Professional Documents
Culture Documents
com/
Teaching Research
Published by:
http://www.sagepublications.com
Additional services and information for Language Teaching Research can be found at:
Subscriptions: http://ltr.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://ltr.sagepub.com/content/18/3/294.refs.html
What is This?
LANGUAGE
TEACHING
Article RESEARCH
Aptitude-treatment interaction
2014, Vol. 18(3) 294319
The Author(s) 2013
Reprints and permissions:
effects on explicit rule learning: sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1362168813510381
A latent growth curve analysis ltr.sagepub.com
Fenfang Hwu
University of Cincinnati, USA
Wei Pan
Duke University, USA
Shuyan Sun
University of Maryland at Baltimore County, USA
Abstract
Finding the match between individuals and educational treatments is the aim of both educators
and the aptitude-treatment interaction research paradigm. Using the latent growth curve analysis,
the present study investigates the interaction between the type of explicit instructional approaches
(deductive vs. explicit-inductive) and the level of foreign language aptitude (high vs. low) in the learning
of explicit grammar rules. The results indicate that on the whole the two equally explicit instructional
approaches did not differentially affect learning performance. However, when the level of language
aptitude, measured by grammatical sensitivity, associative memory, and memory for text (with the
last variable being the best measure), was taken into account, low-aptitude learners performed
significantly better with the deductive instruction, in the sentence-correction tests. The interaction
effects of equally explicit instructional approaches suggest the need for considering aptitude-treatment
interaction to maximize learners potential for success in second language learning.
Keywords
Aptitude-treatment interaction, deduction, domain knowledge, grammar, induction, language aptitude
IIntroduction
The question of how second language (L2) grammar is best learned has generated a lot of
debate among researchers in the fields of second language acquisition (SLA) research and
Corresponding author:
Fenfang Hwu, Department of Romance Languages and Literatures, PO Box 210377, University of Cincinnati,
Cincinnati, OH 45221, USA.
Email: fenfang.hwu@uc.edu
applied linguistics. There is now a considerable amount of evidence that lends support to
explicit types of instruction as opposed to implicit types of instruction (e.g. de Graaff,
1997; DeKeyser, 1995; Doughty, 1991; Reber, Kassin, Lewis, & Cantor, 1980; Robinson,
1997; Scott, 1989; for meta-analyses of the effects of instructed SLA, see Norris & Ortega,
2000, 2001). Additionally, evidence supporting a positive relationship between the extent
of explicitness of an instructional condition and learner performance has also been gath-
ered (Leow, 1998; Robinson, 1997; Rosa & Leow, 2004; Rosa & ONeill, 1999).
Although a considerable body of work has accumulated in support of the positive
effect of explicit types of instruction on L2 acquisition, little is yet known concerning
how individual differences (IDs) at the level of language aptitude interact with various
equally explicit learning conditions to affect the success of the learning of grammar
rules. Almost any L2 teacher or researcher will agree that individual learners differ in
their readiness to benefit from a particular instructional approach and the appropriate
combination of instructional approaches matched with the appropriate learners is what
really produces success (Skehan, 1989). For that reason, whether independent groups of
learners, who share the same aptitude score profile, will differentially benefit from dif-
ferent but equally explicit instructional approaches is a question of great significance for
both L2 teachers and researchers. Answers to this question can shed light on two impor-
tant questions in SLA. First, how can form-focused instruction be used to match learner
characteristics with instructional characteristics (Robinson, 2001; Sawyer & Ranta,
2001)? Second, how do we account for variation in language learning success under
particular instructional conditions (Robinson, 2001; see DeKeyser, 2009)?
The present study investigates the effects of two equally explicit instructional condi-
tions, deduction and an explicit type of induction, hereafter explicit-induction, on the
success of learning the construction with the Spanish psych verb gustar (to like; to be
pleasing), known for being difficult to acquire by native English-speaking learners
(Gascn, 1998; Lpez Jimnez, 2003; Marras & Cadierno, 2008; Montrul, 1997;
VanPatten, 1986; Zyzik, 2006). The study hypothesized that the effectiveness of deduc-
tion and explicit-induction would vary across the level of language aptitude. Specifically,
learners who share the same aptitude score profile would differentially benefit from
equally explicit instructional approaches.
induction means that examples are encountered before rules are inferred (DeKeyser,
1995). Furthermore, deduction is the process that goes from consciously formulated
rules to the application in language use, while induction is the process that involves real
language use, from which patterns and generalizations emerge (Decoo, 1996). While
deductive learning is inevitably explicit, with concurrent awareness of what is being
learned, induction can be either implicit or explicit (DeKeyser, 1995). The debate over
the superiority of one technique as opposed to the other has a long history in L2 teaching
(Erlam, 2003; Fischer, 1979; Haight, Harron, & Cole, 2007, Hammerly, 1975; Herron &
Tomasello, 1992; Seliger, 1975; Shaffer, 1989). Further complicating the debate is that
the techniques of deduction and induction used in L2 teaching are not uniform (Decoo,
1996). While the didactic strategies of deduction may vary in the degree of explicitness
or elaborateness (DeKeyser, 1995), induction has taken on many forms. Decoo (1996)
identified at least four types of induction:
He further maintained that each of these four modalities of induction and even the deduc-
tion could be further refined or combined into subtypes.
Given the diverse ways that induction and deduction are used in L2 teaching (Decoo,
1996), the diversity of research design with regard to treatment conditions comparing
deduction and induction (Haight et al., 2007), and the different ways that explicit instruc-
tions, including deduction and induction, are operationalized in research studies (Norris &
Ortega, 2000), it is not surprising that the results of studies comparing the effectiveness of
deduction and induction to date have been mixed. For example, three studies, which uti-
lized only immediate measurements, reported no significant differences between these
approaches (Abraham, 1985; Rosa & ONeill, 1999; Shaffer, 1989). Seliger (1975) also
reported no significant difference between these approaches; however, the deductive
group showed superiority in the three-week delayed posttest. Abuseileek (2009) reported
that for simple structures there was no difference between these two methods; however,
for complicated structures, the deductive method was better than the inductive method.
Although Fotos and Ellis (1991) did not intend to compare deduction with induction, they
reported no significant difference between the group that received an induction-based
approach (consciousness raising tasks) and the group that received a deduction-based
approach (traditional grammar lessons)1; however, the deduction-based group performed
significantly better in the two-week delayed posttests. In a later study, Fotos (1994)
reported that the group that received the induction-based approach, which was not identi-
cal to that of Fotos and Ellis (1991), and the deduction-based group both made similarly
significant gains, which were maintained after a two-week period. Furthermore, Sjberg
and Trop (1969) reported that the deductive learning was more effective but that the
advantage disappeared after five weeks. Erlam (2003) and Robinson (1997) also reported
that the deductive condition was more effective. On the contrary, Herron and Tomasello
(1992) concluded that guided inductive learning was superior in the learning of certain
grammatical structures. Further, Vogel, Herron, Cole, and York (2011) found that guided
inductive learning was superior in the short-term learning but not in the long-term learn-
ing. Finally, Haight et al. (2007) and Leow (1998) reported the superiority of induction
over deduction in terms of both short-term and long-term retention.2
The use of deductive and inductive methods is also common outside of L2 teaching.
Educational psychologists such as Bruner (1961, 1973) suggested that a discovery
method (as used in induction) can lead to more orderly, integrative, and viable organiza-
tion, transformation, and use of knowledge (Ausubel, 1963, p. 160). This is because the
very attitudes and activities that characterize figuring out or discovering things for
oneself also seems to have the effect of making material more readily accessible in mem-
ory (Bruner, 1973, p. 412). Bruner derived this view largely from psychological research,
which suggested that the key to retrieval of stored information is organization, or know-
ing where to find information and how to get there. On the other hand, Ausubel (1963)
argued that the act of discovery per se does not lead to the organizing and integrative
effects of learning by discovery. It accomplishes such effects only to as much extent as
the learning situation is highly structured, simplified, and skillfully programmed to
include a large number of diversified exemplars of the same principle, carefully graded
in order of difficulty (Ausubel, 1963, p. 160). The above views suggest that a highly
structured and simplified discovery learning environment may lead to learned materials
more readily accessible in memory. In other words, induction may be superior to deduc-
tion. Nevertheless, a differing view was expressed by educational psychologist Anderson
(1967). When summarizing evidence on the general effect of various orders of rule and
example, Anderson indicated that the literature on discovery learning (which follows the
examplerule order) generally shows that ruleexample procedures (as used in deduc-
tion) result in speedier acquisition and better retention than discovery methods.
As a significant amount of evidence from L2 studies has lent strong support to more
explicit over less explicit types of instruction, a logical prediction derived from such
evidence is that learning conditions sharing a comparable level of explicitness will have
a similar effect on learning performance. In other words, despite the conflicting evidence
as to the effectiveness of deduction and induction, deductive and inductive approaches
are expected to produce similar outcomes when they hold comparable degrees of explic-
itness. Taking this prediction into account, to ensure that the deductive and the inductive
approaches used in the current study are comparable in explicitness, they were both
operationalized to draw learners attention to grammatical forms and provide correct
grammar rules, even though at different points during the learning process.
Skehan (1980, 1982, 1989) noted that two aspects of memory can predict language
learning success: response integration, the ability to recall a set of words in an unknown
language without being given word associates in the first language, and memory for
text, the ability to analyse text, to extract its propositional content, and remember such
content (1989, p. 31). To measure memory for text, Skehan (1980, 1982) used
Indonesian grammar rules, each of which contained a different number of propositions
(see Appendix 1). As he explained, what the memory-for-text predictor seems to do is
measure how adept people are at bringing their general knowledge of meaning to bear
on newly presented materials; to see relationships between the elements involved; and
to relate them to existing knowledge (1989, p. 31). This view is consistent with many
psychological research findings, which suggest that prior knowledge influences future
learning. Those who know more learn better. This view will be discussed more in the
next section.
Drawing on the skill acquisition theory, knowledge about L2 grammar can start out
in the declarative form (DeKeyser, 1998, 2001; see Chi & Rees, 1983). It is plausible
that the memory for meaningful material and connected text, as investigated by Skehan
(1980, 1982), has a facilitative effect on the learning of explicit rules. This view seems
to coincide with Robinsons (2001) hypothesis, already mentioned above. That is, text
memory (subsumed under memory for continent text) is related to explicit rule
learning.
Taking the results of previous studies into account, the present study, which used the
latent growth curve analysis (Duncan, Duncan, & Strycker, 2006), hypothesized a priori
that three indicator variables memory for text, grammatical sensitivity, and associative
memory were able to measure a latent variable, language aptitude. The extent of the
relationships between the latent variable and the three indicators were measured by fac-
tor loadings.
speed for well-known information in long-term memory and these differences are related
to measures of intellectual ability, most notably verbal ability (p. 432). Given the views
presented above, it appears reasonable to assume that an instructional approach is able to
accommodate individual learners in domain knowledge base and the organization and
retrieval of information in memory, consequently enhancing their ability to acquire new
knowledge.
Skehan (1980, 1982) indicated that the factor analysis of the test he used to meas-
ure memory for text (used in the present study as well) suggested that it measured
simultaneously verbal intelligence, trained language ability (the amount of education
and previous grammar/language analysis instruction), and presumably memory for
natural language. Thus, it is reasonable to assume that learners who score low on such
a test have a low knowledge base about grammar and verbal intelligence. As the
deductive approach presents grammar explanations and rules to the learners at the
outset and provides them with the opportunity to practice the rule immediately, it is
plausible that, compared to the explicit-inductive approach, it can better remediate
these learners weaknesses. On the contrary, it is reasonable to assume that learners
who score high on this test have a high knowledge base about grammar and verbal
intelligence. It is likely that the explicit-inductive approach, which allows informa-
tion to be organized in such a way that it can be retrieved more easily (Bruner, 1973),
can better capitalize on their strength of not needing much information at the outset
to support learning.
IVMethod
1 Participants
The initial pool of participants consisted of approximately 400 students enrolled in
21 classes of a first-quarter Spanish course at a public university in the USA.
Learners were randomly assigned to two explicit learning conditions: deductive
(DE) and explicit-inductive (EI). The final sample contained 93 native English
speakers from 21 classes who finished the learning activities (46 males, 47 females;
nDE = 42, nEI = 51).3
2 Grammatical structure
The construction with the Spanish psych verb gustar (to like; to be pleasing) was cho-
sen to be the target structure for the following five reasons.
First, Spanish psych verb constructions are considered very difficult to acquire by
native English-speaking learners (Gascn, 1998; Lpez Jimnez, 2003; Marras &
Cadierno, 2008; Montrul, 1997; VanPatten, 1986; Zyzik, 2006). One of the likely factors
underlying the difficulty of its acquisition is the ways that this construction differs from
English. As shown in examples (1) and (2), in English the entity that experiences the
emotion, the experiencer, is coded as the subject of the sentence (I, students), and the
stimulus that causes the emotion as the direct object of the sentence (strawberries, sum-
mer). In contrast, in Spanish the experiencer is coded as the indirect object (me, les)
while the stimulus is coded as the subject (las fresas, el verano). Nevertheless, in both
languages the experiencer is placed at the beginning of a sentence and the stimulus at the
end of a sentence.4
Since native English-speaking learners tend to process the first noun or pronoun they
encounter in a sentence as the subject or agent (VanPatten, 2003), correctly interpreting
sentences containing this verb is not an issue for them. Conversely, correctly producing
sentences containing this verb is a challenge. Because learners can successfully derive
the correct meaning of a sentence in spite of overlooking the case of the noun or pronoun
that occupies the beginning position in a sentence and/or the verb inflection, the gram-
matical features related to this construction lack communicative value (see VanPatten,
1985). The manner in which these two languages differ may have contributed to some
common errors made by native English-speaking learners, well known to Spanish
instructors. They include subjectobject confusion (Gascn, 1998; VanPatten, 1986), the
use of singular verb for plural subject (Gascn, 1998), the omission of object pronouns
(Gascn, 1998), object pronoun errors (Gascn, 1998), and the omission of personal a
(to) when inclusion is required (Gascn, 1998). The following two examples demon-
strate the errors that learners may make with regard to sentence (1) above.
The following examples demonstrate the errors that learners may make with regard to
sentence (2) above.
Second, using the psych verb gustar to construct a sentence is a complex task because
more than one grammatical concept has to be taken into account in order to arrive at a
correct form in language production. The difficulty of its acquisition, the lack of com-
municative value of grammatical features, and the task complexity involved suggest that
the likelihood of spontaneous noticing and processing will be low and explicit instruc-
tion, rather than an instruction based on comprehension, will be more beneficial (see de
Graaff, 1997; Hulstijn & de Graaff, 1994).
Third, the rules for this syntactic structure are 100% regular. Therefore, explicit rule-
based instruction is likely to be beneficial.
Fourth, to date, very few experimental research studies have used this syntactic
structure as the target structure (two exceptions include Bowles, 2008; Lpez Jimenez,
2003).6
Fifth, this grammatical structure was not formally introduced to learners outside the
exposure period during the academic term in which the study was conducted.
3 Instructional materials
All instruction was presented in an electronic format via the internet. Since the partici-
pants who received treatment came from 21 different classes, the online environment
allowed all students to receive the same instruction presented in the same way, eliminat-
ing teacher differences. The online materials consisted of five lessons containing a total
of approximately 4050 minutes of learning materials.
The first three lessons taught three areas of concepts related to the target structure: the
subject of the sentence (subjectverb agreement), the indirect object pronoun, the per-
sonal a (to) before a common/proper noun. The types of pages contained in each of the
first three lessons of the DE condition were as follows:
1. Introduction;
2. Presentation of the grammar concept/s;
3. Exemplars accompanied by explanations;
4. Multiple-choice (with only two selections for each question) or fill-in-the-blank
exercises that required learners to apply the concept learned, with corresponding
try-again and good-job feedback.
The types of pages contained in each of the first three lessons of the EI condition were as
follows:
1. Introduction;
2. Highly structured and simplified activities (Ausubel, 1963) that required learners
to observe exemplars and indicate what was observed through multiple-choice
questions (with only two selections for each question), with corresponding try-
again and good-job feedback;
3. Multiple-choice questions (with only two selections for each question) that
required learners to indicate the pattern/s discovered, with corresponding try-
again and good-job feedback.
To ensure that the participants of both conditions would receive a comparable amount of
exposure to the targeted structure, the researcher provided (1) the same set of exemplars
in both conditions and (2) a comparable amount of grammar explanations in both condi-
tions, through either instructional (DE) or feedback (EI) pages.
In the fourth lesson, the concepts presented in the first three lessons were organized
into a sequence of mental and observable steps that learners could follow when they cre-
ated sentences containing the target verb, from left to right (see Appendix 2). These
steps, created by one of the authors by following an information-processing analysis
(Gagn, Wager, Golas, & Keller, 2005, p. 153), were expected to facilitate in the process
of creating sentences in real time (for a discussion on the need of grammar rules that
learners can use to produce sentences, see Garrett, 1986; Randall, 2007). Learners in the
DE condition received these sequenced rules and were asked to memorize them before
writing them down from memory. Learners in the EI condition were asked to derive their
own sequence of rules by considering the exemplars provided. Afterwards, their sequence
of rules and the researchers were displayed side by side on the screen to facilitate com-
parisons by them.7
The fifth lesson contained five post-instructional exercises, in which learners of both
conditions practiced identifying and correcting errors in sentences using the grammar
rules they learned earlier, and producing the target structure. These activities aimed to
help anchor the grammar rules solidly in the learners consciousness, in declarative form
(see DeKeyser, 1998).8
Because the learning tasks took place online, several program control features were
integrated into the computer program. To ensure that learners paid attention to the learn-
ing of the target structure, learners were required to have obtained perfect scores on the
multiple-choice or fill-in-the-blank learning activities during the first three lessons
before they were allowed to access subsequent activities.9 Additionally, to ensure that
learners would receive the instructional treatment and in the manner intended, they were
required to perform learning activities sequentially without skipping any pages.
4 Aptitude measures
Two measures of memory and one measure of analytical language ability were obtained
via three group-administered tests given in two class periods, one week and two weeks
before the learner accessed the online learning program. They included:
memory for text, assessed through the task involving recall of fifteen Indonesian
grammar rules based on Skehan (1982), each of which contained a different
number of propositions (M = 14.98, SD = 6.70, Maximum Score = 40)
(see Appendix 1);
grammatical sensitivity, assessed by Modern Language Aptitude Test (MLAT), IV
Words in Sentences (Carroll & Sapon, 1959); and
associative memory, measured by MLAT, V Paired Associates (Carroll & Sapon,
1959). The reliabilities for Words in Sentences (M = 17.96, SD = 5.59, Maximum
Score = 45) and Paired Associates (M = 15.85, SD = 5.44, Maximum Score = 24),
using KuderRichardson, were .81 and .89 respectively.
5 Assessment
A written sentence-production test and a written sentence-correction test were chosen to
measure the participants knowledge about the targeted structure (see Appendix 3). The
production task, consisting of four questions in English, enabled learners to engage in
using the language to communicate likes or dislikes. The participant had to answer two
questions that solicited his/her responses about his/her own likes or dislikes and two
questions that solicited his/her responses about the likes or dislikes of someone else. For
the correction task, the test presented the participant with 17 sentences: five grammatical
and 12 ungrammatical. Grammatical sentences were included in this task so that no clue
to the correctness of any sentence was revealed through this test.10 The participant had to
decide if a sentence was grammatical, not grammatical, or I dont know. If the par-
ticipant indicated that the sentence was not grammatical, he or she was asked to correct
the sentence so that it would be grammatical. The production task was administered
before the correction task to avoid learners gaining exposure to the targeted structure
through the correction task and using that knowledge to create sentences during the
production task.
Four versions of the production task differing in the Spanish vocabulary that the
participant was required to use and four versions of the correction task, differing in the
Spanish vocabulary used in the sentences and the order of question types were pre-
pared. None of the instances used in the learning program appeared in the tests so that the
participants were being tested on their ability to transfer what they had learned to new
but parallel situations. Participants were given a maximum of 10 minutes to complete
both tasks.
6 Scoring
In scoring the production and correction tasks, the number of grammar concepts that the
participant had applied correctly was counted. Two concepts (each with two sub-con-
cepts) were identified for sentences containing the first-person experiencer (e.g. I like)
and three concepts (two with sub-concepts) were identified for sentences containing the
common/proper noun experiencer (e.g. students like, John likes). These concepts and the
point assigned for each of them is included in Appendix 4. A maximum of 10 points was
possible for each task, 4 of which were related to sentences containing the first-person
experiencer and 6 of which were related to sentences containing a common/proper noun
experiencer.
For the production task, the participants sentences were checked for the correctness
of the application of rules required by the questions. If the application of a rule was
required twice and the participant applied it correctly twice, full credit for the rule
was assigned. If the participant applied it correctly only once, one-half credit for the rule
was assigned. Since one particular rule was required once in the test, if the participant
applied it correctly, full credit for such rule was assigned. For the correction task, for
each grammatical or I do not know circled, no points were assigned. For those not
grammatical circled correctly, the participants corrections were checked for accuracy.
Since each type of error tested the participants knowledge about a rule and occurred at
least three times during this task, if the participant was able to correct one type of error
correctly twice, he or she was given full credit for the rule. If the participant only cor-
rected the error correctly once, he or she was assigned half-credit for the rule.
7 Procedure
Participants in the instructional groups took the language aptitude tests in two class peri-
ods. All participants took the pretest in class and logged into the research website to
answer a background survey. One week after the pretest, participants in the instructional
groups accessed the learning materials and completed the immediate posttest online.11
Approximately three weeks after the immediate posttest, all participants took the first
delayed posttest in class. Two weeks later they took the second delayed posttest in class.
VResults
Descriptive analysis
Descriptive statistics are summarized in Table 1. It shows that the learners scores
improved considerably from the pretests to the immediate posttests, dropped from the
immediate posttests to the second posttests, and increased slightly from the second post-
tests to the third posttests. This empirically confirms the assumption that the learners
performance development over time is non-linear.
Research question 1
To answer research questions 1 and 2, a LGCA was conducted for the production and the
correction tests, respectively (see Table 2 and Figure 1). The deductive condition was
coded as 0, while the explicit-inductive condition as 1. In other words, a positive path
coefficient would indicate that the explicit-inductive condition produced better results,
whereas a negative path coefficient would indicate that the deductive condition was
superior. Excellent model fit was achieved for both models. Namely, both models pro-
vided a good description of the data.
Note that in a LGCA model, an intercept refers to the value at the beginning of the
growth process and a slope refers to the average rate of growth. For the production test,
the latent mean of the intercept was 3.31 (p < .001) and the latent mean of the slope was
4.91 (p < .001); this suggests that the learners began with an average production score of
3.31 and gained an average score of 4.91, multiplied by the corresponding coefficient
(.00, 1.00, .44, and .53), at each time point. For the correction test, the latent mean of the
intercept was 2.13 (p < .001) and the latent mean of the slope was 4.94 (p < .001); this
1 2 3 4 5 6 7 8 9 10 11 Deductive
M SD
1. Memory for text .381* .466** 0.106 .312* 0.178 0.178 .313* .426** 0.221 .380* 15.1 6.85
2. Grammatical sensitivity .580** 0.301 0.112 0.284 0.214 0.188 0.236 .326* 0.195 .393* 17.83 6.04
3. Associative memory .500** .429** 0.226 0.296 0.3 0.234 0.264 .326* 0.161 .309* 15.64 4.67
4. Production: pre 0.175 0.276 0.215 0.19 .368* .452** .671** 0.125 .317* .334*
5. Production: post 1 .419** .292* .347* .422** .515** .405** .310* .756** .378* .586**
6. Production: post 2 0.257 0.193 0.242 .359** .562** .700** .621 ** .363* .678** .705**
7. Production: post 3 0.209 0.167 0.166 .521** .567** .815** .623 ** .380* .569** .814**
8. Correction: pre 0.141 0.246 0.211 .689** .426** .438** **
.515 0.279 .618** .639**
9. Correction: post 1 .453** 0.274 0.212 .311* .735** .591** .580** .438** .412** .603**
10. Correction: post 2 .362** .310* .319* .441** .595** .842** .777** .565** .669** .694**
11. Correction: post 3 .353* 0.272 0.271 .422** .585** .851** .864** .540** .680** .929**
Notes. n = 42 students in deductive instruction [above diagonal]; n = 51 students in explicit-inductive instruction [below diagonal]. * p < .05; ** p < .01.
307
308 Language Teaching Research 18(3)
(Production) (Correction)
.94*
1.79*** 2.25***
e1 e2 e3 e4 e1 e2 e3 e4
1.00 1.00
1.00 1.00 .53 .59 .68
1.00 1.00
1.00 .44 MeanI = 2.13*** 1.00
MeanI = 3.31*** 1.00 .00 MeanS = 4.91*** 1.00 .00 MeanS = 4.94***
VarI = 3.29*** VarS = 4.14** VarI = 5.13*** VarS = 7.10**
2.27* 3.43*
Learning Aptitude Learning Aptitude
Condition Condition
Figure 1. Latent growth curve models for production and correction with learning condition
and language aptitude as covariates ( p < .10, * p < .05, ** p < .01, *** p < .001).
suggests that the learners began with an average correction score of 2.13 and gained an
average score of 4.94, multiplied by the corresponding coefficient (.00, 1.00, .59, and
.68), at each time point.
Note that in a LGCA model the variance of the intercept and that of the slope show
the extent to which learners had different initial levels and rates of growth over time. For
the production test, the variance for the intercept and that of the slope were 3.29 (p <
.001) and 4.14 (p < .01), respectively, suggesting a large variation in growth curves
among learners. For the correction test, a large variation in growth curves among learners
was also evident; 5.13 (p < .001) for the intercept and 7.10 (p < .01) for the slope.
As shown in Figure 1, the learners assigned to two learning conditions did not differ
significantly in the intercepts; this is evidenced by the non-significant path coefficients
from the treatment to the intercept, i.e. .21 for the production test and .20 for the correc-
tion test. Further, no statistically significant differences in treatment effects were found
between two learning conditions. This is evidenced by the non-significant path coeffi-
cients from the treatment to the slope, namely, .18 for the production test and .30 for the
correction test. To sum up, research question 1 is answered affirmatively.
Research question 2
To find out which indicator variable had the strongest relationship with the latent varia-
ble, language aptitude, the standardized factor loadings from language aptitude to three
indicator variables were obtained. For the production test, the standardized factor load-
ings from language aptitude to memory for text, grammatical sensitivity, and associative
memory were .76, .61, and .63, respectively. For the correction test, the standardized
factor loadings from language aptitude to memory for text, grammatical sensitivity, and
associative memory were .81, .60, and .59, respectively. Note that Comrey and Lee
(1992) suggest that factor loadings in excess of .71 (50% variance) are considered excel-
lent, .63 (40% variance) very good, .55 (30% variance) good, .45 (20% variance) fair,
and .32 (10% variance) poor. Thus, the factor loadings reported above are between excel-
lent (.81) and good (.59). In summary, for both types of tests, memory for text had the
strongest relationship with language aptitude.
To interpret the relationship between the latent variable and the three indicator varia-
bles, the unstandardized loadings had to be used (see Figure 1). For the production test,
the factor loadings of memory for text, grammatical sensitivity, and associative memory,
were 1.00, .67, and .68, respectively. In other words, for each 1 unit increase in language
aptitude, an individuals score on memory for text increases 1 unit point, whereas his/her
scores on grammatical sensitivity and associative memory will increase .67 of 1 unit and
.68 of 1 unit, respectively. For the correction test, the factor loadings of memory for text,
grammatical sensitivity, and associative memory were 1.00, .63, and .60, respectively.
As shown in Figure 1, significant language aptitude effects were detected in both
models. For the production test, the coefficients from the language aptitude to the inter-
cept and from the language aptitude to the slope were .11 (p < .05, b = .30) and .10 (p <
.10, b = .24), respectively, suggesting that higher aptitude positively affects both baseline
test score and rate of growth over time. The same conclusions can be drawn for the cor-
rection test as well. For this test, the coefficients from the language aptitude to the inter-
cept and from the language aptitude to the slope were .13 (p < .01, b = .30) and .20 (p <
.10, b = .37), respectively. In summary, language aptitude had a positive effect on both
the initial level of performance and the rate of growth. Research question 2 is thus
answered affirmatively.
Research question 3
To answer this research question, a multi-group LGCA by language aptitude with a
median split (median factor score = .036) on language aptitude was conducted for both
the production and the correction tests. As explained earlier, a better model fit obtained
from the unconstrained model, vs. the constrained model, would suggest that an ATI
effect was present. The results show that the unconstrained model had a significantly
better model fit, as revealed by the nested chi-square test (for the production test, 2 =
11.413, df = 2, p < .003; for the correction test, 2 = 15.386, df = 2, p < .001). In other
words, an ATI was detected. Figure 2 shows that a good model fit was achieved for the
sub-models of the multi-group LGCA analyses.
Note that in these LGCA models, the deductive condition was coded as 0, while the
explicit-inductive condition as 1. In other words, a positive path coefficient means that
the learners in the explicit-inductive condition performed better, whereas a negative path
coefficient means that the learners in the deductive condition did better. The results show
that for the learners of low language aptitude in the correction test, the unstandardized
(Production) (Correction)
1.30*
2.53** 2.90**
e1 e2 e3 e4 e1 e2 e3 e4
.13 .002
.43
1.11
Learning Learning
Condition Condition
4.16** 5.33
.56
1.18
.97
e1 e2 e3 e4
e1 e2 e3 e4
1 1 1 1
0 0 0 0
CORR0 CORR1 CORR2 CORR3
PROD0 PROD1 PROD2 PROD3
(Low aptitude)
1.00 1.00
1.00 .59 .67
1.00 1.00 .44 .50 1.00
1.00 MeanI = 2.08*** 1.00 MeanS = 5.47***
MeanI = 3.38*** 1.00 1.00
.00
.00 MeanS = 5.25*** VarI = 2.21 VarS = 7.15*
VarI = 1.97 VarS = 3.20
1 e5 ICEPT SLOPE e6
e5 ICEPT SLOPE e6
.58 1.62*
.53
.82
Learning
Learning
Condition
Condition
.43
.12
ChiSquare = 21.603, df = 12, p = .042
ChiSquare = 10.462, df = 10, p = .401
NFI = .883, CFI = .942, RMSEA = .094
NFI = .926, CFI = .996, RMSEA = .023
Figure 2. Latent growth curve models for production and correction with learning condition
as a covariate moderated by language aptitude (p < .10, *p < .05, **p < .01, ***p < .001).
path coefficient from the learning condition to the slope was 1.62 (p < .05), indicating
that those receiving the deductive instruction had a significantly higher slope, or average
rate of growth, than those receiving the explicit-inductive instruction. The standardized
regression coefficient b was .29, revealing a near medium effect. In summary, learners
of low language aptitude learned significantly better under the deductive condition when
(Production) (Correction)
10.00 10.00
8.00 8.00
(High aptitude)
6.00 6.00
4.00 4.00
2.00 2.00
.00 .00
Pretest Posest 1 Posest 2 Posest 3 Pretest Posest 1 Posest 2 Posest 3
10.00 10.00
8.00 8.00
(Low aptitude)
6.00 6.00
4.00 4.00
2.00 2.00
.00 .00
Pretest Posest 1 Posest 2 Posest 3 Pretest Posest 1 Posest 2 Posest 3
Figure 3. Latent growth curves for production and correction moderated by language
aptitude.
they were measured by the correction test. When these learners were measured by the
sentence production test, those receiving the deductive instruction also had a higher
slope (unstandardized path coefficient .53). However, the difference was not signifi-
cant. On the other hand, among learners of high language aptitude, those receiving the
explicit-inductive instruction had higher slopes than those receiving the deductive
instruction (unstandardized path coefficient .13 for the production test and .002 for the
correction test). Nevertheless, the differences were not significant. The mean plots in
Figure 3, estimated from the LGCAs, provide alternative representations of both the
significant and insignificant interaction effects in the form of growth curves. Further,
they show that for the case of a significant interaction, the effect appeared at the immedi-
ate posttest, which endured throughout the two delayed posttests. In conclusion, research
question 3 is partially supported.
VIDiscussion
The main purpose of this study was to discover how individual differences at the level of
language aptitude interact with equally explicit learning conditions to affect the success
VIIConclusions
Although most researchers and teachers agree that learners are not homogeneous, many
L2 research studies to date have focused on determining if one instructional approach is
superior to the other; very few studies have investigated if different learners benefit from
different instructional approaches (exceptions include Abraham, 1985; DeKeyser, 1993;
Gallegos, 1968; Hauptman, 1971; Nation & McLaughlin, 1986; Wesche, 1981; Zampogna
et al., 1976). Our findings indicate that different but equally explicit instructional
approaches can have differential effects on the learning performance of learners at differ-
ent aptitudinal levels. In view of these findings and the existing evidence concerning the
superiority of explicit types of instruction as opposed to implicit types, to maximize
learners potential for success as well as account for variations in language learning suc-
cess under different instructional conditions (see DeKeyser, 2009; Robinson, 2001), it is
necessary for L2 educators to investigate the effects of equally valid grammar instruc-
tional approaches on different types of learners. One learner characteristic that may
deserve their attention is domain knowledge base and verbal intelligence. However, to
uncover at which point learners may become sensitive to different instructional
approaches, researchers may need to examine different combinations of instructional
approaches and learner characteristics. The present study has strengthened the signifi-
cance in understanding the interaction between learner characteristics and instructional
conditions. Such knowledge is surely indispensable to the design of learning activities on
textbooks or electronic materials.
Funding
This study was in part supported by Charles Phelps Taft Research Center.
Notes
1. Because the students in the traditional grammar lesson in Fotos and Ellis (1991) were not
asked to search for rules, the authors consider this as a deduction-based approach.
2. The authors consider the learner-centered approach (the problem-solving technique) used in
Leow (1998) as a type of induction and the teacher-centered approach as a type of deduction.
3. The online lessons kept track of the participants progress (e.g. login time, time of comple-
tion). When a student logged on, the Task Access and Information page informed the par-
ticipant of his/her progress. The researcher obtained such information through the designers
account and used only the data of those who had completed the lessons for the study.
4. Although Spanish is more flexible in word order compared to English, the most frequent
and unmarked word order for the Spanish gustar-construction is objectverbsubject (Lpez
Jimnez, 2003). Since only the unmarked word order is introduced in basic language and
grammar textbooks, only sentences with this word order were included in the study.
5. These errors and others have appeared in the data of the current study.
6. In Lpez Jimenez (2003) grammar explanation and practice (deduction) was used to teach
this structure, whereas in Bowles (2008) a problem-solving task (a type of induction) was
used.
7. The computer log showed that learners in both conditions were able to summarize the
sequenced rules or sequence the rules inferred previously.
8. Although the experiencers in different persons were introduced in the learning program, only
the likes and dislikes of the 1st-person and 3rd-person experiencers were included in the
References
Abraham, R.G. (1985). Field independence-dependence and the teaching of grammar. TESOL
Quarterly, 20, 689702.
Abuseileek, A.F. (2009). The effect of using an online-based course on the learning of grammar
inductively and deductively. ReCALL, 21, 319336.
Anderson, R.C. (1967). Educational psychology. Annual Review of Psychology, 18, 129164.
Ausubel, D.P. (1963). The psychology of meaningful verbal learning: An introduction to school
learning. New York: Grune & Stratton.
Baron, R.M., & Kenny, D A. (1986). The moderator-mediator variable distinction in social psycho-
logical research: Conceptual, strategic, and statistical considerations. Journal of Personality
and Social Psychology, 51, 11731182.
Bors, D.A., & MacLeod, C.M. (1996). Individual differences in memory. In E.L. Bjork & R.A.
Bjork (Eds.), Handbook of perception and cognition: Volume 10: Memory (pp. 411441).
San Diego, CA: Academic Press.
Bowles, M.A. (2008). Task type and reactivity of verbal reports in SLA. A first look at a L2 task
other than reading. Studies in Second Language Acquisition, 30, 259387.
Bruner, J.S. (1961). The act of discovery. Harvard Educational Review, 31, 2132.
Bruner, J.S. (1973). Beyond the information given: studies in the psychology of knowing. New
York: W.W. Norton.
Carroll, J.B., & Sapon, S.M. (1959). Modern language aptitude test (MLAT). New York: The
Psychological Corporation.
Chi, M.T.H., & Rees, E. (1983) A learning framework for development. In M.T.H. Chi (Ed.),
Trends in memory development research (pp. 71107). New York: Karger.
Chiesi, H.L., Spilich, G.J., & Voss, J.F. (1979). Acquisition of domain-related information in rela-
tion to high and low domain knowledge. Journal of Verbal Learning and Verbal Behavior,
18, 257273.
Comrey, A.L., & Lee, H.B. (1992). A first course in factor analysis. 2nd edition. Hillsdale, NJ:
Lawrence Erlbaum.
Corno, L., Cronbach, L.J., Kupermintz, H., Lohman, D.F., Mandinach, E.B., Porteus, A.W., &
Talbert, J.E. (2002). Remaking the concept of aptitude: Extending the legacy of Richard E.
Snow. Mahwah, NJ: Lawrence Erlbaum.
Cronbach, L.J., & Snow, R.E. (1977). Aptitudes and instructional methods: A handbook for
research on interactions. New York: Irvington.
Decoo, W. (1996). The induction-deduction opposition: Ambiguities and complexities of the
didactic reality. International Review of Applied Linguistics in Language Teaching, 34,
95118.
de Graaff, R. (1997). The eXperanto experiment: Effects of explicit instruction on second lan-
guage instruction. Studies in Second Language Acquisition, 19, 249297.
DeKeyser, R. (1993). The effect of error correction on L2 grammar knowledge and oral profi-
ciency. The Modern Language Journal, 77, 501514.
DeKeyser, R. (1995). Learning second language grammar rules: An experiment with a miniature
linguist system. Studies in Second Language Acquisition, 19, 196221.
DeKeyser, R. (1998). Beyond focus on form. Cognitive perspectives on learning and practicing
second language grammar. In C. Doughty & J. Williams (Eds.), Focus on form in classroom
second language acquisition (pp. 4263). Cambridge: Cambridge University Press.
DeKeyser, R. (2001). Automaticity and automatization. In. P. Robinson (Ed.), Cognition and sec-
ond language instruction (pp. 125151). Cambridge: Cambridge University Press.
DeKeyser, R. (2009). Variable interaction in SLA: Much more than a nuisance. Unpublished paper
presented at Second Language Research Forum 2009, Michigan State University, MI, USA.
Doughty, C. (1991). Second language instruction does make a difference: Evidence from an
empirical study of SL relativization. Studies in Second Language Acquisition, 13, 431469.
Duncan, T.E., Duncan, S.C., & Strycker, L.A. (2006). An introduction to latent variable growth
curve modeling: Concepts, issues, and applications (2nd ed.). Mahwah, NJ: Lawrence
Erlbaum.
Eggins, J.A. (1979). The Interaction between structure in learning materials and personality type
of learners. Unpublished PhD dissertation, Indiana University, IN, USA.
Erlam, R. (2003). The effects of deductive and inductive instruction on the acquisition of direct
object pronouns in French as a second language. Modern Language Journal, 87, 242260.
Fischer, R.A. (1979). The inductive-deductive controversy revisited. Modern Language Journal,
63, 98105.
Fotos, S. (1994). Integrating grammar instruction and communicative language use through gram-
mar consciousness-raising tasks. TESOL Quarterly, 28, 323351.
Fotos, S., & Ellis, R. (1991). Communicating about grammar: A task-based approach. TESOL
Quarterly, 25, 605628.
Gagn, R M., Wager, W.W., Golas, K.C., & Keller, J.M. (2005). Principles of instructional design.
Belmont, CA: Wadsworth/ Thomson Learning.
Gallegos, A.M. (1968). A study and comparison of experimenter pacing and student pacing of
programmed instructor. Journal of Education Research, 61, 339342.
Garrett, N. (1986). The problem with grammar: What kind can the language learner use? Modern
Language Journal, 70, 133148.
Gascn, C.D. (1998). The Spanish psych verb construction: Beginning and intermediate learners
patterns of usage. Texas Papers in Foreign Language Education, 3, 6984.
Haight, C., Harron, C., & Cole, S. (2007). The effects of deductive and guided inductive instruc-
tional approaches on the learning of grammar in the elementary foreign language college
classroom. Foreign Language Annals, 40, 288310.
Hammerly, H. (1975). The deduction/induction controversy. Modern Language Journal, 75,
1518.
Hancock, G.R., & Lawrence, F.R. (2006). Using latent growth models to evaluate longitudinal
change. In G.R. Hancock & R.O. Mueller (Eds.), Structural equation modeling: A second
course (pp. 171196). Greenwich, CT: Information Age.
Hauptman, P.C. (1971). A structural approach vs. a situational approach to foreign-language
teaching. Language Learning, 21, 235244.
Herron, C., & Tomasello, M. (1992). Acquiring grammatical structures by guided induction.
French Review, 65, 708718.
Hulstijn, J.H., & de Graaff, R. (1994). Under what conditions does explicit knowledge of a second
language facilitate the acquisition of implicit knowledge? A research proposal. AILA Review,
11, 97113.
Kyllonen, P.C., Tirre, W.C., & Christal, R.E. (1991). Knowledge and processing speed as
determinants of associative learning. Journal of Experimental Psychology: General, 120,
5779.
Leow, R.P. (1998). The effects of amount and type of exposure on adult learners L2 development
in SLA. Modern Language Journal, 82, 4968.
Little, R.J.A. (1988). A test of missing completely at random for multivariate data with missing
value. Journal of the American Statistical Association, 83, 11981202.
Lpez Jimnez, M.D. (2003). Grammar instruction and the acquisition of gustar-type verb
by English-speaking learners of Spanish. Estudios de Lingstica Inglesa Aplicada, 4,
255279.
Marras, V., & Cadierno, T. (2008). Spanish gustar vs. English like: A cognitive analysis of the con-
structions and its implication for SLA. In A. Tyler, Y. Kim, & M. Takada (Eds.), Language
in the context of use: Discourse and cognitive approaches to language (pp. 233254). New
York: Mouton de Gruyter.
McLachlan, J.F.C., & Hunt, D.E. (1973). Differential effects of discovering learning as a function
of conceptual level. Canadian Journal of Behavioural Science, 5, 152160.
Montrul, S. (1997). On the parallels between diachronic change and interlanguage grammars: The
L2 acquisition of the Spanish dative case system. Spanish Applied Linguistics, 1, 87113.
Nation, R., & McLaughlin, B. (1986). Experts and novices: an information processing approach to
the good language learner problem. Applied Psycholinguistics, 7, 4156.
Norris, J., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantita-
tive meta-analysis. Language Learning, 50, 417528.
Norris, J., & Ortega, L. (2001). Does type of instruction make a difference? Substantive findings
from a meta-analytic review. Language Learning, 51, 157213.
Oltman, P.K., Raskin, R., & Witkin, H.A. (1971). Group embedded figures test. Palo Alto, CA:
Consulting Psychologists Press.
Randall, M. (2007). Memory, psychology, and second language learning. Amsterdam: John Benjamins.
Reber, A.S., Kassin, S.M., Lewis, S., & Cantor, G.W. (1980). On the relationship between implicit
and explicit modes in the learning of a complex rule structure. Journal of Experimental
Psychology: Human Learning and Memory, 6, 492502.
Resnick, L.B., & Neches, R. (1984). Factors affecting individual differences in learning ability.
In R. J. Sternberg (Ed.), Advances in the psychology of human intelligence: Volume 2 (pp.
275323). Hillsdale, NJ: Erlbaum.
Robinson, P. (1997). Individual differences and the fundamental similarity of implicit and explicit
adult second language learning. Language Learning, 47, 4599.
Robinson, P. (2001). Individual differences, cognitive abilities, aptitude complexes and learning
conditions in second language acquisition. Second Language Research, 17, 368392.
Rosa, E., & Leow, R.P. (2004). Computerized task-based exposure, explicitness, type of feedback,
and Spanish L2 development. Modern Language Journal, 88, 193217.
Rosa, E., & ONeill, M.D. (1999). Explicitness, intake, and the issue of awareness: Another piece
to the puzzle. Studies in Second Language Acquisition, 21, 511556.
Sawyer, M., & Ranta, L. (2001). Aptitude, individual differences, and instructional design. In
P. Robinson (Ed.), Cognition and second language instruction (pp. 319353). Cambridge:
Cambridge University Press.
Scott, V.M. (1989). An empirical study of explicit and implicit teaching strategies in French.
Modern Language Learning Journal, 73, 1422.
Seliger, H.W. (1975). Inductive method and deductive method in language teaching: A re-exami-
nation. International Review of Applied Linguistics in Language Teaching, 13, 118.
Shaffer, C. (1989). A comparison of inductive and deductive approaches to teaching foreign lan-
guages. The Modern Language Journal, 73, 395403.
Sjberg, K., & Trop, B. (1969). The value of external direction and individual discovery in learn-
ing situations: The learning of a grammatical rule. Scandinavian Journal of Educational
Research, 13, 233240.
Skehan, P. (1980). Memory, language aptitude, and second language performance. Polyglot, 2,
117.
Skehan, P. (1982). Memory and motivation in language aptitude testing. Unpublished PhD dis-
sertation, University of London, UK.
Skehan, P. (1989). Individual differences in second-language learning. London: Edward
Arnold.
Snow, R.E. (1991). Aptitude-treatment interaction as a framework for research on individual dif-
ferences in psychotherapy. Journal of Consulting and Clinical Psychology, 59, 205216.
Tisak, J., & Meredith, W. (1990). Descriptive and associative developmental models. In A. von
Eye (Ed.), Statistical methods in developmental research: Volume 2 (pp. 387406). San
Diego, CA: Academic.
Tomlinson, P.D., & Hunt, D.E. (1971). Differential effects of ruleexample order as a function of
conceptual level. Canadian Journal of Behavioural Science, 3, 237245.
VanPatten, B. (1985). Communicative value and information processing in second language
acquisition. In P. Larson, E. Judd, & D.S. Messerchmitt (Eds.), On TESOL: A brave new
world for TESOL (pp. 89100). Washington, DC: TESOL.
VanPatten, B. (1986). Second language acquisition research and the learning/teaching of Spanish:
Some research findings and implications. Hispania, 69, 202216.
VanPatten, B. (2003). Input processing in SLA. In B. VanPatten (Ed.), Processing instruction:
Theory, research, and commentary. Mahwah, NJ: Erlbaum.
Vogel, S., Herron, C., Cole, S., & York, H. (2011). Effectiveness of a guided inductive versus
a deductive approach on the learning of grammar in the intermediate-level college French
classroom. Foreign Language Annals, 44, 353380.
Wesche, M.B. (1981). Language aptitude measures in streaming, matching students with methods,
and diagnosis of learning problems. In K. Diller (Ed.), Individual differences and universals
in language learning aptitude (pp. 119139). Rowley, MA: Newbury House.
Zampogna, J., Gentile, R.J., Papalia, A., & Gordon, R. (1976). Relationships between learning
styles and learning environments in selected secondary modern language classrooms. Modern
Language Journal, 60, 443448.
Zyzik, E. (2006). Learners overgeneralization of dative clitics to accusative contexts: Evidence
for prototype effects in SLA. In C.A. Klee & T.L. Face (Eds.), Selected proceedings of the 7th
conference on the acquisition of Spanish and Portuguese as first and second languages (pp.
122134). Somerville, MA: Cascadilla Proceedings Project.
Appendix 1
Sample instructions for memory for text test (Skehan, 1982)
In this test, you are going to read some of the rules of Indonesian grammar. There are 15
rules to read. You will have 5 minutes to study them. Here is a typical rule:
Your task is to write down from memory the rest of the rule. In this case:
Appendix 3
Sample test
Production. A typical student tends to feel a certain way about certain things. Using the
verb gustar, indicate whether a typical student (el estudiante tpico) likes Friday classes
(las clases de los viernes).
Correction. Los jvenes les gustan las fiestas. Young people like parties.
Grammatical Not Grammatical I dont know
Make corrections here: Los jvenes les gustan las fiestas.