The idea is older than the app

StepText's core move can sound like a gimmick: take a text in a language you know, and gradually replace parts of it with the language you're learning, increasing the proportion as you go. Words first, then phrases, then whole clauses. It feels novel because no mainstream app does exactly this. But the idea is not ours, and it is not new. It has a name, a date of birth, and a lineage worth knowing — partly because that lineage is what makes us trust the method, and partly because being honest about its limits is how a serious tool earns trust.

1968: an "outlandish proposal"

In 1968 the anthropologist Robbins Burling published a paper with the wonderful title Some Outlandish Proposals for the Teaching of Foreign Languages. In it he described taking an English text and rewriting it so that, page by page, French vocabulary and then French syntax crept in, while the text stayed understandable throughout. He called it the "diglot weave." He reported it felt faster and easier than anything he'd tried.

It's worth being precise: Burling's paper was a proposal with an anecdote attached, not a controlled experiment. He named a mechanism and showed it was plausible. He did not prove it worked better than the alternatives. Fifty-five years later, that's still roughly where the hard evidence sits — more on that below. But he put his finger on the exact thing StepText does, and he did it before the modern language-app industry existed.

The orthodoxy it argued against

To see why a blended text was "outlandish," you have to remember what it was arguing with. For most of the twentieth century, the dominant idea in language teaching was that the learner's first language was the enemy — interference to be banished from the classroom. From the Direct Method through audiolingual drills and into communicative "target-language-only" immersion, the instruction was the same: don't translate, don't lean on what you already know, stay in the new language at all costs.

Deliberately printing the two languages in the same sentence was heresy against that orthodoxy. And the orthodoxy, it turns out, was overstated.

The quiet rehabilitation of the first language

A few dissenting strands rebuilt the case for using what you already know. In 1960s Wales, C. J. Dodson's "bilingual method" used the mother tongue as a tool on purpose, including the sandwich technique: say the foreign phrase, give a quick gloss in the known language, then repeat the foreign phrase — meaning delivered instantly, without breaking the flow. The applied linguist Wolfgang Butzkamm spent decades developing this into what he called "enlightened monolingualism": the mother tongue, used well, lays the cognitive foundations for everything that follows, because it lets you grasp meaning and form at the same time.

Two more ideas finished the rehabilitation. Vivian Cook's notion of multicompetence (1992) reframed the bilingual mind not as two deficient half-speakers but as one integrated system with its own value — so using both languages together isn't cheating, it's working with the brain you actually have. And Ofelia García and Li Wei's translanguaging (2014) made the modern case that a bilingual draws on one unified repertoire, moving between languages as a resource rather than a failure. Add Krashen's comprehensible input — the principle that we acquire language by understanding it — and you have a coherent reason why a blended text should work: it keeps the input comprehensible while the difficulty climbs.

The same instinct, out in the wild

The idea keeps being reinvented by people who clearly never read Burling, which is usually a sign there's something real in it. The Russian-language readers of Ilya Frank embed a literal translation after each phrase, then reprint the passage clean. The "all-Japanese-all-the-time" crowd informally "sandwich" new phrases. And a small cluster of startups now sell Burling's diglot weave directly, often AI-generated. The instinct — let me read something I can mostly understand, with the new language mixed in — is durable and grassroots.

The honest part: the evidence is thin

Here is what we won't do: quote you a big number and pretend it settles things. The best controlled study we know of, by Christensen, Merrill & Yanchar (2007), found a computer-based diglot reader was as effective as a sophisticated drill program for vocabulary — and that learners liked it more. That's a genuinely nice result, but "as good as drilling, and more pleasant" is a modest claim, not a revolution. You will also see a widely repeated figure that the technique is "almost twice as effective." It traces back to a single small study of sixty high-schoolers over five sessions, measuring vocabulary recall only. Treat it with the caution it deserves.

So the fair summary is: blended text rests on a respectable theory (comprehensible input, judicious use of the first language, multicompetence) and a thin, mostly vocabulary-focused evidence base. It is promising and pleasant, not proven and transformative. We'd rather you knew that.

What StepText borrows, and what's actually new

We'll happily credit our inheritance. The core mechanic is Burling's diglot weave; the justification is Krashen, Butzkamm, Cook, and García & Wei. We're not claiming to have invented the wheel.

What's genuinely different is the engineering. Burling wrote his weaves by hand, and that was always the method's fatal flaw: producing well-graded blended text, tuned to a particular reader and ramped over time, is brutal manual labour, which is why it never scaled past enthusiasts. Software changes that. StepText can generate the weave, blend at the level of words and phrases inside a sentence (not as a separate translation column the way parallel-text apps do), and raise the proportion of the target language as you go — automatically, for your pair and your level, fresh every day. The old idea finally has a body that fits it.

That's the bet: a fifty-year-old technique with a sound theory and an impractical delivery, made practical. We think it's a good bet. We'll keep telling you exactly how good the evidence is as we learn more.

Sources

1. Burling, R. (1968). Some Outlandish Proposals for the Teaching of Foreign Languages. Language Learning 18.
2. Christensen, E., Merrill, P. & Yanchar, S. (2007). Second Language Vocabulary Acquisition Using a Diglot Reader or a Drill-and-Practice Program. CALL 20(1).
3. Nemati, A. & Maleki, E. (2014). Teaching Vocabulary through the Diglot-Weave Technique. Procedia.
4. Dodson, C. J. — the bilingual method & sandwich technique (overview).
5. Butzkamm, W. & Caldwell, J. (2009). The Bilingual Reform: A Paradigm Shift in Foreign Language Teaching.
6. Cook, V. (1992). Multicompetence (overview).
7. García, O. & Wei, L. (2014). Translanguaging (overview).
8. Krashen, S. The Case for Comprehensible Input.
9. Frank, I. The reading method.