You are on page 1of 24

Applied Linguistics 2016: 37/5: 669–692 ß Oxford University Press 2014

doi:10.1093/applin/amu058 Advance Access published on 29 October 2014

One Size Fits All? Learning Conditions


and Working Memory Capacity in
Ab Initio Language Development

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


1,
*CRISTINA SANZ, 2HUI-JU LIN, 3BEATRIZ LADO,
4
CATHERINE A. STAFFORD and 5HARRIET W. BOWDEN
1
Georgetown University, 2Chung Yuan Christian University, 3Lehman College, CUNY,
4
University of Wisconsin-Madison, and 5University of Tennessee-Knoxville
*E-mail: sanzc@georgetown.edu

The article summarizes results from two experimental studies (N = 23, N = 21)
investigating the extent to which working memory capacity (WMC) intervenes
in ab initio language development under two pedagogical conditions [  gram-
mar lesson + input-based practice + explicit feedback]. The linguistic target is the
use of morphosyntax to assign semantic functions in Latin. Results suggest that
with the more traditional pedagogical approach [+ grammar lesson], WMC does
not predict learner outcomes. In contrast, in an approach that immediately im-
merses learners in meaning-focused practice with the same explicit feedback,
WMC predicts gains on interpreting aural and written input. Thus, it appears
that pre-practice grammar explanation ‘levels the field’ for learners of varying
WMC, while WMC may play more of a role when metalinguistic information is
limited to reactive feedback. These results extend previous research (Robinson
2002, 2005b; Erlam 2005; Goo 2012; Li 2013; Sagarra and Abbuhl 2013a, 2013b;
Tagarelli et al. 2014), and further call attention to the importance of interactions
between pedagogical tools and individual differences in explaining language
development.

INTRODUCTION
Aptitude for second language acquisition (SLA) is a construct of obvious inter-
est and importance to language researchers, practitioners, and learners.
However, it is not clear exactly what aptitude is, or precisely how it affects
language acquisition. Skehan (2002) provides a framework for understanding
and investigating aptitude in SLA. Aptitude is seen as a multi-dimensional
factor, with several subcomponents which may affect different processes that
are part of language acquisition. As identified by Carroll (1958) and subse-
quently enshrined in the Modern Language Aptitude test (MLAT; Carroll and
Sapon 1959) these subcomponents have been posited to include phonemic
coding ability, grammatical sensitivity, inductive language learning ability,
and associative memory. Skehan includes more abilities, such as perceptual
speed, attention management, working memory (WM), retrieval ability, and
automatization. This view of aptitude is largely cognitively oriented, and
670 LEARNING CONDITIONS AND WORKING MEMORY

although it does not take into account other individual differences that may be
equally or more important in SLA, such as affective or motivational variables
(Winke 2013), research has supported the predictive validity of this construct
for L2 learning.
There have been differing positions on the contexts in which aptitude may
play a role in SLA, however. For instance, some have posited that aptitude may

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


only be important at lower levels of language acquisition (Hummel 2009), or
that different aptitude abilities are needed at different stages of learning
(Robinson 2005a, 2007). In addition, some researchers have posited that
aptitude should only play a role in instructed SLA and have no bearing on
naturalistic acquisition (Krashen 1981, 1982; Reber 1989, 1993). In fact, how-
ever, Skehan (1989a) suggests that it may be argued that aptitude might be
more relevant for uninstructed contexts, given that in such contexts, learners’
abilities to analyze the input and make generalizations are arguably put under
greater pressure to decode and assimilate the new linguistic system. Along
these same lines, even within an instructed context, more explicit and implicit
approaches might be expected to interact in different ways with different com-
ponents of aptitude.
Some SLA research has investigated such aptitude-by-treatment interactions
(see Skehan 2002, DeKeyser 2012, and Vatz et al. 2013 for reviews). For ex-
ample, Wesche (1981) investigated the effects of matching or mismatching
students’ aptitude profiles with teaching methods, and found that analytic
learners performed better and were more satisfied with an analytic method-
ology, while learners with high memory ability did well in a memory-oriented
methodology. Reves (1983) compared the effects of aptitude on learning in
formal and informal contexts, and found it to be important in both contexts.
De Graaff (1997) reported that, similarly, aptitude measures correlated signifi-
cantly with learning in both explicit and implicit conditions. Robinson (1997)
also found correlations for aptitude and learning in instructed, rule-search, and
implicit—but not incidental—conditions. Interestingly, he found in another
(2005b) study that aptitude was significantly correlated with learning under
incidental conditions, although significant results were limited to performance
on a delayed production test. Thus, a number of studies have shown that
aptitude may influence language learning in a variety of contexts, although
more evidence is needed regarding the interaction between treatment type and
aptitude in order to gain a fuller understanding.
With regard to the subcomponents of language aptitude, several researchers
have pointed to WM as a core component (e.g. Skehan 2002; Dörnyei and
Skehan 2003; Robinson 2005a). WM is generally conceived of as a short-dur-
ation, limited-capacity memory system which simultaneously stores and pro-
cesses incoming perceptual (including linguistic) information in the service of
complex cognition (Baddeley and Hitch 1974; Baddeley 2000, 2007).
Extracting form-meaning connections from linguistic input may be considered
one form of complex cognition referred to here.
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 671

To date, a large body of research has identified a relationship between work-


ing memory capacity (WMC) and L2 learning or performance (see Juffs and
Harrington 2011 and Williams 2012 for recent reviews). There are a large
number of studies that provide evidence that greater WMC is favorable for
L2 performance in a number of areas including sentence processing, spoken
output, reading, and writing (e.g. Harrington and Sawyer 1992; Miyake and

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Friedman 1998; Mackey et al. 2002, 2010; Abu-Rabia 2003; Erlam 2005;
Robinson 2005b; Sagarra 2007; Sagarra and Herschensohn 2010; Mackey
and Sachs 2012). However, some studies have not found significant relation-
ships between WMC and L2 processing or learning (e.g. Sagarra 2000; Juffs
2004; Trofimovich et al. 2007), especially in studies in which proficiency was
measured by standardized tests. These results should not be entirely surprising.
After all, L2 development is a long-term process potentially affected by many
variables, so, even if a learner has relatively low WMC, she may be able to
compensate by many other means, including effort, practice or metalinguistic
awareness. The lack of consistency of the findings even on more discrete
measures and tasks is no doubt also due in part to the inconsistency of
WM measurements implemented, which include word span tests, digit
span tests, reading or listening span tests, computational digit span tests, and
non-word recall (see Williams 2012 for a critique of these different
measurements).
Importantly, however, a number of studies have found evidence that WMC
mediates the effects of a variety of learning conditions on language develop-
ment (Ando et al. 1992; Mackey et al. 2002, 2010; Erlam 2005; Sagarra 2007;
Goo 2012; Li 2013; Sagarra and Abbuhl 2013a; Tagarelli et al. 2014).
These studies are part of a growing research trend that takes an interactionist
perspective—one that examines not only the external (such as learning
conditions) or internal variables (such as WMC), but also their interaction—
to explain the variation in rate of acquisition and final attainment character-
istic of SLA (Sanz 2005). Accordingly, the present study examines
whether WMC moderates the effects of two different learning conditions
on the ab initio development of morphosyntax involved in semantic
role assignment in a non-primary language.1 The two conditions are operatio-
nalized as presence [+grammar lesson] vs. absence [-grammar lesson] of
grammatical information provided before input-based practice with explicit
feedback (i.e. feedback with grammar rules). Below we examine in more
detail relevant studies investigating the interaction of WMC and L2 learning
conditions.

WMC and learning conditions


While some studies (see above) have focused on whether participants with
higher WMC outperform those with lower WMC on a single instructional
treatment, others (Robinson 2002, 2005b; Erlam 2005; Sagarra and Abbuhl
2013a, 2013b; Goo 2012; Li 2013; Tagarelli et al. 2014) have examined the
672 LEARNING CONDITIONS AND WORKING MEMORY

mediating role of WMC for L2 development among groups of learners


exposed to different instructional conditions. Results of several of these studies
have suggested that WMC is associated with learning in less explicit conditions,
that is, in the absence of grammar explanations and instructions to search for
rules.
Robinson (2002, 2005b) investigated the role of individual differences in IQ,

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


aptitude, and WMC in both incidental learning of Samoan and degree of suc-
cess in the completion of explicit and implicit learning tasks involving artificial
grammars. Significant correlations were found between performance in the
incidental condition and WMC on immediate and one-week delayed aural
(but not written) grammaticality judgment post-tests, and one-week and
six-months delayed (but not immediate) guided production post-tests.
Robinson speculated that superior performance on delayed tests by high
WMC participants was due to their ability to allocate attention to L2 input
during immediate post-tests such that the completion of the post-tests served
as an additional learning opportunity that was not available to the low WMC
participants.
Erlam (2005) investigated the relationship between aptitude (including
WMC) and learning of L2 French object pronouns. In this classroom study,
60 L1 English secondary school students were assigned to one of four groups:
control, deductive instruction [+ rule presentation + production practice
input practice], inductive instruction [ rule presentation + production
practice + input practice], and structured input instruction [+ rule presentation
production practice + input practice]. Learning was assessed by means of
immediate and delayed tests of oral and written comprehension and produc-
tion. In addition to other measures of aptitude, Erlam administered a written
word recall test as a measure of WMC. Of the 90 (30  3 experimental groups)
correlations run between aptitude component scores and various gain scores,
nine were statistically significant. Of these, five related to performance on
written production post-tests, with four of these obtaining for the structured
input group and one obtaining for the inductive instruction group. To further
examine the clustering of significant correlations for the structured input
group, Erlam conducted stepwise regression, with results indicating that
scores on the Words in Sentences subtest of the MLAT accounted for nearly
24 per cent of variance in performance on the immediate written production
post-test, and the scores for the WM and Words in Sentences subtests in
combination accounted for nearly 48 per cent of variance in performance on
the delayed written production test. In her interpretation of this latter finding,
Erlam suggested that WMC played a role in the production performance spe-
cifically of the structured input group because their instruction did not involve
output practice. She also speculated that WMC played a role specifically in
delayed production because the longer time interval allowed for consolidation
of relevant knowledge. Regarding the lack of association of WMC and devel-
opment in the inductive group, but especially the deductive group, Erlam
concluded following Skehan (1989b) that the combination of rule presentation
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 673

and output practice in the latter group served to neutralize the effects of indi-
vidual differences in language aptitude. However, limitations of this study
suggest caution in interpreting its results. These limitations include the assess-
ment of WMC, as it lacks a processing component, though Erlam noted that
her WM measure is similar to one used by Baddeley (1999) and colleagues
(1975). Additionally, the use of multiple correlations to test for statistical

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


significance, while common in studies of WM in SLA, calls for extreme caution
in interpretation, including consideration of the likelihood that a lack of
significant correlation may result from small sample size (ns ranged from 19
to 21 in Erlam’s experimental groups) and/or low reliability of the measures
used (Juffs and Harrington 2011). Finally, Erlam’s design is characterized by
large differences in the pedagogical variables under study, namely, type of
practice (input vs. output), presence or absence of explicit instruction, and
the nature of this instruction (explanation alone in deductive instruction
but combined with ‘consciousness raising activities’ in structured input); all
conditions included feedback, although the amount and its nature were not
controlled for.
The results of other studies suggest that modality of aspects of instruction
may also interact with effects of WMC on learning. Sagarra and Abbuhl
(2013a) examined the role of WMC in the development of accuracy in
gender and number agreement in L2 Spanish. L1 English speakers interacted
with a computer-based fill-in-the-blank activity in which they were required
to supply correct forms of adjectives to complete Spanish sentences. During the
instructional treatment, experimental groups received one of six types of com-
puter-delivered feedback: written or oral utterance rejection, written or oral
unenhanced recasts, or written or oral enhanced (typographically or prosodi-
cally) recasts. A control group did not receive feedback. Analysis of groups’
performance on written and oral post-tests completed immediately and be-
tween one week and two months after treatment revealed that for those
who had received oral recasts (both enhanced and unenhanced), WMC was
significantly related to written and oral accuracy, and to the production of
target-like modified output during a face-to-face interactional post-test. In a
(2013b) follow-up study, the researchers found no similar role for WMC in
performance by groups who during the same instructional treatment received
written enhanced recasts, written unenhanced recasts, written unenhanced
recasts combined with a pre-task grammar explanation, and no feedback.
They attributed the lack of effects to the fact that in the follow-up study all
feedback had been provided in the written modality.
The effects of oral recasts were also found to be mediated by WMC in an
interaction study conducted by Goo (2012) on the acquisition of the English
that-trace filter. The 54 Korean EFL learners in this study were assigned to one
of three groups—oral recast, oral metalinguistic feedback, or control—and
completed two one-way information gap activities with their native-speaking
teacher. Pre- and immediate post-tests included a grammaticality judgment
task (GJT; in which participants had to correct items they judged as
674 LEARNING CONDITIONS AND WORKING MEMORY

ungrammatical) and a written production test. The results indicated that both
experimental groups performed similarly and better than the control group.
Additionally, WMC was found to mediate learning (on both measures) in the
oral recast group, but not in either the oral metalinguistic feedback group or
the control group. Limitations of this study include the small sample size of the
recast group (n = 14) and the lack of delayed post-tests.

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Li (2013) also compared a recast, a metalinguistic feedback, and a control
group in a study on the acquisition of an opaque linguistic structure (Chinese
perfective –le). Seventy-eight participants completed a video narration and an
interview task. Li investigated whether the effects of the feedback were
mediated by language analytic ability and WMC, measured respectively via
the Words in Sentences subtest of the MLAT and a listening span test. Results
indicated that in the metalinguistic feedback group, WMC correlated negatively
with performance on the delayed GJT, while language analytic ability was
positively correlated with the same test scores. Assuming that learners with
high language analytic ability were better at processing the metalinguistic in-
formation, Li suggested that learners with stronger WMC but weaker language
analytic skills were better at collecting language exemplars, but were not able
to process the metalinguistic information. In addition, a near-significant posi-
tive correlation between WMC and scores on the immediate GJT in the recast
group (r = 0.34, p = 0.07) led the author to conclude that recasts may have
benefited learners with greater WMC, which would coincide with results of
Goo (2012) and Sagarra and Abbhul (2013a).
Tagarelli et al. (2014) investigated the relationship between learning condi-
tions and WMC in a laboratory study contrasting incidental (n = 29) and in-
tentional (n = 26) conditions for learning a semi-artificial language with an
untrained control group (n = 15). In the treatment, both experimental groups
were asked to listen to 120 grammatical sentences (60 semantically plausible
and 60 implausible) with three specific verb-order rules. Whereas the inciden-
tal group was asked to repeat each sentence after a prompt and to judge
whether it was plausible or not, the intentional group was asked to discover
the rules that determined the word order of the sentences. The GJT tests con-
sisted of 30 grammatical and 30 ungrammatical plausible sentences that had
not been seen during treatment. WMC was measured by two complex span
tasks, an operation-word span task and a letter-number ordering task. The
results for the GJT indicated that both experimental groups outperformed
the control, and that the intentional group outperformed the incidental
group. Taking all participants together, results indicated no significant correl-
ation between accuracy on the GJT and performance on either WMC measure.
When analyzed by group and type of item, however, a positive correlation was
found between scores on both WM measures and accuracy on grammatical
items only for participants in the intentional group. Importantly, this study
did not include delayed tests, where WMC seemed to play a significant role
under incidental conditions in Robinson’s study.
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 675

Taken together, the results of these studies suggest that learning under con-
ditions in which participants are provided with grammar rules prior to or
during practice tends not to be affected by WMC as measured by word recall
or sentence span tests (as in Erlam, when explanation was combined with
production practice). Under these conditions, WMC may in fact be negatively
correlated with development, as in Li. However, under learning conditions

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


that do not provide explicit information on the targeted grammar (as in
Robinson’s incidental group, or the aural recast groups in Goo, in Sagarra
and Abbuhl, and in Li), WMC was positively associated with learning, espe-
cially on delayed tests when they were administered (Robinson 2005b).
Somehow contrary to these conclusions, however, Tagarelli and colleagues
suggest that merely asking learners to look for rules, as in their intentional
group, triggers the involvement of WMC. Additionally, the results for the
structured input in Erlam (2005) suggest that WM may have been relevant
in what the group had not practiced (i.e. production), which provides more
evidence to support the literature that has revealed how WM mediates
learners’ performance differently depending on the type of outcome measure
(Révész 2012).
Previous research isolated learning conditions that were variously cogni-
tively motivated (i.e. intentional or incidental learning, as in Robinson and
Tagarelli et al.) or pedagogically motivated (i.e. provision of corrective feed-
back, or inductive, deductive, or structured input instruction, as in Erlam, Goo,
Li, and Sagarra and Abbuhl) in relation to the learner’s cognitive capacity.
However, these conditions have combined a number of key variables whose
roles in L2 development still need to be isolated. These variables include input-
and/or output-based practice; presence or absence of rule explanation; type of
input (structured vs. non-structured); and type and modality of feedback. Our
goal is to add to this line of research with a more controlled, finer-grained view
that examines whether WMC moderates the effects of a traditional pedagogical
variable—presence or absence of a grammar explanation session prior to prac-
tice—while tightly controlling for all other variables, including feedback. The
general research question guiding this research is the following: What is the
role of WMC in the effects of learning conditions that combine input-based
practice with explicit feedback but differ in the presence or absence of a pre-
practice grammar lesson on the ab initio development of Latin morphosyntax?
Here we report the results from two independent, but related, experiments
designed to answer this question. In Experiment 1, the computer-delivered
treatment consists of a grammar lesson followed by input-based practice
with explicit feedback. In Experiment 2, the treatment is similar to
Experiment 1, minus the grammar lesson. WMC is operationalized as the abil-
ity to simultaneously process and store linguistic information and is measured
by a listening span task. Morphosyntactic development is operationalized as
increased ability to correctly assign semantic functions to noun phrases in
Latin, which involves the processing of case and agreement morphology.
676 LEARNING CONDITIONS AND WORKING MEMORY

EXPERIMENT 1
Method
The present study is one in a series conducted under the Latin Project paradigm
(Sanz et al. 2009; Lenet et al. 2010; Stafford et al. 2012; Lado et al. 2013; Sanz
et al. 2015) developed to study multilingual development in different popula-

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


tions under different learning contexts. The method described here follows
that of previous Latin Project research.

Participants
From an initial sample of 57 young adults, we included 23 who properly fol-
lowed instructions, and completed all tests and questionnaires with a pre-test
cutoff score of 10 or below out of 12. Participants were all native English
speakers, so that they could follow all instructions. For homogeneity, they
were all early L2 learners of the same language, Spanish, and had no prior
knowledge of case marking languages, as this may have provided an uncron-
trolled advantage at learning Latin. All participants were undergraduates en-
rolled in a second-semester Spanish course at a private university on the East
Coast of the United States; the age range was 18–32. Participants received a
combination of extra points in their Spanish classes and gift certificates for
campus stores to compensate them for their time.

Target structure
For our target, we chose the assignment of semantic functions to noun phrases,
that is, deciding ‘who did what to whom’ in Latin. To successfully assign se-
mantic functions in Latin, case morphology is the strongest cue—the most
available and reliable (Bates and MacWhinney 1982; see also MacWhinney
2005)—while verb agreement is less strong, as it is not always informative;
word order is the weakest of all three cues. The opposite applies to the par-
ticipants’ L1, English, while in the participants’ L2, Spanish, subject–verb
agreement is the strongest cue. Thus, native speakers of English are likely to
rely on word order (their L1’s strongest cue) in the early stages of L2 and L3
acquisition and must learn which cues are more informative for the target
language. All items in the practice and testing sessions consisted of sentences
with a transitive verb and two animate nouns (see examples below), such that
agent and patient roles could not be derived from semantics. The cues of word
order, agreement, and case were manipulated so that we could observe if and
when participants would come to rely on case morphology, the strongest cue
in Latin but unfamiliar to participants, rather than the other more familiar cues
from their L1 or L2. The following examples show the variation in the inform-
ativeness of these different potential processing cues for participants. We also
use these examples (with both English and Spanish translations given) to
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 677

illustrate that knowledge of Spanish would be of little advantage to learners at


the lexical and the morphosyntactic levels.
(1) Potentissimi salutant stultum.
king-nom.pl. greet-3rd pl. fool-acc.sing
The kings greet the fool.

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Los reyes saludan al bufón.

(2) Stultum salutant potentissimi.


fool-acc.sing. greet-3rd pl. king-nom.pl
The kings greet the fool.
Al bufón lo saludan los reyes.

(3) Stultum salutat potentissimus.


fool-acc.sing. greet-3rd sing king-nom.sing.
The king greets the fool.
Al bufón lo saluda el rey.
Sentences (1)–(3) above reflect the informativeness of subject–verb agree-
ment, word order, and case marking for L1 English learners of Latin. In (1), all
three cues point to the same interpretation, whereas in (2) and (3), learners
must rely on subject–verb agreement and/or case marking (2), or solely on
case marking (3) to assign agent and patient roles.

Treatment
All participants completed a computer-administered treatment that lasted ap-
proximately 70 minutes and consisted of (i) vocabulary presentation and test-
ing, (ii) a grammar lesson, and (iii) task-essential practice with explicit
feedback. The grammar lesson and practice share some characteristics with
Processing Instruction materials (VanPatten 2005); specifically, the lesson pro-
vides information on processing strategies, and practice is input-based with
manipulated input. It was purposefully designed to push the learners to rely
more on noun and verb morphology to assign semantic functions to noun
phrases (NPs) in a sentence and rely less on strategies based on word order
(typical of their L1 English processing).
The vocabulary lesson presented nominative, accusative, singular, and plural
forms of Latin nouns (n = 35) and infinitive forms of Latin verbs (n = 11).
Words were presented one at a time both aurally and visually, with each
word accompanied by its translation and a picture (see Figure 1 for an
678 LEARNING CONDITIONS AND WORKING MEMORY

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Figure 1: Vocabulary lesson example

example). The lesson did not label lexical forms according to function (i.e.
accusative or nominative). Immediately after the lesson participants completed
a vocabulary test. Those who scored 60 per cent or higher reviewed the missed
items and retook the test until they reached 100 per cent accuracy; those who
scored less than 60 per cent repeated the whole vocabulary lesson, then retook
the test until reaching 100 per cent accuracy.
An interactive grammar lesson followed. The lesson combined explanations
with examples (as in Figure 2). It started with an explanation of what case
morphology is and how it works, followed by practice with noun declen-
sions—singular and plural masculine subject and object markers first, then
feminine forms—in a multiple-choice format with feedback. The lesson then
advanced to a description of word order rules followed by presentation of
singular and plural verb endings. The session concluded with practice with
noun declensions and verb endings and a summary of the major points to
be remembered.
Finally, participants completed a practice section consisting of a series of
six tasks, three of which provided aural input, and three of which provided
written input; each task included 9 or 10 items. Note that all tasks were input-
based and so did not require production of the target form. In Tasks 1 and 4,
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 679

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Figure 2: Grammar lesson examples

participants chose (via key press) the correct English translation of the sen-
tence they read (1) or heard (4) from two possibilities, as in the following
example:
1 Procus basiat parvulam
(b) The suitor kisses the girl.
(c) The girl kisses the suitor.

In Tasks 2 and 5 participants read (2) or listened (5) to a sentence and chose
which of two pictures the sentence described. In Task 3 participants decided
whether or not a sentence they heard correctly described an onscreen picture,
and in Task 6 they decided which of two written sentences correctly described
a picture.
The practice was designed with two principles in mind. First, practice was to
be task-essential so both meaning and form had to be processed [i.e. ‘the task
cannot be successfully performed unless the structure is used’ (Loschky and
Bley-Vroman 1993: 132)]. Thus, for each item in the input-based practice, the
response options hinged on correctly interpreting the agent and patient of the
sentence, and participants could not rely solely on lexical knowledge to success-
fully perform the task. Secondly, case morphology was the only cue that was
always present and reliable for correctly assigning semantic functions. Thus,
sentences were designed so that reliance on subject–verb agreement and/or
SVO order was not sufficient for all items. To control for amount of feedback
provided to all participants, practice included feedback after every response, and
was positive or negative depending on the accuracy of the participant’s answer
(Figure 3); this feedback also included metalinguistic information.

Testing
In order to elicit a more complete picture of language development, the design
includes three tests—written interpretation, aural interpretation, and written
grammaticality judgment—that differ according to the requirements they
680 LEARNING CONDITIONS AND WORKING MEMORY

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Figure 3: Example of negative feedback during practice

make on processing mechanisms. Sanz (1997) argued for a differentiation


among tests along the lines of demands they place on WM based on modality
and amount of information conveyed. The aural interpretation test satisfies
differentiation by modality as time constraints put more demands on cognitive
resources compared with the written interpretation test. In contrast to
interpretation tests, the GJT requires exclusive attention to form.
Each test included 12 critical items and 8 distractors. For the interpretation
tests, critical items required interpretation of the relevant morphology to
decide which of two pictures correctly described the sentence they read/
heard. For critical interpretation test items, the two picture choices illustrated
the same event, but with the two nouns depicted in opposite agent and patient
roles (e.g. ‘the fool helps the kings’ and ‘the kings help the fool’). The distrac-
tors could be answered using lexical knowledge alone. In the grammaticality
judgment task, participants read a sentence and decided whether or not it was
a correct depiction of the activity, including the role of the NPs in the sentence.
Prior to each test there was a mini vocabulary review and quiz with relevant
lexical items. At each of the three testing sessions (pre-, post-, and delayed), we
administered one of three parallel versions of the battery of language tests,
randomized across participants, and with random order of test presentation.
Items within each test were randomized also. Finally, in contrast with the
binary-choice design of the practice task items, for every test item, a third ‘I
don’t know’ option was available. To score each test, one point was given for
each correct answer to the 12 critical items.

Experimental design
All treatments, tests (including the WM measures), and questionnaires were
delivered by a web-based computer application developed for The Latin Project
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 681

Table 1: Experimental design


1st session (pre-test)
1. Background questionnaire
2. WM test
3. Vocabulary lesson and test
4. Language pre-tests (each preceded by a vocabulary test)

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


2nd session (one week later: treatment and immediate post-test)
1. Treatment: [grammar lesson] + practice with explicit feedback
2. Language post-tests (each grammar test was preceded by a vocabulary
test)
3rd session (delayed post-test, two weeks later)
1. Delayed language post-tests (each preceded by a vocabulary test)
2. Debriefing questionnaire

that combined Flash and Cold Fusion programming tools. Responses were
automatically uploaded to a database. The use of multimedia facilitated deliv-
ery of simultaneous visual and aural input, thus making the lesson attractive
and accommodating to different learning styles (Lee and VanPatten 2003).

Working memory tests’ design, scoring, and analyses


We used a listening span test adapted from Waters and Caplan’s (1996) reading
version of the original Daneman and Carpenter (1980) test (see Williams 2012
for a review). The test evaluated the two functions posited to comprise work-
ing memory—processing and storage—and required participants to listen to
sets of three, four, or five sentences (four sets of each length for a total of 48
sentences). As they listened, participants made two judgments for each sen-
tence, one of its grammatical acceptability and one of its semantic congruity.
Ungrammatical sentences included grammar violations concerning subject–
verb agreement, verb tense, and/or article use, while semantically incongruous
sentences included a semantic anomaly (e.g. animacy violation). At the end of
each set of sentences, participants were asked to verbally recall the final word
(always a concrete noun) of each sentence in that set. To familiarize partici-
pants with the test’s format, the application delivered on-screen instructions
followed by practice items. Scoring was straightforward: one point was
awarded for each correct grammatical and semantic judgment and for each
correctly recalled final word. A composite WM score was computed with a
maximum score of 144 points (48 sentences  3 subscores). The mean score
for the WM test was 111.35 (STD = 12.44, range = 76–129).
682 LEARNING CONDITIONS AND WORKING MEMORY

Table 2: Experiment 1 language test scores (N = 23)


Min Max M STD

Written interpretation
Pre 4 10 7.17 1.825
Post 3 12 10.17 1.946

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Delayed 5 12 9.61 1.877
Aural interpretation
Pre 3 10 6.26 1.864
Post 5 12 9.87 1.817
Delayed 4 12 8.91 1.975
GJT
Pre 0 10 5.09 2.109
Post 4 12 9.57 2.293
Delayed 1 12 7.04 2.654

Table 3: Experiment 1 Pearson’s correlations between scores of WM test and


gain scores on the three language tests (N = 23)
Pre-post r (p value) Pre-delayed r (p value)

Written interpretation .109 (.621) .109 (.622)


Aural interpretation .279 (.197) .012 (.957)
GJT .159 (.468) .179 (.414)

Results
Separate repeated-measures ANOVAs on raw scores from each of the three
language tests yielded significant main effects for time [written interpretation
F(2,44) = 29.08, p < .01, aural interpretation F(2,44) = 33.83, p < .01, and GJT
F(2,44) = 31.98, p < .01]. Comparisons between pre-test scores2 and delayed-
test scores showed learning was maintained two weeks after the treatment,
with significant results for all three measures [written interpretation
F(1,22) = 37.66, p < .01, aural interpretation F(1,22) = 27.98, p < .01, and GJT
F(1,22) = 12.50, p < .01]. (See Table 2 for a summary of raw scores).
Following previous studies (Ando et al. 1992; Harrington and Sawyer 1992;
Juffs 2004; Erlam 2005; Robinson 2005b), we ran correlations between
composite WM scores and language test gain scores (results are reported in
Table 3). Correlational analyses between language test gain scores (pre-post,
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 683

and pre-delayed) computed for each of the three language tests and composite
WM scores on the WM test did not yield any statistically significant
relationship.

EXPERIMENT 2

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Motivated by Robinson (2005b), Erlam (2005), Sagarra and Abbuhl (2013a,
2013b), Goo (2012), and Li (2013), who suggested that explicit conditions
could diminish the role of individual differences in language development,
we conducted a second experiment with a comparable sample of participants
but a different treatment to compare the role of WMC on language develop-
ment under learning conditions that manipulated the timing of provision of
explicit information. The second treatment consisted of the same task-essential
practice with explicit feedback, but included no pre-practice grammar lesson.
In order to match time on task in the absence of the grammar lesson, the
practice session was repeated. In designing this second experiment, we strived
to avoid the pitfalls of previous research on effects of more and less explicit
instruction while maintaining comparability with the first experiment (see
Norris and Ortega 2000 for a description of the problem and Stafford et al.
2012 for a solution parallel to ours). Conditions differed only in the nature
and timing of the information presented. Specifically, in Experiment 2 partici-
pants were exposed to more Latin exemplars and more instances of metalan-
guage, but the metalinguistic information was presented piecemeal and was
therefore less structured than in the pre-practice grammar explanation in
Experiment 1. Also, metalinguistic information was provided prior to practice
as well as reactively in Experiment 1, but only reactively in Experiment 2.
The original tests were maintained intact in number and presentation of
items, and the critical and distractor items were the same, but two changes
were made. As in the previous study, participants could choose between a, b
and c, where a and b were pictures or sentences and c was ‘neither a nor b’.
Upon follow-up exchanges with participants in Experiment 1, the wording for
c was changed in Experiment 2 to ‘I don’t know’.
Secondly, with the intention of getting an even better picture of language
development, we incorporated a production test that would demand transfer
of skills given that the treatment was exclusively input-based. In the written
sentence production task, participants saw a picture on the screen and were
asked to form a sentence that correctly described the picture by putting to-
gether noun and verb roots and endings that were provided to the learner. The
roots provided always corresponded to the two nouns and verb depicted in the
image, and the endings were the full complement of case and verb morphemes
as described above. The sentence production test included 10 critical items;
each answer was awarded 1, 2, or 3 points depending on the number of ac-
curate endings: 1 point for a correct verb ending (number) and 2 points for
each correct noun ending (both case and number had to be correct, otherwise
0 points were awarded for the noun endings). All other aspects of the
684 LEARNING CONDITIONS AND WORKING MEMORY

procedure remained unaltered from Experiment 1. Test effects would have


been unexpected even for the production test given that this is a study of ab
initio language development.

Method
Participants

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Every effort was made to follow the same procedures and make this sample as
comparable as possible in every way (age, language background, student
status, cut-off scores) to that of Experiment 1. Of an original pool of 71 par-
ticipants we included 49; however, due to a computer glitch, WM data were
missing for some participants, leaving 21 participants in the final sample. The
mean WM score for participants in Experiment 2 was 102 (STD = 16.17,
range = 55–125), or less than 10 points below the mean for participants in
Experiment 1 and with larger variability (mean = 111.35: STD 12.44;
range = 76–129). We used a one-way ANOVA to compare WM scores from
the two experiments, and there was a significant difference between groups,
F(1,42) = 4.48, p < .05. However, correlations depend on position of scores, in
this case WM and gain scores, relative to each other, rather than absolute value
or range. In addition, comparisons between pretest scores on the three Latin
tests yielded no significant between-groups differences [written interpretation:
F(1,42) = 3.89, p = .055; aural interpretation: F(1,42) = 2.33, p = .134; GJT:
F(1,42) = 1.05, p = .311]. Thus, it is still experimentally valid to investigate cor-
relations for both groups, although they have differing WM mean scores.

Results
The descriptive statistics for the language tests are presented in Table 4.
Separate repeated-measures ANOVAs on raw scores from each of the four
language tests yielded significant main effects for time [written interpretation,
F(2,38) = 36.12, p < .01; aural interpretation, F(2,40) = 39.76, p < .01; GJT,
F(2,40) = 6.91, p < .01; written production, F(2,38) = 19.03, p < .01].
Comparisons between pre-test scores and delayed test scores showed that
learning was maintained two weeks after the treatment, with significant re-
sults for three of the four measures [written interpretation, F(1,20) = 15.16,
p < .01; aural interpretation, F(1,20) = 39.03, p < .01; written production,
F(1,19) = 20.07, p < .01]. On the contrary, the GJT yielded no significant
results, F(1,20) = 1.33, p = .26. These results confirm that a condition charac-
terized by input-based, task-essential practice with explicit feedback is enough
to promote language development, as shown by significant increases in ability
to interpret and produce the target form, and does not require additional
explicit discussion of grammar prior to practice.
To investigate the relationship between WMC and language development,
correlational analyses were run on gain scores (pre-post and pre-delayed)
computed for each of the four language tests and composite scores on the
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 685

Table 4: Experiment 2 language test scores


Min Max M STD

Written interpretation
Pre 3 9 6.14 1.621
Post 9 12 10.65 1.040

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Delayed 4 12 8.71 2.533
Aural interpretation
Pre 2 8 5.48 1.504
Post 4 12 9.67 2.477
Delayed 5 12 9.10 2.234
GJT
Pre 0 9 5.76 2.256
Post 5 10 7.76 1.729
Delayed 2 10 6.52 2.421
Written production
Pre 9 28 15.81 3.894
Post 11 30 22.71 6.627
Delayed 8 30 21.10 6.340

Note: The maximum possible score is 12 for the interpretation and grammaticality judgment tests.
The maximum possible score is 30 for the written production test.

Table 5: Experiment 2 Pearson’s correlations between scores of WM test and


gain scores on the four language tests (N = 21)
Pre-post r (p value) Pre-delayed r (p value)

Written interpretation .252 (.284) .483 (.027)*


Aural interpretation .440 (.046)* .171 (.459)
GJT .176 (.445) .284 (.213)
Written production .256 (.262) .033 (.886)

Note: N = 20 on the delayed test of written production.


*p < .05

WM test). Results show significant correlations between WM scores and pre-


post gain scores from the aural interpretation test [r(21) = .441, p < .05] and
the pre-delayed gain scores from the written interpretation test [r (21) = 483,
p < .05]. See Table 5 for a summary of the correlation results.
686 LEARNING CONDITIONS AND WORKING MEMORY

Taken together, the results from Experiments 1 and 2 show a complex pic-
ture of the relationship between WMC and pedagogical conditions. WMC is
significantly and positively correlated with participants’ ab initio Latin learning
under certain conditions but not others. More specifically, when learners are
directly exposed to task-essential practice with manipulated input and meta-
linguistic feedback, but no explicit grammar lesson prior to practice, WMC is

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


positively associated with increases in participants’ ability to interpret semantic
functions of noun phrases in aural and written Latin sentences via noun and
verb morphology. However, when providing learners with a pre-practice
grammar lesson, WMC does not predict language development.

DISCUSSION
The current study examined the role of WMC in the ab initio development of a
non-primary language under learning conditions that differed in the structure
and timing of linguistic information that was provided (whether or not a
structured, explicit pre-practice lesson was provided vs. undergoing practice
with piecemeal metalinguistic feedback twice). In line with results of previous
studies (Robinson 2005b; Erlam 2005; Goo 2012; Li 2013; Sagarra and Abbuhl
2013a, 2013b), our analyses identified positive correlations between WMC and
areas of language development under one condition but not the other.
Crucially, the relationships identified were limited to the less structured and
explicit pedagogical treatment, where more new incoming information had to
be processed without the benefit of an advanced organizer, thus supporting
Skehan’s (2002) view of aptitude as a variable contributing to differential suc-
cess, especially under conditions that provide less scaffolding and leave lear-
ners more to their own devices.
Our study adds depth to the growing literature on the topic of aptitude-by-
treatment interactions by focusing on two learning conditions that varied ex-
clusively in the presence or absence of a grammar lesson prior to practice. Both
conditions included input-based practice that required attention to both form
and meaning and provided explicit feedback, and both were effective in driv-
ing development. As the results from Experiment 1 suggest, the provision of
grammar explanation prior to practice largely cancelled out any advantage for
high-capacity WMC learners, while in Experiment 2, the results from the aural
and written interpretation tests indicate that in the absence of a grammar
lesson, the higher the learner’s WMC, the more he/she benefitted from prac-
tice of this nature, at least in the context of performance on interpretation
tests. Interestingly, this advantage for higher WMC learners emerged later, in
performance on the written interpretation test two weeks after the treatment,
a result that somewhat aligns with the findings for the structured input group
in Erlam’s (2005) classroom study, where she found significant correlations
between WMC and gain scores in delayed production measures. Similarly, the
relationship between WMC and increased ability to process aural input in our
study aligns with Robinson’s (2005b) results.
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 687

Different patterns of correlations were found for the four language tests,
showing that an investigation of the role of individual differences in moderat-
ing the effects of different pedagogical techniques needs to carefully consider
the nature of the tests employed to measure language development. In this, we
agree with Révész’s arguments in a recent study (Révész 2012), which sug-
gested that the observed effectiveness of recasts is influenced by the type of

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


outcome measure used, and that different aspects of WMC are differentially
associated with learners’ performance on various outcome measures. Counter
to Erlam’s study, who found WM scores to be correlated with gain scores on
production tests for the structured input group (which had not undergone
production practice), in the present study WMC appeared to be related to
growth, at least for this target and this level, on those tasks that aligned
with practice, namely, the aural and written interpretation tasks. A different
picture emerges in our study from tasks that demand transfer of skills from
input-based practice to production, or when, as in the GJT, the task requires a
focus on form rather than processing form/meaning connections, despite sig-
nificant improvements in accuracy across the sample in the GJT and written
production tasks.
An even more complex picture appears when we turn our attention from
the measurement tools to the learning conditions. In most SLA research in this
vein, learning conditions are operationalized as provision of grammar instruc-
tion at different times and in different amounts; that is before and/or during
instruction. These studies (Li, Goo, Sagarra and Abbuhl, Erlam) coincide with
ours in showing that more explicit instruction levels the playing field for
learners of different cognitive capacities. In contrast, Tagarelli et al.’s experi-
ment manipulates task requirements, specifically focus on form (operationa-
lized as instructions to find the rule and labeled intentional) vs. focus on
meaning (operationalized as instructions to memorize sentence stimuli and
labeled incidental). They find that WMC plays a role in predicting development
only for the rule-search group, which showed significantly more learning than
the incidental group. However, Tagarelli et al.’s intentional group was exposed
to input under conditions that included neither feedback nor explicit rule
presentation. In this sense, Tagarelli et al.’s results do not depart from results
from pedagogically motivated research that includes ‘more implicit’ conditions,
that is, no rule presentation prior to or during practice, conditions in which
learners are pushed to rely on their own devices to process new, incoming
information without the benefits of an advanced organizer. Even in those cases
when instruction is provided, as in our Experiments 1 and 2, the timing and
nature rather than the amount of instruction may be key: as pointed out by
one of our reviewers, it makes sense that WMC, which reflects learners’ ability
to process and store linguistic information, plays a role when instruction is
provided as part of feedback in a piecemeal fashion rather than in bulk prior to
practice.
Results of our two experiments together with findings from others that look
at provision of grammatical information prior to or during practice suggest that
688 LEARNING CONDITIONS AND WORKING MEMORY

more grammar information, especially when provided prior to practice, de-


creases the chances for learners with higher WMC to outperform their coun-
terparts, thus identifying WMC, operationalized as the ability to
simultaneously process and store linguistic information, as a key factor asso-
ciated with success in less explicit language learning conditions. Vatz et al.
(2013) call for more research like the present study that leads to pedagogical

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


implications, as they outline a number of them. Beyond the implications Vatz
and colleagues propose, it is possible to think of numerous pedagogical deci-
sions that could be guided by research results such as ours. For example, for
students similar to our participants, assigning homework that provides a struc-
tured grammar lesson prior to practice may benefit students with lower WMC
who might otherwise struggle with explicit input that may move their atten-
tion away from processing for meaning during practice. In this vein, a recent
Latin project study (Cox 2013) suggests that, given a negative relationship
between aging and WMC, a teacher working with older learners may optimize
learning by offering a traditional grammar lesson that acts as an advanced
organizer and avoid explicit feedback during practice, as this places a heavy
load on WMC.

CONCLUSION
Of utmost interest to researchers and practitioners alike, the question we posed
at the outset of this study is not simply whether WMC and language develop-
ment are related, or whether a variety of external learning conditions differ-
entially affect language development. Rather, we were interested in
investigating the idea that individual differences such as WMC are involved
in how learners approach learning under different conditions and in consider-
ing the potential pedagogical implications of the finding that certain conditions
favor particular kinds of learners or equalize instruction across types of lear-
ners. This study contributes to a picture that is more complex and less clear-cut
than perhaps expected.
In line with previous pedagogically motivated research, we conclude that
the more traditional pedagogical approach characterized by grammar explana-
tion, followed by practice with explicit feedback levels the field for all learners,
reducing differences that may result from the cognitive advantage of higher
WMC. In contrast, WMC plays a role in explaining variation in success at ab
initio language development when grammar instruction is limited to feedback.
We add to previous literature, and especially to the results of Erlam’s (2005)
and Robinson’s (2005b) studies, by extending their results to ab initio devel-
opment of semantic role assignment promoted by two learning conditions—
with or without a grammar lesson prior to meaningful practice with the lan-
guage—and the relationship between development and WMC as measured by
a listening span test.
Our conclusions are limited to the ab initio development of Latin morpho-
syntax. As part of our long-term investigation (Lenet et al. 2010; Stafford et al.
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 689

2012; Lado et al. 2013), we are looking into alternative measures of WMC and
a wider range of pedagogical conditions (presence vs. absence of feedback, type
of feedback, and type of practice) with the goal of taking a broader and more
in-depth view of the involvement of WMC in the learning and retention of
learning resulting from different types of pedagogical interventions.

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


ACKNOWLEDGEMENTS
The study is part of The Latin Project, developed by Cristina Sanz, Catherine Stafford, and Harriet
Bowden; funding to Sanz from the Spencer Foundation and Georgetown’s Graduate School made
it possible. We would like to thank our three reviewers for their insightful and painstakingly
detailed comments on earlier versions of this manuscript. Any remaining errors are exclusively
ours.

NOTES
1 Ab initio means in the absence of pre- presented in SVO word order,
vious knowledge, as learners had never which may have aided educated gues-
been in contact with Latin. sing on the pre-test, thereby inflating
2 Although percentages may seem high, scores.
half of the test sentences were

REFERENCES
Abu-Rabia, S. 2003. ‘The influence of working Baddeley, A., N. Thompson, and
memory on reading and creative writing pro- M. Buchanan. 1975. ‘Word length and the
cesses in second language,’ Educational structure of short-term memory,’ Journal of
Psychology 23: 210–22. Verbal Learning and Verbal Behaviour 14:
Ando, J., N. Fukunaga, J. Kurahachi, 575–89.
T. Suto, T. Nakano, and M. Kage. 1992. ‘A Bates, E. and B. MacWhinney. 1982.
comparative study on the two EFL teaching ‘Functionalist approaches to grammar’
methods: The communicative and the gram- in E. Wanner and L. R. Gleitman (eds):
matical approach,’ Japanese Journal of Language Acquisition: The State of the Art.
Educational Psychology 40: 247–56. Cambridge University Press, pp. 173–218.
Baddeley, A. 1999. Essentials of Human Memory. Carroll, J. B. 1958. ‘A factor analysis of two for-
Psychology Press. eign language aptitude batteries,’ Journal of
Baddeley, A. 2000. ‘The episodic buffer: A new General Psychology 59/1: 3–19.
component of working memory?,’ Trends in Carroll, J. B. and S. M. Sapon. 1959. Modern
Cognitive Sciences 4/11: 417–23 Language Aptitude Test. The Psychological
Baddeley, A. 2007. ‘Working memory: Multiple Corporation/Harcourt Brace Jovanovich.
models, multiple mechanisms’ in H. Cox, J. 2013. ‘Bilingualism, Aging, and
L. Roediger, Y. Dudai, and S. M. Fitzpatrick Instructional Conditions in Non-Primary
(eds): Science of Memory: Concepts. Oxford Language Development,’ Unpublished PhD
University Press, pp. 151–3. Dissertation, Georgetown University.
Baddeley, A. and G. Hitch. 1974. ‘Working Daneman, M. and P. A. Carpenter. 1980.
memory’ in G. H. Bower (ed.): The Psychology of ‘Individual differences in working memory
Learning and Motivation: Vol. 8. Advances in Research and reading,’ Journal of Individual Learning
and Theory. Academic Press, pp. 47–90. and Verbal Behavior 19: 450–66.
690 LEARNING CONDITIONS AND WORKING MEMORY

DeKeyser, R. 2012. ‘Interactions between indi- Language Learning: Conditions, Processes, and
vidual differences, treatments and structures in Knowledge in SLA and Bilingualism.
SLA,’ Language Learning 62/(Suppl. 2): Georgetown University Press, pp. 73–84.
189–200. Li, S. 2013. ‘The differential roles of two aptitude
De Graaff, R. 1997. ‘The experanto experiment: components in mediating the effects of two
Effects of explicit instruction on second lan- types of feedback on the acquisition of an
guage acquisition,’ Studies in Second Language opaque linguistic structure’ in C. Sanz and

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


Acquisition 19/2: 249–76. B. Lado (eds): Individual Differences, L2
Dörnyei, Z. and P. Skehan. 2003. ‘Individual Development & Language Program Administration:
differences in second language learning’ From Theory to Application. Cengage Learning,
in C. Doughty and M. Long (eds): The pp. 32–52.
Handbook of SLA. Blackwell, pp. 589–630. Loschky, L. and R. Bley-Vroman. 1993.
Erlam, R. 2005. ‘Language aptitude and its rela- ‘Grammar and task-based methodology’
tionship to instructional effectiveness in second in G. Crookes and S. Gass (eds): Tasks and
language acquisition,’ Language Teaching Language Learning: Integrating Theory and
Research 9/2: 147–71. Practice. vol 1. Multilingual Matters,
Goo, J. 2012. ‘Corrective feedback and working pp. 123–67.
memory capacity in interaction-driven L2 Mackey, A. and R. Sachs. 2012. ‘Older learners
learning,’ Studies in Second Language Acquisition in SLA research: A first look at working
34: 445–74. memory, feedback, and L2 development,’
Harrington, M. and M. Sawyer. 1992. ‘L2 Language Learning 62: 704–40.
WMC and L2 reading skill,’ Studies in Second Mackey, A., R. Adams, C. A. Stafford, and
Language Acquisition 14/1: 25–38. P. Winke. 2010. ‘Exploring the relationship
Hummel, K. 2009. ‘Aptitude, phonological between modified output and working
memory, and second language proficiency in
memory capacity,’ Language Learning 60:
nonnovice adult learners,’ Applied
501–33.
Psycholinguistics 30: 225–49.
Mackey, A., J. Philp, A. Fujii, T. Egi, and
Juffs, A. 2004. ‘Representation, processing and
T. Tatsumi. 2002. ‘Individual differences in
working memory in a second language,’
working memory, noticing of interactional
Transactions of the Philological Society 102/2:
feedback, and L2 development’
199–225.
in P. Robinson and P. Skehan (eds):
Juffs, A. and M. Harrington. 2011. ‘Aspects of
Individual Differences in L2 Learning.
working memory in L2 learning,’ Language
Benjamins, pp. 181–208.
Teaching 44: 137–66.
MacWhinney, B. 2005. ‘New directions in the
Krashen, S. D. 1981. ‘Aptitude and attitude in
Competition Model’ in M. Tomasello and D.
relation to second language acquisition and
I. Slobin (eds): Beyond Nature-Nurture: Essays
learning’ in K. C. Diller (ed.): Individual
in Honor of Elizabeth Bates. Lawrence Erlbaum
Differences & Universals in Language Learning.
Associates, pp. 81–110.
Newbury House, pp. 155–75.
Miyake, A. and N. F. Friedman. 1998.
Krashen, S. D. 1982. Principles and Practice in
‘Individual differences in second language pro-
Second Language Acquisition. Prentice-Hall.
Lado, B., H. W. Bowden, C. A. Stafford, and ficiency: Working memory as language apti-
C. Sanz. 2014. ‘A fine-grained analysis of the tude’ in A. F. Healy and L. E. Bourne (eds):
effects of more vs. less explicit feedback on Foreign Language Learning: Psycholinguistic
language development,’ Language Teaching Studies on Training and Retention. Erlbaum,
Research 18/3: 320–44 pp. 339–64.
Lee, J. and B. VanPatten. 2003. Making Norris, J. M. and L. Ortega. 2000. ‘Effectiveness
Communicative Language Happen. McGraw Hill. of L2 instruction: A research synthesis and
Lenet, A. E., C. Sanz, B. Lado, quantitative meta-analysis,’ Language Learning
J. H. Howard Jr, and D. V. Howard. 2010. 50/3: 417–528.
‘Aging, pedagogical conditions, and differential Reber, A. S. 1989. ‘Implicit learning and tacit
success in SLA: An empirical study’ in C. Sanz knowledge,’ Journal of Experimental Psychology
and R. P. Leow (eds): Implicit and Explicit 118/3: 219–35.
C. SANZ, H.-J. LIN, B. LADO, C. A. STAFFORD AND H. W. BOWDEN 691

Reber, A. S. 1993. Implicit Learning and Tacit language learning,’ The Modern Language
Knowledge: An Essay on the Cognitive Journal 97: 196–216.
Unconscious. vol. 19. Oxford University Press. Sagarra, N. and R. Abbuhl. 2013b. ‘Computer-
Reves, T. 1983. ‘What makes a good language delivered feedback and L2 development: The
learner? Personal characteristics contributing role of explicitness and working memory’
to successful language acquisition’. in C. Sanz and B. Lado (eds): Individual
Unpublished Ph.D. dissertation, Hebrew Differences, L2 Development, & Language Program

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


University. Administration: From Theory to Application.
Révész, A. 2012. ‘Working memory and the Cengage Learning, pp. 53–70.
observed effectiveness of recasts on different Sagarra, N. and J. Herschensohn. 2010. ‘The
L2 outcome measures,’ Language Learning 62: role of proficiency and working memory in
93–132. gender and number agreement processing in
Robinson, P. 1997. ‘Individual differences and L1 and L2 Spanish,’ Lingua 120/8: 2022–39.
the fundamental similarity of implicit and ex- Sanz, C. 1997. ‘Experimental tasks in SLA re-
plicit adult second language learning,’ Language search: Amount of production, modality,
Learning 47/1: 45–99. memory, and production processes’ in W.
Robinson, P. 2002. ‘Effects of individual differ- R. Glass and A. T. Pérez-Lerroux (eds):
ences in intelligence, aptitude and working Contemporary Perspectives on the Acquisition of
memory on incidental SLA: A replication and Spanish. vol. 2. Cascadilla Press, pp. 41–56.
extension of Reber, Walkenfield and Hernstadt Sanz, C. 2005. ‘Adult SLA: The interaction be-
(1991)’ in P. Robinson (ed.): Individual tween internal and external factors’ in C. Sanz
Differences and Instructed Language Learning. (ed.): Mind and Context in Adult Second Language
John Benjamins, pp. 211–66. Acquisition. Georgetown University Press,
Robinson, P. 2005a. ‘Aptitude and second lan- pp. 3–20.
guage acquisition,’ Annual Review of Applied Sanz, C., H. J. Lin, B. Lado, H. W. Bowden,
Linguistics 25: 45–73. and C. A. Stafford. 2009. ‘Concurrent verbal-
Robinson, P. 2005b. ‘Cognitive abilities, chunk- izations, pedagogical conditions, and reactivity:
strength, and frequency effects in implicit arti- Two CALL studies,’ Language Learning 59/1:
ficial grammar and incidental L2 learning: 33–71.
Replications of Reber, Walkenfeld, and Sanz, C., H. I. Park, and B. Lado. 2015. ‘A
Hernstadt (1991) and Knowlton and Squire functional approach to cross-linguistic influ-
(1996) and their relevance for SLA,’ Studies in ence in ab initio L3 acquisition,’ Bilingualism:
Second Language Acquisition 27: 235–68. Language and Cognition 18/2: 236–51.
Robinson, P. 2007. ‘Aptitude, abilities, contexts, Skehan, P. 1989a. Individual Differences in Second
and practice’ in R. M. DeKeyser (ed.): Practice Language Learning. Edward Arnold.
in a Second Language: Perspectives from Applied Skehan, P. 1989b. ‘Individual differences in
Linguistics and Cognitive Psychology. Cambridge second language learning,’ Studies in Second
University Press. Language Acquisition 13: 275–98.
Sagarra, N. 2000. ‘The longitudinal role of work- Skehan, P. 2002. ‘Theorising and updating apti-
ing memory on adult acquisition of L2 gram- tude’ in P. Robinson (ed.): Individual Differences
mar,’ Unpublished doctoral dissertation, in Instructed Language Learning. John
University of Illinois at Urbana-Champaign. Benjamins, pp. 69–93.
Sagarra, N. 2007. ‘From CALL to face-to-face Stafford, C. A., H. W. Bowden, and C. Sanz.
interaction: The effect of computer-delivered 2012. ‘Optimizing language instruction:
recasts and working memory on L2 develop- Matters of explicitness, practice, and cue learn-
ment’ in A. Mackey (ed.): Conversational ing,’ Language Learning 62/3: 741–68.
Interaction in Second Language Acquisition: A Tagarelli, K. M., M. Borges Mota, and
Series of Empirical Studies. Oxford University P. Rebuschat. 2014. ‘Working memory,
Press, pp. 229–76. learning conditions, and the acquisition of L2
Sagarra, N. and R. Abbuhl. 2013a. ‘Optimizing Syntax’ in W. Zhisheng, M. Borges Mota, and
the noticing of recasts vis computer-delivered A. McNeill (eds): Working Memory in Second
feedback: Evidence that oral input enhance- Language Acquisition and Processing: Theory,
ment and working memory help second Research and Commentary. Multilingual Matters.
692 LEARNING CONDITIONS AND WORKING MEMORY

Trofimovich, P., A. Ammar, and Waters, G. S. and D. Caplan. 1996. ‘The


E. Gatbonton. 2007. ‘How effective are re- measurement of verbal WMC and its relation
casts? The role of attention, memory, and ana- to reading comprehension,’ Quarterly Journal of
lytical ability’ in A. Mackey (ed.): Experimental Psychology, Section A: Human
Conversational Interaction in Second Language Experimental Psychology 49A/1: 51–79.
Acquisition: A Series of Empirical Studies. Oxford Wesche, M. 1981. ‘Language aptitude measures in
University Press, pp. 171–95. streaming, matching students with methods,

Downloaded from https://academic.oup.com/applij/article/37/5/669/2196675 by NILE user on 11 October 2023


VanPatten, B. 2005. ‘Processing instruction’ and diagnosis of learning problems’ in K. Diller
in C. Sanz (ed.): Mind and Context in Adult (ed.): Individual Differences and Universals
Second Language Acquisition: Methods, Theory, in Language Aptitude. Newbury House,
and Practice. Georgetown University Press, pp. 119–54.
pp. 267–81. Williams, J. N. 2012. ‘Working memory and
Vatz, K., M. Tare, S. Jackson, and SLA’ in S. Gass and A. Mackey (eds):
C. Doughty. 2013. ‘Aptitude-treatment inter- Handbook of Second Language Acquisition.
action studies in second language acquisition: Routledge, pp. 427–41.
Findings and methodology’ in G. Granena and Winke, P. 2013. ‘An investigation into second
M. H. Long (eds): Sensitive Periods, Language language aptitude for advanced Chinese lan-
Aptitude, and Ultimate L2 Attainment. John guage learning,’ Modern Language Journal 97/
Benjamin, pp. 273–92. 1: 109–30.

You might also like