An unfamiliar word sits in the middle of a 740-word article. The reader recognises the topic, sees a contrast marker, and makes a plausible guess. Two paragraphs later, that guess bends the argument in the wrong direction. Context has done useful work, though its answer remains provisional. For a B1 or B2 reader in Taiwan, the sound routine is simple: infer enough to keep reading, mark the confidence level, then verify words that carry the writer's claim.
A guess can feel complete
Lexical inferencing means using the surrounding sentence, grammar, word parts, and prior knowledge to estimate an unknown word's meaning. It keeps a reader moving when stopping for every unfamiliar item would destroy the thread. The danger comes from fluency of thought. Once an interpretation fits the sentence, the mind may stop testing it. That early sense of fit deserves a second source of evidence before it becomes a saved vocabulary note. Hossein Nassaji examined think-aloud reports from 21 intermediate learners of English as they inferred word meanings. Their overall success rate was low even though they used several available strategies and knowledge sources. Success depended more on the quality of strategy use than on the number of strategies attempted. The 2003 TESOL Quarterly study was small and observed an elicited inferencing task, so it does not supply a universal failure rate for ordinary reading. It does show why the instruction "use context" leaves too much unspecified.
One clue rarely settles the matter.
Uncertainty is part of the record. Writing "probably an adjective meaning cautious, confidence 2/4" preserves both the proposed sense and the reason to revisit it when a later sentence supplies stronger evidence. Take the invented word "brindle" in this sentence: "The committee gave a brindle response, approving the budget while delaying every purchase." The grammar identifies an adjective. The second clause points toward a mixed or qualified response. Neither clue tells us whether brindle actually carries that meaning in English. In fact, brindle usually describes streaked animal colouring. A tidy local interpretation can still be false.
Context clues also differ in strength. A direct definition nearby carries more weight than the general mood of a paragraph, while a familiar prefix helps only when the reader knows how that particular word uses it. Topic knowledge can narrow the possibilities and can also create confirmation bias, especially when the writer is challenging a belief the reader already holds. A careful inference combines at least two independent clues, such as sentence grammar and a later example, then records uncertainty when those clues pull in different directions. The result should remain a testable proposal.
Choose which words deserve a stop
Verification costs time, so readers need a selection rule. A word deserves attention when it changes who did what, carries the author's evaluation, returns several times, or blocks the relation between two claims. An adjective describing the wallpaper in a long story may wait. A verb describing what a company did with NT$37,810 cannot. I would mark guesses with three symbols in the margin: a dot for low stakes, a question mark for uncertain and important, and a tick for verified. The symbols separate "I can continue" from "I know this word." They also keep dictionary use narrow. Seven lookups in a 12-minute reading session may be reasonable. Thirty-seven usually turn the article into a sequence of isolated entries.
The reader should first write a short proposed meaning, such as "reduce sharply," and identify the evidence.
Was it a prefix, a nearby example, a contrast word, or knowledge of the topic? Then check a learner dictionary for the sense, part of speech, pronunciation, and one example. If the entry offers six meanings, return to the sentence and choose the one that fits its grammar and topic. Copying the first Chinese translation gives no such check.
Repetition helps after meaning is available
Jan Hulstijn, Merel Hollander, and Tine Greidanus studied 78 advanced learners of French who read a short story under one of three conditions: marginal first-language glosses, access to a bilingual dictionary, or neither. Sixteen target words appeared once or three times. Repetition supported learning more when readers received or looked up the meanings. Without that support, readers sometimes ignored a word or inferred it incorrectly. The 1996 Modern Language Journal article tested advanced Dutch university students reading French, and its immediate word-recall task cannot represent every English reading goal. The result sets a useful boundary. Meeting a word three times may reinforce a correct form-meaning connection. Three encounters can also rehearse the same wrong guess. Exposure counts start to matter after the learner has a defensible meaning to attach to the form.
Verification alone also leaves a shallow trace.
A dictionary entry viewed for four seconds can vanish by the next page. After checking, reread the original sentence and make one small retrieval attempt: cover the entry, say the meaning, and explain which phrase in the sentence supports it. For a word worth keeping, add a card later. The card should ask for recall in both directions, following the logic in a vocabulary card needs a way back to English.
A four-minute classroom routine
Give learners a 170-word paragraph containing four underlined targets. Two targets should be inferable from strong local clues, one should contain a misleading familiar word part, and one should remain genuinely ambiguous. Ask for a proposed meaning and a confidence score from 1 to 4 before any dictionary opens. Pairs compare the clue they used. Agreement earns no automatic credit. They have to point to syntax, a cohesive marker, a definition, or a concrete example in the text. After 90 seconds, the teacher releases dictionary access. Learners record the verified sense and revise the sentence summary if the new meaning changes it. The ambiguous item matters because a good reader sometimes has to write "insufficient evidence."
This routine suits dense exam passages and workplace reports. It is excessive for extensive reading chosen mainly for pleasure, where frequent checking can damage pace and attention. In that setting, verify only recurring words and items that block the plot. Different purposes justify different lookup budgets.
Tomorrow, take one GEPT reading passage and circle exactly three unknown words. Beside each one, write the guessed meaning, the clue, and a confidence score. Open the dictionary only after all three guesses are visible on paper.