Introduction

‘Sometimes maligned for its allegedly behaviorist connotations but critical for success in many fields from music to sport to mathematics and language learning, practice is undergoing something of a revival in the applied linguistics literature’ (Long & Richards Reference Long, Richards and DeKeyser2007, p. xi). This research timeline provides a systematic overview of the contributions of computer-assisted language learning (CALL) to the role, nature, and development of individual practice in language learning. We focus on written language practice in Tutorial CALL, corrective feedback and language awareness-raising in Intelligent CALL (ICALL), and individualization of the learning process through tailoring of learning sequences and contingent guidance.

The origins of CALL can be traced back to the early 1960s. Early attempts in CALL followed the contemporary concept of Programmed Instruction (PI), proposed by the behaviourist psychologist B. F. Skinner. PI delivers information to be learned in small chunks, is self-paced by the learner, and provides immediate, corrective feedback. Moreover, PI in CALL was largely based on the assumption that students learn the language in the classroom and practise it on the computer. For this reason, computer feedback limited to basic knowledge-of-result (Annett Reference Annett1969) was deemed sufficient.

Almost all research and development in CALL in the 1960s to 1990s can be labelled Tutorial CALL. Tutorial CALL describes the interaction of language learners with the computer in contrast to computer-mediated communication, another yet much larger area of CALL, in which learners, instructors, and speakers of the second language (L2) interact with each other via the computer. In Tutorial CALL, the computer is commonly used for sentence-based practice activities in the form of multiple choice or multiple select, fill-in-the-blank, match or rank answers, and reassemble or translate small chunks of text. When it comes to academic enquiry in Tutorial CALL, the rapid pace of technological development from dumb terminals and large mainframe computers to microcomputers and other small devices entailed an iterative re-focus on changing technologies. Although the curricular and pedagogic context of foreign language learning changed comprehensively and rapidly through a new focus on different conceptualizations of communicative competence (Hymes Reference Hymes and Bright1966; Canale & Swain Reference Canale and Swain1980; Krashen Reference Krashen1982), this is not always evident in the pedagogic design of Tutorial CALL. The same, or very similar, learning activities were rewritten for new computer hardware and operating systems. Their storage capacity and processing speed increased and the user interface, including the handling of diacritics and non-Roman scripts, improved greatly. Just-in-time delivery across vast networks and the addition of genuine multimedia capabilities, such as audio and video clips embedded in activities and spoken learner input for multiple-choice exercises, became possible. The essential language learning activities with simplistic knowledge-of-result feedback, however, remained largely unchanged. The proliferation of multimedia-capable machines after the launch of the Macintosh in 1984 and Microsoft's Windows 3.1 in 1992 contributed to the pedagogic innovation stagnating. Similarly, these learning activities were ported to web servers with the advent of the World Wide Web in 1993 with less sophisticated techniques of input processing and less informative corrective feedback than in comparable stand-alone programs. Activity formats and feedback algorithms remained the same although second language acquisition research and pedagogic practice had moved on. In their communicative classrooms, students were taught less about linguistic structures and were thus often ill-equipped to learn during individual practice with a Tutorial CALL program. It was largely in the field of ICALL that research focused on enriching the learning experience by providing contextual metalinguistic feedback and instructional guidance based on the complex linguistic processing of students’ textual input. However, the changing role of language practice in Tutorial CALL, from reinforcing and automatizing language skills learned in the classroom to developing language skills and learning about language structures outside of the communicative classroom, is as yet under-researched.

In early programmed learning approaches in Tutorial CALL, learners were provided with mainly mechanical practice (DeKeyser Reference DeKeyser2007) of selected grammatical phenomena and lexical items. The programs were limited to simplistic knowledge-of-result feedback generated by monolithic string-matching algorithms and binary holistic answer processing (Right vs. Wrong, Try again) in addition to more sophisticated matching to regular expressions (‘Regular expression’ 2013). More detailed feedback on error location was achieved by leaving the anticipated character substring of the student answer intact and highlighting or deleting the substring that deviated from the model answer.

Changes that did occur within the research and development efforts of Tutorial CALL focused on the individualization of the learning experience. Tutorial CALL packages paid particular attention to branching, the individualization of practice sequences. Based on calibrated item difficulty—a tenet of classical testing theory, which depends on scores for a completed activity—students would be presented with more difficult practice material after achieving a high score and less difficult and remedial practice material after obtaining a low score. The main challenge for Tutorial CALL activities that rely on more or less sophisticated string-matching algorithms to provide students with contextual corrective feedback and/or additional instructional guidance is that they have to be based on the anticipation of each and every possible correct and erroneous answer in order for them to be robust and comprehensive. Of course, the number of well-formed utterances in a language is infinite and the number of erroneous utterances is then infinitely infinite. In the absence of comparing the student's input string to the strings of anticipated answers, the computer needs to be capable of a linguistic analysis of student input to detect errors and to provide corrective feedback and contextual instructional guidance. Broadly speaking, this research approach is taken in ICALL.

ICALL is the label of predominantly Tutorial CALL applications at the nexus of Artificial Intelligence (AI) and CALL. ICALL relies on natural language processing (NLP), student modelling, and expert systems. NLP produces a formal linguistic representation of learners’ linguistic input with the goal to provide corrective feedback and instructional guidance. The record of this information over time, which is maintained in student profiles, provides the basis for inferring interrelated facets of the student's cognitive belief system about the learned language, that is, the construction of a student model. Information from the student model, in turn, provides some basis for the tailoring of learning sequences and contingent guidance. Expert systems model the learning domain, such as aspects of the grammar of a language, and are a rich source of structured (linguistic) knowledge that can guide and scaffold the students’ learning processes. They enable learners to query this knowledge base during task completion thus serving as a comprehensive reference tool in learner-computer interactions.

Most research in ICALL between 1978 and the early 2000s was aimed at grammar checking for language learners, error detection and correction of learner input, and the provision of more informative, metalinguistic learner feedback. After thorough discussions of the NLP-hard problems of analysing learner language and error classification and their annotation in learner corpora, ICALL research has focused increasingly on the implementation of robust human language technologies such as lemmatizers (i.e., algorithms which aim to remove inflectional endings to return the base form or dictionary entry of a word), part-of-speech taggers, and parsers to analyse well-formed texts and augment them with additional information that is made available and useful to language learners (Schulze & Heift Reference Schulze, Heift, Thomas, Reinders and Warschauer2013). The information (e.g., conjugational paradigms of verbs in a reading text and/or highlighted less salient word classes such as prepositions) focuses the learners’ attention, helps them notice linguistic patterns by raising their language awareness, and/or scaffolds their language use.

Our belief about the pedagogic importance of corrective feedback, instructional guidance, and language awareness-raising in language learning practice and the nature of the developmental trend described above led us to include fewer publications on Tutorial CALL based on simple string-matching algorithms for the analysis of learner language. The literature on these is also scarce. Authoring tools for language learning practice, for example, Camsoft's Fun with Texts (www.camsoftpartners.co.uk/fwt.htm) and the Wida suite of software, as well as general instructional authoring software such as Toolbook (http://tb.sumtotalsystems.com/) or software for designing tests such as the Question Mark products (www.questionmark.com) were very popular among language instructors and were presented and discussed at workshops, but hardly ever featured in research articles. Accordingly, we identified the following themes in the publications provided in our timeline:

• • Survey and review articles

  • ○ Overviews (O) summarize the state of the art at that time and focus on research questions or challenges pertinent to Tutorial CALL;

  • ○ Background articles (B) discuss important processes and variables in Tutorial CALL without necessarily focusing on one particular project.

• • Pedagogy articles (P) discuss general aspects of second language development and language teaching methodology. Two important specific aspects are identified:

  • ○ Corrective Feedback (F) subsumes form-focused feedback on individual items of practice activities, written texts and dialogues, and generally the recognition and processing of learner errors;

  • ○ Language Awareness (A)—understood very broadly—encompasses scaffolding resources for students’ perceptive activities and guidance that focuses their attention, facilitates noticing, and generally helps learners to sequence the necessary steps in a complex problem-solving process.

• • Research and development projects

  • ○ ICALL project (I);

  • ○ Tutorial CALL project—no NLP (T).