Computer Assisted Language Learning (CALL)

Computer-assisted language learning (CALL) is, in the standard definition, “the search for and study of applications of the computer in language teaching and learning” (Levy, 1997). In practice it covers everything from a 1960s mainframe drilling Russian vocabulary to the app on your phone scheduling today’s review. What ties these together is not a single technology but a single idea: that a machine can present language, elicit a response, and react to it, giving a learner interactive practice without a teacher present for every minute of it. The technology has changed beyond recognition over sixty years; the underlying question — can a computer genuinely help someone learn a language, and if so, how — has stayed the same.

The three phases of CALL

The most influential way of organising this history comes from Mark Warschauer and Deborah Healey, who in the late 1990s divided CALL into three overlapping phases, each tied to a level of technology and a theory of how languages are learned. The framework matters because it makes a point that recurs throughout this article: the pedagogy behind a program matters more than the hardware it runs on.

Behaviouristic CALL (conceived in the 1950s, implemented through the 1960s and 1970s) treated the computer as a tireless drill master. Grounded in behaviourist learning theory, programs presented a stimulus and judged the learner’s response, offering endless repetitive practice on the assumption that language is a set of habits built by reinforcement. This was the era of drill-and-practice on mainframes. Warschauer later renamed this phase “structural CALL” to describe its focus on grammatical form rather than its psychological roots.

Communicative CALL (late 1970s into the 1980s) emerged as behaviourism fell out of favour and communicative language teaching rose. Its advocates argued that computers should give learners something to communicate about and let them generate original utterances, not just fill blanks. Programs of this era — text reconstruction, simulations, exploratory games — used the computer as a stimulus and a tool, with the skill practised implicitly rather than the form drilled explicitly.

Integrative CALL (from the 1990s onward), enabled by multimedia and the Internet, sought to integrate the various language skills — reading, writing, listening, speaking — into authentic tasks, and to embed technology into the ordinary flow of learning rather than treating a trip to the computer lab as a separate event. Multimedia CD-ROMs and, soon after, networked communication made it possible for a learner to read, listen, record their voice and write to a real audience within one activity.

From PLATO to the CD-ROM

CALL has a definite birthplace. PLATO, built at the University of Illinois from 1960, was the first generalised computer-assisted instruction system, and among its thousands of lessons were substantial foreign-language courses in Russian, French, German, Spanish and others. Its terminals offered interactive drills, immediate feedback and self-paced study — the template language software still follows. A parallel 1970s project, TICCIT (developed by the MITRE Corporation and Brigham Young University), took a different design route but shared the same drill-and-tutorial DNA; together PLATO and TICCIT defined what early institutional CALL looked like. Both were expensive, mainframe-bound and confined largely to universities and the military.

The personal computer changed the audience. Through the 1980s, CALL migrated from mainframes to Apple II and IBM PC machines, and language exercises could finally sit on a floppy disk in a classroom cupboard. The decisive leap in ambition, though, came with the CD-ROM in the late 1980s and 1990s, which could hold sound and video and so turned the computer into a genuine multimedia tutor. This is the moment CALL reached ordinary consumers: the era of Tell Me More and Rosetta Stone, the first commercial language products to sell CALL as a boxed course you could buy and run at home. For the first time, learning a language with a computer was something you did in your living room rather than a research lab.

The web, mobile learning and data-driven learning

The arrival of the World Wide Web dissolved the boundary of the CD-ROM. Content no longer had to fit on a disc; it could be updated, linked and shared, and learners could reach real texts, native speakers and each other. The 2000s brought Web 2.0 — podcasts, wikis, blogs and social networks pressed into language teaching — and virtual learning environments such as Moodle became the ordinary infrastructure of a language course.

Then the phone arrived. Mobile-assisted language learning (MALL), a recognised sub-field of CALL, moved practice off the desk and into spare minutes on a bus or in a queue. Its defining traits are portability and short, frequent sessions, and it is the form in which most people now encounter CALL at all: the modern language app is MALL in everyday clothes.

A quieter but influential strand runs alongside the mainstream. Data-driven learning (DDL), a term coined by the applied linguist Tim Johns in the early 1990s, turns the learner into a “language detective”: instead of being told a rule, they query a large corpus of real texts through a concordancer and infer patterns from the many examples the software lays out side by side. DDL uses the same tools professional linguists use and rests on the idea that authentic data, discovered rather than dictated, teaches usage better than invented sentences. It has never been the dominant model, but its influence — on how apps surface real examples in context — is wide.

Today the frontier is intelligent CALL (ICALL) and AI. ICALL applies natural-language processing so a program can parse what a learner actually wrote and diagnose the specific error rather than merely marking an answer wrong — the long-promised “intelligent” feedback that early CALL researchers argued about. Large language models now make conversational practice, on-demand explanation and automatic generation of exercises cheap and fluent. They also reintroduce an old caution in a new form: a system that produces confident, plausible language can produce confidently wrong language too, which is why current work focuses as much on correcting and constraining AI as on generating with it.

What the research says

Does any of this actually work? The honest summary from decades of study is: yes, modestly, and mostly because of the teaching, not the technology. Meta-analyses that pool many studies typically find that computer-supported language instruction is at least as effective as instruction without it, and often somewhat better, with effect sizes in the small-to-moderate range. A widely cited synthesis put the median advantage at roughly half a standard deviation on language tests — a real but not miraculous gain. Mobile learning shows similar or slightly stronger effects in recent reviews.

The more important finding is what these studies say about why. When researchers look closely, the technology itself is rarely the active ingredient; what predicts a good outcome is sound pedagogy — meaningful practice, useful feedback, spaced and repeated exposure, tasks matched to the learner’s level. A well-designed exercise on plain hardware beats a badly designed one on expensive hardware. This is the empirical version of Warschauer and Healey’s point: the same device can host behaviouristic drilling or communicative exploration, and it is that choice, not the silicon, that moves the needle. It is a useful corrective to every wave of hype — videodisc, CD-ROM, the web, now AI — that arrives promising that the new medium will transform learning by itself.

What this means for language learning

Sixty years of CALL have settled a few things. A computer really can help you learn a language — that was proved on PLATO in the 1970s and confirmed by every meta-analysis since. The first commercial products put that capability in the home; Duolingo (2011) and the app generation put it in everyone’s pocket. But the research is equally clear that the medium is not the method: a tool helps to exactly the degree that it delivers good pedagogy — interactive practice, honest feedback, and exposure timed so you meet a word again just as you are about to forget it. That last piece, memory scheduling, is what the earliest CALL systems lacked and what modern spaced-repetition apps add. Taalhammer’s method is built on exactly this reading of the evidence: use the computer for what it is genuinely good at — relentless, personalised, well-timed practice on real sentences — and let the pedagogy, not the novelty, do the work.

Frequently asked questions

What is the difference between CALL and CALI?

CALI (computer-assisted language instruction) was the earlier name, common through the 1970s. It fell out of favour because “instruction” implied a teacher-centred, drill-based model, and as the field embraced communicative and learner-centred approaches the term “learning” (CALL) was adopted to reflect the shift. They refer to the same broad field at different stages of its thinking.

Is a language app like Duolingo an example of CALL?

Yes. Any use of a computer to support language learning falls under CALL, so a phone app is squarely within it — specifically the mobile-assisted (MALL) branch. Modern apps are the mass-market, pocket-sized descendants of the mainframe courseware that CALL began with; the lineage runs directly from PLATO through the CD-ROM era to today.

Does using a computer actually help you learn a language faster?

On average, modestly yes: pooled research finds computer-supported instruction is at least as effective as, and often a bit better than, teaching without it. But the gain comes from the quality of the exercises and feedback, not from the device itself. A thoughtfully designed program helps; a flashy but shallow one does not. Judge a tool by its pedagogy, not its technology.

Sources:

1. Mark Warschauer & Deborah Healey, “Computers and language learning: an overview”, Language Teaching, vol. 31, 1998, pp. 57–71: https://education.uci.edu/uploads/7/2/7/6/72769947/computers_and_language_learning-_an_overview.pdf
2. Mark Warschauer, “CALL for the 21st century” / the three-phase framework, with the first phase later renamed “structural” (ICT4LT overview): http://www.ict4lt.org/en/warschauer.htm
3. Computer-assisted language learning — Wikipedia (history, phases, ICALL, MALL): https://en.wikipedia.org/wiki/Computer-assisted_language_learning
4. Yong Zhao, “A meta-analysis of effectiveness studies on computer technology-supported language learning” and related syntheses (median effect ≈ 0.5 SD): https://dr.lib.iastate.edu/server/api/core/bitstreams/fd7d2c1e-b61c-4522-aa36-ea9c6ce0fc9e/content
5. Data-driven learning (Tim Johns, concordancing) — Wikipedia: https://en.wikipedia.org/wiki/Data-driven_learning