5 Psychological Perspectives

The psychology of second language (L2) task performance refers to

factors relating to ‘the mental experiences, processes, thoughts, feel-
ings, motives, and behaviours of individuals involved in language
learning’ (Mercer, Ryan and Williams, 2012, p. 2). The psychological
dimension of task-based language teaching (TBLT) thus defined
includes the learner characteristics that are, in Snow’s (1991) terms,
‘propaedeutic’ (‘required as preparation for a learning condition’)
(p. 205) to a learning goal, including affective (feelings and emotions),
conative (motivation) and cognitive (reasoning and memory) vari-
ables. In this chapter, we discuss the role of these variables in affecting
task performance and the effects of task-based instruction by elabor-
ating the theoretical underpinnings for the role of psychological
factors and synthesizing the research on these factors.
The factors are broadly divided into cognitive and affective factors,
following Robinson (2011). In line with the current mainstream second
language acquisition (SLA) literature (Dörnyei, 2005; Mercer et al.,
2012; Ellis and Shintani, 2014) these variables are collectively referred
to as individual difference variables. Special attention is given to lan-
guage aptitude and working memory in the cognitive domain and
motivation and anxiety in the affective domain, due to their importance
for TBLT, the relatively clear definitions and operationalizations of the
constructs, and the availability of a body of relevant empirical research.
The synthetic review of each of these variables starts with an overview
of the construct in L2 research in general, including, but not limited to,
the conceptualization, operationalization and methods of the research,
followed by a more specific discussion of the TBLT research.

Theoretical Issues
The theoretical basis of the role of individual differences in language
learning can be found in Robinson’s Cognition Hypothesis (CH)

129

[Link] Published online by Cambridge University Press

 130 Theoretical Perspectives

(2001, 2011; see Chapter 2). Robinson’s theory posits a three-

component framework: task complexity, task conditions and task
difficulty, which concern the conceptual/cognitive, interactional and
perceptual demands of tasks, respectively. Among the three dimen-
sions, task difficulty relates to individual difference variables, includ-
ing affective (e.g. motivation and anxiety) and cognitive (e.g. language
aptitude and working memory) variables, which account for inter-
learner variation in task performance. Individual differences constitute
a key component of Robinson’s triadic framework that interacts with
the other two components in affecting learners’ task performance.
The CH makes the following predictions about how individual
difference variables interact with the other two groups of variables.
First, individual differences in the affective domain are implicated
when tasks are performed under different conditions leading to emo-
tions and interpersonal relationships coming into play. For example,
monologic tasks (e.g. narratives) that require public reporting may
lead to more anxiety than dialogic tasks performed in pairs or small
groups; tasks that require equal contribution from all participants
are likely to be more motivating than those that only require some
participants to contribute. Second, individual differences in cognitive
abilities are related to performance along different dimensions of task
complexity. For example, reasoning ability is important for successful
performance in tasks that are complex along the resource-directing
dimension which pose greater processing demands; attention control is
predictive of performance in tasks manipulated along the resource-
dispersing dimension. Third, the role of individual differences is
more evident in complex tasks that are more demanding of cognitive
abilities than simple tasks that require less mental effort and fewer
cognitive resources.
It can be seen that the CH affords a theoretical basis for the role of
individual difference factors in task performance involving learners’
existing knowledge and skills, but it does not spell out how these
variables impact L2 development or learning (gains in new knowledge
and skills). However, as we will see, empirical studies have investi-
gated both performance (e.g. Ahmadian, 2012) and learning (e.g. Li,
2013a, 2013b). Another caveat is that the matching of different types
of individual difference variables on the one hand and variables relat-
ing to task condition and task complexity on the other may not be as
transparent as predicted by the CH. For example, anxiety is an affect-
ive variable that is postulated to be drawn upon when tasks are
performed under different conditions, but it is also possible to posit
a logical link between anxiety and the procedural aspects of tasks,
such as ‘with or without planning time’ (Mak, 2011) or ‘with or

[Link] Published online by Cambridge University Press

 Psychological Perspectives 131

without task structure’ (Trebits, 2014). Nevertheless, the CH is the

only theory that has attempted to map the complicated relationships
among the three groups of task variables and foreground the import-
ance of learner factors in accounting for variation of task performance
in the triadic framework.

Language Aptitude
Overview
According to Carroll (1981), language aptitude is a componential
construct that consists of three cognitive abilities, namely phonetic
coding ability, language analytic ability (which entails grammatical
sensitivity and inductive learning) and rote memory, which correspond
to the learning of pronunciation, grammar and vocabulary respect-
ively. Language aptitude is considered to be (1) domain specific in the
sense that it is only important for learning a foreign language, (2)
distinct from other individual difference variables such as motivation
and anxiety, and (3) not subject to change. While some of these
characteristics have been empirically confirmed, others remain contro-
versial. Gardner and Lambert (1965) found that foreign language
learners’ scores on the subtests of the MLAT (Modern Language
Aptitude Test) (Carroll and Sapon, 1959) loaded on different factors
from their scores on the subtests of the PMA (Primary Mental Abil-
ities) – a test of academic intelligence – suggesting that language
aptitude involves distinct abilities for other academic subjects. How-
ever, in a meta-analysis on the construct validity of language aptitude
(Li, 2016), aptitude was found to overlap with intelligence. The meta-
analysis also found that aptitude was unrelated to motivation and
negatively correlated with anxiety. Regarding whether aptitude is
subject to change, there is no clear answer. While some studies found
that learners with more language learning experience had higher
aptitude scores than those with less experience (e.g. Einstein, 1980),
it is possible that those learners with more experience may have had
high aptitude to begin with. Therefore, there is a need for research to
show (1) the higher scores of those with more experience are not due
to their higher aptitude, and (2) the improvement in the same learners’
aptitude scores is only attributable to study experience instead of
maturation effects.
Language aptitude has been measured via test batteries consisting of
multiple subtests that tap the three components, and the most influen-
tial test is the MLAT. The MLAT was validated with more than 5,000
foreign language learners and therefore has strong predictive validity,

[Link] Published online by Cambridge University Press

 132 Theoretical Perspectives

but it has been criticized on a number of accounts. First, it was

validated in the 1950s using traditional audiolingual classes charac-
terized by rote learning and mechanical practice. Thus, whether it is
relevant to current meaning-oriented approaches that emphasize the
importance of exposure to authentic linguistic materials and incidental
learning is questionable. Second, it was developed based on observa-
tions of what happened in language classes, not on SLA theories, and
therefore it lacks theoretical basis. Third, the five subtests do not
correspond with the three hypothesized aptitude components, making
it difficult to interpret the related findings. Fourth, these abilities are
only important for learning the formal aspects of language and for
learning language as discrete items and they do not account for how
the pragmatic and contextual aspects of a language are learned
(Skehan, 2002). Despite the criticisms levelled against the MLAT, it
is still the most dominant aptitude test in current research. Recent
developments include the Hi-LAB (High-Level Language Aptitude
Battery Test) (Linck et al., 2014), which targets high-level learners,
and tests of implicit aptitude (see Wen et al., in press).

Aptitude and TBLT

Aptitude research falls into two major categories: predictive and inter-
actionist, and TBLT falls into the latter. Predictive research aims to
investigate the associations between aptitude and learning rate regard-
less of learning conditions. Interactional studies, which are mainly
based on Robinson’s triadic framework (2011) and his Aptitude Com-
plexes Hypothesis (2002), seek to ascertain whether the role of apti-
tude or different aptitude components varies as a function of different
learning conditions such as the following four (see Robinson, 2002):
(1) short-term classroom treatments developed on the basis of a set of
pedagogical constructs, such as deductive vs. inductive instruction
(Hwu et al., 2014);
(2) laboratory-based treatments defined and operationalized in terms
of the degree of explicitness such as explicit, implicit and inciden-
tal (de Graaf, 1997);
(3) instructional treatments involving interactional corrective feed-
back (CF) (Sheen, 2007);
(4) specific instructional approaches such as communicative teaching
(Ranta, 2002) or immersion (Harley and Hart, 1997).
Among the four streams of research, (1) and (2) are not entirely conducted
with instructional tasks; (3) concerns how focused tasks containing CF
facilitates L2 development; and (4) caters to some of the fundamental

[Link] Published online by Cambridge University Press

 Psychological Perspectives 133

principles of TBLT (Ellis, 2003): the primary focus is on meaning, the

tasks relate to the real world and task outcomes are non-linguistic.
The focus of this section is the studies in the third and fourth categories.
Aptitude and CF. As discussed in Chapter 2, there has been exten-
sive research on the role of CF in SLA because feedback embedded
within meaning-oriented tasks caters to an important principle of
TBLT – focus on form (Ellis, 2003; Spada et al., 2014; Long, 2015).
One line of feedback research concerns whether the effectiveness of
feedback is constrained by individual differences in language aptitude.
For example, Sheen (2007) conducted a classroom study where Eng-
lish as a Second Language (ESL) learners received recasts and metalin-
guistic feedback in learning English indefinite articles a/an. She
reported that language analytic ability only predicted the effects of
metalinguistic feedback, not those of recasts. Yilmaz (2013a) reported
that explicit correction was more effective than recasts only when
learners had high analytic ability. These two studies suggest that
aptitude is more clearly relevant in explicit learning conditions.
However, two studies (Trofimovich, Ammar and Gatbonton, 2007;
Sachs, 2010) that investigated computerized feedback reported that
aptitude was also important in implicit learning conditions such as
when no feedback or implicit feedback (recasts) was provided. How-
ever, the recasts in Trofimovich et al.’s study, which included a correct
model regardless of whether the utterance was correct, are not really
implicit. An explanation for Sachs’ finding is that the instructional
treatment required the learners to process the linguistic target to
complete the task, which drew on their analytic ability.
The influence of aptitude in CF also depends on the nature of
the linguistic target. Li (2013a, 2013b) investigated the three-way
interaction between feedback type, language aptitude and the linguis-
tic target. He found that in the learning of Chinese classifiers, analytic
ability was correlated with the effects of recasts but not metalinguistic
correction. In contrast, the data for perfective -le showed that
the reverse was true: analytic ability predicted the effects of metalin-
guistic correction, but not recasts. Li attributed this discrepancy to the
different linguistic properties of the two structures: the classifier is
syntactically and semantically simple, so the provision of metalinguis-
tic explanation levelled off the role of analytic ability. However, in the
recast condition where metalinguistic explanation was unavailable,
analytic ability came into play. The perfective -le is an opaque
structure that involves complicated form-meaning mapping, and
understanding the metalinguistic explanation poses challenges for
analytic ability. When metalinguistic information was absent, the
learners were unable to learn this complicated structure despite

[Link] Published online by Cambridge University Press

 134 Theoretical Perspectives

support in the form of recasts (which were not effective in learning the
aspect marker). In this case, analytic ability did not play a role because
learners were unable to learn the linguistic target by relying on their
analytic and they were unable to benefit from the instruction.
Aptitude and meaning-focused language teaching. As mentioned,
the MLAT – the most influential aptitude test – was validated in
traditional audiolingual classes involving rote learning and mechan-
ical drills, which led to questions regarding whether it is relevant in
more meaning-oriented approaches such as communicative language
teaching (CLT) or immersion. Ehrman and Oxford (1995) stated that
the suspicion was unfounded because their study showed that aptitude
was the strongest predictor of learning in foreign language classes
which were ‘heavily influenced by the communicative teaching trends’
(p. 77). However, the classes that contributed the data were from
state-funded intensive programmes that, as the researchers admitted,
partly relied on drilling, and therefore the extent to which they were
communicative is uncertain. Stronger support for the relevance of
aptitude in CLT comes from Ranta’s (2002) study, which showed that
aptitude was significantly correlated with learning outcomes on mul-
tiple measures in classes judged to be communicative based on obser-
vations and interviews with the teachers. One caveat about Ranta’s
study is that aptitude was measured by means of a metalinguistic test,
not a validated measure such as the MLAT, although first language
(L1) metalinguistic knowledge has been shown to be related to lan-
guage analytic ability (Alderson, Clapham and Stee, 1997).
Harley and Hart’s studies (1997, 2002) show that aptitude was
implicated in French immersion classes for young learners where the
L2 was learned through exposure to the language. However, these
studies reported an interaction between age and aptitude components,
that is, the learners whose initial age of exposure was younger relied
on memory and the later starters on analytic ability. Harley and Hart’s
findings demonstrate that (1) aptitude is not only important in form-
based instruction but also in meaning-based instruction, (2) aptitude is
drawn on by young learners (10th and 11th graders), and (3) learners
of different age groups or at different stages of learning may draw on
different aptitude components.

Summary
The three types of studies discussed allow us to reach the following
tentative conclusions. First, the feedback research indicates that apti-
tude is more likely to be drawn on in tasks with an explicit focus on
form, which disadvantages low-aptitude learners. However, because

[Link] Published online by Cambridge University Press

 Psychological Perspectives 135

overall explicit feedback has proven more effective than implicit

feedback (Ellis, Loewen and Erlam, 2006; Li, 2010), at least in the
short term, it is advisable to make the corrective intention known to
the learner when CF is used as a form-focusing device in TBLT. (See
Chapter 2 for more detailed discussion of CF in TBLT.) Second, despite
the need for more research, the findings to date suggest that aptitude is
relevant in meaning-based instruction such as CLT and immersion
classes. Furthermore, the finding that younger and older learners draw
on memory and language analytic ability respectively suggests that a
heavy dose of form-focused instruction is not ideal for young learners.

Working Memory
Overview
Working memory refers to the ability to simultaneously store and
process incoming information. Baddeley (2007) proposed a compon-
ential model where working memory consists of a central executive
and three slave systems – a phonological loop, a visuospatial sketch-
pad and an episodic buffer. The central executive coordinates different
components, controls attentional shifts between meaning and form
and between information retrieval and task performance, and inhibits
irrelevant information (Miyake and Friedman, 1998; Juffs and Har-
rington, 2012). The phonological loop is responsible for storing and
rehearsing verbal information. The visuospatial sketchpad deals with
visuospatial information such as images, shapes and locations. The
episodic buffer integrates information from the slave systems and
long-term memory. Although working memory has been argued to
be a component of language aptitude, research has shown that it is
separate from aptitude (Li, 2017), probably because working memory
is a domain-general cognitive device that is essential for learning in
general, not just language learning.
Working memory has been measured in two ways – by using simple
tasks that only tap the storage component and complex tasks that
gauge both the storage and processing components (Conway et al.,
2005). Simple tasks include the word span or digit span tests that
require learners to repeat series of unrelated words, non-words or
digits. A complex task typically consists of two parts: one that requires
the learner to conduct some sort of information processing and one
that requires the learner to recall an element of the item in question.
For example, in a typical reading or listening span test, the learner
reads or hears sentences divided into sets of two to seven sentences
(called span sizes), judges their semantic or syntactic plausibility

[Link] Published online by Cambridge University Press

 136 Theoretical Perspectives

(e.g. ‘The man standing in his office was bitten by a wall’), and at the
end of each set, recalls the final word of each item in that set. In
addition to listening or reading span tests, other measures of complex
working memory that have been used in the literature include oper-
ation span tests that ask the learner to perform some mathematical
computation and remember the letter or word that follows the equa-
tion in the item (e.g. ‘10/2 – 2 = 5 Q’) and backward digit span tests
where the learner is presented with sets of unrelated digits and asked
to recall the digits in the reverse order. Forward digit span is con-
sidered a simple task and backward digit span a complex task.

Working Memory and TBLT

We will now consider the research that has investigated how working
memory is implicated in task-based instruction. This research falls into
two broad categories: studies examining the effects of working
memory on task performance under different conditions and studies
exploring how working memory mediates the learning that results
from interactional feedback. These studies are based on three theoret-
ical models of TBLT and SLA: the Limited Attention Capacity
Hypothesis (LACH) (Chapter 3), the CH (Robinson, 2011) and the
Interaction Hypothesis (IH) (Long, 1996, 2015) – see Chapters 2 and
3. The LACH posits a central role for working memory – a limited
capacity device – in affecting learners’ task performance, which is
often assessed through measures of complexity, accuracy and fluency
(CAF). The LACH draws on Levelt’s (1989) theory of speech produc-
tion, which holds that the production of spoken language undergoes
three stages: conceptualizing the message, formulating the language
representation (selecting the linguistic forms for the message) and
articulating the message. In Levelt’s model, the role of working
memory is restricted to message conceptualization, and formulation
and articulation are ‘underground processes’ (p. 22) that happen
without awareness and that are beyond attention control. However,
while it is perhaps true that formulation and articulation are auto-
matic in L1 oral production, L2 oral production often relies heavily on
attention control in all three phases, not only during message concep-
tualization. This suggests a more crucial role for working memory in
L2 production.
The CH states that complex tasks involving resource-directing
variables divert the learner’s working memory resources to the ‘input
that complex tasks promote’ (p. 19), and therefore the role of
working memory should be more evident in complex tasks than
simple tasks.

[Link] Published online by Cambridge University Press

 Psychological Perspectives 137

The theoretical justification for a mediating effect of working

memory on the effects of interactional feedback can be found in the
IH (Long, 1996, 2015). This emphasizes the role of selective attention
when focusing on linguistic forms in meaning-oriented tasks as
learners switch attention between form and meaning, necessitating a
heavy reliance on working memory. In the case of CF, the learner must
attend to and temporarily hold the information contained in the
feedback and retrieve information from long-term memory in order
to process the available negative and/or positive evidence. At the same
time, the learner needs to maintain the continuation of the ongoing
discourse. Some feedback types such as output-prompting feedback
(Ellis, 2010) push the learner to modify their output, and this also
requires working memory resources.
Working memory and task performance. Studies investigating the
role of working memory in performing tasks have examined how it
interacts with planning, learner proficiency, +/ task structure and
task complexity. With regard to planning, researchers (R. Ellis, 2005)
distinguish pre-task or strategic planning and within-task planning
(i.e. whether learners are pressured to perform the task rapidly or
are given time to think about the information to be communicated
and the language needed).
Within-task planning, then, can be studied by determining the time
learners are given to perform a task, as in Ellis and Yuan (2004).
However, in many task-based studies within-task planning was either
not controlled or there is a lack of information about whether or not it
was controlled. Ahmadian (2012) is one of the few studies investi-
gating the role of working memory in careful online planning. The
study showed that working memory as measured through a listening
span test was significantly correlated with accuracy and fluency but
not complexity. Guará Tavares (2011) examined the effect of working
memory in pre-task planning and found that in the planning condi-
tion, the learners with high working memory outperformed those with
low working memory in terms of complexity and fluency but not
accuracy. Working memory did not affect the task performance of
the no-planning group. One problem with this study, however, is
that the working memory was measured using a speaking span test
where the learners were asked to create grammatically and semantic-
ally acceptable sentences with given words, which might be considered
more like a speaking test than a memory test. Also, there was no
information about whether and how online planning was restricted.
These findings are derived from separate studies conducted in dif-
ferent settings and using varying methods. One study that investigated
both pre-task and unpressured within-task planning with learners

[Link] Published online by Cambridge University Press

 138 Theoretical Perspectives

from the same instructional context is Li and Fu (2018). The study

found significant correlations between working memory (measured
via an operation span test) and accuracy and fluency in the within-
task planning condition, but no significant correlations were found for
the pre-task planning condition. Also, the majority of the correlations
for the within-task planners were positive while those for the pre-task
planners were mostly negative, suggesting that larger working
memory capacity did not help in pressured performance.
The role of working memory may also be constrained by learners’
general L2 proficiency, as found by Gilabert and Munoz (2010). In
this research, adult L2 English learners at a Spanish university were
divided into high- and low-proficiency groups based on their scores on
the Oxford Placement Test. They performed a video narrative task
where they watched a video twice and retold the story with no pre-task
planning and no time limit for task performance. Learners’ working
memory scores on a reading span test were found to be only predictive
of lexical complexity for the high-proficiency learners’ oral perform-
ance and there were no significant results for the low-proficiency
group. One possible explanation is that at the lower-proficiency level
it was the learners’ linguistic proficiency rather than their working
memory capacity that affected task performance.
Kormos and Trebits (2011) reported a study examining whether
working memory had differential effects on structured vs. unstruc-
tured tasks. In the structured task, learners were asked to tell a story
based on a set of cartoon pictures sequenced in the correct order. In
the unstructured task, they had to invent a story based on a set of
unrelated pictures. In both tasks, learners were allowed two minutes to
plan before starting the narratives, but it is not clear whether there was
a time limit for task performance. Significant effects for working
memory were found for the structured task but not the unstructured
task, but the relationship between working memory and task perform-
ance was non-linear. For example, learners with high working
memory capacity outperformed those with lower memory abilities in
terms of clause length, but the latter performed better in terms of
subordination. The researchers speculated that although the struc-
tured task was assumed to be simpler, it may have turned out to be
more cognitively demanding because the learners had no choice but to
follow the provided storyline, which posed a greater challenge than in
the unstructured task where they had more freedom to draw on their
own linguistic resources. This might explain why working memory
was only predictive of the performance under the structured task. The
researchers further pointed out that working memory may not always

[Link] Published online by Cambridge University Press

 Psychological Perspectives 139

be beneficial because those with higher working memory may try to

attend to too many aspects of performance, which can have a detri-
mental effect on their performance.
Finally, there is one study (Crespo, 2011) that investigated the
interaction between working memory and task complexity operation-
alized as +/ reasoning. Adult L1 Spanish EFL (English as a Foreign
Language) learners performed two versions of the same decision-
making task, the more complex version requiring learners to figure
out the relationships between more elements, consider more factors
when making decisions and have access to fewer resources. The study
included measures of three aspects of working memory: phonological
short-term memory, attention control – one function of the central
executive – and working memory as a global construct. Surprisingly,
neither working memory nor attention control showed strong correl-
ations with performance in either of the two task conditions. Phono-
logical short-term memory, however, was significantly correlated with
a number of outcome measures for both the simple and complex tasks.
The study failed to confirm Robinson’s prediction that complex tasks
are more likely to draw on working memory. It suggests that increas-
ing task complexity along the resource-directing dimension may not
increase the processing load. The study also suggests that despite the
putative links between working memory and L2 task performance,
phonological short-term memory, which has received little attention in
task-based research, may prove to be critical in speech production.
To sum up, the studies on the impact of working memory on L2
task performance showed the following:

(1) Working memory seems to be implicated in unpressured perform-

ance during within-task planning, while its role during pressured
performance after pre-task planning is inconclusive;
(2) the role of working memory is greater for advanced learners;
(3) tasks that provide a clear structure for performance may tax
learners’ working memory resources to a greater extent than tasks
without a clear structure (contrary to what is commonly assumed);
(4) there may be a non-linear relationship between working memory
and task performance and greater working memory capacity may
have adverse effects in some task conditions;
(5) complex tasks along the resource-directing dimensions do not
necessarily draw more on working memory than simple tasks;
(6) the role of phonological short-term memory in oral task perform-
ance may be of particular significance – see the section ‘Working
Memory and CF’.

[Link] Published online by Cambridge University Press

 140 Theoretical Perspectives

Working Memory and CF

Mackey et al. (2002) was the first study to explore the role of working
memory in noticing interactional feedback embedded in communica-
tive tasks and facilitating L2 development. Thirty ESL learners whose
L1 was Japanese were paired with native speakers of English and
performed three communicative tasks during which the learners
received recasts on errors relating to English question formation. The
results showed that learners with high working memory capacities
reported more noticing of feedback but that those with low working
memory scores manifested greater development initially. However,
those learners with high scores did better in the delayed post-test.
Kim et al. (2015) confirmed Mackey et al.’s findings. Working
memory was predictive of ESL learners’ noticing of recasts and the
effects of recasts on question formation in dyadic interaction. Kim
et al. also examined the +/ reasoning variable, reporting that task
complexity was not a significant predictor of the noticing of feedback
or learning gains. However, more learners with high working memory
in the complex task advanced to higher stages of question formation
than in the simple task.
Révész (2012) reported a complex interface between working
memory and outcome measures in a study of the effects of recasts on
learning. The study included two measures of phonological short-term
memory (digit span and non-word repetition) and one measure of
working memory (reading span). The effects of the recasts were meas-
ured by means of an oral description task, a written production task
and a grammaticality judgement task. It was found that phonological
short-term memory was correlated with gains in accuracy in oral
production and working memory with gains on the written tests.
Révész argued that phonological short-term memory facilitates the
acquisition of proceduralized/implicit knowledge whereas working
memory is more useful for the development of declarative/explicit
knowledge. The hypothesis about the differential roles of different
types of working memory in facilitating the acquisition of different
types of knowledge is important and needs to be investigated further.
Goo (2012) and Yilmaz (2013a) probed the interface between
working memory and feedback type, both studies investigating both
explicit feedback – metalinguistic feedback in Goo’s study and explicit
correction in Yilmaz’s – and implicit feedback – recasts in both studies.
However, they obtained different results. Goo found working memory
to be a significant predictor of the effects of implicit feedback while
Yilmaz reported it was correlated with the effects of the explicit
feedback. Goo explained that the learners who received recasts utilized

[Link] Published online by Cambridge University Press

 Psychological Perspectives 141

their working memory to notice the linguistic target, whereas the

metalinguistic feedback did not pose attentional demands. Yilmaz also
resorted to the concept of noticing when interpreting his results,
arguing that the recasts in his study were not explicit enough to trigger
the learners’ working memory for conscious learning. However, an
explanation is still needed for the conflicting findings for the explicit
feedback types. One possibility is that the explicit feedback in Goo’s
study contained rule explanation, which does not require heavy use of
the storage function, whereas the explicit feedback in Yilmaz’s study
took the form of explicit correction, necessitating storage of the cor-
rections in order to induce the general rule.
Li (2013a, 2013b) investigated the interaction between working
memory, feedback type and the nature of the linguistic target. Li
reported that working memory was drawn upon when learners received
metalinguistic feedback in learning both Chinese classifiers – a simple
structure and the perfective -le – a complex structure, but not when they
received recasts. However, one striking finding was that working
memory was a positive predictor of the effects of the explicit feedback
in the learning of classifiers but a negative predictor for the perfective
-le. In other words, learners with high working memory capacity
benefited less from metalinguistic feedback when learning a complex
linguistic structure. Li referred to Newport’s Less Is More Hypothesis
(1990) when interpreting the results, that is, with high working memory
capacities tend to store linguistic input as large chunks and ignore the
detail, while those with smaller working capacities engage in deeper
processing of the linguistic input. Li’s studies again show the contingent
relationship between cognitive resources and task conditions.

Summary
It seems that learners make heavy use of their memory resources when
planning their speech during unpressured performance. Allowing pre-
task planning may alleviate the burden on working memory. Further-
more, given the positive effect of unpressured online planning on task
performance, allowing both pre-task planning and unpressured
within-task planning can be expected to have even greater effects on
task performance. One important implication from this line of
research is that tasks that are assumed to be simple may turn out to
be complex and consequently be more taxing on working memory
resources. A further finding is that working memory is implicated in
processing the online feedback embedded in communicative tasks. To
date, however, there has been no study investigating the role of
working memory in delayed, offline feedback.

[Link] Published online by Cambridge University Press

 142 Theoretical Perspectives

Motivation
Overview
Motivation is considered a primary determinant of L2 success, which
explains why it has been one of the most extensively studied individual
difference factors. Dörnyei (2005) explained that the importance of
motivation lies in that fact that it ‘provides the primary impetus to
initiate L2 learning and later the driving force to sustain the long and
often tedious learning process’ (p. 65). Motivation, according to Ellis’
(2015b) synthesis, is a complex construct consisting of three compon-
ents: (1) the reasons why a learner wants to learn an L2, (2) the effort
one invests in the learning process and how it is influenced by the
immediate context, and (3) the impact of the evaluation of the
outcome and progress of learning on subsequent behaviour.
Of these three aspects of motivation, those in (1) constitute general-
ized, macro motives that relate to the general goal to be achieved and
the general orientation towards the language, culture and speech
community. These include the traditional integrative and instrumental
motivation in Gardner’s (1985) model, with the former referring to
motives arising out of positive attitudes towards the speakers of the
target language and the desire to integrate and identify with the
community, and the latter to pragmatic motives such as getting a job
or promotion. Those involved in (2) and (3) can be regarded as the
specific, micro aspects of motivation that relate to the process of
learning or the ongoing learning tasks.
The tripartite framework also incorporates the dynamic, situated
model of motivation proposed by Dörnyei and his associates (Dörnyei
and Ottó, 1998; Kormos and Dörnyei, 2004; Dörnyei, 2005; Dörnyei
and Ushioda, 2009), which differs from the traditional static model,
where motivation is viewed as a trait that correlates with the ultimate
learning outcomes. In this model, motivation is (1) subject to temporal
variation and (2) influenced by multiple contextual factors such as the
school, the course, the class and the target language. Dörnyei’s ideas
are well represented in the so-called ‘process model’ (2005, p. 84)
where different conglomerates of motives are drawn on at different
stages of learning. At the pre-actional stage prior to the start of the
learning process, learners’ motivation is generated and the goal is set.
This is called choice motivation, and it relates to learners’ general
dispositions or the macro factors in Gardner’s model. During the
actional stage or the learning process, the general motivation is influ-
enced by various supporting factors as well as factors that inhibit
distractions and supress unfavourable behaviours (e.g. off-task

[Link] Published online by Cambridge University Press

 Psychological Perspectives 143

behaviours). These factors influence executive motivation. In the post-

actional stage when the learning process/task is completed, the learner
makes retrospective evaluations of what transpired during the learning
process, and the results of the evaluation in turn affect subsequent
actions. This concerns the attributional dimension of motivation.
One motivation theory proposed by Dörnyei (2005) that has had a
profound influence on recent research is the L2 self system. This entails
three dimensions of motivation: Ideal L2 self, ought-to L2 self and L2
learning experience. The Ideal L2 self refers to the motivation driven
by the desire to reduce the discrepancy between one’s current state and
the future state to be reached (e.g. ‘I often imagine myself speaking
English fluently’). The ought-to L2 self concerns the motivation that
prompts the learner to study an L2 to avoid the negative consequences
(e.g. ‘If I don’t study English, others will be disappointed’). L2 Learning
Experience relates to the motives associated with the immediate con-
texts of learning as determined by the curriculum, the course materials
and the teacher (e.g. ‘I always look forward to English classes’).
Whereas the two types of self-related motives are associated with the
macro dimensions of motivation drawn upon in the pre-actional stage,
the motives relating to learning experience concern the micro aspects of
motivation and are involved in the actional stage and the post-actional
stage (although Dörnyei was not clear about which of the three types
of motivation are important in the post-actional stage).

Motivation in TBLT
Task motivation encompasses all motives that may affect task per-
formance or engagement at any of the three stages of a task cycle (pre-
task, main task and post-task). It will be influenced by both general
motives such as the Ideal Self and the Ought-to Self but in particular
by the more specific motives relating to the task-as-workplan and task-
as-process (see Chapter 1), such as attitudes towards the task, percep-
tions about the difficulty or complexity, how the task is implemented
and the other participants, all of which are subsumed under the L2
Learning Experience component of the L2 Self System. In this section,
we discuss what the little empirical research has shown about task
motivation and how the concept of motivation has been investigated
in relation to TBLT.
Dörnyei (2002) was one of the first to explore the multifaceted
nature of task motivation in dyadic interaction. The study included
measures of different levels of motivation, including generalized dis-
positions such as integrative and instrumental motivation, as well as

[Link] Published online by Cambridge University Press

 144 Theoretical Perspectives

motivation relating to the immediate contexts such as attitudes

towards the course and the task. The outcome measure was task
engagement, which was operationalized as the number of words and
turns produced by the learners (forty-four Hungarian secondary
school EFL students) in an argumentative task. It was found that all
motivation variables except for integrative motivation were signifi-
cantly correlated with task engagement. However, a different picture
emerged when the learners were divided into two groups based on
their task attitudes. Whereas both generalized and task-specific
motives were predictive of high-task attitudes and learners’
engagement, the engagement of learners with low-task attitudes was
only predicted by course attitudes. This suggests that their lack of
interest in the task was compensated for by their positive attitudes
towards the course. The study also found that the task engagement of
the learners with low-task attitudes was affected by their interlocutors’
motivation. Taken together these results indicate that both task-
specific and general motives contribute to task motivation and that
task motivation is co-constructed.
Drawing on the same data as Dörnyei (2002), Kormos and Dörnyei
(2004) found that although the learners’ task attitudes predicted the
quantity of their oral production (number of words and turns), this
factor did not have a positive effect on the quality (the linguistic
accuracy and complexity) of their production. Course attitudes were
found to be significantly correlated with both quantitative and quali-
tative aspects of production. In a study by Al Khalil (2011), learners’
general motivation measured by Gardner’s Attitude/Motivation Test
Battery (AMTB) (1985) was significantly correlated with the CAF of
the oral production of forty-four L2 Arabic learners in the United
States when engaged in dyadic communication with a native speaker.
Integrative motivation was significantly correlated with the noticing of
recasts, an index of task engagement. These two studies seem to
indicate that task-specific motivation affects task engagement, and
that more general types of motivation may affect both task
engagement and the linguistic aspects of task performance. Following
Dörnyei and Ushioda (2009), we can suggest motivation is best con-
ceived as not directly related to achievement but as the antecedent of
action that may have indirect effects on achievement. It would follow
that measures of motivated behaviours such as task engagement are
better measures of motivation.
Dembovskaya (2009) examined whether pre-task motivation-
enhancing strategies lead to more positive perceptions of learners’ task
experience and whether pre-task cognitive strategy training improves
students’ task performance in terms of the CAF of their oral

[Link] Published online by Cambridge University Press

 Psychological Perspectives 145

production. L2 French learners at a US university performed an

information-gap task where they received a list of objects and clues
found in the apartment of a crime suspect and were asked to work in
groups to come up with a description of the suspect. To motivate one
group of learners the teacher informed them about the value of the
task (e.g. by telling them it would help them improve their communi-
cative competence), enhanced their interest (e.g. by telling students
that other students had performed the task and found it enjoyable) and
promoted their self-confidence (e.g. by telling students that they could
do it well). The students in the other group were equipped with the
linguistic and strategic tools needed to complete the task, such as
activating their schematic knowledge and informing them what infor-
mation to look for to identify a suspect. The results showed that the
group that received the pre-task motivational instruction perceived the
task to be more interesting and valuable and themselves as more
autonomous than the groups that did not receive motivational
training, but the finding was only true of Year 2 students, not Year
3 students. Also, there was no difference in the two groups’ perform-
ance of the task. Noteworthy is the fact that although the pre-task
phase can be thought of as corresponding to the pre-actional stage in
Dörnyei’s process model as it relates to so-called ‘choice motivation’,
the pre-actional stage in Dörnyei’s conceptualization concerns L2
learning in general and therefore is seen as involving generalized
motives. In this study, the pre-actional stage concerns a task and,
consequently, the choice motivation that the learners received training
for relates more clearly to task-specific motivation.
Jauregi et al. (2012) investigated whether learners’ motivation can
be improved through authentic, video-web communicative tasks.
A group of L2 Dutch learners in Czech attended three 30-minute
virtual interaction sessions with some Dutch pre-service language
teachers. A thirteen-item questionnaire was utilized to measure differ-
ent dimensions of motivation – attitudes towards interacting with
native speakers, attitudes towards the course and attitudes towards
the L2 culture. Significant effects were found and the effects were more
prominent for the beginner group than the more advanced learners.
What is unique about this study is that motivation was examined as a
dependent variable that served as the ‘effect’ rather than ‘cause’, and it
(both integrative motivation relating to the culture and speech com-
munity and more specific motivation relating to the course) was found
to be improved through task-based interaction.
One commonality between these two studies is that learners’ motiv-
ation seems to more easily influenced at the beginning stage of L2
learning than at more advanced stages. One possible explanation is

[Link] Published online by Cambridge University Press

 146 Theoretical Perspectives

that the attitudes and perceptions of more advanced learners about the
learning process or tasks are entrenched and not easily swayed. It can
also be speculated that learners that choose to proceed to higher stages
of learning are more motivated to begin with. In any case, it would
seem more important to enhance learners’ motivation at the beginning
stages of learning, so the interface between motivation and learner
proficiency seems to be a promising area of research.

Summary
The few studies that have investigated task motivation show that it is a
dynamic, complex and multi-componential construct. It consists of
motives relating to different facets of the learning task, including
general learning goals, course motivation and motives to do with the
performance of the task per se. These motives have been found to be
significant predictors of learners’ task engagement and performance.
Pre-task motivational strategies can enhance learners’ motivation to
perform a task while participation in meaning-oriented tasks can
improve learners’ course and integrative motivation. Finally, initia-
tives to increase learners’ motivation seem to work better for low-level
learners than high-level learners, suggesting that practitioners should
make a special effort to stimulate and maintain beginning L2 learners’
motivation when implementing TBLT.
Motivation is one of the most promising areas of TBLT research
and Robinson’s (2011) triadic framework provides a useful frame-
work for so doing. For example, while increasing the cognitive
demands of a task may enhance learners’ task performance, increasing
task complexity beyond a certain threshold may have a harmful effect
on learners’ motivation, which may in turn have adverse effects on
their task performance and engagement. Second, with regard to the
variables relating to task condition, research on whether and how
factors pertaining to participatory structure and participant character-
istics affect motivation may provide valuable insights for the imple-
mentation of TBLT. For example, Dörnyei’s (2002) finding that
learners’ motivation was affected by their partners’ motivation sug-
gests that it is advisable to pair up learners’ with different levels of
motivation. However, Dörnyei and Kormos (2000) found that the
relationship between task participants was predictive of task
engagement when the task was performed in their L1 but not when
the task was performed in the L2. This suggests that participants’
relationships may not be as important as teachers have assumed but
clearly more research is needed. Third, it is surely important to exam-
ine how motivation in contrast to and in combination with other

[Link] Published online by Cambridge University Press

 Psychological Perspectives 147

individual difference factors affects task performance but to the best of

our knowledge, there has been no research in this regard.

Anxiety
Overview
Horwitz, Horwitz and Cope (1986) defined anxiety as ‘the subjective
feeling of tension, apprehension, nervousness, and worry associated
with an arousal of the autonomic nervous system’ (p. 125). Three
types of anxiety have been identified in the literature: trait anxiety,
state anxiety and situation anxiety (Ellis, 2015b). Trait anxiety is a
personality variable that refers to the general disposition, state anxiety
relates to one’s emotional condition at a particular moment and
situation anxiety is associated with what one experiences in particular
contexts. Trait anxiety accounts for interpersonal variation, that is,
certain individuals are inherently more anxious than others. State and
situation anxiety can be considered as intrapersonal variables in the
sense that the same individual may experience different levels of
anxiety at varying moments in a particular situation and in different
situations. Language learning anxiety is a type of situation anxiety,
and it occurs when a leaner produces or comprehends an L2. Horwitz
et al. (1986) argued that language learning anxiety is principally
derived from three sources: spontaneous communication, fear of nega-
tive evaluation and test anxiety. In L2 research, anxiety has been by
default been associated with speaking, and measures of anxiety –
typically questionnaires – primarily consist of speaking-related items
(Phillips, 1992; Aida, 1994), although anxieties for other skills such as
listening (Elkhafaifi, 2005), writing (Cheng, Horwitz and Schallert,
1999) or reading (Saito, Horwitz and Garza, 1999) have also been
investigated. Furthermore, anxiety can be debilitative or facilitative,
that is, while too much anxiety may have a negative influence on task
performance or learning outcomes, a certain amount of anxiety may
play a positive role. However, the distinction has not received much
attention in empirical research.
In general, anxiety has been found to have negative effects on
language learning (e.g. Ewald, 2007), which is in line with the harmful
effect of anxiety on general academic performance (r = 0.25),
according to a meta-analysis of 126 studies (Seipp, 1991). Tobias
(1985) attributed the adverse effect of anxiety to cognitive interfer-
ence, that is, anxiety-prone learners have to split their cognitive
resources between task-relevant and task-irrelevant processes, thereby
affecting their task performance. Drawing on Tobias’s model,

[Link] Published online by Cambridge University Press

 148 Theoretical Perspectives

MacIntyre and Gardner (1994) posited that anxiety causes interfer-

ence at all three stages of learning: (1) when learners receive linguistic
input, (2) when learners try to organize and store input, and (3) when
learners are required to produce previously learned material.
MacIntyre and Gardner developed the Input, Processing and Output
Anxiety (IPOA) scale to measure the types of anxiety for the three
stages of learning, with six items for each of the three stages. Theoret-
ically sound as it might be, some of the items in the questionnaire do
not seem to be clear measures of the three types of anxiety. For
example, the items for input and processing are not clearly distinguish-
able, and one item in the output section is about test anxiety. Finally,
while the mainstream view is that anxiety is a cause for low achieve-
ment, Sparks and Patton (2013) contended that it is the consequence
of learning difficulties or lack of aptitude.

Anxiety and TBLT

The research on the role of anxiety in TBLT has revolved around three
themes. One is the correlation between anxiety and task complexity.
As discussed, the CH (Robinson, 2011) predicted that stronger correl-
ations between anxiety and task performance can be expected when
task complexity is increased along either resource-directing or
resource-dispersing dimensions. Put in another way, as the processing
demands of tasks increase, the negative impact of anxiety becomes
more evident. A second line of research has focused on whether
learners’ anxiety levels vary as a function of the modality of inter-
action. For example, computer-mediated (CM) task-based instruction
may alleviate learners’ anxiety in the absence of the pressure and
interaction demands that characterize face-to-face communication.
A third stream of research focuses on the mediating effects of anxiety
on the learning that results from CF provided in communicative tasks.
In the following, we discuss these studies in more detail.
Task complexity. There have been two studies investigating the
correlations between anxiety and the resource-directing variable of
‘with or without reasoning demand’. In Robinson’s (2007c) study,
forty-two L1 Japanese university EFL students formed twenty-one
dyads, each performing three narrative tasks at different levels of
reasoning demand. Each dyad was given a set of jumbled pictures.
One learner narrated the story based on the sequence he/she decided
on and the other put the pictures in the sequence based on the
speaker’s narrative. Anxiety was measured using MacIntyre and
Gardner’s (1994) IPOA scale. Output anxiety was found to be nega-
tively correlated with syntactic complexity, and with the increase of

[Link] Published online by Cambridge University Press

 Psychological Perspectives 149

task complexity the correlations became stronger. However, one inter-

esting finding that was not discussed in detail is that processing anxiety
was significantly and positively correlated with the accuracy of the
production under the simple task condition.
Kim and Tracy-Ventura (2011) examined the mediating role of
anxiety in affecting the learning of the English past tense morphology
under three task conditions that differed in terms of the presence of
reasoning demand (simple vs. complex) and number of elements
(complex vs. more complex). The study involved 128 Korean EFL
learners who performed four dyadic interaction tasks within a two-
week period. The researchers found that the learners with low anxiety
(measured through a six-item questionnaire) outperformed their high-
anxiety peers in all task conditions. They concluded that there was no
interaction between anxiety and task complexity because the role of
anxiety did not vary across task conditions.
Whereas these two studies concern a resource-directing variable that
involves information manipulation, Trebits (2014) examined a
resource-dispersing variable relating to the procedural dimension of
task complexity: single vs. dual task. In this study, which is based on
the same data as Kormos and Trebits (2011) (discussed in the section
on ‘Working Memory and TBLT’), the learners performed a cartoon
description task where they told a story following a given sequence,
and a picture description task where they told a story based on unre-
lated pictures that must be sequenced logically during the narrative.
The cartoon task was easier in terms of content organization than the
picture task – a dual task condition where the learners had to attend to
both content organization and language formulation (selection of
linguistic forms). However, the author argued that the cartoon task
was more challenging in terms of language formulation than the
picture task because the former required the learners to select linguistic
forms to match the prescribed content while the latter allowed learners
the flexibility of tailoring the content to match their linguistic reper-
toire. As in Robinson (2007c), anxiety was measured via the IPOA
battery. It was found that: (1) in the cartoon task, processing anxiety
correlated positively with lexical and syntactic complexity, and (2) in
the picture task, output anxiety correlated negatively with accuracy.
As can be seen, it is difficult to draw unequivocal conclusions about
the interface between anxiety and task complexity due to the conflict-
ing findings, which in turn may have resulted from the methodological
differences between the studies. However, these studies did show some
interesting patterns. First, anxiety does have some negative effects on
task performance and L2 development. As Robinson (2007a) and
Trebits (2014) showed, it had a negative impact on task performance,

[Link] Published online by Cambridge University Press

 150 Theoretical Perspectives

particularly in complex tasks. Kim and Tracy-Ventura (2011) showed

that low-anxiety learners consistently outperformed high-anxiety
learners while learning the English past tense. Second, it would seem
that anxiety may be facilitative of some aspects of oral production in
simple tasks where the burden of content organization is alleviated
and where anxiety may serve as an impetus for more attention to the
linguistic aspects of their task performance. This is evidenced by the
finding in both Robinson and Trebits’s studies that significant, positive
correlations were found for anxiety in the simple task conditions.
Third, Kim and Tracy-Ventura found that low-anxiety learners out-
performed high-anxiety learners in both simple and complex task
conditions. However, the data also showed an advantage for complex
tasks in comparison with simple tasks in enhancing L2 development
regardless of anxiety – a finding that was not discussed due to the
focus of the study. This raises the question of whether priority should
be given to the effectiveness of instruction or the concern over the
possible negative consequence of incurring more anxiety by using
complex tasks.
Task modality. Computer-mediated (CM) communication in the
form of text chat has been assumed to be effective in easing learners’
anxiety in comparison with face-to-face communication because of the
opportunity for more online planning and lack of requirement for
public performance in the former mode. However, Baralt and
Gurzynski-Weiss (2011) showed that there was no difference between
the two modes of communication in terms of the amount of anxiety the
learners experienced. What is unique about this study is that it exam-
ined state anxiety, that is, whether there was any change in learners’
anxiety measured during and after task performance. Twenty-five
fourth-semester Spanish learners from a large public university in the
United States performed an information-gap task where they engaged in
dyadic interaction with a native-speaker interlocutor in CM and face-
to-face communication. The results revealed no difference in the
learners’ anxiety between the two modes of communication either
during or after task performance. The researchers explained that this
might be because the novelty of engaging in online chat in the foreign
language led to anxiety levels comparable to face-to-face communica-
tion. While this stands to reason, an alternative speculation is that the
study happened in a laboratory setting that did not require public
performance, which is often the direct cause of anxiety.
In another study, Satar and Ozdener (2008) compared the effects of
text and voice chat on the development of Turkish EFL learners’ oral
proficiency and on learners’ anxiety. Ninety high school students
were divided into three groups: text chat, voice chat and control.

[Link] Published online by Cambridge University Press

 Psychological Perspectives 151

The learners in the experimental groups were paired up and performed

two tasks in each of four treatment sessions. The task types included
information gap, problem-solving, jigsaw and decision-making.
The control group followed the normal curriculum and did not receive
any treatment. The learners also answered the Foreign Language
Classroom Anxiety Scale (FLCAS) anxiety questionnaire (Horwitz
et al., 1986) before and after performing the tasks. The results showed
that (1) both chat groups outperformed the control group in their
speaking performance as a result of the treatment, and (2) the text
chat group showed significantly lower anxiety scores after the treat-
ment, although the anxiety level of the voice chat group was also
reduced (the change was non-significant). The researchers explained
that online chat led to lower anxiety because it was a secure environ-
ment without peer pressure. However, surprisingly, the control group
showed consistently higher anxiety scores than both experimental
groups both before and after the study even though they did not
participant in any chat sessions.
In these two studies, interaction happened either between a native-
speaker interlocutor and a learner or between two learners. A recent
study by Côté and Gaffney (2018) examined the impact of anxiety on
learners’ public performance in front of a large group. The researchers
claimed that this setting is more representative of a real classroom and
therefore the findings have higher ecological validity. The study
involved sixty-one beginning learners of French at a Canadian univer-
sity who completed a grammar and vocabulary lesson followed by two
production activities in a face-to-face mode and a CMC (computer-
mediated communication) mode. The learners also answered a revised
version of the FLCAS questionnaire. The study revealed that the
learners experienced significantly less anxiety and produced signifi-
cantly more turns and words in the CMC mode than the face-to-
face mode.
Overall, the limited research on the effects of modality of instruction
delivery on learner anxiety show that learners indeed suffer from
lower anxiety in the CMC mode compared with the face-to-face mode.
However, the outcomes investigated in the studies relate primarily to
the process aspects of task-based instruction, such as the number of
words produced, and more research is necessary whether anxiety
affects the product dimensions or learning gains in the two modes of
instruction in different ways.
Task-based feedback. There have been two studies on the mediating
role of anxiety in affecting the effectiveness of CF embedded in com-
municative tasks (Sheen, 2008; Rassaei, 2015). In Sheen’s study, sixty-
one learners from a large ESL programme in a US community college

[Link] Published online by Cambridge University Press

 152 Theoretical Perspectives

were divided into two groups: recast and control, each subdivided into
high and low based on their anxiety, which was measured through a
six-item questionnaire. The recast group performed two narrative
tasks, each including a practice stage where they worked in groups,
followed by a reporting stage where each student produced a few
sentences before passing the speaker role to another group member.
The teacher corrected their errors on English articles a/the by
using recasts. The results revealed that low-anxiety learners not only
benefited more from recasts in learning the target structure but also
produced more modified output (responses after feedback) than high-
anxiety learners.
Rassaei carried out a study with Iranian EFL learners from a private
language teaching institute following Sheen’s procedures, but the
researcher included two types of feedback: recasts and metalinguistic
correction, aiming to see whether anxiety has differential impacts on
the effects of the two types of feedback. The study found that high-
anxiety learners benefited more from recasts, and low-anxiety learners
benefited from both recasts and metalinguistic feedback. However, a
closer inspection of the results showed that in the recast group, high-
and low-anxiety learners were similar in their scores across all three
tests of treatment effects, but in the metalinguistic group, low-anxiety
learners performed consistently better than high-anxiety learners. It
would seem that anxiety did not play a role when recasts were pro-
vided – a finding that is different from Sheen’s finding that anxiety
played a negative role in affecting the effects of recasts. However, one
piece of useful information to take away from this study seems to be
that metalinguistic feedback – an explicit form of correction – does
have an adverse effect on learning outcomes.

Summary
To conclude this section, we would like to point out that the amount
of research on the role of anxiety is in disproportion to its importance
given its putative connections with TBLT. The limited research seems
to suggest that anxiety is unfavourable for speech performance under
complex task conditions, and that there is a possibility that simple
tasks favour anxious learners by freeing up their cognitive resources
and diverting their attention to the linguistic aspects of their perform-
ance. In terms of its role in affecting the learning of new linguistic
knowledge, anxiety seems to have a harmful effect, regardless of task
complexity. The research on task modality suggests that in laboratory
settings text-based CM interaction seems to lessen learners’ anxiety
compared with face-to-face interaction, and that within CMC, text

[Link] Published online by Cambridge University Press

 Psychological Perspectives 153

chat has a positive effect in reducing anxiety compared with voice

chat. The research on CF indicates that explicit feedback in the form of
metalinguistic correction disadvantages highly anxious learners while
mixed findings are obtained for whether anxiety affects the effects of
recasts.

Final Comments
In general, the research on aptitude was conducted to ascertain
whether aptitude mediates the effects of different instructional treat-
ments or whether it is implicated in different learning conditions. The
research has shown that traditional aptitude is a set of cognitive
abilities that are most likely drawn on in tasks with an explicit focus
on form. One promising area of research that has been recently
initiated is identifying those abilities that are important in implicit or
unconscious learning (Granena, 2013, 2015), which TBLT is claimed
to facilitate. It would be interesting to ascertain whether a task-based
approach draws more on implicit aptitude than traditional aptitude,
which has been found to be relevant to explicit learning.
The research on working memory explored whether it affected
learners’ performance under different task conditions and how it
mediated the effectiveness of interactional feedback. One general find-
ing supported by several studies is that the role of working memory is
more evident in unpressured performance where learners have oppor-
tunities to plan the content and language of their speech. The role of
working memory in mediating the effects of CF has been attributed to
its function in noticing the corrective force of feedback. One objective
of this stream of research should be to identify tasks that facilitate task
performance or L2 learning but do not pose heavy processing
demands on working memory resources. For tasks that enhance task
performance or learning outcomes but are taxing on working memory
resources, it is important to find ways to support learners by alleviat-
ing the processing load.
The few studies of task motivation investigated the predictive power
of generalized and task-specific motives for task engagement and
performance and ways to improve task motivation. Unlike the
research on the two cognitive variables, studies on motivation are less
uniform and the findings are less robust, reflecting in part the lack of
theorization about its role in L2 acquisition (Dörnyei, 2005; Ellis,
2015b) and the uncertainty over the nature, composition and
measurement of the construct. However, as pointed out, the investi-
gation of task motivation may provide important insights, given that
students’ lack of motivation has been considered a major hindrance to

[Link] Published online by Cambridge University Press

 154 Theoretical Perspectives

the implementation of TBLT in some contexts – for example, in

foreign language as opposed to L2 settings (Ellis, 2003).
The research on anxiety affords a complex, interesting picture that
to some extent undermines the commonly held beliefs about this
affective factor. For example, while it did have some adverse effects
under some task conditions, there was also evidence for a possible
positive effect for anxiety. One principle that should be adhered to is
to (1) prioritize task or instruction type when facing the choice
between tasks that enhance learning and those that cause less anxiety,
but (2) find ways to adapt the aspects of the task or instruction that
disadvantages highly anxious learners. For example, if complex tasks
consistently show larger effects than simple tasks in improving
learners’ L2 knowledge for both high- and low-anxiety learners, as
Kim and Tracy-Ventura’s (2011) data showed, then it would be
advisable to use complex rather than simple tasks even though com-
plex tasks may cause more anxiety. However, teachers then should
take steps to reduce learners’ anxiety by manipulating the procedural
aspects of task, such as by allowing learners to plan before performing
a task. However, to date, there has been no research on the role of
anxiety in different planning conditions.
Finally, we would like to call for more research into the role of
individual difference variables in affecting task performance and learn-
ing gains. It is a mistake in tacitly (and mostly) assuming that tasks
work in the same way for all people, and this concern has been borne
out by the limited research we synthesized in this chapter.

[Link] Published online by Cambridge University Press