can't believe that he gave her the ring. 1b. I can't believe that he gave her a ring. The two sentences are almost completely alike except in their use of articles, the vs. a. Yet, what seems to be a trivial difference engenders a subtle but significant pragmatic difference. Sentence 1a conveys the meaning that the speaker and the hearer have shared knowledge of the ring, while 1b suggests absence of such shared knowledge, and conveys a subtle evolving relationship between “him” and “her.” The classic line in Abraham Lincoln's Gettysburg address demonstrates the value of function words to the fullest:

10

HAN AND KANG

2… that this government of the people, by the people, and for the people, shall not perish from the earth. Pennebaker (2013: 34, 37) wrote: All function words work similarly in that they are tied to the personal relationship between the speaker and the listener. … The ability to use them, then, is a marker of basic social skills. On the other hand, talking about nouns and verbs demands the ability to understand culturally shared categories and definitions. Nouns and regular verbs generally translate across languages fairly smoothly. It is the function words that can cause the biggest problems. These statements underscore that content words convey basic, universal concepts, but function words convey language‐ and culture‐specific meanings. Consider the cross‐linguistic example below: 3a Spanish: (Yo) estoy triste. 3b English: I'm sad. In Spanish, the subject pronoun is often dropped, but in English it is canonically compulsory. Pennebaker observed that the dropping of subject pronoun tends to occur in languages from tightly knit collectivist cultures, whereas languages from individualist societies retain it. Given the insights both from extant CH research finding lexical load a quite consistent correlate of task complexity and from psychology research into function words, it would seem reasonable to tie lexical load to the breadth of thought and functional load to the depth of thought. Take Sample A, a comment by a native speaker of English on an op‐ed article in the New York Times, and see how the breadth and depth of thought play out respectively in terms of lexical density (i.e., ratio of lexical to total number of words) and functional density (i.e., ratio of functional to total number of words).

Sample A Excellent, thought‐provoking points by David, and some very insightful comments from all sides. But where is the discussion about the role of special interest money in our politics and by extension in the views of so many of our citizens on all sides of the spectrum who remain woefully uninformed and unsophisticated? I worry less about an informed, articulate and thoughtful person with whom I disagree than I do about the lemmings on either side who blindly accept sound bites and extremist revisionism as gospel. When money becomes the driver of what passes for national discourse, no wonder we have lost faith and have no direction. This sample comprises four terminal (T) units, as follows: T1. Excellent, thought‐provoking points by David, and some very insightful comments from all sides. T2. But where is the discussion about the role of special interest money in our politics and by extension in the views of so many of our citizens on all sides of the spectrum who remain woefully uninformed and unsophisticated? T3. I worry less about an informed, articulate and thoughtful person with whom I disagree than I do about the lemmings on either side who blindly accept sound bites and extremist revisionism as gospel. T4. When money becomes the driver of what passes for national discourse, no wonder we have lost faith and have no direction. The four T‐units are of unequal length. The second and third T‐units stand out as stretching longer than both the first and the fourth. Table 2 provides a breakdown by the number of words of each T‐unit as well as its lexical and functional densities.

11

HAN AND KANG

TABLE 2

Sample A by T‐unit, word total, lexical density, and functional density

T‐unit

Word total

Lexical density

Functional density

T1

14

0.71

0.29

T2

39

0.41

0.59

T3

33

0.52

0.48

T4

21

0.48

0.52

107

2.12

1.88

0.53

0.47

Total Mean

26.5

If we are right in tying lexical density to the breadth of thought, the following order obtains – T1 > T3 > T4 > T2, but a different one for the depth of thought – T2 > T4 > T3 > T1, counting functional density. By this analysis, longer T‐units do not necessarily encode more complex thoughts, nor monolithically. Take T2, the longest T‐unit, as an example, “But where is the discussion about the role of special interest money in our politics and by extension in the views of so many of our citizens on all sides of the spectrum who remain woefully uninformed and unsophisticated?” This T‐unit does not surpass T1 (the shortest by word count) in the breadth of thought, though it does by a substantial margin in depth. In other words, T1 has greater breadth, but T2 has greater depth. Subject to a more systematic investigation of lexical density (LD) and functional density (LD) as indexes of thought complexity, one seeming advantage of having a differentiated indexation than a monolithic one is that it may better capture the influence of task demands, which, likewise, are not monolithic, as the CH has captured in positing two dimensions of task demands, the resource‐directing and the resource‐dispersing. Still, finding out what provides a reasonable index of thought complexity is one thing, but applying it to L2 learners can be quite another, a crucial takeaway from examining extant CH research.

7

|

B RI D G I N G A C H A S M I N L 2 L E A RN E R S

Proficiency in the L2 can get in the way of thought expression. Higher proficiency portends a higher correspondence between thought and language (cf. Skehan, 2015). Consider Sample B, an intermediate L2 learner's output on a narrative task (Tarone & Swierzbin, 2009). Sample B One day in the supermarket at 12:00 o'clock, two women and a little child, to talk about of the shopping in this moment. They are friends and good persons. But they lost to attention in the child, so, the child is taking a bottle milk and she's introducing in the bag of the older woman. Probably the older woman will be arrested because she don't know that in your bag is a bottle milk and she will has to pay for her. This sample breaks down into four T‐units, as follows:

T1. One day in the supermarket at 12:00 o'clock, two women and a little child, to talk about of the shopping in this moment. T2. They are friends and good persons. T3. But they lost to attention in the child, so, the child is taking a bottle milk and she's introducing in the bag of the older woman.

12

HAN AND KANG

TABLE 3

Sample B by word total, lexical density, and functional density

T‐unit

Word total

Lexical density

Functional density

T1

23

.52

.48

T2

6

.83

.17

T3

27

.56

.44

T4

28

.71

.29

Total

84

2.62

1.38

Mean

21

.66

.34

T4. Probably the older woman will be arrested because she don't know that in your bag is a bottle milk and she will has to pay for her. Using lexical density to index the breadth of thought and functional density for the depth, Table 3 gives the results. As shown, the four T‐units are of differential thought complexity, not necessarily correlated with their respective lengths. In terms of the breadth of thought, the following order obtains: T2 > T4 > T3 > T1; in terms of the depth of thought, the order changes toT1 > T3 > T4 > T2. A comparison of T3 and T4 is particularly revealing: These twoT‐units are similar in length, and yet T4 shows greater breadth of thought, T3 greater depth. This is seen in T4 (breadth) invoking multiple characters (“they,” “the child,” “the older woman”), while T3 (depth) is mostly about one character (“the older woman”). Sample C came from a different L2 learner performing the same narrative task. Sample C Once upon a time, there is a old lady. She went to market because she need to buy some things. In the market, she talked with another lady who has a little girl. The little girl take it a bottle and she introduced in mom's bag. Her mom doesn't didn't know the fact. Obviously, wasn't pay it. I think that the little girl's mom tell her “Don't do that.” Sample C contains seven T‐units, as follows: T1. Once upon a time, there is a old lady. T2. She went to market because she need to buy some things. T3. In the market, she talked with another lady who has a little girl. T4. The little girl take it a bottle and she introduced in mom's bag. T5. Her mom doesn't didn't know the fact. T6. Obviously, wasn't pay it. T7. I think that the little girl's mom tell her “Don't do that.” As with Sample A and B, Sample C comes with differentials in lexical and functional densities, or differences in the breadth and depth of thought, as summarized in Table 4. A comparison of Sample B and C, as in Table 5, is instructive. As indicated in Table 5, Sample B is, overall, longer than Sample C. Sample B also appears to have greater breadth of thought (0.66 vs. 0.43), Sample C greater depth (0.57 vs.0.34). Sample B feels all over the place, but Sample C sticks more closely to the task at hand, reads tighter, with more clarity – what Robinson (1995) sought in terms of precision when envisioning a positive relation between task complexity and linguistic complexity. Taken together, Sample A, B, and C illustrate the potential of using lexical density and functional density to index the breadth and depth of thought complexity. They combine to reveal that the extent to which these indexes are

13

HAN AND KANG

TABLE 4

Sample C by word total, lexical density, and functional density

T‐unit

Word total

Lexical density

Functional density

T1

9

0.44

0.56

T2

11

0.45

0.55

T3

13

0.46

0.54

T4

13

0.54

0.46

T5

9

0.33

0.67

T6

5

0.40

0.60

T7

13

0.38

0.62

Total

73

3

4

Mean

10.42

0.43

0.57

TABLE 5

Sample B and Sample C compared Word total

T‐units

Mean length

Mean lexical density

Mean functional density

Sample B

84

4

21

0.66

0.34

Sample C

73

7

10.4

0.43

0.57

effective can be influenced by proficiency in the language in question – more reliable for native speaker output than for non‐native speakers'. Furthermore, Sample B and C converge on showing that for L2 learners, the extent to which these indexes work is predicated on the proficiency level of the learner. It is probable that higher proficiency predicts greater validity of the indexes. Conversely, limited proficiency may hinder thought expression, perhaps more in depth than in breadth, because of the greater and the ubiquitous challenge learners experience in acquiring L2 function words (Han, 2011). Despite that lexical and functional indexes of thought complexity are likely to be moderated by learners' level of proficiency, hence compromising their reliability, they can be used in two ways in the CH research: as a design index – using them with native speakers to help establish task complexity before investigating its impact on L2 learners and L2 output – something akin to independent validation of task complexity (a); and as a developmental index ‐ using them to gauge L2 development, in particular the learner's developing ability to map meaning onto form, or thought onto language (b). In closing, researching thought as a product of cognition – what we advocate ‐ and attentional allocation as a process ‐ what extant CH research has focused on, likely provides complementary perspectives on the relationship between task demands and their cognitive consequences, both necessitating a two‐step process in a full‐blown validation of the CH: piloting before implementing.

ENDNOTES 1

The model comprises three levels: Conceptualizer (i.e., where thoughts or content of communication is generated), Formulator (i.e., where thoughts are coded linguistically), and Articulated (i.e., the linguistically coded thoughts are externalized).

2

Three text samples are used in this article solely for the purpose of illustration of our constructs of lexical and functional density. They neither come from, nor serve to exemplify, a full‐blown empirical study.

3

Incidentally, we ran the three text samples on the Lexile® framework (MetaMetrics, 2015), which measures text complexity, following the lead of Douglas and Miller (2016), and the results yielded the rank ordering of Sample A > Sample B > Sample C, corresponding to our measure of breadth (i.e., lexical density) of thought complexity, but not of depth (i.e., functional density). Future research on thought or content complexity should look more closely at how depth can be reliably captured, analytically.

14

HAN AND KANG

ORCID ZhaoHong Han

http://orcid.org/0000-0001-9139-441X

RE FE R ENC ES Brown, R., & Bellugi, U. (1964). Three processes in the child's acquisition of syntax. Harvard Educational Review, 34, 133–151. Candlin, C. (1984). Syllabus design as a critical process. In C. Brumfit (Ed.), General English syllabus design (pp. 29–46). Oxford: Pergamon. Candlin, C. (1987). Towards task‐based language learning. In C. Candlin, & D. Murphy (Eds.), Language learning tasks (pp. 5–22). London: Prentice Hall. Douglas, Y., & Miller, S. (2016). Syntactic and lexical complexity of reading correlates with complexity of writing in adults. International Journal of Business Administration, 7(4), 1–10. Ellis, R. (2003). Task‐based language learning and teaching. Oxford: Oxford University Press. Givón, T. (1985). Function, structure, and language acquisition. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition: Vol I (pp. 1008–1025). Hillsdale, NJ: Lawrence Erlbaum. Givón, T. (2009). The genesis of syntactic complexity. Amsterdam: John Benjamins. Han, Z.‐H. (2011). Fossilization – a classic concern of SLA research. In S. Gass, & A. Mackey (Eds.), The handbook of second language acquisition (pp. 476–490). New York: Routledge. Han, Z.‐H. (2008). On the role of meaning in focus on form. In Z.‐H. Han (Ed.), Understanding second language process (pp. 45– 79). Clevedon: Multilingual Matters. Ishikawa, T. (2011). Examining the influence of intentional reasoning demands on learner perceptions of task difficulty and L2 monologic speech. In P. Robinson (Ed.), Second language task complexity: Researching the Cognition Hypothesis of language learning and performance (pp. 307–330). Amsterdam: John Benjamins. Jackson, D. O., & Suethanapornkul, S. (2013). The cognition hypothesis: A synthesis and meta‐analysis of research on second language task complexity. Language Learning, 63, 330–367. Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press. Long, M. (1985). A role for instruction in second language acquisition: Task‐based language teaching. In K. Hyltenstam, & M. Pienemann (Eds.), Modeling and assessing second language development (pp. 77–99). Clevedon: Multilingual Matters. Long, M. (1989). Task, group, and task‐group interactions. University of Hawai'i Working Papers in ESL, 8, 1–25. Long, M. (2015). Second language acquisition and task‐based language teaching. New York: Wiley‐Blackwell. Long, M., & Crookes, G. (1992). Three approaches to task‐based syllabus design. TESOL Quarterly, 26, 27–56. Lu, X. (2010). Automatic analysis of syntactic complexity in second language writing. International Journal of Corpus Linguistics, 15(4), 474–496. Meisel, J. M. (1987). Reference to past events and actions in the development of natural second language acquisition. In C. Pfaff (Ed.), First and second language acquisition processes (pp. 206–224). Rowley, MA: Newbury House. MetaMetrics. (2015). Lexile infographic. Retrieved on 16 April 2018 from https://www.lexile.com/about‐lexile/lexile‐overview/lexile‐infographic/ Michel, M. C., Kuiken, F., & Vedder, I. (2007). The influence of complexity in monologic versus dialogic tasks in Dutch L2. International Review of Applied Linguistics in Language Teaching, 45(3), 241–259. Nunan, D. (1989). Designing tasks for the communicative classroom. Cambridge: Cambridge University Press. Pennebaker, J. (2013). The secret life of pronouns: What your words say about us. New York: Bloomsburry Press. Préfontaine, Y., & Kormos, J. (2015). The relationship between task difficulty and second language fluency in French: A mixed methods approach. The Modern Language Journal, 99, 96–112. Révész, A. (2014). Towards a fuller assessment of cognitive models of task‐based earning: Investigating task‐generated cognitive demands and processes. Applied Linguistics, 35, 87–92. Révész, A., Michel, M., & Gilabert, R. (2016). Measuring cognitive task demands using dual‐task methodology, subjective self‐ ratings, and expert judgments: A validation study. Studies in Second Language Acquisition, 38(4), 703–737. Robinson, P. (1995). Task complexity and second language narrative discourse. Language Learning, 45, 283–331. Robinson, P. (2001). Task complexity, task difficulty, and task production: Exploring interactions in a componential framework. Applied Linguistics, 22, 27–57. Robinson, P. (2005). Cognitive complexity and task sequencing: Studies in a componential framework for second language task design. IRAL, 43, 1–32.

HAN AND KANG

15

Robinson, P. (2009). Syllabus design. In M. H. Long, & C. Doughty (Eds.), Handbook of second language teaching (pp. 294–310). Oxford: Blackwell. Robinson, P. (2011a). Second language task complexity, the Cognition Hypothesis, language learning, and performance. In P. Robinson (Ed.), Second language task complexity: Researching the Cognition Hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Robinson, P. (2011b). Task‐based language learning: A review of issues. Language Learning, 61, 1–36. Sasayama, S. (2016). Is a ‘complex’ task really complex? Validating the assumption of cognitive task complexity. The Modern Language Journal, 100, 231–254. Sasayama, S., Malicka, A., & Norris, J. (2015, September). Primary challenges in cognitive task complexity research: Results of a comprehensive research synthesis. In Paper presented at the 6th Biennial International Conference on Task‐Based Language Teaching (TBLT). Leuven: Belgium. Skehan, P. (1996). A framework for implementation of task‐based instruction. Applied Linguistics, 17, 38–62. Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University Press. Skehan, P. (2015). Limited attention capacity and cognition: Two hypotheses regarding second language performance on tasks. In M. Bygate (Ed.), Domains and directions in the development of TBLT: A decade of plenaries from the international conference (pp. 123–156). Amsterdam: John Benjamins. Tarone, E., & Swierzbin, B. (2009). Exploring learner language. Oxford: Oxford University Press. Tavakoli, P. (2009). Investigation task difficulty: Learners' and teachers' perceptions. International Journal of Applied Linguistics, 19, 1–25. Van den Branden, K., Norris, J., & Bygate, M. (Eds.) (2009). Task‐based language teaching: A reader. Amsterdam: John Benjamins. VanPatten, B. (1996). Input processing and grammar instruction: Theory and research. Norwood, NJ: Ablex. Widdowson, H. (1990). Aspects of language teaching. Oxford: Oxford University Press.

How to cite this article: Han ZH, Kang EY. Revisiting the Cognition Hypothesis: Bridging a gap between the conceptual and the empirical. Int J Appl Linguist. 2018;1–15. https://doi.org/10.1111/ijal.12209