Should You Look Up, Infer, Save, or Ignore a New Word?

A reader reaches the end of a purpose-written illustration containing three invented target forms: naver, doral, and selwick.

The first occupies a position that could change the conclusion, the second sits inside a strongly suggestive sentence, and the third adds detail about a gate. All three create the same small urge: stop, translate, save.

Although officials initially blamed the drought, the review did not naver the contractor. Inspectors found the storage bed doral after weeks without rain, beside a selwick iron gate.

These are invented forms, not vocabulary to learn. They let us compare three consequences without assuming that the reader already knows the words.

The impulse to treat all three alike is understandable. Once a lookup has demanded attention, saving the result can feel like the best way to preserve that effort. Meanwhile, unfamiliarity is immediately visible; judging a word’s importance to the paragraph or future usefulness takes another decision.

But readers can be selective without directing every consultation toward the main point. One small study included 34 female Japanese university students who read and summarized three authentic English passages with access to an online bilingual dictionary.Source

In that study, 75% of dictionary consultations involved a frequent or study-designated useful word, or a word in a main-point clause; 25% involved an infrequent word in supporting details.Source

When an unfamiliar word appears during meaning-focused reading, how can a reader decide whether to infer it, look it up, save it, or ignore it without unnecessarily losing the story or argument?

Three words, but not one decision

The three positions in the opening illustration create different risks. Misinterpreting naver could change the conclusion about responsibility. The sentence around doral strongly narrows its plausible local meaning. Removing selwick leaves the event and argument intact.

The 75%/25% result does not establish a universal lookup cutoff or prove that a supporting-detail consultation was wasted.Source

The categories depended partly on frequency lists and the researcher’s coding of main-point clauses, and the study did not experimentally test whether reducing lookups would improve learning.Source

Before assigning actions to the three invented forms, however, consider a tempting alternative: perhaps one strategy—guessing or dictionary use—is generally best.

What happened when a study assigned both strategies

The Alahmadi and Foltz study included 61 Arabic-L1 Saudi senior undergraduate English majors.Source

Participants guessed underlined target words in some texts and used a bilingual dictionary for targets in others.Source

The task had no time restriction, required responses for the target words, and included two comprehension questions after each text.Source

The comparison found no evidence that guessing produced a larger delayed vocabulary-learning effect than lookup: t=0.59, df=119.85, p=.55, and Cohen’s d=0.11.Source

Larger measured vocabulary size was associated with larger gains in both conditions, while successful guessing was additionally associated with gains for targets handled through inference.Source

These are associations involving measured learner characteristics, not proof that vocabulary size or guessing success alone caused the outcomes.Source

The result does not establish that guessing and lookup are always equivalent in ordinary web reading.Source

The participants were senior English majors, the targets were underlined, and the task required responses.Source Self-directed reading is less controlled, and the consequences of misinterpreting different words are not uniform.

Back in the illustration, the study makes doral look like a reasonable candidate for provisional inference, but it does not make guessing naver prudent when a wrong interpretation could reverse the conclusion.

Eye-tracking research adds a related caution. One study included 28 advanced learners of English reading texts containing 12 incidental-learning targets.Source

Participants spent more time processing unknown pseudowords than matched controls, and longer looking was associated with a greater probability of recognizing a pseudoword on an unannounced vocabulary post-test.Source

This association does not prove that deliberately looking longer causes vocabulary learning; fixation duration was observed rather than assigned.Source

The useful question, then, is not which strategy wins everywhere. It is what would happen to the current passage if this particular interpretation were wrong.

A two-stage checkpoint protects both meaning and attention

Use a checkpoint at the end of the smallest coherent unit—normally a sentence or short paragraph—unless the word prevents you from parsing the sentence or immediately controls a warning, instruction, quantity, negation, heading, or other high-stakes meaning.

The two-stage, four-action checkpoint is Stickly Editorial’s evidence-informed synthesis, not a diagnostic algorithm validated as a whole by the consulted studies.Sources Sources

Those studies compare assigned strategies and observed lookup patterns; they do not test the complete sequence presented here.Sources Sources

At the checkpoint, make two separate decisions: first protect the meaning of the passage, then decide whether the word has enough future value to save. The second decision is not an automatic consequence of the first.

First decide what the word does to the current passage; only then decide whether it deserves future review. Editorial illustration generated for this article.

Stage 1: What changes if the interpretation is wrong?

Look up a hinge word. In this editorial model, a hinge word can change who did what, the author’s position, polarity, cause, contrast, quantity, referent, or an action you must take. A repeated topic term can also become a hinge when later sentences depend on it.

Infer a constrained word provisionally. Do this when grammar, surrounding meaning, and relevant knowledge converge on one plausible interpretation—and when a mistake would not materially change the passage. “Provisionally” matters: you are carrying a working interpretation, not declaring that you have verified or learned the word.

Ignore a peripheral word for now. Mentally remove it. If the event, argument, relevant tone, and connection to the next sentence remain coherent, continue reading. Reclassify it if it recurs or later becomes consequential.

A compact decision rule is:

Finish the sentence or short paragraph when possible.
Ask, “If I am wrong about this word, what changes?”
If the claim, event, relation, quantity, referent, or required action changes, look it up.
If context strongly constrains the answer and little changes if you are slightly wrong, infer provisionally.
If nothing material changes, ignore it until the text gives you a reason to reconsider.

For legal, medical, examination, translation, or safety-related reading, lower the lookup threshold. This is editorial risk-management advice rather than an outcome tested by the cited vocabulary studies.

Stage 2: Will this word earn future attention?

After the passage makes sense, ask a different question: “Am I likely to need this word again?” Consider saving it when at least one of these signals is credible:

it has already recurred;
it is central to a subject you read regularly;
you expect to use it in speaking or writing;
it names a distinction you repeatedly need to understand;
its context is memorable enough to make later review useful.

These future-value signals are editorial prioritization, not experimentally proven saving thresholds.Sources Sources

Evidence does support keeping different vocabulary outcomes conceptually separate. In one controlled story-reading study, data from 23 L2 participants were included in the analysis after eye-tracking exclusions.Source

After eight exposures to target nonwords, the L2 readers recognized 86% of the forms, recognized 75% of the meanings, and recalled 55% of the meanings.Source

These percentages came from repeated nonwords in one controlled story and do not prescribe a universal encounter count for ordinary web reading.Source

A meta-analysis synthesized 359 effect sizes from 42 studies with 3,802 participants.Source

Across the included studies, glossed reading produced vocabulary-learning rates of 45.3% immediately and 33.4% after a delay, compared with 26.6% and 19.8% for nonglossed reading; results also varied with treatment, testing, comprehension, and learner characteristics.Source

Researcher-provided glosses are not the same intervention as asking learners to infer a word or consult a dictionary.Sources Sources

Together, the studies support a modest interpretation: accessible lexical information can support vocabulary learning without making durable, recallable knowledge automatic.Sources Sources

Return to the paragraph

Now the opening illustration produces different actions.

Naver: look it up because its position controls the conclusion about responsibility. Save the real word you discover only if it passes the separate future-value test.
Doral: infer a provisional meaning because the storage bed and absence of rain strongly constrain the interpretation. Verify it if later information makes precision consequential.
Selwick: ignore it because the property of the gate does not change the event or argument. Reconsider it if later sentences make that property important.

The initial impulse was attractive because each unknown form presented the same visible uncertainty. It failed when it treated present comprehension, lexical curiosity, and future study value as one problem. The replacement is a two-stage allocation of attention: classify the word by its consequence for current meaning, then apply a separate future-value test.

The exact thresholds remain uncertain and should vary with proficiency, genre, reading purpose, available context, and the cost of misunderstanding.Sources Sources

Ignoring a peripheral word is not itself a vocabulary-learning technique. In this editorial model, it is a way to continue when the passage remains coherent, rather than automatically converting every uncertainty into a lookup and every lookup into future review.

On your next page, wait for the end of a coherent unit when possible and ask what would change if your interpretation were wrong. Repair consequential meaning first; decide what deserves a place in future study only after the passage is stable.

A restrained role for reading tools

As checked on August 11, 2026, Stickly’s first-party page says that a small action appears after deliberate selection.Source

The page also says that selection translation, remembered-word highlights, and inline practice are separate controls that can be changed independently.Source

These documented controls do not establish that the editorial checkpoint improves learning or determine which words deserve attention.Source

Sources

AI-assisted research and automated checks by Stickly Editorial

Stickly Editorial uses AI-assisted research and writing tools. Every published article passes automated source, product-accuracy, and quality checks.