Applied Linguistics 2015: 36/3: 290–305 ß Oxford University Press 2014

doi:10.1093/applin/amu071 Advance Access published on 8 December 2014

The Effectiveness of Processing Instruction

in L2 Grammar Acquisition: A Narrative
Review

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

*,1ROBERT DEKEYSER and 2GORETTI PRIETO BOTANA
1
University of Maryland at College Park and 2Franklin and Marshall College
*E-mail: rdk@umd.edu

The past two decades have seen ample debate about processing instruction (PI)
and its various components. In this article, we first describe what PI consists of
and then address three questions: about the role of explicit information (EI) in
PI, the difference between PI and teaching that incorporates production-based
(PB) practice, and various factors that may moderate the impact of these treat-
ments. Our review shows that while many studies find little difference between
PI with and without EI and between PI and PB, the results vary depending on
whether the comprehension practice (structured input; SI) is task-essential or
not and whether the PB is communicative or not. Furthermore, PI tends to favor
comprehension abilities and PB production abilities. This review also shows that
almost all PI research so far has been very short-term and limited to a narrow
range of structures and populations, typically American college students learn-
ing a foreign language. Therefore the implications that some have drawn from
the PI literature, namely, that neither EI nor production practice help beyond SI,
are still far from generalizable.

Processing instruction (PI) is a grammar teaching technique that focuses on

altering learners’ default processing strategies and promotes effective proper
processing of target forms. It is based on the insights of research on how
learners process input: the way in which learners process linguistic data con-
stitutes the departure point for the design of instruction that fosters proper
form-meaning connections. PI offers a number of claims that state what me-
diates input processing, and which serve as the basis for devising more effective
processing strategies.
The basic problem that research on input processing has documented is that
learners often do not notice or process elements in the input, particularly
grammar structures that are communicatively redundant because meaning is
clear from lexical or pragmatic cues. Moreover, even when there are no such
cues, students’ L1 processing routines sometimes seem to prevent them from
paying enough attention to the morphosyntactic elements of the input. For
example, L1 native speakers of English tend to rely heavily on word order to
interpret who is the agent or the patient, even when a language they are
learning provides that information through a variety of morphological cues
 R. DEKEYSER AND G. PRIETO BOTANA 291

(see MacWhinney 1997; Gass et al. 2003; VanPatten 2012; Sagarra and Ellis
2013; among many others). This last ‘processing strategy’ is the most fre-
quently taken as an example in the PI literature, in particular the difficulties
encountered by English-speaking learners of Spanish L2, who struggle with
the Object-Verb-Subject (OVS) word order, in comprehension as well as pro-
duction, even after several semesters of instruction including a substantial

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

amount of time spent on this grammar point.
Based on this premise that input processing tends to be inadequate, and that
this is a major impediment to L2 learning, PI aims to alter learners’ processing
mechanisms where necessary and proposes three steps to accomplish this. The
first component of PI is explicit information (EI): learners are provided with
the properties and rules pertaining to the target form. Importantly, in EI, the
relationship between the form and its meaning should be overtly explained.
Secondly, PI incorporates information about processing strategies, designed to
steer learners away from acting according to default processing that would
result in incorrect form-meaning mappings. Finally, PI’s third component,
structured input (SI), consists of activities purposefully manipulated to contain
meaning-bearing input and which are deliberately designed to push learners to
actively process the target form and connect it to its function. SI comprises a
series of sentences or phrases where the learner has to make the correct choice
between different meanings by focusing on the crucial forms (referential activ-
ities). To use the example of OVS word order in Spanish L2 again, learners see
or hear sentences such as ‘la llama mi padre’ (my father is calling her) and
have to pick the one picture among several that depicts this and, crucially, not
the one in which a woman is calling the father (the interpretation students
would choose when misled by the English word-order processing strategy).
This activity is followed by a series of sentences that require learners to express
how they feel about the content of sentences containing the structure at issue
(affective activities). For example, in the case of the OVS structure, to say ‘me
alegra’ (I am happy about it) or ‘no me gusta’ (I don’t like it). Since PI seeks to
work on focus on and deepen the phase of acquisition when input is processed
(rather than the phase during which the internalized data are retrieved), SI
activities do not require learner output.1
This technique has shown promising results with structures that, in spite of
their frequency, are rarely correctly incorporated into communicative use of
the language, even after several years of instruction. Because of the funda-
mental nature of this problem, and given that PI does not require any invest-
ment of time or money compared with the alternative treatments, this is
potentially a very important finding for L2 instruction; it is crucial, therefore,
to evaluate these findings carefully. At the same time, PI research, if conducted
carefully, can provide better understanding of basic psycholinguistic questions
about second language acquisition (the role of enhanced input, explicit learn-
ing processes, or different kinds of practice in skill development).
Dozens of studies have been published, providing data on the learning of at
least 14 structures in at least 6 languages, with mostly college students, in a
292 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

few cases middle or high school learners, and in one case primary school stu-
dents. The participants in these studies were native speakers of at least seven
different languages. Almost all of these studies have provided data on learning
outcomes through both receptive and productive tests. The design has varied:
some have compared PI with a form of instruction that includes practice in
production (variably labeled traditional instruction (TI), meaning-based output

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

instruction, communicative output, or dictogloss); others have compared PI
with and without EI or PI with and without affective activities.
The outcome measures have also varied. Comprehension tests have some-
times consisted of matching pictures to the prompt sentences, while at other
times learners had to choose the words or phrases that best captured the
meaning conveyed by the grammatical structure in a given sentence; produc-
tion tests have sometimes consisted of a cued fill-in-the-blanks test, and at
other times have required learners to match first and second halves of
sentences.
Given this variety of empirical studies, the time is ripe for an assessment of
the effectiveness of PI, or more precisely of what effect the different compo-
nents of this set of techniques have on the learning of L2 structures.
We structured our review of the literature around three questions:
1. Is there any difference in the effects of PI and SI (i.e. with and without
explicit instruction) on L2 grammar acquisition?
2. Is there any difference in the effects of PI and production-based instruc-
tion on L2 grammar acquisition?
3. What factors moderate the effects of PI in the above comparisons?
We systematically examined the studies from the list agreed on with the
authors of the meta-analysis in this issue (Shintani et al. 2013); a few other
studies are mentioned in our text for contextualization only. We tabulated the
contents of the articles on a number of points (L1, L2, structure targeted,
number of participants, age of participants, treatments, outcome measures,
findings) and added additional comments that seemed important for the inter-
pretation (mainly weaknesses in design, participant recruitment, validity of the
outcome measures, or unexplained patterns in the data). The text below dis-
cusses the extant research and its findings for each of the three questions
separately.

PI WITH AND WITHOUT EI

From the earliest days of PI research, debate has surrounded the question of
how much the EI component contributes to students’ learning. VanPatten and
Cadierno (1993) already sought to examine the effects of different types of
explicit instruction in the acquisition of grammar. The paper reported on an
experiment comparing traditional methods of instruction2 to PI, where two
experimental groups (PI and TI) received instruction on Spanish OVS
sentences.3 Posttests revealed superior performance for PI over both TI and
 R. DEKEYSER AND G. PRIETO BOTANA 293

control groups in an interpretation task and no difference between PI and TI in

production, both performing significantly better than the no-instruction
control group. Although these results were interpreted by the authors as a
clear indication of the superiority of PI, critics pointed out that the study
confounded two variables: treatments for TI and PI groups differed not only
in the type of practice but also in the type of EI administered, as the TI group

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

received no strategies. In order to determine whether the recorded advantage
for PI lies in EI, VanPatten and Oikkenon (1996), also targeting OVS sentences,
carried out a study with three groups: a PI group (receiving EI, with informa-
tion on processing strategies and SI) was compared with an a SI group (which
differed from the former group in that participants received neither EI nor
information on processing strategies), and a third group that received EI and
strategies4 only and acted as control. Results revealed no significant differences
between the PI and SI groups, although both groups performed significantly
better than the EI-only group. These results indicated that the gains the PI
group exhibited did not originate in EI, thereby validating VanPatten and
Cadierno’s (1993) claim that PI is superior to TI. By extension, they also
appeared to suggest that SI is superior to TI.
Fueled by the possibility that a demonstrably more effective way to teach
grammar may have been identified, the field saw a proliferation of studies
seeking to replicate SI gains and the apparent lack of need for EI. Moving
beyond Spanish, Benati (2004a, 2004b) and Wong (2004) probed the robust-
ness of PI benefits in Italian and French. Using designs identical to VanPatten
and Oikkenon (1996), Benati (2004a, 2004b) featured the Italian gender
agreement and future tense, respectively, as the target linguistic unit. In
Benati (2004a), three outcome measures all revealed an improvement of PI
and SI groups compared with EI only, with no difference between the two
higher scoring groups. Similarly, results from Benati (2004b) showed that PI
and SI groups outperformed the EI-only group, exhibiting no difference be-
tween each other, and with gains being maintained a month after treatment.
Wong (2004) focused on the French negative particle de and also reported
significantly better results for SI and PI over EI and control groups. In line
with VanPatten and Oikkenon (1996), then, all three studies suggest that the
gains recorded in PI were to be attributed to the SI, EI being neither necessary
nor beneficial.
The first study to gauge the suitability of PI for a structure of higher
complexity was Farley (2004), targeting the Spanish subjunctive,5 and show-
ing that both PI and SI groups made significant gains. Farley’s results, however,
also revealed greater gains obtained by the PI group. This suggested that
although SI was sufficient for gains to occur, its benefits were not equal to
those generated by the full PI treatment (i.e. including EI), and invited specu-
lation of a possible facilitative role for EI for structures in which form-meaning
mapping is difficult to determine.
Fernández (2008) followed up on Farley’s results, conducting a study that
tracked students’ behavior while engaged in practice. Featuring the classical
294 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

PI vs. SI-only vs. EI-only design, Fernández targeted OVS sentences in add-
ition to the Spanish subjunctive. While results aligned with VanPatten and
Oikkenon (1996), revealing no difference between the two treatment groups
in OVS processing, the EI group started to process the subjunctive forms
significantly sooner than both PI and SI groups, lending credence to
Farley’s (2004) claim that EI may be beneficial with certain linguistic

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

phenomena.
In light of these results, Culman et al. (2009) and Henry et al. (2009) con-
ducted conceptual replications of Fernández (2008), this time with German
OVS sentences as target structure. The first study featured two different groups,
+EI and EI, and like Fernández (2008), it showed that the +EI group took
significantly fewer trials to reach criterion (Henry et al. 2009). This was also the
case for Culman et al. (2009), where two PI groups of different proficiencies
were also found to start processing OVS sentences faster than the analogous SI
groups, suggesting that EI provision had a beneficial effect. In other words,
these findings are in line with Farley (2004) and Fernández (2008), showing a
beneficial effect of EI, in this case even for OVS. Both Henry et al. (2009) and
Culman et al. (2009) argued that Spanish OVS required more complex explan-
ations than German OVS, and that this may have been taxing for learners, thus
making an EI effect less likely to surface.
Further evidence of the same effect comes from Russell (2012), another
study on the subjunctive. It compared the effects of SI and PI alone with
their effects when combined with visually enhanced input and added a TI
group to serve as control. In line with previous studies on the subjunctive,
results offered some evidence of the facilitative effect of EI, manifested in this
case by the superiority of the enhanced PI group over the non-enhanced SI
group, as well as a lack of difference between enhanced and non-enhanced PI
conditions. As in previous studies, production outcomes revealed significant
improvement after treatment, albeit this affected all groups alike. In general,
however, the main question that the subjunctive studies raised pertains to the
specific processing problems that are more amenable to SI-oriented treatment,
and the role of EI in those environments where SI alone may fall short. In the
following section, we review a number of relevant research endeavors that
shed light on this issue.

WHAT FACTORS MODERATE THE CONTRIBUTION OF EI

OVER SI?
Task-essentialness as a mediating variable
A study of particular interest to this discussion is Sanz and Morgan-Short
(2004). Designed to examine the individual and combined effects of EI (E)
and explicit feedback (F), this study featured four groups. The +E +F and +E
F groups received explicit instruction, one receiving feedback, the other one
not. Two groups did not receive explicit instruction, again one with and one
 R. DEKEYSER AND G. PRIETO BOTANA 295

without feedback. Although all groups improved from pretest to posttest, the
gains obtained were not significantly different across conditions for any of the
contrasts, suggesting that when SI is incorporated neither EI nor feedback is
necessary for OVS acquisition.
The importance of this study is twofold: first, Sanz and Morgan-Short
(2004) provide a somewhat novel definition of SI in that unlike VanPatten’s

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

and that of other authors, it overtly states that activities should be task-
essential. Coined by Loschky and Bley-Vroman (1993), task-essentialness
refers to practice conditions where the targeted item is essential for successful
task completions and where ‘the grammatical point itself is the ‘‘essence’’ of
what is being attended to’ (Loschky and Bley-Vroman 1993: 139). Sanz and
Morgan-Short (2004) attributed the lack of difference across conditions to the
task-essential nature of their treatments. Second, that lack of difference ex-
tended to the E F group, suggesting SI alone is not only capable of making
up for lack of EI (a result found before), but also that it is as effective as both EI
and feedback combined. A retrospective look at research of PI vs. SI paradigm
will reveal that all studies reporting a lack of EI effect implemented SI that
consisted of task-essential activities. This prompts the question as to whether
the task-essential nature of referential activities and the high level of control
involved in making a given form essential could favor autonomous rule
induction. The question then becomes whether dispensing with EI may only
be viable under practice conditions that favor rule induction. As DeKeyser
(2003) argues, the combination of relevant practice and feedback may success-
fully lead learners themselves to come up with the rules that are explicitly
provided to the full PI groups. An additional factor that may have conditioned
the role of EI in PI studies is that research implementing a PI vs. SI paradigm
generally allows participants a single chance to go over EI throughout the
duration of the experiment. This presents problems of ecological validity and
makes any positive effects of EI contingent upon participants’ memory and
degree of attention. Studies such as Marsden (2006) and Marsden and Chen
(2011) offer valuable findings in this respect.
Marsden (2006) reported on two experiments that compared a PI group,
which received EI and practice that required form-meaning connections,
with a group who also received EI but whose practice did not require form-
meaning connections. In one of the experiments, the PI group made significant
gains, whereas the group practicing in non-task-essential conditions did not.
The second experiment obtained results along the same lines, although the
group practicing under non-task-essential conditions did improve in some of
the outcome measures, which the author attributed to greater explicit know-
ledge of participants.
Marsden and Chen (2011) compared the performance of a PI group whose
practice consisted of referential activities, which were task-essential, with that
of a PI group whose activities were affective, and therefore exposed learners to
the targeted form without requiring it for task completion. Results suggested
that, unlike the referential activities, affective ones yielded no gains.
296 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

Both these studies lent credence to the critical nature of task-essentialness,

but exposed learners to EI on a single occasion, which leaves us wondering
what may happen if learners were able to rely on EI throughout the treatment.
Prieto Botana (2013) sought to address this issue through a study that isolated
the effects of task-essentialness and provided access to EI on several occasions
throughout the training phase. The experiment featured four experimental

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

groups, where the variables task-essential and EI were manipulated to create
the following conditions: EI and task-essential SI, EI and non-task-essential
SI, +EI and task-essential SI, as well as +EI and non-task-essential SI. A control
group was also incorporated to obtain a baseline. One hundred and thirty
learners of Spanish randomly assigned to each condition completed a com-
puter-delivered treatment on the ser/estar copula distinction and OVS struc-
tures. Participants in the +EI groups were exposed to the EI five times during
the training phase, and on each of those occasions they were warned that a
question on the information they read would follow. Results showed that
without TE, EI was necessary for any gains to obtain, both in interpretation
and production tasks, and for OVS and ser/estar alike. This supports the claim
that when rules cannot be induced easily, they need to be provided for learn-
ing to ensue. The study also revealed that whereas both EI groups were able to
transfer gains to production regardless of the target structure, the group receiv-
ing EI and task-essential SI was only able to do so for ser/estar, not OVS. In
addition, the +EI and task-essential SI condition scored significantly higher
than the EI and task-essential SI groups in one of the two ser/estar interpret-
ation tasks, suggesting that for certain grammatical structures EI can further
facilitate acquisition even when task-essential SI practice is at play. Finally,
delayed posttest results revealed that EI yielded more durable gains for OVS,
the only one of the target units for which significant gains were maintained,
underscoring once again that even when rule deduction is possible, EI can still
be helpful.
In summary, it appears that dismissing the potential of EI without further
qualification may be too hasty. Only rarely does input, whether in classroom
or naturalistic situations, lend itself to rule induction. While in those rare
occasions learners may be able to do without it, there is evidence to suggest
that in all other cases EI is a much-needed facilitative agent, without which
learning may simply not happen.

Strategies as a mediating variable

By all accounts, the conception of EI in the PI framework was and continues to
be novel for the attention it pays to the particular cognitive processes learners
engage in upon encountering the target form in discourse. White and DeMil
(2013) acknowledge this and explore the relative contribution of strategy-less
EI with a study featuring the classic PI and SI conditions, in comparison with a
group who received form-related explicit information, but no strategies or
practice. After receiving training in Spanish OVS sentences, results revealed
 R. DEKEYSER AND G. PRIETO BOTANA 297

that only PI and SI groups, but not the group receiving strategy-less EI, made
significant gains in interpretation tasks. Importantly, PI retained significantly
more gains than SI in the delayed posttest administered three weeks after
treatment.
In addition to looking at the effect of EI, this study also examined whether
any of the treatment forms were more likely to result in transfer of knowledge

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

to dative clitic pronouns. Interestingly, results showed that whereas both the
PI and SI groups were able to generalize, in this case the SI condition
maintained its gains significantly better than the PI groups, which suggests
that, clearly, these two approaches should not be seen as better or worse
than one another, but rather that each may be appropriate under different
circumstances.
To conclude this section, we emphasize that these findings not only reiterate
the facilitative effect of EI but also suggest that the extent to which EI can be
beneficial inevitably depends on its nature and quality. EI that does not have
default processing strategy instruction as its starting point will hardly resonate
with learners, a fact that may lead us to the erroneous conclusion that EI is of
no help in the language learning process. In this respect, the EI espoused by PI
frameworks and its overt focus on the strategies learners bring to the task of
processing language, constitutes a step forward that calls for further sophisti-
cation in the EI creation process rather than its omission. In the following
section, we shall also see that this notion of EI quality bears close ties with
our learners’ capacity to decode its content, in other words, with their gram-
matical sensitivity and general language learning aptitude.

Aptitude as a mediating variable

An issue of great importance with regard to the positive impact of EI is whether
its provision may benefit certain learners more than others. Given its metalin-
guistic nature, a rather common and largely unanswered question is whether
perhaps learners with higher aptitude may profit from these types of treat-
ments substantially more than low-aptitude learners. VanPatten et al. (2013)
delved into this issue by analyzing the correlation between grammatical
sensitivity and the outcomes for +EI and EI (i.e. PI and SI) groups across
four experiments involving French, German, Spanish, Russian. Three of the
four experiments revolved around the OVS, with the fourth targeting the
causative faire in French. Results varied across languages: interestingly,
Spanish and Russian exhibited parallel results, giving evidence of neither an
effect for EI nor a substantial correlation with grammatical sensitivity. In the
case of German, a beneficial EI effect was found, echoing Culman et al. (2009)
and Henry et al. (2009), and accompanied by significant correlations of out-
comes with grammatical sensitivity, albeit for the +EI group only. The French
learners who received EI started processing causative faire sentences correctly
significantly sooner, but no significant correlation was found for grammatical
sensitivity. VanPatten et al. (2013) put forth a complexity-based argument to
298 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

account for some of the differences in results. They argue, for example, that
the surface forms in the Russian OVS resulted in EI that required greater pro-
cessing in real time. In the case of German and French, authors claim that EI
was less complicated as compared with Russian EI, and was, therefore, access-
ible. Here too, we might speculate that French EI was perhaps simpler than
German EI, making it equally accessible for those of high and low grammatical

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

sensitivity alike.
VanPatten et al.’s (2013) study reaffirms that EI is beneficial, at the very least
in as much as it expedites learning. Given that after completion of the entire
treatment EI and +EI groups did not differ from each other, it may be tempt-
ing to argue that EI is, at the end of the day, not essential, or at least not to the
unhurried learner. Yet, that view seems to ignore the fact that the sooner the
learner makes the form-meaning mappings, the sooner he/she will further
notice the target form in subsequent input, whether in or outside the class-
room. Considering that input is scarce and bearing in mind the role that fre-
quency of noticing plays in language learning, the EI contribution to a faster
connection between form and meaning appears to be an invaluable asset.
At any rate, the results from this study should be interpreted with caution,
as VanPatten et al. (2013) still implemented the classic single-EI-exposure
paradigm. It seems reasonable to think that given a single chance to read
the rules and strategies, the full potential of EI is far from realized in their
study, as it is in the vast majority of PI studies. In conclusion, it would be
premature, to say the least, to conclude from the existing PI research compar-
ing +EI and EI treatments that EI is superfluous.

PI WITH AND WITHOUT OUTPUT PRACTICE

For our second question, we examined studies comparing a PI group with a
group that had output practice (hereafter PB for production-based); in most
cases there was also a control group.
When the outcome measures were receptive, about half of these PI vs. PB
studies reported an advantage for the PI group on the posttests: Farley (2001a);
Benati (2005, 2009); Benati et al. (2008a, 2008b) and Uludag and VanPatten
(2012). Roughly the other half of the studies yielded no significant difference
between PI and PB for receptive measures: Farley (2001b); Buck (2006);
Morgan-Short and Bowden (2006); Lee and Benati (2007a, 2007b); Keating
and Farley (2008); Qin (2008) and VanPatten et al. (2009); Farley and Aslan
(2012). Finally, Toth (2006) found no difference using a grammaticality
judgment test as receptive outcome measure. No study found a significant
advantage for PB over PI with receptive measures.
When the outcome measures were productive, the picture was roughly the
inverse: where PI > PB or PI = PB for receptive measures, PI < PB or PI = PB for
productive measures. The majority of the studies showed no significant differ-
ence. The studies that did show a significant advantage of PB over PI with
productive measures were Morgan-Short and Bowden (2006); Toth (2006);
 R. DEKEYSER AND G. PRIETO BOTANA 299

Keating and Farley (2008); Farley and Aslan (2012). The only study that
showed an advantage for PI even in production was Benati et al. (2008b).
There is a strong tendency for studies with a significant advantage for PI in
comprehension not to show a significant difference in production; conversely,
when there is a significant advantage for PB in production, then there is usu-
ally no significant difference in comprehension. In other words, the advantage

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

of PI is never stronger in production than in comprehension and the advantage
of PB is never stronger in comprehension than in production. It appears then,
that everything else being the same, the relative advantage of PI is clearer for
comprehension than on production measures, and even questionable for
production measures. This is not surprising, as PI provides comprehension
practice only.
Furthermore, if there is an advantage for either treatment at the immediate
posttest, this usually diminishes or disappears on delayed posttests.
Interestingly, the four studies that show an advantage for PB over PI for pro-
duction and no significant difference for comprehension (Morgan-Short and
Bowden 2006; Toth 2006; Keating and Farley 2008; Farley and Aslan 2012) all
stressed the more communicative implementation of their production treat-
ments than in most other PI studies.
One more element needs to be discussed in this context. Recent PI literature
has introduced a new question, about transfer from the structure taught and
practiced (called primary) to structures that are not trained but presumably
draw on the same processing strategy (called secondary effects) and to struc-
tures that call for different strategies (called cumulative effects). These studies
tend to show little transfer from the primary structure. Benati (2009) found an
advantage for the primary structure only (in this case negation in present tense
in Japanese), not the secondary structure (in this case past tense); similarly,
Benati et al. (2008b) found an advantage for the primary structure only
(imparfait in French), not the secondary (subjunctive) or cumulative structure
(a structure that draws on a different processing principle, in this case the
causative).
The overall PI vs. PB results should be interpreted with caution because, as
was pointed out by several researchers, in the comprehension measures stu-
dents are tested on the skill that was stressed in PI (and often in a very similar
format); the same situation also applies in the other direction. It is only logical
that students improve more in the areas (closest to the ones) they were tested
in. From the point of view of skill acquisition theory, the learners in these
experiments acquire declarative knowledge about form-meaning mapping,
which they proceduralize in very specific ways depending on the nature of
the practice they get (cf. DeKeyser and Sokalski 1996, 1997, 2015).
On the other hand, as all the studies that made a PI–PB comparison incor-
porated EI in both treatments, including information about how to process a
sentence to arrive at the correct form-meaning mapping, the often docu-
mented significant effect of PI for comprehension suggests that the information
about processing helped with proceduralization of declarative knowledge.
300 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

Factors moderating the differential effectiveness of the PI and

PB treatments
Besides the difference in effectiveness of PI depending on the outcome meas-
ures (comprehension vs. production), do other factors play a role in the +/ EI
or PI/PB differences, such as specific variants of the treatments, age or aptitude

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

of the learner, or the nature of the structure being taught? Answering this
question from the available data is not easy, because to this day there has been
remarkably little variation among PI studies, especially taking into account
their large number.
The variation in structure taught and in treatment implementation and
design is somewhat larger than in the early days of PI instruction, but still
limited. Of the +/ EI studies reviewed here, seven deal with OVS and five
with the subjunctive in Spanish, whereas Italian noun-adjective gender agree-
ment and future tense, Spanish ser/estar copula distinction and French nega-
tive particle de and causative sentences are the focus of one study each. Among
the PI/PB studies, two used OVS as primary structure (one in Spanish and one
in German), five the subjunctive (two in Spanish, one in French, one in
Italian, and one in English), four the past tense (two in English, one in
French, and one in Japanese), and seven a variety of structures (two negation
in Japanese present tense, one OVS in Spanish, one OVS in German, one
gender agreement in Italian, one present progressive in English, and one pas-
sive in English). As these structures are represented by only one to seven
studies each, it is hard to come to any conclusions about the role of the struc-
ture taught in the effectiveness of PI +/ EI or of PI vs. PB.
With regards to treatment implementation, the variation comes mostly from
the PB side. Several publications stress that their PB treatment was as commu-
nicative as the PI treatment. It is interesting that four of these studies found an
advantage for PB on at least some measures (three found no significant differ-
ences; none found an advantage for PI). This corroborates Toth’s (2006) hypoth-
esis that the positive results for PB were due to the more ‘communicative’
nature of the output practice in these studies compared with most PI research
(or conversely, that the limited effectiveness of output-based instruction in the
PI literature may be due to the mechanical-drill nature of most practice).
Further limitations are the scant information about the role of aptitude and
the limited variation in age of the learners. Aptitude of any kind is almost
never included in the design, not even as a covariate (exceptions being
Culman et al. 2009; Henry et al. 2009; VanPatten et al. 2013), making any
conclusions about differential effectiveness of the contrasting methodologies
with different kinds of learners impossible so far. In the same vein, almost all
studies were conducted with young adult learners (college); very few have
dealt with adolescents (secondary school) or with children (primary school);
this clearly limits the generalizability of the findings to younger learners.
Finally, almost all PI research has been carried out with foreign language
learners with little or no access to input beyond the classroom.
 R. DEKEYSER AND G. PRIETO BOTANA 301

SUMMARY AND CONCLUSIONS

The body of research in this area clearly shows that PI can be an effective
instructional technique, in the sense that all studies reviewed find a significant
difference between PI and a control group without instruction for the structure
at issue. With regard to our two main research questions, however, the answer

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

is less clear. Our review of the literature paints a more nuanced picture than
the generalizations that are often made about the role of EI in PI and about the
contribution of production practice beyond the SI inherent in PI:
1. While EI did not make a significant difference in many studies, that may
have been because of incomplete learning of the EI on the one hand, or
because of the task-essentialness of the SI practice in the absence of EI at
the start, which encouraged induction of explicit knowledge, on the
other. Explicit knowledge from good inductive learning is better, of
course, than lack of it from deductive learning. What matters, ultimately,
is how much (explicit) knowledge there is to be proceduralized and
automatized, rather than what the ultimate source of that knowledge
is. Be that as it may, the studies reviewed show that EI can be beneficial
when provided regularly, especially with structures of greater complex-
ity, such as the subjunctive, and that gains derived from it may be more
robust than those obtained with rule induction (in the sense that they
are more likely to be found not only on immediate, but also on delayed
posttests). For easier structures (such as OVS), the SI or other forms of
practice may be enough to allow successful rule induction for most
learners.
2. While it has often been argued that PI is better than PB for comprehen-
sion ability and that there is no difference for production ability, our
review shows that the PI advantage for comprehension is found in
only half of the relevant studies, and that PB can be more effective
than PI for productive outcome measures, depending mainly on how
the output practice is implemented. The lack of advantage of PB docu-
mented in several studies may be the result of the non-communicative,
drill-like nature of the practice they provided.
What stands out the most from our literature review, however, is how all
these questions are in need of more systematically varied structures, treat-
ments, and learner characteristics.
It should be noted also that, with the exception of Prieto Botana (2013) and
Marsden and Chen (2011), none of the studies mentioned included an assess-
ment of how well students had learned the EI, nor did they control whether
the EI had even been attended to. Clearly one would want to isolate the effect
of incomplete learning of EI, especially as the EI was, again with the exception
of Prieto Botana (2013), only presented once, at the beginning of the treat-
ment. Moreover, most studies in this paradigm had inadequate or marginal
sample sizes, which may be one of the main reasons why so many comparisons
302 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

were insignificant. Most studies did not have a design to exclude teacher effects
either. Exceptions were Marsden (2006), where teachers alternated between
conditions, and Marsden and Chen (2011) and Sanz and Morgan-Short
(2004), where the instruction was delivered by computer.
Perhaps most importantly, all the studies in the systematic review had treat-
ments of very short duration (as is the case for most experimental studies on

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

Second Language Acquisition (SLA) of course) and used outcome measures
that allowed a considerable degree of monitoring. In other words, they docu-
mented the extent of (partially) proceduralized declarative knowledge, not of
automatized, let alone implicit knowledge. The research reviewed here, there-
fore, needs to be supplemented with studies of much longer duration, and with
outcome measures that allow less monitoring. (It is interesting in this regard
that one of the few studies to use a more communicative outcome measure (a
structured interview), VanPatten and Sanz (1995), found only a small differ-
ence comparing PI with no instruction for the structure at issue. Without re-
search of longer duration and without more communicative measures, the PI
literature will remain very skewed toward short-term gains of largely explicit/
declarative knowledge, at a time when the SLA literature as a whole is putting
more emphasis on testing with more communicative tasks and/or more impli-
cit knowledge.
Finally, where the relative effect of input vs. output practice is concerned, it is
interesting to compare the findings here with those of Shintani et al. (2013),
who concluded from their broader meta-analysis on comprehension-based vs.
PB instruction that the apparent short-term advantage of input practice over
output practice was due to the studies carried out in the PI paradigm. The rela-
tive contribution of EI, input practice, and output practice is far from settled.

NOTES
1 Although the definition of SI does not meaningful, communicative; Paulston
make overt mention of feedback re- 1972). Note that the practice component
quirements, referential activities usu- in PI (i.e. SI) comprises purely receptive
ally include feedback after processing practice.
of each sentence to let learners know 3 Although the canonical order in the
whether their response was correct or target language is SVO, due to its rich
not, without providing any information morphology, Spanish exhibits a fairly
as to why. flexible word order. Structures featur-
2 This study, as well as subsequent studies ing the direct object pronoun in sen-
in the same paradigm, operationalized tence initial position, and in which
TI as instruction that progressed from the subject appears post-verbally, are
provision of metalinguistic information both grammatical and common. It is a
to practice that involved production of documented phenomenon that such
the targeted form. In most PI studies, structures are problematic for language
as discussed in more detail in the last learners both in L1 and L2 acquisition.
section, this meant practice limited to Evidence from various studies indicates
the first stages of MMC (mechanical, that both first (Bever 1970; Bates et al.
 R. DEKEYSER AND G. PRIETO BOTANA 303

1984) and second language learners two separate items, in studies such as
(LoCoco 1987; Gass 1989) tend to VanPatten and Cadierno (1993) and
interpret the first noun phrase (NP) in VanPatten and Oikkenon (1996) as
any sentence as the subject well as in all their replications, he
(VanPatten’s First Noun Principle). does use the label ‘EI’ to refer to both
When parsing OVS structures, the strat- practices combined.
egy of parsing the first NP as the subject 5 Although the claim that the subjunct-

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

results in the erroneous allocation of ive is of greater complexity is rather
agent and theme roles. uncontroversial and is rationalized in
4 While VanPatten (1996) clearly con- Farley (2004), there is no evidence to
siders EI and strategy suppliance as date to support this claim empirically.

REFERENCES
Bates, E., B. MacWhinney, C. Caselli, in A. Benati and J. Lee (eds): Grammar
A. Devescovi, F. Natale, and V. Venza. Acquisition and Processing Instruction: Secondary
1984. ‘A cross- linguistic study of the develop- and Cumulative Effects. Multilingual Matters,
ment of sentence interpretation strategies,’ pp. 121–57.
Child Development 55: 341–54. Bever, T. 1970. ‘The cognitive basis for
Benati, A. 2004a. ‘The effects of processing in- linguistic structures’ in J. Hayes (ed.):
struction and its components on the acquisi- Cognition and the Development of Language.
tion of gender agreement in Italian,’ Language Wiley, pp. 279–362.
Awareness 13/2: 67–80. Buck, M. 2006. ‘The effects of processing instruc-
Benati, A. 2004b. ‘The effects of structured input tion on the acquisition of English progressive
activities and explicit information on the aspect,’ Estudios de Lingüı́stica Aplicada 43:
acquisition of the Italian future tense’ 77–95.
in B. VanPatten (ed.): Processing Instruction: Culman, H., N. Henry, and B. VanPatten.
Theory, Research, and Commentary. Laurence 2009. ‘The role of explicit information in in-
Erlbaum Associates, pp. 207–26. structed SLA: An on-line study with processing
Benati, A. 2005. ‘The effects of processing in- instruction and German accusative case inflec-
struction, traditional instruction and mean- tions,’ Die Unterrichtspraxis/Teaching German
ing-output instruction on the acquisition of 42/1: 19–31.
the English simple past tense,’ Language DeKeyser, R. 1997. ‘Beyond explicit rule learn-
Teaching Research 9/1: 67–93. ing: Automatizing second language morpho-
Benati, A. 2009. Japanese Language Teaching: A syntax,’ Studies in Second Language Acquisition
Communicative Approach. Continuum. 19/2: 195–221.
Benati, A., J. Lee, and S. D. Houghton. 2008a. DeKeyser, R. 2003. ‘Implicit and explicit learn-
‘From processing instruction on the acquisition ing’ in M. Long and C. Doughty (eds): The
of English past tense to secondary transfer- Handbook of Second Language Acquisition.
of-training-effects on English third person sin- Blackwell, pp. 313–48.
gular present tense’ in A. Benati and J. Lee DeKeyser, R. 2015. ‘Skill acquisition theory’
(eds): Grammar Acquisition and Processing in B. VanPatten and J. Williams (eds): Theories
Instruction: Secondary and Cumulative Effects. in Second Language Acquisition: An Introduction
Multilingual Matters, pp. 188–21. (2nd edition). Routledge, pp. 94–112.
Benati, A., J. Lee, and C. Laval. 2008b. ‘From DeKeyser, R. M. and K. J. Sokalski. 1996.
processing instruction on the acquisition of ‘The differential role of comprehension and
French imparfait to secondary transfer- production practice,’ Language Learning 46/4:
of-training effects on French subjunctive 613–42.
and to cumulative transfer-of-training effects Erlam, R. 2005. ‘Language aptitude and its rela-
with French causative constructions’ tionship to instructional effectiveness in second
304 THE EFFECTIVENESS OF PI IN L2 GRAMMAR ACQUISITION

language acquisition,’ Language Teaching T. R. Dvorak, and J. F. Lee (eds):

Research 9/2: 147–71. Foreign Language Learning: A Research Perspective.
Farley, A. 2001a. ‘Authentic processing instruc- Newbury House, pp. 119–29.
tion and the Spanish subjunctive,’ Hispania 84/ Loschky, L. and R. Bley-Vroman. 1993.
2: 289–99. ‘Grammar and task-based learning’
Farley, A. 2001b. ‘Processing instruction and in G. Crookes and S. Gass (eds): Tasks and
meaning-based output instruction: a com- Language Learning: Integrating Theory and

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

parative study,’ Spanish Applied Linguistics 5/2: Practice. Multilingual Matters, pp. 123–67.
57–93. MacWhinney, B. 1997. ‘Second language acqui-
Farley, A. 2004. ‘Processing instruction and the sition and the competition model’ in A. M.
Spanish subjunctive: is explicit information B. de Groot and J. F. Kroll (eds): Tutorials in
needed?’ in B. VanPatten (ed.): Processing Bilingualism. Psycholinguistic Perspectives.
Instruction: Theory, Research, and Commentary. Erlbaum, pp. 113–42.
Erlbaum, pp. 227–39. Marsden, E. 2006. ‘Exploring input processing in
Farley, A. and E. Aslan. 2012. ‘The relative ef- the classroom: an experimental comparison of
fects of processing instruction and meaning- processing instruction and enriched input,’
based output instruction on L2 acquisition of Language Learning 56/3: 507–66.
the English subjunctive,’ ELT Research Journal Marsden, E. and H.-Y. Chen. 2011. ‘The roles
1/2: 120–41. of structured input activities in processing
Fernández, C. 2008. ‘Reexamining the role of instruction and the kinds of knowledge they
explicit information in processing instruction,’ promote,’ Language Learning 61/4: 1058–98.
Studies in Second Language Acquisition 30: Morgan-Short, K. and H. W. Bowden. 2006.
277–305. ‘Processing instruction and meaningful
Gass, S., I. Svetics, and S. Lemelin. 2003. output-based instruction: effects on second
‘Differential effects of attention,’ Language language development,’ Studies in Second
Learning 53/3: 497–545. Language Acquisition 28/1: 31–65.
Gass, S. M. 1989. ‘How do learners resolve lin- Qin, J. 2008. ‘The effect of processing instruction
guistic conflicts?’ in S. Gass and J. Schacter and dictogloss tasks on acquisition of the
(eds): Linguistic Perspectives on Second Language English passive voice,’ Language Teaching
Acquisition. Cambridge University Press, Research 12/1: 61–82.
pp. 183–99. Paulston, C. B. 1972. ‘The sequencing of struc-
Henry, N., H. Culman, and B. VanPatten. tural pattern drills,’ TESOL Quarterly 6: 197–208.
2009. ‘More on the effects of explicit informa- Prieto Botana, G. 2013. ‘The role of task-
tion in SLA,’ Studies in Second Language essentialness and explicit information in pro-
Acquisition 31: 559–75. cessing instruction,’ Unpublished Ph.D. disser-
Keating, G. D. and A. Farley. 2008. ‘Processing tation, University of Maryland, College Park.
instruction, meaning-based output instruction, Russell, V. 2012. ‘Learning complex grammar in
and meaning-based drills: Impacts on class- the virtual classroom: a comparison of process-
room L2 acquisition of Spanish object pro- ing instruction, structured input, computerized
nouns,’ Hispania 91/3: 639–50. visual input enhancement, and traditional in-
Lee, J. F. and A. Benati. 2007a. Second Language struction,’ Foreign Language Annals 45/1: 42–71.
Processing. An analysis of Theory, Problems, and Sagarra, N. and N. C. Ellis. 2013. ‘From seeing
Possible Solutions. Continuum. adverbs to seeing verbal morphology: language
Lee, J. F. and A. Benati. 2007b. ‘Comparing experience and adult acquisition of L2 tense,’
modes of delivering processing instruction Studies in Second Language Acquisition 35/2:
and meaning-based output instruction on 261–90.
Italian and French subjunctive’ in J. F. Lee Sanz, C. and K. Morgan-Short. 2004. ‘Positive
and A. G. Benati (eds): Delivering Processing evidence versus explicit rule presentation and
Instruction in Classrooms and in Virtual Contexts. explicit negative feedback: a computer-assisted
Equinox, pp. 99–136. Study,’ Language Learning 54: 35–78.
LoCoco, V. 1987. ‘Learner comprehension of oral Shintani, N., S. Li, and R. Ellis. 2013.
and written sentences in German and Spanish: ‘Comprehension-based versus production-
The importance of word order’ in B. VanPatten, based grammar instruction: A meta-analysis
 R. DEKEYSER AND G. PRIETO BOTANA 305

of comparative studies,’ Language Learning communicative tasks’ in F. Eckman,

63/2: 296–329. D. Highland, P. Lee, J. Milcham, and R.
Toth, P. 2006. ‘Processing instruction and a role R. Weber (eds): Second Language Acquisition
for output in second language acquisition,’ Theory and Pedagogy. Lawrence Erlbaum
Language Learning 56/2: 319–85. Associates, pp. 169–85.
Uludag, O. and B. VanPatten. 2012. ‘The com- VanPatten, B., E. Collopy, J. E. Price,
parative effects of processing instruction and S. Borst, and A. Qualin. 2013. ‘Explicit infor-

Downloaded from https://academic.oup.com/applij/article/36/3/290/2422464 by NILE user on 11 October 2023

dictogloss on the acquisition of the English mation, grammatical sensitivity, and the first-
passive by speakers of Turkish,’ IRAL 50: noun principle: a cross-linguistic study in
189–212. processing instruction,’ The Modern Language
VanPatten, B. 1996. Input Processing and Journal 97/2: 506–27.
Grammar Instruction: Theory and Research Ablex. VanPatten, B., D. Inclezan, H. Salazar, and
VanPatten, B. 2012. ‘Input processing’ A. Farley. 2009. ‘Processing instruction
in S. Gass and A. Mackey (eds): The Routledge and dictogloss: a study on object pronouns
Handbook of Second Language Acquisition. and word order in Spanish,’ Foreign Language
Routledge, pp. 268–81. Annals 42/3: 557–75.
VanPatten, B. and T. Cadierno. 1993. ‘Explicit White, J. and A. DeMil. 2013. ‘Transfer of non-
instruction and input Processing,’ Studies in training effects in processing instruction,’
Second Language Acquisition 15: 225–44. Studies in Second Language Acquisition 35:
VanPatten, B. and S. Oikkenon. 1996. 519–44.
‘Explanation vs. structured input in processing Wong, W. 2004. ‘Processing instruction in
instruction,’ Studies in Second Language French: the roles of explicit Information and
Acquisition 18: 495–510. structured input’ in B. VanPatten (ed.):
VanPatten, B. and C. Sanz. 1995. ‘From input Processing Instruction: Theory, Research, and
to output: processing instruction and Commentary. Erlbaum, pp. 187–205.