PROCESSING INSTRUCTION AND EXPLICIT INFORMATION: REVIEW OF THE LITERATURE

Processing instruction is a pedagogical technique based on the principles of the model of input processing proposed by VanPatten (Reference VanPatten1996, Reference VanPatten2002, Reference VanPatten and VanPatten2004). The principles outlined by this model are based on the default processing strategies that learners' linguistic processors apply to make sense of the input; for example, the lexical preference principle states that learners tend to rely on lexical items and not on grammatical form to get meaning when both encode the same semantic information (VanPatten, Reference VanPatten2002, Reference VanPatten and VanPatten2004). The goal of PI is to alter the nonoptimal default processing strategies that lead L2 learners not to process forms or to process them incorrectly; this is accomplished in part by manipulating the input provided to learners. The modified input, called structured input (SI), pushes learners to change their default processing strategies in favor of other strategies that are more effective. A complete PI treatment promotes such a strategy change by (a) giving EI, (b) giving information about an inappropriate input processing strategy, and (c) exposing learners to SI.

Explicit information is focused on one linguistic form or structure at a time, and it informs learners about how the form or the structure works. It also includes information about a particular default—but inappropriate—processing strategy and how this strategy does not lead learners to correctly process certain forms or structures. Table 1 provides an example of how EI might look in PI. The target form in this example is the third person singular of the Spanish subjunctive in expressions of doubt. EI tells learners what the subjunctive looks like, what its location is in the sentence, and when it is used. The “Be careful!” section also includes information to alert learners about a faulty input processing strategy—in this case, the strategy described in the lexical preference principle (VanPatten, Reference VanPatten2002, Reference VanPatten and VanPatten2004).

Table 1.

Example of explicit information in PI

From “The Relative Effects of Processing Instruction and Meaning-Based Output Instruction on L2 Acquisition of the Spanish Subjunctive,” by Farley, Reference Farley2000. Unpublished doctoral dissertation, University of Illinois, Urbana-Champaign. Reprinted with permission.

The SI is presented in activities called structured input activities, which are of two types: referential and affective. Referential activities always have a correct answer, whereas affective activities do not require a correct answer; rather, learners respond with what is true or false for them, with their opinion, or with what applies for them. A referential activity with the third person Spanish subjunctive in expressions of doubt is presented in (1) and an affective activity also using the subjunctive is presented in (2).

1. (1)

   

2. (2)

   

In both types of activities, learners must attend to the input and process it correctly in order to understand what they are reading or listening to. In (1), the verbs in the aural sentences are inflected either in the subjunctive or in the indicative mood, and the written options convey the meaning of either doubt (e.g., No creo que … “I don't believe that…”) or certainty (e.g., Yo sé que… “I know that…”). Learners must attend to the inflection of the verb in the aural phrases in order to correctly select the meaning that the speaker is conveying (i.e., doubt or certainty). In (2), learners need to understand both the phrase No creo que… “I don't believe that…” and the subordinate clauses to select the one(s) with which they agree. Note that the input has been structured so that the inflected verbs appear in initial position to make them more salient. Because these are input-based activities, learners are not asked to produce language (see Wong, Reference Wong and VanPatten2004a, for a detailed description of PI).

VanPatten and Oikkenon (Reference VanPatten and Oikkenon1996) began to investigate the role of EI in PI by considering whether the effectiveness of PI demonstrated in an earlier study by VanPatten and Cadierno (Reference VanPatten and Cadierno1993) was due to SI only or to EI. Unlike VanPatten and Cadierno, VanPatten and Oikkenon separated out EI in their treatment and considered it as an independent variable. Three groups were compared in VanPatten and Oikkenon's study: a PI group that received EI, SI activities, and feedback (exactly as in VanPatten and Cadierno's study); a SI group that only received SI activities and feedback on whether their answers were correct or incorrect (but not why; i.e., they did not receive EI in any part of the treatment); and an EI group that only received information about the target structures and the particular input processing strategy that negatively affects their correct interpretation. The target structures were Spanish object-verb-subject (OVS) sentences with clitic pronouns (e.g., lo escucha la abuela “The grandma listens to him,” in which lo is the object, escucha is the verb, and la abuela is the subject). The assessment consisted of an immediate posttest that included both a sentence-level interpretation task and a sentence-level production task. The results showed that both the PI group and the SI group made gains on the interpretation task, and the groups were not different from each other. The EI group showed no improvement at all on the interpretation task. On the production task, all groups improved. The PI group improved significantly more than the EI group. The SI group scores were lower than the PI group scores and higher than the EI group scores, but these differences were not statistically significant. The authors interpreted these findings as indicating that SI alone is sufficient to cause improvement on interpretation and production tasks and that EI is not necessary or even beneficial for PI. Although the EI group improved and had statistically equal gains as the SI group on the production task, the authors argued that these results show that SI alone is sufficient to cause gains, because the SI group performed as well as the PI group in both of the modalities.

Wong (Reference Wong and VanPatten2004b) conducted a study to investigate whether the results of VanPatten and Oikkenon's (Reference VanPatten and Oikkenon1996) study would generalize to other linguistic structures that do not communicate new information and thus present different processing problems. The target forms in her study were the French de/d' forms used in negative constructions. Following the lexical preference principle (VanPatten, Reference VanPatten1996, Reference VanPatten2002, Reference VanPatten and VanPatten2004), learners will skip the grammatical de/d' forms in negative sentences because they compete with the lexical form ne … pas “not” to communicate negation. The de/d' forms, then, are redundant. Wong examined whether SI alone would be sufficient to make learners process these forms or if EI would be beneficial or even necessary. She divided participants into four groups: a PI group that received EI plus SI activities, a SI group that received only SI activities, an EI group that received only EI (as in VanPatten & Oikkenon), and a control group that did not receive any instruction. The treatment for the PI group consisted of giving learners EI on the meaning and the function of the target structures and on the learners' tendency to overlook the forms de/d' in negative structures. After EI, learners did SI activities designed to push them to attend to de/d' or un/une “a” and not to rely on the construction ne … pas “not” to determine whether sentences were negative or affirmative. The SI group did not receive EI; they only did the SI activities and received feedback on whether their responses on each item were right or wrong (but not why). The EI group received the information on the target structures and on the faulty processing strategy but did not do the SI activities. The assessment consisted of an immediate posttest that included both a sentence-level interpretation task and a sentence-level production task. The results showed that on the interpretation test, both the SI and PI groups improved equally and more than the EI and control groups. The EI group improved, but the results were not different from the control group. On the production task, all treatment groups improved. The PI group and the SI group were different from the control group but were not different from each other. The PI group was different from the EI group, and the EI group was not different from the control group. The SI group and the EI group were not different from each other. Unlike in VanPatten and Oikkenon's study, the EI group improved in both of the modalities, which suggests that EI might have helped learners correctly interpret and produce the target structure used for this study. However, Wong argued that because the EI group was not different from the control group, the EI contribution was minimal, particularly when compared with the improvements made by the SI and PI groups. She concluded that SI is sufficient to cause gains in production and interpretation of the target forms and that EI plays no important role.

Benati (Reference Benati and VanPatten2004) set out to replicate VanPatten and Oikkenon's (Reference VanPatten and Oikkenon1996) study, investigating the role of EI in the interpretation and production of Italian third person future forms. As in Wong (Reference Wong and VanPatten2004b), the processing problem these structures present is stated in the lexical preference principle (VanPatten, Reference VanPatten1996, Reference VanPatten2002, Reference VanPatten and VanPatten2004). In this case, learners tend not to attend to the verb ending that denotes future; instead, they rely on adverbial phrases (tomorrow, next week, etc.) to get the meaning. Benati divided 38 college students of Italian into three groups: a PI group that received EI plus SI activities, a SI group that only received SI activities, and an EI group that only received EI. The treatment for the PI group consisted of EI on the form and function of the target forms as well as on learners' tendency to rely on adverbial phrases to get the future tense meaning. After EI, the PI group did SI activities. The SI group did the activities without any EI. The EI group only received the information on the target structures and on the faulty processing strategy. There was an immediate posttest, and a delayed posttest was given 1 month later. Both of the posttests consisted of sentence interpretation and production tasks. For interpretation, the results showed that all groups improved. The EI group was different from both the PI and the SI groups. The PI group and the SI group improved more than the EI group and were not different from each other. The same results were maintained 1 month later. For the production test, all groups improved. The PI group improved more than the EI group but was not different from the SI group, and the pattern of results was maintained 1 month later. Unlike in VanPatten and Oikkenon and in Wong, this time the SI group was different from the EI group on both the interpretation and production tests. Benati concluded that SI is sufficient for learners to correctly process and produce the target structures. However, EI might have an effect—albeit minimal if compared to the improved performance of both the PI and SI groups—as seen in the EI group's improvement in the two modalities.

Sanz and Morgan-Short (Reference Sanz and Morgan-Short2004) examined the effects of computer-delivered EI and explicit feedback on the acquisition of Spanish OVS word order through PI treatment. Participants were randomly assigned to four groups: a PI group that received explicit feedback (+EI, +F), a PI group that did not receive explicit feedback (+EI, −F), a SI-only group that received explicit feedback (−EI, +F), and a SI-only group that did not receive explicit feedback (−EI, −F). The treatment was taken from VanPatten and Cadierno's (Reference VanPatten and Cadierno1993) paper-and-pencil materials and adapted to a computer-assisted language learning format. There were 56 input practice items: Some required learners to read a sentence and choose the picture it referred to, whereas others required learners to do other tasks, such as reading a text or interpreting what a short article was about based on its title. The computer-delivered EI was given to the PI groups right before the activities. Feedback was delivered immediately after each response; it was explicit for the +F groups (e.g., it explained how the Spanish word order works) or in the form of “sorry, try again” or “OK” for the −F groups. Participants also completed an interpretation and production pretreatment test and an interpretation and production posttreatment test based on the VanPatten and Cadierno materials. An additional production posttest, not included in the VanPatten and Cadierno materials, consisted of a written video-retelling task that required learners to produce the target form at the suprasentential level. The results of the study showed that all groups increased in their ability to interpret and produce OVS-type sentences after treatment, but no effect was found for explanation or for feedback on any of the tests. Hence, the amount of EI provided both before the tasks and during the tasks (as explicit feedback) did not have any impact on participants' ability to interpret and produce Spanish OVS-type sentences. Sanz and Morgan-Short claimed that the fact that the SI activities were task-essential (i.e., learners needed to process the critical forms to complete the task) was the reason why supplemental information was not necessary or even beneficial.

Farley (Reference Farley and VanPatten2004a) undertook another replication of VanPatten and Oikkenon's (Reference VanPatten and Oikkenon1996) study in an attempt to determine whether the results observed in this and later PI studies would also obtain with a form, such as the Spanish subjunctive, known to be difficult for learners.Footnote ¹ Farley tested whether SI activities alone bring about improved performance on interpretation and production of third person singular Spanish subjunctive forms in expressions of doubt and, if so, whether the gains would be the same for a SI-only group as for a group receiving SI plus EI (i.e., a PI group). Unlike VanPatten and Oikkenon, Wong (Reference Wong and VanPatten2004b), and Benati (Reference Benati and VanPatten2004), Farley's study only included two experimental groups: a PI group and a SI-only group. The SI was designed to (a) push learners to notice and process the meaning of the main clause (i.e., doubt or certainty) by requiring them to attend to the verb inflection to complete the task and (b) increase verb salience by placing the inflected verbs at the beginning of the sentence. The EI explained how the subjunctive works. Results showed that both the PI and the SI groups improved significantly from the pretest to the posttest, but, unlike the previous studies, the PI group improved more than the SI group in both modalities. Because participants had to rely on the subjunctive forms to complete the tasks, the SI group not performing as well as the PI group was not due to a lack of task essentialness; Farley explained this difference in outcome as due to the nature of the target form, arguing that whereas other linguistic features such as tense or agency are readily transparent in meaning, mood as a feature might not be as clear to learners. The range of expressions that convey doubt (it is possible, I'm not sure, etc.) may not communicate uncertainty as clearly as other linguistic features, hence the beneficial effect of EI. Farley concluded that “EI might have helped learners see the form-meaning connections in the SI activities more quickly and reduce the item by item analysis the SI participants might have undertaken” (p. 238). The fact that the SI group also improved indicates that SI is sufficient to have learners process subjunctive forms. These findings suggest that some types of linguistic structures or processing problems might require more exposure to SI or perhaps more explicit treatments.Footnote ²

THE PRESENT STUDY

Given that previous studies on the effects of EI in PI seem to indicate that EI is not necessary for PI, is it possible that EI is nevertheless somehow beneficial for learners? Is it possible that EI assists acquisition in some way that is not measurable by traditional pretesting and posttesting research designs? Is it possible that PI groups that get EI before being engaged in tasks somehow have an easier time processing grammatical features when working through those tasks? Do learners who get EI start making form-meaning connections earlier than those who do not get EI?

To determine whether EI does have a beneficial effect for learners, the present study compares various processing groups with and without EI and examines how learners perform during tasks designed to promote acquisition. An online analysis allows for observation of changes in behavior as learners try to process linguistic forms and structures with and without EI. Through comparisons between PI groups that receive EI and SI-only groups that do not receive EI, this study attempts to observe (a) whether providing EI to learners helps them process two different linguistic targets earlier in a sequence and also more correctly than learners who do not receive EI on the target structures, (b) whether EI helps learners process the linguistic targets faster (i.e., the time it takes them to submit their answer), and (c) whether EI helps learners begin and maintain correct processing strategies in a series of SI items. Instruction was delivered via a computer that tracks response time and response accuracy over time. The criterion for correct input processing was established as the point when learners began and continued to provide correct responses for at least four items—three target items and one distracter—in a row.Footnote ³ The number of SI items encountered by learners before this criterion point of correct processing is referred to as trials to criterion in the analysis.

Two linguistic variables are used to test the effects of EI in PI: Spanish OVS-type sentences with clitic object pronouns and Spanish subjunctive in expressions of doubt. These two linguistic targets were chosen because (a) they represent two different processing problems that require certain processing strategies for their acquisition and (b) previous PI studies with these forms have shown different outcomes regarding the effects of EI. Experiment 1 investigates the effect of EI on the processing of OVS-type sentences, and Experiment 2 investigates the effect of EI on the processing of the Spanish subjunctive. There are two experimental groups: the PI group (which receives EI before engaging in SI activities) and the SI group (which does not receive EI). One might ask why the study does not include an EI-only group—as used in Benati (Reference Benati and VanPatten2004), Wong (Reference Wong and VanPatten2004a), and VanPatten and Oikkenon's (Reference VanPatten and Oikkenon1996) studies—but, given that the question is not whether EI is necessary, it was determined that EI-only treatment groups were not needed for the purpose of the present study.Footnote ⁴ Of relevance here is the relative effect of EI during the acquisition phase, and the basic question is not whether EI is necessary (as measured by posttests in previous research) but what it does for learners as they go about incorporating particular grammatical features into their linguistic systems. Thus, the relevant comparison is between groups that differ only with respect to EI, holding additional exposure constant: SI plus EI for the PI group versus SI without EI for the SI-only group.

EXPERIMENT 1

Method

Participants

A large number of college-level L2 learners of Spanish from third-semester (i.e., beginning-intermediate level) Spanish classes were solicited for this study; however, only 84 of the 170 who met the study criteria volunteered to participate. These 84 volunteers comprised the final subject pool. The criteria for inclusion were (a) English as L1, (b) no reading or hearing impairment, and (c) inability to correctly process Spanish OVS-type sentences (as demonstrated through the screening test). Volunteers were paid US$10.00 for their participation.

Target Form

The target form was Spanish OVS-type sentences with third person clitic object pronouns. Spanish is a language that has a subject-verb-object (SVO) word order, as shown in (3). However, Spanish also allows OVS word order when a clitic pronoun appears in preverbal position to replace the noun in the object position, as in (4):

1. (3)

   

2. (4)

   

One of the cues that speakers of Spanish have when interpreting sentences like (4) is that agency, or the subject, can be identified by the verb ending, which must agree with the subject.Footnote ⁵ In this way, whether the subject is explicitly stated or not and regardless of its position in the sentence, it will be identified by the verb ending. For example, the verb ending –a in compra “buys” reliably denotes third person singular.

VanPatten's (Reference VanPatten2002, Reference VanPatten and VanPatten2004) model of input processing proposes the first noun principle, which posits that L2 learners will process the first noun or pronoun in the sentence as the subject or agent of the sentence. They will rely on a processing strategy that uses word order cues—rather than a strategy that uses verb ending cues—to interpret Spanish OVS sentences like (4). The result of this processing strategy is that learners will assign the role of subject to what is actually the object, thereby incorrectly processing and interpreting the sentence.

Materials

The materials used in this study included a paper-and-pencil pretreatment packet in addition to a set of instructions and SI activities for the treatment. The pretreatment packet contained a consent form, a language background questionnaire, and a screening test that tested learners' ability to process OVS-type sentences correctly. For the screening test, participants listened to 15 sentences once and choose the picture (either A or B) that illustrated the sentence they heard. If they were not sure, they could mark letter C as a third option on the scantron sheet. The test included seven target items (OVS sentences) and eight distracters (three subject-object-verb [SOV] and five SVO sentences). Participants were given 8 s between sentences to mark their choice. Figure 1 illustrates two examples of items on the screening test. For the screening test, participants could not have more than three target items correct out of the seven (i.e., 43% maximum score) in order to be included in the study.

Figure 1.

Sample screening test items.

The instructions, the EI activities, and the SI activities for the treatment were taken from the materials in both the VanPatten and Cadierno (Reference VanPatten and Cadierno1993) and the VanPatten and Oikkenon (Reference VanPatten and Oikkenon1996) studies and were adapted to be shown on a computer screen. Only referential-type SI activities were included in the treatment because these have a right or wrong answer. Affective-type SI activities generally used in PI were not included because they do not have a correct or incorrect answer (i.e., learners respond with their opinion or with what applies to them), and it is not possible to correct an opinion or a choice based on personal preference. For the PI group, the treatment materials contained (a) EI about OVS-type sentences, (b) information about the nonoptimal default processing strategy L2 learners use when trying to understand OVS-type sentences, and (c) referential-type SI items. For the SI group, the treatment materials were only referential-type SI items. There were a total of 30 SI items: 20 target (OVS) items and 10 distracters (with SOV or SVO word order). The distracters were placed every two or three target items in the series. The first two items in the series were target items (see Appendix B).

Design and Procedure

The 84 participants who volunteered for the study were randomly divided in two groups: 42 comprised the PI group and 42 comprised the SI group. In both groups, each participant wore headphones and was taught individually through a computer program. The EI and the SI items were adapted to a computer program designed with the experiment generator E-prime (http://www.pstnet.com/products/E-prime/default.htm). Participants in the PI group first read the instructions and then read the EI. The EI was presented across seven frames that appeared automatically every 10–45 s, depending on the length of the information. The amount of time that each frame remained on the screen was established based on a pilot test. Participants could not control the pace of the pages on the screen or backtrack. This ensured that all participants had the same time of EI exposure and thus prevented an intervening variable that could affect the interpretation of the results. Once the EI presentation concluded, participants indicated that they were ready to start by pressing the space bar. They listened to the first SI item once (e.g., Lo llaman sus padres por teléfono “His parents call him”) at the same time that two pictures appeared on the screen. Picture A depicted two parents calling a boy, and picture B depicted a boy calling his parents (see Appendix B). Participants had to select the picture that depicted what they heard by pressing either the key with the number 1 (for picture A) or the key with the number 2 (for picture B) on the keyboard.Footnote ⁶ The SI group went through the same procedure as the PI group, except that these participants started with the SI items immediately after reading the instructions (i.e., they did not have EI). Both groups could take as much time as they needed to answer, and all participants received feedback right after each response (i.e., “correct!” or “incorrect” on the screen). The accuracy of the learners' responses as well as their response times were tracked. After the 30 SI items, a screen appeared to indicate that the treatment was over and to thank participants.

Data Scoring

Trials to criterion, response time, and accuracy after criterion were measured for the treatment tasks. To score the trials to criterion, the researcher counted the number of SI items that participants completed until they reached criterion—that is, the number of items that participants completed up to the point when they correctly answered three target items and one distracter item in a row. Response time is the period of time from the onset of a stimulus to the response of the participant (Schneider, Eschman, & Zuccolotto, Reference Schneider, Eschman and Zuccolotto2002). In this study, it was the period of time from the onset of both the visual (i.e., the pictures) and the aural (i.e., the sentence) stimuli to the time a key was pressed indicating a response. Response time was recorded in milliseconds. Accuracy after criterion was calculated by counting all of the correct SI items after criterion and then dividing them by the total number of SI items after criterion. For example, if a participant started to reach criterion on item 20 (out of 30), the subsequent correct items were counted and then divided by 10 (i.e., the number of items in the sequence after criterion). If a participant had six items correct (i.e., four of criterion plus two more), his or her percentage of accuracy would be 60%.

Results

Number of Participants Who Reached Criterion

Not all of the 84 participants reached criterion; there were participants who did not correctly answer three target items and one distracter in a row. In the PI group, 26 of the 42 participants reached criterion, whereas in the SI group, 21 of the 42 participants reached criterion. A test of differences between proportions was performed to measure whether the difference in the number of participants was significant. The proportion of learners from the PI group who reached criterion was .61, whereas the proportion of learners from the SI group who reached criterion was .5. The difference in proportions was not significant (z = 1.1, p > .05).

Trials to Criterion

The means and standard deviations for the treatment conditions are shown in Table 2. An independent measures t test revealed no significant difference in trials to criterion between participants in the PI condition and participants in the SI condition, t(45) = 1.118, p = .269.

Table 2.

Mean scores and standard deviations for the processing of OVS sentences

Response Time

The means and standard deviations for the time it took participants to respond are shown in Table 2. An independent measures t test revealed no significant difference in response time between participants in the PI condition and participants in the SI condition, t(45) = −1.427, p = .160.

Accuracy After Criterion

The means and standard deviations are shown in Table 2. An independent measures t test revealed no significant difference in accuracy after criterion between participants in the PI condition and participants in the SI condition, t(45) = −0.730, p = .469.

EXPERIMENT 2

Results

Number of Participants Who Reached Criterion

As in Experiment 1, not all participants reached criterion while processing the subjunctive; however, this time the number of participants in the PI group who reached criterion (n = 32) was considerably higher than the number of participants who reached criterion in the SI group (n = 21). A test of differences between proportions was performed to measure whether the difference in the number of participants was significant. The proportion of learners from the PI group who reached criterion was .76, whereas the proportion of learners from the SI group who reached criterion was .5. This time, the difference in proportions was significant (z = 2.55, p < .05).

Trials to Criterion

The means and standard deviations for the treatment conditions are shown in Table 3. An independent measures t test revealed that the participants in the PI group took significantly fewer trials to reach criterion than the participants in the SI group, t(51) = −2.381, p < .05.

Table 3.

Mean scores and standard deviations for the processing of subjunctive

Response Time

The means and standard deviations for the time it took participants to respond are shown in Table 3. An independent measures t test revealed that the participants in the PI group responded significantly faster than the participants in the SI group, t(51) = −2.907, p < .05.

Accuracy After Criterion

The means and standard deviations are shown in Table 3. An independent measures t test revealed that the participants in the PI group responded significantly more accurately after having reached criterion than the participants in the SI group, t(51) = 2.682, p < .05.

GENERAL DISCUSSION

The explanation for the contrasting results of Experiments 1 and 2 might be found in (a) the nature of the online tasks or (b) the nature of the processing problem. Regarding (a), remember that the online task for processing the subjunctive required participants to attend to the verb inflection (i.e., the less noticeable form) and match it with the corresponding trigger clause presented on the screen. EI gave information about the regular verb endings and some of the most frequent irregular forms triggered by main clauses that convey doubt, also explaining that the meaning of doubt was expressed in the main clause and giving some examples of clauses that trigger the subjunctive. Furthermore, PI-group participants were informed that doubt was redundant and that it was important not to miss the verb when attending to messages. Therefore, these participants came to the task knowing that the inflection of the verb changes based on the meaning the main clause conveys. If learners know what (lexical and grammatical) forms go together, they will start to match them sooner and faster than if they do not know. In contrast, the SI group not only had to keep guessing what in the aural stimuli indicated which clauses to select but also which of the different verb endings needed to be matched with the clauses.

On the other hand, the task for processing OVS-type sentences required learners to listen to a sentence and interpret its meaning in order to match it with the corresponding picture. There is a difference between linking a grammar form to an expression of doubt and linking a whole sentence to a picture. Although EI informed learners about the difference between objects and subjects in OVS-type sentences in Spanish as well as the tendency of learners to rely on the first noun or pronoun to assign agency, participants still had to understand the relationships between the components of the sentence to perform the task accurately. Participants in the PI group might have recognized the elements in the sentence they heard; nonetheless, they still had to assign each component its grammatical role. With the subjunctive, it was necessary to notice and process one grammatical form. Therefore, the nature of the online task might have been one of the reasons for the difference between the experiments. EI seems to be beneficial when the task is to notice and process a single form, but it seems not to play any role when the task is to assign different grammatical roles in sentences. Further investigation is merited to confirm this claim.

An interesting phenomenon was that although EI informed learners about the forms triggered by the meaning of particular clauses when processing the subjunctive, participants did not start to match the forms correctly right away; rather, it took them a few trials to begin to answer correctly. It seems that EI helped learners notice forms that might not have been noticed or that might have taken more trials to notice, but it did not obviate or reduce the need for minimal exposure time; that is, correct processing was not instantaneous. Perhaps learners were familiarizing themselves with the task, and this affected their performance at the beginning. Further investigation is necessary to understand why the PI group did not start to process the input correctly at the beginning of the task.

Additionally, the nature of the processing problem might have contributed to the different effects of EI on the two target structures. For the Spanish subjunctive, native speakers of English have to process a redundant, less noticeable form in order to acquire it. This linguistic processing strategy was already built in when they learned their L1, because redundant forms in the input are also a feature of English. There are situations in which two forms encode the same meaning and occur in the same sentence—for example, “he talks” or “I missed class yesterday.” Thus, English-speaking L2 learners of Spanish do not have to build a new processing strategy for processing the Spanish subjunctive; they only learn to apply the knowledge they already possess to new linguistic contexts. When processing Spanish OVS sentences, however, native speakers of English have to refrain from using the first noun strategy, which works well in English, and must adopt a new strategy that allows processing of a more flexible system. They must learn a new strategy that does not require them to rely on word order cues but rather has them look for other indications of object, subject, and verb forms to confirm any preliminary assignment of a noun to subject or object role. English-speaking learners of Spanish thus cannot use existing strategies to interpret OVS-type sentences, because they will end up with an incorrect interpretation. It might be that EI has no effect in building a new processing strategy. Although learners have the pieces of the puzzle figured out if they receive EI (e.g., objects vs. subjects, OVS- and SOV-type sentences, L1 strategies), the processing strategy that they are lacking, and that they need to process correctly, can only be built with frequent exposure to input coupled with feedback that indicates whether what was listened to was understood correctly.

LIMITATIONS, DIRECTIONS FOR FUTURE RESEARCH, AND CONCLUSIONS

This study has several limitations that need to be taken into consideration when making generalizations about the findings. One of them is in the use of the instructional approach chosen. The findings should be applied to PI and not to other approaches to grammar teaching. Additionally, even within PI, the reader should keep in mind that participants did not receive the complete set of activities used for PI, such as affective-type SI activities. A second limitation is that both experiments were conducted in laboratory settings that do not reflect classroom dynamics where learners interact with input in a variety of ways (through affective-type SI activities, text enhancement, interaction, recasts, etc.) and have ample access to EI. A third limitation of this study is that it does not address why some participants were able to correctly process at least four items in a row within 30 trials, whereas some participants were not able to do so. Certainly, the role of individual differences while processing input is an issue that needs further investigation in studies of PI and other types of instructional techniques.

The findings of this study suggest that in some cases EI does not help learners process certain structures but it is beneficial for processing of nonsalient forms, depending on the task, the processing problem, or both. In order to investigate this issue further, it will be necessary to conduct more online studies that compare two forms that require applying the same processing strategies and having learners do similar tasks. A potential study could use the preterit and the simple future tense in Spanish as linguistic targets; the lexical principle affects the processing of both of these forms and SI activities for these forms could be structured in similar ways. If it is observed that learners start to make correct form-meaning connections for both structures early in a sequence of SI items, we will have further evidence that EI is beneficial for structures that are processed according to the lexical principle, as shown in the present study.

This study did not include posttests because the goal was to observe learners' behavior during input processing in an effort to discover possible EI effects that might not have been observed in previous PI studies due to the nature of their treatments. It is hoped that by using online treatments to investigate the role of EI during the process of making form-meaning connections, this study has contributed to a better understanding of the role of EI in PI and in instructed SLA.