CONCURRENT ANDLR ETROSPECTIVE ERBAL EPORTS INTERNATIONAL JOURNAL OF APPLIED INGUISTICS , Vol.V13, No.R2, 2003 201 Concurrent and retrospective verbal reports as tools to better understand the role of attention in second language tasks JOAQUIM CAMPS Universtiy of Florida This study investigates how the use of think-aloud protocols, both concurrent and retrospective, can contribute to the study of the role of attention in second language acquisition. It is based on the analysis of think-aloud protocols produced by 74 first-year learners of Spanish during and immediately after a reading and multiple-choice activity. The activity consisted of a text with direct object pronouns and 16 blanks for which a choice of three possible antecedents was given. The key to making the right choice was attending to both form and meaning in the input. The data in the think-aloud protocols was classified into mentions of the pronouns in the text as well as references to gender and number agreement. The results showed that mention of the targeted structure in the think-aloud protocols was related to better performance on the task for second-semester students, but not for first-semester students. There was some difference in the results for the concurrent and retrospective protocols. The possible complementary nature of these two sources of data is discussed. Introduction Recent studies have shown the usefulness of think-aloud protocols in understanding learners’ cognitive processes as they perform tasks designed to help them make form–meaning connections when processing input (Alanen 1995; Jourdenais et al. 1995; Jourdenais 1998; Leow 1997, 1998a, 1998b, 2000, 2001; Rosa and O’Neill 1999). Think-aloud protocols, described by Ericsson and Simon (1993), have been used in psychology for over a century. In the past few © Blackwell Publishing Ltd. 2003, 2003 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA 202 JOAQUIM CAMPS decades they have also been applied to a variety of areas in second language acquisition, such as writing (Cumming 1989, 1994; Raimes 1994), reading (Davis and Bistodeau 1993; Hosenfeld 1977), strategies for language use and interpretation (A.D. Cohen 1984; Mangubhai 1991), and discourse (A.D. Cohen and Olshtain 1993). Jourdenais (2001), Kasper (1998), and Kormos (1998) offer recent discussions of the application of verbal reports to L2 research. In concurrent think-aloud protocols, participants are asked to verbalize their thought processes at the same time as they are carrying out a certain activity (reading, writing, a problem-solving task, etc.). As Ericsson and Simon (1993: xiii) indicate: “It is important to note that subjects verbalizing their thoughts while performing a task do not describe or explain what they are doing – they simply verbalize the information they attend to while generating the answer” (emphasis in the original). Despite the fact that think-aloud protocols have been commonly used in other areas of second language acquisition (SLA), their use in analyzing the role of attention in SLA is relatively recent. Empirical studies on the effect of instruction and the role of attention in SLA have usually relied on a design consisting of pretest, instructional treatment and posttest(s). The assumption was that if the results in the posttest were significantly better than those in the pretest, we could safely conclude that learners had paid attention to the structures targeted during the instructional treatment. The introduction of think-aloud protocols in studies on the role of attention in SLA has enriched the experimental designs in this line of research by offering the possibility of collecting online evidence of the actual processes learners go through when they are exposed to linguistic input. By collecting data through verbal reports, we can obtain more accurate and detailed information that was impossible to obtain by means of the pretest–treatment–posttest experimental design mentioned above.1 Studies that have used think-aloud protocols to investigate the role of attention in SLA have analyzed the data provided in the verbal reports that the participants produced while they performed an experimental task. This was done in order to assess the level of awareness of the participants in a task designed to help them attend to certain aspects of the input. This information has allowed researchers to compare the posttest results of different groups of learners based on the levels of awareness they showed in their verbal reports. This methodological improvement has strengthened the empirical support for the “noticing hypothesis” (Schmidt 1990, 1994, 1995).2 Think-aloud protocols in studies on the role of attention to input Think-aloud protocols have been used in studies on the role of attention to the input in second language acquisition in order to analyze the effect of a variety of experimental tasks or treatments. Most of these studies suggest that the learner’s awareness of the target forms is positively related to recognition and © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 203 learning of those forms. Alanen (1995) studied the acquisition of a modified version of Finnish and reported that awareness of the target forms (as reflected in think-aloud protocols) was positively correlated to learning in some, but not all, experimental groups. Another example is provided by Leow (1997, 1998a, 1998b), where learners performed think-aloud protocols while doing a crossword puzzle involving irregular forms of the Spanish preterit. Leow found that those participants who showed a higher awareness of the target forms in their protocols performed better in the later recognition and production of those forms. Similar results were obtained by Rosa and O’Neill (1999) in a study that used multiple-choice jigsaw puzzles with five different treatment groups using conditional sentences in Spanish. In that study, learners whose protocols indicated levels of awareness classified as understanding obtained better results than those who showed awareness at the level of noticing or no awareness at all. Rosa and O’Neill applied the distinction between noticing and understanding proposed by Schmidt (1990, 1995). Their study analyzed intake but not production data. Leow (2001) also found a significant positive relationship between reported noticing of the target form (Spanish formal commands) during a reading activity and its subsequent recognition. His study found no difference between the enhanced and the unenhanced condition. Overall, the studies mentioned above support Schmidt’s position that noticing, that is, attention to a particular form in the input, together with some level of awareness of that form, allows for further processing of that input. Concurrent and retrospective verbal reports Most of the research on the role of attention in SLA that used verbal reports has relied on concurrent protocols. However, concurrent protocols are not the only type of verbal report available to researchers: the collection of retrospective reports is another interesting option. Retrospective protocols are typically performed immediately after the completion of a task, and participants are asked to report on the thoughts they had while they were completing the task. While concurrent think-aloud protocols reflect thought processes as they occur, retrospective protocols may draw from short- and long-term memory depending on the length of the task performed and the amount of time between the end of the task and the beginning of the verbal report. In their discussion of thinkaloud protocols, Ericsson and Simon (1993) favor the use of concurrent protocols, collected as the processing task is being performed, but they also recognize the benefits of retrospective protocols as complementary data and recommend the use of both. They indicate that even though in long cognitive tasks (beyond two minutes in length) the retrospective reports may be incomplete, having data from two sources may be very helpful. An important methodological point they emphasize is that retrospective reports need to be collected as soon as possible after the cognitive task. This would minimize the possibility that participants may start relying on inferences rather than reporting what happened. © Blackwell Publishing Ltd. 2003 204 JOAQUIM CAMPS Next, we will discuss a few studies that have used retrospective measures to assess the learners’ level of awareness of the particular target forms analyzed. P. Robinson (1996, 1997a, 1997b) asked participants a few questions, in writing, regarding their level of awareness of the rules and target forms they had been exposed to. This allowed him to classify his participants according to whether they noticed any rules, were looking for rules, or were able to verbalize the rules. In his (1997a) study of dative alternation in English, only instructed learners were able to refer to the key features of the rule, but many incidental and enhanced learners mentioned the position of direct and indirect objects in the sentence. On the other hand, few implicit learners verbalized any aspect of the rules. The methodology used by P. Robinson departs from that typically used in verbal reports, as described by Ericsson and Simon (1993), in that the reports were written, rather than oral, and the questions he asked were very direct (some were yes/no questions). This may have prompted specific responses from the participants. Ericsson and Simon favor more open-ended methods of letting participants retrospect on their experience. In a study on the role of output in promoting noticing, Izumi and Bigelow (2000) conducted retrospective interviews with four of the nine participants in their experimental group (they do not provide details on how the interviews were carried out). They attested great individual variation in what learners found problematic in their production and in what they noticed in the input they were subsequently exposed to. To assess noticing, they asked the learners to underline those elements that they thought would be necessary for their subsequent production. Kormos (2000) followed more closely the methodology proposed by Ericsson and Simon for retrospective reports. The goal of Kormos’ study was to analyze the self-correction behavior in the oral production of Hungarian learners of English as a foreign language. In her study, participants listened to a recording of their production, which included examples of self-correction, and were encouraged to stop the tape and comment on their production whenever they had something to say. The researcher also paused the tape when she felt there was something worth discussing. A few studies on the role of attention in SLA have complemented the use of concurrent think-aloud protocols with retrospective measures. In addition to using concurrent think-aloud protocols, Alanen (1995) measured her participants’ awareness by means of a questionnaire in which they were asked to state the rules that governed the use of the linguistic forms in the input they received. The participants who received textual enhancement were also asked whether they had noticed the highlighting used. Leow (2000) is a recent study that combined a variety of measurements of learners’ awareness. He conducted concurrent think-aloud protocols not only during the exposure task but also during the post-exposure assessment task. Before and after the assessment task, participants answered, in writing, two probe questions to further assess their awareness of the target forms (the irregular preterit in Spanish). Finally, three weeks after the completion of the experiment, Leow interviewed participants © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 205 who had been classified as being unaware of the target forms and asked them to elaborate on their answers to the probe questions and to explain their choices in the assessment task. Although the measures Leow used in addition to the concurrent think-aloud protocols would not match the ideal for retrospective reports described by Ericsson and Simon (written responses, a long delay between the end of the experiment and the interviews), they are an interesting example of how to gather data from a variety of sources in order to increase the reliability of the results obtained. Several studies in SLA have successfully implemented the use of both concurrent and retrospective verbal reports. Mangubhai (1991) conducted a study that investigated and classified the different behaviors learners engage in when they extract meaning from the input they receive. Learners received individualized Total Physical Response instruction in Hindi while wearing lapel microphones that recorded everything they said. They were encouraged to constantly verbalize whatever was going through their minds as they performed the TPR activities. Mangubhai collected concurrent think-aloud protocols during 20 teaching sessions. At some points during a lesson, when a comprehension problem was observed, participants were probed to report on those problems. These reports produced immediate retrospective data. Further retrospective verbal reports were conducted at the end of some of the teaching sessions. M. Robinson’s (1991) goal was to evaluate verbal reports as a means of investigating interlanguage pragmatics knowledge. She gathered concurrent think-aloud protocols from 12 female native speakers of Japanese as they planned how to fill out written discourse completion items using refusals. Participants were instructed not to think aloud while they were writing but only before they started and whenever they paused to plan what to write. Immediately after the task the researcher interviewed each participant, basing the discussion on the tape recording of the think-aloud protocol the participant had just produced. M. Robinson concluded that the combination of concurrent and retrospective reports enhances the informative value of verbal reports, and that data from verbal reports provide information regarding language-processing strategies that would not be available from the analysis of discourse completion tasks alone. Rationale and research questions In the review presented above, we saw that research on the role of attention in second language acquisition supports a positive relationship between awareness of the target forms and their subsequent recognition and learning, thus favoring Schmidt’s (1990, 1994, 1995) “noticing hypothesis”. This research points to the usefulness of think-aloud protocols as a tool to gather data regarding the actual online processing of input by language learners, and it is mostly based on concurrent verbal reports to establish awareness. However, we also saw how retrospective verbal reports have been successfully used in combination with concurrent verbal reports in studies in other areas of SLA research. The © Blackwell Publishing Ltd. 2003 206 JOAQUIM CAMPS discussion of these different procedures suggests that using multiple datacollection procedures (in this case, concurrent and retrospective verbal reports) may greatly enrich our knowledge of how learners attend to input. By designing a study that requires participants to complete both concurrent and retrospective protocols based on one given task, we will not only benefit from two different sources of information but will also be able to compare the information obtained through each type of verbal report. The present study was designed to explore whether those participants who noticed the targeted forms performed differently in the experimental task than those who did not. In this study, the learners’ noticing was operationalized as mentioning in their verbal protocols the targeted object pronouns, the object pronouns and their agreement features, or just the agreement features as the basis for their choice of answer in a reading and multiple-choice task. The scope of this project was limited to analyzing how learners processed the input present in one particular task. It did not use any post-exposure measures to test the recognition or production of the target forms included in the experimental task, since the aim was to carefully investigate the characteristics of this specific type of task and to analyze how learners processed the input it provided. This study also sought to research a methodological question: whether the concurrent and retrospective protocols, collected during the experimental task and right after it was completed, produced similar results. The following research questions were investigated: a) Will the participants who notice the target forms obtain better scores in the experimental task than those who do not? b) Will the results vary depending on the type of think-aloud protocol considered (concurrent or retrospective)? c) Will the results vary depending on whether students are in first- or secondsemester courses? The experimental study PARTICIPANTS A total of 74 first-year Spanish college students, all native speakers of English, participated in the study. The pool of participants was drawn from five sections of first-semester Spanish classes and seven sections of second-semester Spanish. The final distribution by groups was the following: 29 in the first-semester group and 45 in the second-semester group.3 All participants took part in the study during the same week in July of 1998. The first-semester students performed the task the week after they had been introduced to the target structure as part of their regular Spanish class, while the second-semester students had been introduced to the target structure several weeks earlier, during their first semester of study. Although no independent measures of proficiency were used to compare the two groups, the second group had had many more opportunities © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 207 to be exposed to those forms in the input they received and to produce them during their regular class activities. TARGETED LINGUISTIC FORM The study focused on the processing of third-person object clitic pronouns in Spanish (lo, la, los, las). Sentence (1) is an example of the structure considered, where la refers to una carta: 1) Juan escribe una carta y después la envía por correo. ‘John writes a letter, and then he mails it.’ The structural characteristics of the constructions with clitic pronouns, and especially the fact that clitic pronouns can never appear in isolation, make them an appropriate target for studies that deal with form–meaning connections. The clitic pronouns considered in this study must always have a referent (an antecedent), either overtly present in the linguistic context or understood in a more general, situational context. Thus in example (1), the agreement features (gender and number) of the clitic pronoun la have to match with those of its antecedent una carta. MATERIALS All participants received the same text, a narration of events which contained four examples each of the four third-person direct object clitic pronouns (lo, la, los, las) and 16 blanks. The learners’ task was to read the text and choose one of the three options provided for each blank (see Appendix). Sentence (2) is an example taken from the text, with the possible answers available to the learners: 2) Recoge ____ del mes pasado que tiene por el suelo y las lleva a reciclar. ‘(She) picks up ____ from last month, which are lying on the floor, and takes them out for recycling.’ a. las revistas (‘magazines’) b. los periódicos (‘newspapers’) c. las blusas (‘blouses’) The participants had to check both how well the meaning of a possible answer fit in the context and whether the gender and number of the noun phrase agreed with the clitic pronoun related to it. In the example given, if the learner just focused on meaning, both answers (a) las revistas and (b) los periódicos would appear to be correct, since they can be from last month and can be recycled. However, only las revistas agrees in gender and number with the direct object pronoun las. If the learner just focused on the form, the agreement features of the pronoun las, then both (a) las revistas and (c) las blusas would appear to be correct choices. However, blouses is not a good choice, because the text talks about something from last month that María is going to recycle. © Blackwell Publishing Ltd. 2003 208 JOAQUIM CAMPS The order of the three possible types of answers was varied randomly from item to item. PROCEDURE The entire data collection process took place in a language laboratory where students sat at individual recording booths. After the researcher introduced himself and the general purpose of the study, students read and signed the consent form, which contained detailed information about the study. Next, participants completed a questionnaire with personal and language background information. The researcher read aloud instructions on how to perform think-aloud protocols for the activities that were going to follow. The participants were asked to perform both concurrent and retrospective think-aloud protocols. They were told about the retrospective protocol only after the experimental task was completed. The script used for the concurrent protocol, adapted from Ericsson and Simon (1993), was the following: In this experiment we are interested in what you think about when you perform a reading and multiple-choice activity in Spanish. In order to find out, I am going to ask you to THINK ALOUD while you perform the task. What I mean by “think aloud” is that I want you to tell me EVERYTHING you are thinking while you are doing this activity. I would like you to talk CONSTANTLY while you perform the activity. I don’t want you to try to plan out what you say or to try to explain what you are saying. Just act as if you were alone in the room speaking to yourself. What’s most important is that you keep talking, and talk clearly and loudly enough into the microphone. If you are silent for any period of time I will remind you to talk by saying “What are you thinking?” Similarly, if you begin to speak softly, I may ask you to speak a little more loudly. I will not be able to help you in any way, but you can use the vocabulary list provided. You will be audiotaped while you work on the task. Do you understand what I want you to do? After that, the researcher provided an example of a think-aloud protocol by performing a multiplication task (24 × 26) on the board for the entire class to see. Next, the participants were asked to perform another multiplication task (14 × 34) while thinking aloud and recording their voices on tape. These examples were included in order to provide the participants with some training on the procedure involved in think-aloud protocols, as suggested by Ericsson and Simon (1993). In the next step, participants were given a bilingual vocabulary list with items relevant to the main activity that would follow. They received another page with five different categories, e.g. Cosas que se pueden comer (‘Things you can eat’). Their task was to classify the Spanish words into their corresponding categories. They were asked to perform a think-aloud protocol while doing this task and were given eight and a half minutes to carry it out. This vocabulary activity served a double purpose. On the one hand, it provided additional practice in performing a think-aloud protocol before the main task. On the other hand, it helped to control for vocabulary familiarity by providing a review of the vocabulary participants would need in order to carry out the main task. © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 209 After the vocabulary activity, the participants moved on to the main activity, for which they were asked to provide a think-aloud protocol while completing the reading and multiple-choice activity. They were allowed to check with the vocabulary list as many times as needed and to complete the task at their own pace. Most of the participants finished in 7 to 12 minutes. As each participant finished the main task, they were given the following set of written instructions for a retrospective think-aloud protocol: “Please, talk about what you remember with regard to how you went about performing the multiple-choice activity you just completed.” Most participants completed this final task in approximately 2 minutes. SCORING PROCEDURE AND ANALYSIS In scoring the multiple-choice task, 1 point was given for every correct answer. The maximum score possible was 16. A coding system was developed to classify the different types of comments found in the think-aloud protocols. Two slightly different sets of categories were developed: one for the concurrent think-aloud protocols and one for the retrospective think-aloud protocols. The analysis of the concurrent protocols identified specific mentions of the object pronoun forms which indicated that the choice of answer was based on that pronoun, as well as general mentions of gender and number. The specific mentions of pronouns were further subclassified as making reference to gender or to number features. Below are the categories used in the concurrent thinkalouds, with some examples. Mention of a pronoun: 3) I’ll say lettuce ’cause of la . . . yeah, la. Mention of a pronoun and reference to gender: 4) Let’s see, lo is masculine so it would have to be plato [dish].4 Mention of a pronoun and reference to number: 5) I thought bananas ’cause it’s plural, so it’s las pone [puts them], so it can’t be un tenedor [a fork] . . . and spoon . . . so, that makes sense. General mentions of gender or number: 6) Let’s go back to these. One, feminine, and this is still feminine, oh man . . . that’s not plural though.5 The data in the retrospective protocols were classified in terms of general mentions of object pronouns and agreement (gender and number). These types were considered to be equivalent to the categories used in the concurrent protocols, because these mentions made reference to either the object pronouns or the agreement features that make possible the matching of pronouns and their antecedents. Below are the categories used in the retrospective think-alouds, followed by some examples. © Blackwell Publishing Ltd. 2003 210 JOAQUIM CAMPS Reference to object pronouns: 7) I would . . . look at the direct object pronouns and determine whether it was male, I mean masculine or feminine. Reference to gender: 8) I looked for what went with the masculine and feminine. Reference to number: 9) Usually I had to finish the sentence to see to, like, see what it is they are referring to later on . . . if it’s, like, plural or singular. While in the concurrent protocols, most references were directly related to specific examples of pronouns in the input, although there were also general mentions of gender and number, the comments in the retrospective protocols were more general. They included references to paying attention to pronouns and to gender and number features in order to solve the multiple-choice task, but they did not focus on specific items in the task.6 In view of the limited number of mentions of pronouns and their agreement features, and given the methodological difficulty of assigning specific examples to categories reflecting different levels of awareness (e.g. noticing vs. understanding), a conservative approach to the analysis of the data was taken. Thus the participants were sorted into two groups: those who mentioned pronouns and their agreement features and those who did not. Results The number of participants who mentioned pronouns, pronouns and their agreement features, or just the agreement features in their protocols as the basis for their choice of answer was fairly limited. Only 6 out of 29 students in first semester and 10 out of 45 in second semester did so in the concurrent think-aloud protocols. This represents only about 20% of the participants, as seen in Table 1. The percentage was slightly higher in the retrospective protocols (around 27%). Table 1. Participants who mentioned pronouns and agreement in protocols 1st semester (N = 29) Concurrent Retrospective 2nd semester (N = 45) Total (N = 74) no. % no. % no. % 6 8 20.7 27.6 10 12 22.2 26.7 16 20 21.6 27.0 © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 211 Table 2. 2 × 2 ANOVA for Group × Level based on concurrent and retrospective protocols df SS MS F-value p Concurrent Group Level Group × Level Error 1 1 1 70 9.35 12.36 20.27 248.76 9.35 12.36 20.27 3.55 2.63 3.48 5.70* .109 .066 .020 Retrospective Group Level Group × Level Error 1 1 1 70 5.47 6.09 8.91 269.22 5.47 6.09 8.91 3.85 1.42 1.58 2.32 .237 .213 .133 * statistically significant The scores in the multiple-choice task for each of the 74 participants were computed. Recall that the maximum score possible was 16, corresponding to the 16 blanks. Table 2 shows the results of the 2 × 2 ANOVA performed on the data based on the concurrent and the retrospective protocols. For the concurrent protocols, neither Group (mention vs. no mention) nor Level (1st vs. 2nd semester) were significant, although Level showed a trend towards significance. However, the interaction Group × Level was significant. Similar results were obtained based on the retrospective protocols: neither of the two independent variables showed significance, but in this case the interaction between Group and Level was not significant, either. Table 3 shows a comparison of mean scores and the results of t-tests based on the type of protocol, semester and mention of pronouns. When comparing those who mentioned pronouns and those who did not, there were significant differences in accuracy on the multiple-choice task only for second-semester participants; this was true for both types of protocol, but the effect size was larger for the concurrent protocols.7 Looking at all those who mentioned pronouns, within the concurrent protocols the mean for the second-semester participants was significantly higher than for those in the first-semester, with a large effect size. There were no other significant effects. Table 4 shows the individuals in each group who mentioned pronouns and agreement during at least one of the two types of protocols, how many times they did so, and the types of mentions they made. For instance, participant X27 from the first-semester group mentioned pronouns and their agreement three times in her concurrent think-aloud protocol and two times in her retrospective protocol. We can see that in the first-semester group only three participants mentioned pronouns or their agreement features in both their concurrent and retrospective protocols, while seven of the second-semester students did so. With regard to what features those participants made reference to in their © Blackwell Publishing Ltd. 2003 212 JOAQUIM CAMPS Table 3. Comparison of results on multiple-choice task by type of protocol, mention of pronouns and course level Concurrent No. of students © Blackwell Publishing Ltd. 2003 1st semester Mention No mention 6 23 2nd semester Mention No mention 10 35 Mention 1st semester 2nd semester 6 10 No mention 1st semester 2nd semester 23 35 * = statistically significant; d = effect size Mean score 10.67 11.09 p = .666, d = .20 13.00 10.80 p = .001*, d = 1.27 10.67 13.00 p = .028*, d = 1.27 11.09 10.80 p = .575, d = .15 Retrospective s.d. No. of students 1.97 2.13 8 21 1.76 1.73 12 33 1.97 1.76 8 12 2.13 1.73 21 33 Mean score 10.88 11.05 p = .850, d = .08 12.33 10.91 p = .045*, d = .76 10.88 12.33 p = .151, d = .70 11.05 10.91 p = .797, d = .07 s.d. 2.17 2.09 2.02 1.81 2.17 2.02 2.09 1.89 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 213 Table 4. Frequency of mention of pronouns, gender and number by participant 1st semester Stud. X27 X31 X67 3 X11 X53 X63 3 X7 X9 X21 X35 X49 5 11 Concurrent 3 (O, G) 1 (G) 2 (G, N) 6 Retrospective 2 (O, G, N) + 1 (O, G) + 2 (G, N) = 5 2nd semester Stud. Y11 Y53 Y55 Y57 Y61 Y93 Y105 7 2 (N) 2 (G, N) 1 (O) 5 0 Y13 Y33 Y97 3 0 11 1 (G, N) 1 (G, N) 1 (G) 1 (N) 1 (O, G) 5 10 Y15 Y23 Y79 Y83 Y101 5 15 Concurrent 3 (O, G) 1 (O) 1 (N) 2 (G, N) 3 (O, G) 1 (G) 1 (N) 12 Retrospective 1 (G) – 1 (G) ≠ 1 (G) ≠ 1 (G, N) = 1 (O, G) = 1 (G) = 1 (N) = 7 1 (G) 1 (O) 1 (O) 3 0 0 15 1 (N) 1 (G, N) 1 (G) 2 (G, N) 1 (N) 6 13 O = Object; G = Gender; N = Number; totals in italics + increase in features; – decrease in features; = same features; ≠ different features comments, half of the participants mentioned the same types of elements in both protocols (e.g. X67 referred to both gender and number), two others provided additional information in the retrospective protocol (e.g. X27 had not mentioned number in the concurrent protocol), and only one provided less information in the retrospective report (Y11). All but one of those 10 participants made reference to gender. There were the same number of participants at both levels who made mention of pronouns or agreement in only one of their protocols. Overall, more second-semester students mentioned pronouns or agreement. The total number of mentions of relevant features was slightly higher in the concurrent protocols for both groups.8 When considering the cases where the learners mentioned the pronouns or some of their characteristics in the concurrent think-aloud protocols, it is important to look into the relationship between mentioning a pronoun or its agreement features and getting the corresponding item right. The learners mentioned pronouns on 26 occasions in the concurrent protocols, and on 24 of those occasions they chose the correct answer. That is an accuracy level of © Blackwell Publishing Ltd. 2003 214 JOAQUIM CAMPS 92%, compared to only 69% accuracy for those items that were not mentioned (802 out of 1158). This gives a clear indication that when the learners mentioned a pronoun, they made the right choice of antecedent in the vast majority of the cases. The research questions in this study explored the possible effect of three different factors on the results obtained: (a) the difference between noticing and not noticing the target forms, (b) the effect of the type of protocol considered, and (c) the course level at which the learners were enrolled. The results obtained show an interaction of factors that lead us to address the answers to the three questions in an interrelated manner. In summary, when considering all participants, those who mentioned the object pronouns and their agreement features in their think-aloud protocols did not obtain higher scores than the participants who did not mention the pronouns and their features. However, among the second-semester students, those who mentioned the pronouns and their features did better than those who did not, and that was the case with both types of protocols. Additionally, among all those who mentioned the target forms, the scores for the second-semester group were significantly higher than for the first-semester group when we considered the concurrent protocols. Discussion The results of this study are in line with those of Leow (1997, 1998a, 1998b, 2000, 2001), and Rosa and O’Neill (1999), in that the second-semester students who noticed the targeted forms and mentioned them in their protocols also obtained the best results in the experimental task. However, that was not the case with the first-semester students. This finding, together with the lack of difference between first- and second-semester students within the group of those who did not mention pronouns or their features, may lead us to postulate a difference in the ability of the two groups to make use of the information they attended to. This could be due to their ability to rehearse in short-term memory the information they had just attended to (P. Robinson 1995). We could argue that the first-semester students who mentioned the pronouns did not benefit much from paying attention to that particular aspect of the input because they were not able to process the information provided deeply enough to activate its rehearsal in short-term memory. On the other hand, the second-semester students who mentioned the pronouns may have been able to process the information at a deeper level and activate its rehearsal in short-term memory. This may have helped them obtain significantly different results from those of the second-semester students who did not mention the pronouns in the input. The first-semester students’ more limited processing capacity (VanPatten 1996; Wickens 1989) may have prevented them from using the input they noticed with the same success as the second-semester students.9 Regarding the comparison of the results based on the concurrent and the retrospective reports, in the statistical comparisons the results for concurrent © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 215 and retrospective protocols agreed in all but two of the measures considered. One of the differences was in the analysis of variance of the whole database, where no interaction was found between Group and Level based on the retrospective protocols. However, in the descriptive analysis we saw that many participants provided as much information, if not more, in the retrospective protocols as in the concurrent protocols. Despite the absence of complete agreement between the two different measures, it seems that using both types of verbal reports in the same study may help us obtain richer data. This is particularly important when we try to assess the attentional processes of the participants in a study. As Leow (2000) points out, exploring the learners’ internal processes by means of multiple data-elicitation measures may offer the necessary evidence to better understand how participants go about completing a specific task. If only concurrent protocols had been used in this particular study, we would have counted as many as ten of the participants among those who did not notice the targeted forms, because they only commented on those forms in their immediate retrospective protocols. In this study, that would have meant undercounting the number of participants who noticed the targeted forms by 37%. A final point regarding retrospective protocols is their level of costeffectiveness. There are various ways to use them. In some research designs researchers may need to meet individually with participants, replay a recording of their performance, and ask them to comment.10 The present study sought to obtain general comments about the task that the participants had performed. Since the data were collected from all the participants at the same time (recording their comments in their individual laboratory booths), the collection of retrospective reports only added three or four minutes to the total datacollection time. An issue that needs to be discussed here is the relatively low frequency of mentions of the target forms that appear in the verbal reports (none of the participants made more than three mentions in the concurrent protocols). A possible explanation is the nature of the task. Although the task consisted of 16 blanks with three choices for each, which could technically be considered similar to the problem-solving tasks in other previous research, such as the crossword puzzle used in Leow (1997, 1998a, 1998b, 2000) and the jigsaw puzzle used in Rosa and O’Neill (1999), there was one crucial difference. In the current study the 16 blanks were embedded in a long text (237 words). While tasks in previous studies generated much more evidence of awareness from the participants, possibly because they may have focused on each element of the puzzles one at a time as discrete units, the learners in the present study may not have felt the need to focus and report on each individual item, since they may have approached the task more globally and not as a series of tasks to be completed one item at a time. A similar explanation is offered in Leow (2001), where a lower frequency of reported noticing was explained by the fact that the target forms were part of a text and not part of a discrete-point problem-solving task. © Blackwell Publishing Ltd. 2003 216 JOAQUIM CAMPS With regard to verbalization of thought processes, it is important to keep in mind a potential limitation of any study involving think-aloud protocols: although such data can provide evidence of awareness of the elements targeted on the part of the learners, lack of verbalization of their thought processes does not mean that learners are not aware of the elements they are processing (A.D. Cohen 1987; Seliger 1983). In connection to this point, the effectiveness of think-aloud protocols may be limited if participants do not provide enough useful data due to a lack of practice with the think-aloud procedure. This study provided learners with training on how to perform think-aloud protocols in order to address this limitation and produce more reliable results. However, it is difficult to assess the degree of familiarity with the procedure the participants achieved with the practice examples carried out before the main task. Another factor to keep in mind when carrying out concurrent think-aloud protocols is the issue of reactivity, that is, the possibility that performing a think-aloud protocol may affect the learners’ thought processes. Jourdenais (2001) offers a discussion of such concerns, while Ericsson and Simon (1993), based on a review of studies in psychology, state that, when carried out properly, think-aloud protocols do not influence the sequence of thought, although they may increase the amount of time needed to complete a task.11 There are other limitations which are more specific to this study. First, it was purposely designed to explore only how learners processed information in one particular experimental task, without measuring the possible effects of exposure to this task in terms of subsequent recognition or production of the targeted forms. The decision to limit the scope of the study was based on the belief that it is important to carefully investigate the characteristics of a specific task before attempting to explore its effects as a possible instructional task in a pretest–posttest experimental design. Another limitation is that the discussion in this article is mostly confined to a quantitative analysis of the data. Future work will benefit from an analysis of the comments in the participants’ verbal reports from a qualitative point of view by comparing the remarks in their concurrent protocols to those in their retrospective reports. This type of analysis should further develop our understanding of the differences between the two types of reports and how they can complement each other in investigating how learners attend to input. Conclusion This study adds further evidence to research that has shown the connection between noticing targeted forms and success in linguistic tasks involving these forms. The results of the study suggest a possible difference between two levels of beginning learners with regard to the benefits they obtain from attending to forms targeted in the input. From a methodological point of view, this study has discussed the differences and similarities between the results obtained from concurrent and retrospective protocols, both at the level of statistical analysis © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 217 and at the descriptive level. The evidence shows that retrospective protocols are an effective complement to the use of concurrent protocols. It would be to the benefit of the field if more studies on the role of attention in SLA implemented the use of both concurrent and retrospective verbal reports in order to more fully assess the potential advantages of using both types together when studying attentional processes. Acknowledgement This research was supported by a grant from the College of Liberal Arts and Sciences at the University of Florida. An early version of this paper was presented at the Conference on Form–Meaning Connections in Second Language Acquisition held in Chicago in February, 2002. I am grateful to Renée Jourdenais and Gillian Lord for their comments on earlier versions of this article. I would also like to thank the two anonymous reviewers for their suggestions. Notes 1. Leow (1998b) pointed out this internal validity issue as something affecting many studies in SLA. Further discussion can be found in Leow (1999). 2. Other on-line measures used to assess learners’ attention to input include eyetracking and some applications of computer-based instruction. See Tanenhaus and Spivey-Knowlton (1996) and Blake (1998) for respective descriptions. 3. In all, data were collected from 155 participants, but 81 of them were exposed to a type of treatment that will not be discussed in this article. 4. Translation of Spanish words is provided in brackets. 5. In some cases, participants made reference to several traits in a given mention, for example, mentioning gender and number at the same time. 6. The two types of protocols also differed in that the retrospective protocols elicited comments on how the participants performed the task, while the concurrent protocols did not. 7. An effect size of .80 or greater is considered a large effect. Values between .50 and .80 indicate a medium effect, while anything below .50 indicates a small effect (Norris and Ortega 2000, following J. Cohen’s 1988 recommendation). 8. See note 5. 9. As one of the reviewers pointed out, the differences in processing may also be due to individual differences between learners. It is plausible that a specific first-semester learner may have benefited more from attending to the input than a second-semester student, despite the fact that the group scores are higher for the second semester. For example, in the analysis based on the participants who mentioned pronouns in their concurrent protocols, the top four learners in the first-semester group had higher scores than the two weakest learners in second-semester group. However, 8 of the 10 learners in the second-semester group had higher scores than all the learners in the first-semester group. 10. An example of this would be the application of stimulated recall methodology (Gass and Mackey 2000). © Blackwell Publishing Ltd. 2003 218 JOAQUIM CAMPS 11. Leow and Morgan-Short (2002) empirically addressed the issue in the context of a reading task in Spanish as a foreign language and found no reactivity effects – that is, no difference between learners who performed think-alouds and those who did not. References Alanen, R. (1995) Input enhancement and rule presentation in second language acquisition. In R. Schmidt, Attention and awareness in foreign language learning (Technical Report #9). Honolulu: University of Hawai’i, Second Language Teaching and Curriculum Center. 259–302. Blake, R.J. (1998) The role of technology in second language learning. In H. Byrnes, Learning foreign and second languages. New York: The Modern Language Association of America. 75–104. Cohen, A.D. (1984) On taking language tests: what the students report. Language Testing 1: 70–81. —— (1987) Using verbal reports in research on language learning. In C. Faerch and G. Kasper, Introspection in second language research. Clevedon: Multilingual Matters. 82–95. —— and E. Olshtain (1993) The production of speech acts by ESL learners. TESOL Quarterly 27: 33–56. Cohen, J. (1988) Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cumming, A. (1989) Writing expertise and second language proficiency. Language Learning 39: 81–141. —— (1994) Writing expertise and second-language proficiency. In A. Cumming, Bilingual performance in reading and writing. Ann Arbor: John Benjamins. 173–221. Davis, J. and L. Bistodeau (1993) How do L1 and L2 reading differ? Evidence from think aloud protocols. The Modern Language Journal 77: 459–72. Ericsson, K.A. and H.A. Simon (1993) Protocol analysis. Cambridge, MA: MIT Press. Gass, S. and A. Mackey (2000) Stimulated recall methodology in second language Research. Mahwah, NJ: Lawrence Erlbaum. Hosenfeld, C. (1977) A preliminary investigation of the reading strategies of successful and nonsuccessful second language learners. System 5: 110–123. Izumi, S. and M. Bigelow (2000) Does output promote noticing and second language acquisition? TESOL Quarterly 34: 239–78. Jourdenais, R. (1998) The effects of textual enhancement on the acquisition of the Spanish preterit and imperfect. Unpublished doctoral dissertation, Georgetown University, Washington, DC. —— (2001) Cognition, instruction and protocol analysis. In P. Robinson, Cognition and second language instruction. Cambridge University Press. 354–75. —— , M. Ota, S. Stauffer, B. Boyson, and C. Doughty (1995) Does textual enhancement promote noticing? A think-aloud protocol analysis. In R. © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 219 Schmidt, Attention and awareness in foreign language learning (Technical Report #9). Honolulu: University of Hawai’i, Second Language Teaching and Curriculum Center. 183–216. Kasper, G. (1998) Analysing verbal protocols. TESOL Quarterly 32: 358–62. Kormos, J. (1998) Verbal reports in L2 speech production research. TESOL Quarterly 32: 353–8. —— (2000) The role of attention in monitoring second language speech production. Language Learning 50: 343–84. Leow, R.P. (1997) Attention, awareness, and foreign language behavior. Language Learning 47: 467–505. —— (1998a) The effects of amount and type of exposure on adult learners’ L2 development in SLA. Modern Language Journal 82: 49–68. —— (1998b) Toward operationalizing the process of attention in second language acquisition: evidence for Tomlin and Villa’s (1994) fine-grained analysis of attention. Applied Psycholinguistics 19: 133–59. —— (1999) The role of attention in second/foreign language classroom research: methodological issues. In J. Gutiérrez-Rexach and F. Martínez-Gil, Advances in Hispanic linguistics. Somerville, MA: Cascadilla Press. 60–71. —— (2000) A study of the role of awareness in foreign language behavior: aware versus unaware learners. Studies in Second Language Acquisition 22: 557–84. —— (2001) Do learners notice enhanced forms while interacting with the L2? An online and offline study of the role of written input enhancement in L2 reading. Hispania 84: 496–509. —— and K. Morgan-Short (2002) To think aloud or not to think aloud: the issue of reactivity in SLA Research methodology. Paper presented at the Conference on Form and Meaning Connections in SLA, Chicago, IL. Mangubhai, F. (1991) The processing behaviours of adult second language learners and their relationship to second language proficiency. Applied Linguistics 12: 268–98. Norris, J.M. and L. Ortega (2000) Effectiveness of L2 instruction: a research synthesis and quantitative meta-analysis. Language Learning 50: 417– 528. Raimes, A. (1994) Language proficiency, writing ability, and composing strategies: a study of ESL college student writers. In A. Cumming, Bilingual performance in reading and writing. Ann Arbor: John Benjamins. 139–72. Robinson, M. (1991) Introspective methodology in interlanguage pragmatics research. In G. Kasper, Pragmatics of Japanese as native and target language (Technical Report #3). Honolulu: University of Hawai’i, Second Language Teaching and Curriculum Center. 29–84. Robinson, P. (1995) Attention, memory, and the “noticing” hypothesis. Language Learning 45: 283–331. —— (1996) Learning simple and complex second language rules under implicit, incidental, rule-search, and instructed conditions. Studies in Second Language Acquisition 18: 27–67. © Blackwell Publishing Ltd. 2003 220 JOAQUIM CAMPS —— (1997a) Generalizability and automaticity of second language learning under implicit, incidental, enhanced, and instructed conditions. Studies in Second Language Acquisition 19: 223–47. —— (1997b) Individual differences and the fundamental similarity of implicit and explicit adult second language learning. Language Learning 47: 45–99. Rosa, E. and M. O’Neill (1999) Explicitness, intake, and the issue of awareness. Studies in Second Language Acquisition 21: 511–56. Schmidt, R. (1990) The role of consciousness in second language learning. Applied Linguistics 11: 129–56. —— (1994) Deconstructing consciousness in search of useful definitions for applied linguistics. AILA Review 11: 11–26. —— (1995) Consciousness and foreign language learning: a tutorial on the role of attention and awareness in learning. In R. Schmidt, Attention and awareness in foreign language learning (Technical Report #9). Honolulu: University of Hawai’i, Second Language Teaching and Curriculum Center. 1–63. Seliger, H.W. (1983) The language learner as a linguist: of metaphors and realities. Applied Linguistics 4: 179–91. Tanenhaus, M.K. and M.J. Spivey-Knowlton (1996) Eye-tracking. Language and Cognitive Processes 11.6: 583–8. VanPatten, B. (1996) Input processing and grammar instruction. Norwood, NJ: Ablex. Wickens, C.D. (1989) Attention and skilled performance. In D.H. Holding, Human skills. New York: John Wiley. 71–105. [Received 7/9/02; revised 10/12/02] Joaquim Camps Department of Romance Languages and Literatures University of Florida 170 Dauer Hall, Box 117405 Gainesville, Florida 32611-7405 USA e-mail: [email protected]fl.edu © Blackwell Publishing Ltd. 2003 CONCURRENT AND RETROSPECTIVE VERBAL REPORTS 221 Appendix Text and multiple-choice task. Read the text and circle the letter of the items below that would better fill in the blank in every case. María tiene mucho trabajo en casa los sábados por la mañana. Primero lava ____(1) y luego la plancha. Después arregla ____(2) porque lo quiere tener en orden. Ordena ____(3) que tiene en su escritorio y los pone en montones distintos. Recoge ____(4) del mes pasado que tiene por el suelo y las lleva a reciclar. Limpia ____(5) que tiene en su escritorio y la cambia de lugar. Al mediodía almuerza. Toma ____ (6) y lo pone en la mesa de la cocina. Saca ____(7) que tiene en el refrigerador, lo pone en el plato, y prepara un sándwich. Toma ____ (8) y la pone en el sándwich. También pone tomate y jamón. A María le gusta comprar ____(9) que hay en el mercado del barrio, porque los venden a buen precio. Toma también ____(10) y las pone al lado del sándwich. Entonces va a la sala de estar, toma ____(11) que está al lado de la mesa, la coloca al lado de la ventana, y se sienta. Mira ____(12) en el balcón, los llama y les da migas de pan. Después llega la hora de ____(13). María las mira todos los días en el canal 4 de televisión. Se sienta delante de ____(14) que le regaló su madre, y lo enciende. Le gusta mucho ver a ____ (15) tan importantes que actúan en esos programas y las admira por su belleza. También le parecen muy guapos ____ (16), y los contempla con emoción. En fin, María pasa una tarde muy agradable los sábados. (1) a. los pantalones b. la cara c. la ropa (5) a. las fotos b. la lámpara c. la mesa (9) a. las manzanas b. los tomates c. los cuchillos (13) a. las tareas b. las telenovelas c. los noticieros (2) a. la habitación b. el cuarto c. el almuerzo (6) a. una botella b. una taza c. un plato (10) a. unas bananas b. unas cucharas c. un tenedor (14) a. la radio b. el periódico c. el televisor (3) a. las cosas b. los teléfonos c. los libros (7) a. la mantequilla b. el vino c. el pan (11) a. una revista b. una silla c. un sillón (15) a. las sirvientas b. las actrices c. los actores (4) a. las revistas b. los periódicos c. las blusas (8) a. (la) lechuga b. (la) cerveza c. (el) atún (12) a. los pájaros b. los dibujos c. las golondrinas (16) a. los actores b. los libros c. las actrices © Blackwell Publishing Ltd. 2003
© Copyright 2025 Paperzz