Zero Uncertainty

The state in which a learner has no unanswered questions about a text. The term is Scott Thornbury's (2011), borrowed from Frank Smith's Understanding Reading, where Smith writes: "We predict to reduce any uncertainty we might have, and therefore to reduce the amount of external information we require... until uncertainty is reduced to zero" (p. 18). Smith never uses the exact phrase; Thornbury coined the label to argue that most coursebook listening treatments stop far short of it, leaving learners with a superficial grasp of the text and very little language uptake.

The critique

Thornbury's polemic targets the now-standard skills-lesson recipe: activate schema, pre-teach a short vocabulary list, set one gist question, set one slightly deeper question, move on. He illustrates with a parodic sequence about an abducted garden gnome in which the gist question ("put the interviewer's questions in order") can be answered without really listening, and the deeper question ("why are these words mentioned?") goes "as deep as it gets."

His complaint is structural, not a matter of taste: this sequence is "like looking at the target language from 30,000 feet. But that's where the learners are already." By leaning on top-down scaffolding (background knowledge, predictive activities, easy gist tasks), the lesson "may delude both learners and teachers into thinking that linguistic information can safely be ignored." After it, learners are no better off than before. The TAVI stage never gives way to sustained bottom-up engagement.

What zero uncertainty actually requires

Thornbury's proposal is to keep probing:

further layers and layers of questions that probe and probe and probe at the learners' emergent understanding, until not a word has been by-passed, not a discourse marker ignored, not a verb ending overlooked, and not a question left unanswered.

The culmination is a word-by-word transcription task, not necessarily of the whole text but of a decent-sized chunk. The transcription does three things at once: it forces decoding, it surfaces the learner's remaining uncertainties, and it converts the recording into a text that can then be mined for form.

Smith's qualifier: comprehension is relative

The concept is not "100 % understanding" in the abstract. Smith is careful that "comprehension is relative, that it depends on the questions that an individual happens to ask" (Understanding Reading, p. 19). Two learners listening to the same weather report may each reach their own zero uncertainty at different points, because they are asking different questions of the text. Thornbury extends this in the comment thread: "the amount of certainty that the learners want to be left with should depend on them," and "it might be better if coursebooks didn't include tasks at all, but just directed the teacher to 'mediate the text to the extent that seems necessary'."

This saves the concept from being a counsel of perfection. Zero uncertainty is not a universal target but a learner-specific end state: no remaining question that the learner wanted to ask.

When to push for it and when not to

Commenter Neil McNeil Mahon's pushback is well taken: an airline announcement demands more certainty than a passing film joke, and even L1 speakers tolerate significant uncertainty in ordinary life. Luiz Otávio Barros adds a diagnostic observation from classrooms: modern learners may be too comfortable with ambiguity, often "assuming they had heard things that were never said" through hyper-activated schemata. In both directions, the zero-uncertainty ideal functions less as a fixed target and more as a corrective against the routine coursebook under-exploitation of text.

Text quality prerequisite

"To withstand the weight of so much probing, I would need a text that was of much more intrinsic interest, educational value, and linguistic capital than one about abducted garden gnomes." A text chosen to bear close attention is a precondition; most coursebook scripts are not. Thornbury's preferred alternatives: unscripted recordings (not necessarily authentic, but made with real speakers rather than actors), teacher-recorded interviews with colleagues, live listening to the teacher, and learner-chosen audio or video from the open web. Sites like ELLLO were cited in the thread.

Techniques that push toward zero uncertainty

• Word-by-word transcription of a short chunk, often set as homework (Tony Lynch's practice, echoed by several commenters who transcribe song lyrics, film scenes, or Skype conversations to learn a language).
• Dictogloss and reconstruction tasks, where learners compare what they wrote with what was said and notice the gap.
• Listen-and-write-exactly micro-tasks: "Listen to these sentences and write down exactly what you hear," "Count the number of words."
• Stopping live listening mid-utterance (Paul Seligson): the teacher pauses, learners whisper a completion to a partner, then the teacher continues. Turns exposure into prediction and noticing.
• Repeated hearings with escalating probes: after the gist pass, each subsequent pass targets a smaller grain (discourse markers, verb endings, connected-speech features).
• Transcript as scaffold: release the transcript only after learners have done their own work, then use it to confirm or revise what they caught.

Connection to TAVI → TALO

Zero uncertainty is the honest end-point of the TAVI phase before a lesson pivots to TALO. Without it, learners reach TALO (language-focus) work on a text they never fully heard, and the language-focus risks becoming pseudo-teaching around a still-opaque text. In Thornbury's Essential 32, the "onion with layers" metaphor names the same idea from the other side: each task peels away a layer until zero uncertainty is reached.

Caveats from the thread

• Not "understand every word in every text." A text chosen for extensive listening or reading is, by definition, not supposed to be plodded through at this level.
• Time cost. Pursuing zero uncertainty on one text means sacrificing coverage of others. Transcription homework offloads some of the cost.
• Learner tolerance. Some learners welcome deep text work; others lose motivation. Task design should let learners choose how far they want to probe.
• Text choice is load-bearing. A dull text does not reward sustained scrutiny. Richard Cauldwell's adjacent point is that the payoff of deep work is "engaging with the sound-substance of speech," which only matters if the speech is worth engaging with.

References

• Smith, F. (2004). Understanding Reading (6th ed.). Lawrence Erlbaum.
• Thornbury, S. (2011). Z is for Zero Uncertainty. An A–Z of ELT [blog post].
• Richards, J. C. Second Thoughts on Teaching Listening. (PDF linked from the original thread.)
• Williams, R. (1986). "Top ten" principles for teaching reading. ELT Journal, 40(1).
• Johns, T. & Davies, F. (1983). Text as a vehicle for information. Reading in a Foreign Language, 1(1), 1–19.