CALL Feedback and iCall

Multimedia glossing has been the focus of much CALL research (for an excellent overview, see Chun, Reference Chun, Ducate and Arnold2006; Lafford, Lafford, & Sykes, Reference Lafford, Lafford and Sykes2007). According to Chun (Reference Chun, Ducate and Arnold2006, p. 78), L2 vocabulary is best remembered when learners look up picture or video glosses in addition to receiving translations of unfamiliar words. But if given the choice, L2 learners tend to choose only simple definitions or translations in a fashion like what WordChamp provides along with audio recordings in an amazing number of languages (for a review, see LeLoup & Ponterio, Reference LeLoup and Ponterio2005). Chun's (Reference Chun, Ducate and Arnold2006, p. 82) study shows that lexical knowledge, rather than grammatical knowledge, is significantly related to both reading and listening comprehension. Following Grabe's (Reference Grabe2004) work, Chun (Reference Chun, Ducate and Arnold2006) correctly pointed out two separate aspects of lexical knowledge that impact reading comprehension: the learning of vocabulary and the fluency (i.e., automaticity) of word recognition. The research shows that explicit instruction, including tutorial CALL (see Lafford et al., Reference Lafford, Lafford and Sykes2007), has a significant impact on vocabulary acquisition, but reading comprehension is more complicated and is affected by a number of factors, especially prior or background knowledge of the material to be read. Chun (Reference Chun, Ducate and Arnold2006, p. 92) counsels teachers to include a large battery of prereading activities to prime students for what they will encounter. Fukkink, Hulstijn, & Simis (Reference Fukkink, Hulstijn and Simis2005) have demonstrated that L2 students retrieve faster and with less variation the unfamiliar words on which they have trained in contrast to those on which they have not trained. Research of this kind indicates obvious areas where tutorial CALL and explicit instruction can make a difference in second language development.

Other more interactive programs that provide feedback to L2 student responses are represented by programs known as chatterbots. Fryer and Carpenter (Reference Fryer and Carpenter2006) described how chatterbot programs can give the appearance of being a sympathetic listener to students by prodding them with follow-up questions such as: “How are you feeling?” “Why are you feeling tired? “Tell me more of last night.” In actual fact, chatterbot programs identify key phrases, nouns, adjectives, verbs, and adverbs recognized only as character strings (i.e., not as parsed linguistic structures) and then match these strings with clichéd responses that give the impression of a normal, give-and-take conversation. Payne (Reference Payne2007) has been experimenting with another type of bot program called BodoBot that provides L2 German learners with an instant messenger (IM) interface that coordinates access to many different online language-learning resources. BodoBot simultaneously tracks students’ learning behavior in order to provide reading recommendations and quizzes tailored to the learner's skill level. Students send messages in either English or German to request translations to or from German, additional examples of German words in context, related links to authentic German texts at the same level of difficulty, a German word-of-the-day, and vocabulary quizzes. Programs such as BodoBot will also provide researchers with new sources of student behavior data because they keep track of all requests and actions.

In the earlier days of CALL, feedback focused more on the grammatical or sentence level. CALL feedback was limited to spellcheckers, grammar checkers, and discrete string or keyword matching. Unfortunately, most spellcheckers assume that the users are already competent speakers of that language. Few programs are designed with L2 learners in mind (see Burston's (Reference Burston1998) review of a French spellchecker). Spellcheckers do well at catching single-letter violations, but often fail to analyze learner's more competence-based errors (Heift & Schulze, Reference Heift and Schulze2007, p. 166–167).

Although string-matching routines, such as what is available from Hot Potatoes (http://hotpot.uvic.ca/), are far better than simple right/wrong responses, this approach still falls short of providing students with CALL materials that can truly be labeled as interactive. As the CALL field has moved away from the model of drill-and-kill tutorials (but see also Hubbard & Bradin Siskin, Reference Hubbard and Bradin Siskin2004), the field has demanded CALL applications that react to student input in ways that appeared to be more context sensitive. As early as the 1980s, Underwood (Reference Underwood1989) described what the ideal intelligent tutoring system (ITS) should be able to do: Act as a real tutor or guardian for L2 students, leading them by the hand to discover more and more about the target language with each response and prescribing the appropriate exercises and metalinguistic explanations about the language that are needed to advance. An ITS, then, involves an expert (usually, some type of grammatical parser, a program capable of separating utterances into phrases, words, parts of speech, and limited semantic interpretations), a student, and a teacher module, frequently circumscribed in a highly defined semantic microworld or domain to minimize potential ambiguities and confusions. At the time, artificial intelligence (AI), a subfield of computer science, held great promise that something like Underwood's forward-looking description of an expert tutor could actually come to pass.

Gradually, the term ITS or expert system has been supplanted by a more modest notion of iCALL, but as Schulze (Reference Schulze, Chambers and Davies2001) explained, a more accurate name would be parser-based CALL. This later term makes it clear that natural language parsers are used to evaluate students’ syntactic and sometimes semantic input. Although Schulze (Reference Schulze, Chambers and Davies2001, p. 117) openly acknowledged that a parser-based CALL cannot account for the full complexity of natural human languages, “it does not mean that interesting fragments or aspects of a given language cannot be captured by a formal linguistic theory and hence implemented in a CALL application.” Obviously, a parser-based approach places an inordinate emphasis on morpho-syntactic errors—not a bad thing seeing that these types of errors appear to constitute the most frequently occurring mistakes that students make in freely produced texts (Schulze, Reference Schulze, Chambers and Davies2001, p. 121). However, the parser approach also produces its share of false acceptances and false alarms. Once again, the more constrained the semantic domain is, the more successful the parser will be at diagnosing errors.

A much more serious problem for the iCALL field has been the fact that despite a great deal of work in artificial intelligence over the years, there still only exist two operational Web-based iCALL programs: E-Tutor for German (Heift, Reference Heift2002; Heift & Schulze, Reference Heift and Schulze2007; http://www.e-tutor.org) and Robo-Sensei for Japanese (formerly called Banzai, Nagata Reference Nagata1996, Reference Nagata2002; http://www.usfca.edu/japanese/RSdemo/preRSfiles/demoTOC.htm; for a review, see Ushida, Reference Ushida2007). A third system, Tagarela for Portuguese (http://tagarela.osu.edu), is promised for spring 2007. The goal of all these programs is to provide error-specific feedback and flexibility in handling student textual input.

In the case of the E-Tutor, a natural language parser for German is combined with a student module that tracks each student's errors and proficiency level, controlling the type of feedback that will be delivered: Namely, for beginners and intermediates, rich metalinguistic feedback messages and remedial exercises (e.g., dictation, build a phrase, which word is different, word order practice, build a sentence), but something much less explicit for the advanced students who need only light guidance.

Heift (Reference Heift2001) acknowledged that building this type of parser-based CALL is very labor and time intensive and requires close cooperation among computer programmers, linguists, and pedagogues. However, error analysis of this ilk can also be very accurate in providing the type of feedback and interactivity that is helpful to the learner, even if it does not meet the lofty ideals that Underwood (Reference Underwood1989) originally imagined. And students really pay attention to well-crafted feedback! Heift (Reference Heift2002) discovered that most students (85%) using E-Tutor revised their sentences without peeking at the answers, with the weaker performers tending to also be the more frequent peekers. Heift (Reference Heift2004) also showed that the more explicit and metalinguistic in nature the feedback is, the more it helps the L2 learners successful complete their L2 tasks.

Why, then, has iCALL not gained more notice in the CALL field? Perhaps what has distracted the field from the slow but sure advances in parser-based error analysis has been to the meteoric surge of interest in CMC and sociocultural/intercultural research. In short, CALL practitioners are looking at human communication as the true source of interactivity; however, much learning still occurs when individuals are alone, working by themselves in isolation at odd times of the day and in nonacademic settings. In these self-study times, iCALL has proven superior to using static workbooks alone (Nagata Reference Nagata1996). When one considers that classroom teachers infrequently correct phonological and grammatical errors, but rather concentrate on discourse, content, and lexical errors, iCALL could have an honored niche in the L2 learner's individual study time. As more parser-based systems become available for a large array of languages, the goals of iCALL still hold considerable promise as a useful tool for L2 learning.

Feedback to textual input is not the only kind of response the computer is capable of; while iCALL has concentrated on syntax and text production, researchers working on automatic speech recognition (ASR) have looked to applying their results to teach pronunciation. As in the case of parsers, ASR applications “perform best when designed to operate in clearly circumscribed linguistic sub-domains” (Ehsani & Knodt Reference Ehsani and Knodt1998, p. 56) in pursuit of executing specific tasks: for example, individual sound practice, word recognition, and short sentence repetition. Although ASR software is good at recognizing short utterances (words, phrases, or short sentences), continuous spoken language input with open responses and multiturn dialogs poses almost insurmountable challenges without access to powerful computers and lots of memory. In addition, the input must be clear and noise-free, which often depends on a good microphone and noise-cancellation baffle.

Despite these obstacles, commercial software does exists that provides specific exercises dealing with mastering linguistic forms at the word or sentence level as opposed to simulating real communicative exchanges. For examples, Tell Me More Pro (Auralog, http://www.auralog.com/english.html; for a review of the Spanish version, see Lafford, Reference Lafford2004) tracks and visually displays the waveform and pitch contours from speech input in nine languages and compares them to those patterned after native speakers; the program then scores the learner's efforts on a scale from 1 to 7. The learner can zero in on and practice particular phones in isolation, if so desired. Tell Me More also provides a 3-D animation that demonstrates how the lips and tongue should move in the mouth to produce the target sound.

With the Auralog sentence exercises, L2 learners choose to speak one of several closed responses with ready-made vocabulary and syntax, which greatly simplifies the speech recognition algorithm, but prohibits learners from actively constructing their own utterances (Eskenazi, Reference Eskenazi1999). As Eskenazi (p. 64–65) observed, “To our knowledge, current speech-interactive language tutors do not let learners freely create their own utterances because underlying speech recognizers require a high degree of predictability to perform reliably.” Eskenazi (Reference Eskenazi1999) reported on some successful research efforts (i.e., the FLUENCY project at Carnegie Mellon University: http://www.lti.cs.cmu.edu/Research/cmt-projects.html) to construct a production task that uses carefully constrained elicitation techniques (patterned after the Modern Language Association's (MLA) audio-lingual method) to guide the L2 learners into speaking only the specifically targeted speech data.

ASR technology, as in the case of parser-based iCALL, may have a significant impact on CALL development in the near future, but what the field needs now is an authoring tool that makes incorporating ASR simple for nonengineers. Undoubtedly, this will be forthcoming soon, but a more serious obstacle to implementing ASR in any given language will be the existence of and access to large corpora of native and nonnative speech data, in both print and conversational formats.

Distance Learning for Language

Few language teachers would dare to admit that they are not interested in technology, at least as a support for the language curriculum, for fear of being classified as outdated or out of touch with best practices. Not surprisingly, most teachers routinely employ Web pages to distribute syllabi assignments, cultural materials, WebQuest instructions, and even lecture notes. But at the point that the faculty is polled about accepting credit for language courses delivered in a hybrid or completely distance-learning format on an equal footing with classroom instruction, the smiles may fade and the positive attitudes toward technology may evaporate. Some FL teachers still harbor deep-seated doubts as to whether or not a hybrid course, much less a completely virtual learning experience, could ever provide L2 learners with an accepted way to gain linguistic proficiency, especially when oral language skills are in question. It may be the case that a percentage of language instructors secretly worry that these new distance-learning (DL) classes will displace them or perhaps force them to drastically change their more traditional and teacher-centered classrooms into educational environments that are more student-centered.

Despite these fears, national education trends increasingly include DL classes in all disciplines. In 2005, the Sloan Consortium (Allen & Seaman, Reference Allen and Seaman2006, pp. 1–2) reported that there were approximately 3.2 million students, or one in six enrolled in higher education, taking at least one course online in the fall of 2006, up from 2.3 million in the previous year, which represents a 40% increase. These online students are overwhelmingly undergraduates and hail almost exclusively from the larger universities (i.e., 15,000 enrollments or more). More than half (51.5%) of the online students are enrolled in 2-year associate institutions (p. 6). In terms of quality, 62% of the chief academic officers at these large universities felt that the online courses are equivalent or even superior to their face-to-face counterparts. Not surprisingly, the most important barriers to the implementation of online courses and programs cited were faculty acceptance, the demands of learning teaching methods, and lack of student discipline in the online learning environment.

The term distance learning has been loosely applied to many different types of learning environments including formats that can be referred to with descriptors such as teleconference, hybrid, blended, or virtual (for a recent overview of DL education, see Fleming & Hiple, Reference Fleming and Hiple2004; Holmberg, Shelley, & White, Reference Holmberg, Shelley and White2005; Stickler & Hauck, Reference Stickler and Hauck2006). Teaching languages through teleconferencing (two-way interactive closed circuit TV) has the longest track record in the field, although this medium does not relieve either students or instructors of the burden of showing up at some specific time and place (i.e., the campus TV studio or satellite campus studio), not unlike the demands imposed by the classroom. Teleconferencing allows a teacher, usually of less commonly taught languages (LCTLs), to reach students who might otherwise have no access to instruction.

O'Dowd (Reference O'Dowd2006, p. 189) separated videoconferencing into three types: teacher-to-class, student-to-student, and class-to-class. All three types of videoconferencing potentially suffer from sound delays and gaps in fluidity in handovers that make the medium critically different from face-to-face interactions, sometimes leading students into a pattern of passive viewing (O'Dowd, Reference O'Dowd2006, pp. 191–193). Practical difficulties can become formidable obstacles to carrying out successful videoconference projects when implemented among institutions that do not share the same calendar, (i.e., semesters vs. quarters) or do not have adequate mechanisms for sharing credit and financial resources.

Ironically, teleconferencing language courses are more readily accepted by faculty curricular committees as being equivalent to the classroom experience despite whatever difficulties might exist in duplicating the same learning experience found in communicative language classroom, the gold standard for best practice in the United States. (Magnan, Reference Magnan2007). This is, no doubt, due to the privileged status of the teacher-driven classroom and the physical presence of the instructor, even if only represented by a video image projected by teleconferencing in real time, which leads the profession to validate the learning experience on a par with what happens in standard classrooms.

Unfortunately, with respect to the question of evaluation, there are only a handful of completely online DL language courses being offered for credit in our universities and even fewer evaluation studies available that could be used to establish a strong track record with which to assess their impact or to argue in favor of offering students this alternative (Zhao, Reference Zhao2003; for an online database of comparative studies for all uses of technology in language classrooms, also see Grgurovic & Chapelle, Reference Grgurovic and Chapelle2007).

Critics of learning language at a distance routinely tend to couch the debate in terms of all or nothing—all distance-learning curriculum versus a traditional classroom one—whereas the real issue is how to obtain the smooth articulation of distance-learning offerings into the overall language curriculum in ways that accommodate the needs of an institution's learners without compromising quality. In other words, DL language courses should be conceived as part of a diversified path, including study abroad opportunities, that FL departments employ to meet different needs of their students.

No language professional, not even the most technologically enamored instructor or CALL researcher, would dispute the notion that in order to reach advanced proficiency levels (Interagency Language Roundtable (ILR) level 3 or higher), L2 students need to interact face-to-face with native speakers, preferably in a country where the language is spoken. How else could an L2 student develop an appropriate sense of pragmatic or sociolinguistic competence, let alone advanced linguistic competence? Davidson's (Reference Davidson2007) study of a database spanning 30 years of students studying Russian abroad (n = 3,500) established without any doubt that the gains toward advanced proficiency are due to study abroad—especially, when carried out in full-year study programs as opposed to just one semester. Unfortunately, of the 45% of entering U.S. freshmen who profess the intent to study abroad, fewer than 3% actually undertake it (Davidson, Reference Davidson2007). This explains why both administrators and the FL profession obviously need to bring more pressure on the university curriculum to include study abroad as a requirement for graduation (see Lincoln Commission Report: http://www.lincolncommission.org/).

The central research question with respect to DL language delivery asks whether a hybrid or a completely virtual format can engender linguistic results comparable to those fostered within a traditional L2 classroom environment that meets 5 days a week, especially with respect to oral proficiency. Rephrased in a broader educational context, do DL classes have a role to play in the FL language curriculum? The answer is “yes,” because DL courses represent one way to begin the journey down the long road to advanced proficiency, a feat that will require a minimum of 700 to 1,320 hours of instruction depending on the target language's complexity (Bialystok & Hakuta, Reference Bialystok and Hakuta1994).

Chenoweth and Murday (Reference Chenoweth and Murday2003) examined the outcomes of an elementary hybrid French course, Elementary French Online (2000), developed at Carnegie Mellon University, and delivered primarily online along with an hour-long, face-to-face class meeting once per week as well as weekly 20-minute individual or small-group meetings with a native speaker tutor. Chenoweth, Jones, and Tucker (Reference Chenoweth, Jones, G. Richard Tucker, Donaldson and Haggstrom2006) and Chenoweth, Ushida, and Murday (Reference Chenoweth, Ushida and Murday2006) provided a project update on the project that includes a Spanish counterpart, Spanish Online, to that of the French Online course. The progress of French and Spanish L2 students in the online group was compared to that of others who attended a traditional class 4 hours per week on tests of oral production, listening comprehension, reading comprehension, grammar knowledge, and written production. The results showed that the scores for the treatment and control groups were not significantly different for oral production and only slightly different in the case of the writing samples, with essays by students in the online group being judged superior to those of the control group on a variety of measures, including grammatical accuracy and syntactic complexity (Chenoweth, Jones, & Tucker, Reference Chenoweth, Jones, G. Richard Tucker, Donaldson and Haggstrom2006, pp. 158–159). Blake and Delforge (Reference Blake, Delforge, Lafford and Salaberry2007) also reported no significant differences in student outcomes for discrete grammatical knowledge from both a hybrid and completely virtual Spanish online as compared to traditional classroom learners.

Chenoweth, Jones, & Tucker (Reference Chenoweth, Jones, G. Richard Tucker, Donaldson and Haggstrom2006) also looked at writing and found that all students were weakest in the use of transitions and general essay organization. Likewise, other measures for listening comprehension, grammar knowledge, and reading comprehension registered minimal statistical differences. In the first study, it was also found that the online students spent approximately 1 hour less per week studying than did those in the traditional class, suggesting that the online course was more efficient because students achieved results similar to those attained by learners in the conventional class with less time expenditure. But the second study that included Spanish showed that students spent an equal amount of time regardless of learning format (e.g., 8.5 hours per week).

Some hybrid courses confine the use of technology to the context of online workbooks. Scida and Saury (Reference Scida and Saury2006), Echávez-Solano (Reference Echávez-Solano2003), Epps (Reference Epps2004), and Walczynski (Reference Walczynski2002) reported on the progress of students using Mallard, a rote drill/quiz exercise and tracking program designed and supported by the University of Illinois at Urbana-Champaign as a way to free up classroom time for more communicative activities as well as increasing the number of sections that each instructor teaches. These three studies show no significant differences in final grades among the experimental (Mallard) group and the control group, with Scida and Saury's (Reference Scida and Saury2006) study registering slightly higher final grades for the group using Mallard. These researchers attributed much of the student success with the Mallard program to the ability for students to continue working on the exercises until reaching 100% accuracy. In essence, the availability of the tutorial CALL program appears to have allowed students to dedicate more time to making more automatic their control of the basic language structures.

Hybrid or blended classrooms are not necessarily constrained by this limited vision of technology (i.e., drill-and-kill), as illustrated by Bañados (Reference Bañados2006), whose blended approach to English as a second language for Chilean students includes (1) dedicated CALL programs; (2) online monitoring; (3) face-to-face EFL teacher-led classes; and (4) conversation classes with native speakers of English, both face-to-face and online. In Bañados's (Reference Bañados2006, p. 539) framework, both teachers and students are challenged by new roles: Teachers are more like guides and collaborators while students must assume a more autonomous participatory status.

With completely virtual language courses, the teachers and students interact exclusively online; there are no face-to-face meetings, which, no doubt, accounts for the negative reactions typically voiced by language faculty with respect to distance-learning language courses, in contrast to the tolerance shown toward the teleconferencing format and the marginal acceptance of hybrid courses. But completely virtual courses come in all shapes and sizes and encompass different combinations of asynchronous and/or synchronous modes. More often than not, virtual language courses take advantage of the authoring features provided by standardized course management Systems: for example, archived lessons, Web links, multimedia libraries, drop boxes, announcements, internal mail, bulletin boards or discussion forums, wikis, chat tools (synchronous textual exchange, sometimes voice exchanges, too), whiteboards, quiz templates, and customized grade books.

With respect to Spanish, Blake (in press) used an automated Spanish oral phone test called Versant (http://harcourtassessment.com/haiweb/Cultures/en-US/Harcourt/Community/PostSecondary/Products/Versant/VersantHome.htm) to verify that online first-year Spanish students—both in hybrid and completely virtual courses—achieved a similar level of oral proficiency to that of students learning in a traditional classroom setting. The students of the completely virtual courses outperformed the other groups, most probably because the online course attracted more mature students from the community who paid for each course they took. More data are needed to fully evaluate this question, especially from students of LCTLs, where access to beginning instruction is especially limited as a result of teacher shortages, low enrollments, and the concomitant financial constraints.

Teacher Training in Using Technology

The last 5 years have ushered in a marked desire in the CALL field to discuss issues of training teachers to use technology in service of the L2 curriculum, as evidenced by the following spate of publications: Beatty (Reference Beatty2003); Blake (forthcoming); Colpaert (Reference Colpaert2006); Davies (Reference Davies2006); Donaldson and Haggstrom (2006); Egbert (Reference Egbert2005); Egbert, Paulus, and Nakamichi (Reference Egbert, Paulus and Nakamichi2002); Felix (Reference Felix2003); Hubbard and Levy (Reference Hubbard and Levy2006a); Kassen, Lavine, Murphy-Judy, and Peters (Reference Kassen, Lavine, Murphy-Judy and Peters2007); Lai and Zhao (Reference Lai, Zhao and Zhao2005); Lomicka and Cooke-Plagwitz (Reference Lomicka and Cooke-Plagwitz2004); Stickler and Hauck (Reference Stickler and Hauck2006). Kassen et al. (Reference Kassen, Lavine, Murphy-Judy and Peters2007, p. ix) divided concerns dealing with teaching training and technology into five broadly defined categories: national frameworks (e.g., Standards, European Common Framework of Reference for Languages, No Child Left Behind), specific contexts (e.g., retraining faculty to teach in e-learning environments, graduate student training), e-communities (also for teachers, for example, MERLOT: http://merlot.org/), toolboxes, and critical reflections. Hubbard and Levy (Reference Hubbard and Levy2006b, p. 11) distinguished four possible roles for teachers with respect to CALL: practitioner, developer, researcher, and trainer. Although teachers minimally are asked to perform as practitioners, it is preferable that they be encouraged to assume other roles as well. In what follows, I will offer some pedagogical reflections on teaching training (see Blake, forthcoming, chap. 6) drawing heavily on Laurillard's (Reference Laurillard2002) insightful book, Rethinking University Teaching.

Laurillard (Reference Laurillard2002, p. 23) conceived of teaching as being fundamentally a rhetorical activity: “seeking to persuade students to change the way they experience the world through an understanding of the insights of others.” Although Laurillard is an engineer, her description of teaching goals would easily match the stated ideals for CALL researchers of either the interactionist or sociocultural school. Teachers need to help guide their students during this process of change so that they can successfully find their own third place as an incipient bilingual. Classroom L2 language learning is the result of indirect experiences or formal learning language—what Laurillard (p. 20) dubbed precepts rather than percepts, as had been the case when acquiring a first language (L1). The challenge that teachers face every day in the classroom is how to impart decontextualized knowledge and precepts about L2 language and culture, given that most students will not experience either in a direct way by living in the target society, but rather only by acting in an academic setting. Laurillard (p. 93) disfavored the lecture or teacher-centered format because it is inherently flawed: How many brilliant lectures can a teacher give in a day? In a week? In a month?

Her solution is to propose using an iterative conversational framework that first reveals and then takes into account what teachers and students alike bring to the learning environment (e.g., their respective theories, ideas, conceptions, observations; Laurillard (Reference Laurillard2002, p. 87). In processing the variation that exists between them, teachers must also imagine themselves as the learner, which will determine the focus of further dialogue and classroom activities (p. 71). As a consequence of this iterative process, there can never to be said to exist one way of teaching something—the formulation of specific lesson plans must be based on this dialogic process, which, in turn, depends on the variation of student responses (i.e., student reception).

Laurillard's (Reference Laurillard2002) approach requires that lesson plans be constantly reviewed, reworked, and adapted based on student reception—which may account for some of the resistance to technology and any change in traditional teaching methods. Teaching in the fashion advocated by Laurillard, with or without technology, becomes a discovery process, not a deductive exercise in hypothesis testing. Laurillard (p. 69, pp. 86–87) called this discovery process a phenomenographic teaching strategy because it asks teachers to qualitatively assess and then incorporate the variation in students’ conceptions into the learning environment, using the basic pathway given in the following list. Be aware that not all technological tools support all of the 12 steps; her list should be viewed as an ideal set of criteria by which to evaluate CALL.

1. 1. Teacher can describe conception.

2. 2. Student can describe conception.

3. 3. Teacher can redescribe in light of student's conception or action.

4. 4. Student can redescribe in light of teacher's conception or action.

5. 5. Teacher can adapt task goal in light student's description or action.

6. 6. Teacher can set task goal.

7. 7. Student can act to achieve task goal.

8. 8. Teacher can set up world to give intrinsic feedback on actions.

9. 9. Student can modify action in light of feedback given.

10. 10. Student can adapt actions in light of teacher's description or student's redescription.

11. 11. Student can reflect on interaction so as to modify redescription.

12. 12. Teacher can reflect on student's action to modify redescription.

Notice that the conversational framework heightens, rather than lessens, the active intervention of teachers, although not necessarily in the role of doing all of the talking. Educational media, depending on their respective affordances, contribute to at least some of the 12 steps of the conversational framework. Print, TV, video, podcasts, and DVDs provide students with narratives; DVDs and Web resources offer students an interactive space; CMC tools (e.g., chat, wikis, blogs) assist communications; simulations and iCALL make available an adaptive environment; and, finally, modeling allows students to become productive agents in their own right. An intelligent language curriculum uses a variety of educational media in ways that enable as many steps from the iterative conversational pedagogy as possible.