The evidence question
Should You Guess an English Word from Context — or Look It Up?
When should you infer an unfamiliar English word from context, and when is checking a dictionary or gloss the better move? Research suggests the best strategy is often to infer first, then verify selectively.
A question worth asking.
09 sources, open to exploreIn this article 16 sections
You are reading in English and hit a word you do not know.
You can still understand the sentence. The surrounding clues are strong enough that you have a rough idea what the word probably means.
Do you keep reading and trust the context?
Or do you stop, open a dictionary, and check?
Language learners are often given opposite advice here. One teacher says never interrupt your reading — learn to guess from context. Another says never learn a word from a guess — look it up so you do not remember the wrong meaning.
Both rules are too simple.
Usually, try to infer the meaning first when the context gives you enough evidence — then verify the word when the exact meaning matters, when your guess is uncertain, or when the word seems worth learning.
Guessing from context is a real reading skill. It keeps attention on the text, makes you use grammatical and semantic clues, and can contribute to vocabulary learning. But contextual guesses are often incomplete or wrong, especially when the surrounding text is weak, the learner's vocabulary is limited, or the word has several plausible meanings. Research comparing inferencing with dictionary consultation does not show that guessing is universally superior. In some studies the two strategies produced comparable learning; in others dictionary use produced stronger retention. Alahmadi & Foltz, 2020 Chen, 2012
The useful question is therefore not guess or look up?
It is what should you do first, and when is checking worth the interruption?
Guessing from context is more than making a random prediction
Researchers usually call this lexical inferencing: using available clues to construct a likely meaning for an unfamiliar word while reading.
Those clues can come from several places. The sentence may tell you what happened before and after the word. Grammar may reveal whether the unknown item is a noun, verb, adjective, or adverb. Prefixes and suffixes may narrow the possibilities. A contrast marker such as but or a cause marker such as because can reveal the relationship between ideas. Your background knowledge may also help.
In an influential study of learners reading English texts, Paribakht and Wesche found that when learners did attend to unknown words, inferencing was a major strategy and drew on multiple kinds of linguistic and contextual knowledge. Paribakht & Wesche, 1999
So inferencing is not simply staring at one sentence and inventing a synonym.
It is evidence-based reading under uncertainty.
One word, two stages
Consider this sentence:
The trail became treacherous after the rain, so we turned back before reaching the summit.
Even if you have never seen treacherous, the sentence gives useful clues.
The trail changed after the rain. The hikers turned back. They did not reach the summit. The unknown word is an adjective describing the trail.
A reasonable first inference is:
treacherous ≈ dangerous / unsafe
That inference is good enough to keep reading.
But now imagine the word appears again later in the article, or you want to use it yourself. Checking a dictionary adds precision: treacherous can mean dangerous because of hidden or unpredictable hazards, and in another context it can describe someone who is disloyal or deceptive.
The context got you into the right neighborhood. The lookup showed you the property boundaries.
That distinction matters because successful comprehension and complete word learning are not the same task.
Context is useful — but it is not guaranteed to reveal the exact meaning
A sentence can constrain meaning without identifying it perfectly.
Suppose you read:
Maya gave a terse reply and returned to her work.
You might infer that terse means rude, annoyed, brief, cold, or unfriendly. Several of those interpretations fit the situation. The actual meaning is closer to brief and using few words, sometimes with an abrupt tone.
Your guess could be sufficient for understanding the scene while still being lexically inaccurate.
This is one reason classic research has warned against treating contextual guessing as a substitute for deliberate vocabulary learning. Context varies enormously in how informative it is, and even apparently helpful passages can license several plausible interpretations. Laufer, 1990
More recent work reaches a less absolute but compatible conclusion: inferencing can support learning, but the quality of the inference matters. In Alahmadi and Foltz's comparison of lexical inferencing and dictionary consultation, both approaches produced significant and broadly comparable delayed vocabulary gains for their university EFL learners. Crucially, learners who made more accurate guesses also learned more from the inferencing condition. Alahmadi & Foltz, 2020
That gives us an important rule:
The learning benefit of guessing depends partly on whether the guess is actually good.
What dictionary and gloss research adds
Looking up a word is sometimes described as the lazy option: you did not struggle, so perhaps you learned less.
The evidence does not support that simple story.
A dictionary or a well-designed gloss can do several useful things at once. It can confirm that you interpreted the sentence correctly, prevent a plausible-but-wrong meaning from becoming attached to the word, make the form–meaning connection more noticeable, and let you return to the text with less uncertainty.
A 2020 meta-analysis by Yanagisawa, Webb, and Uchihara synthesized 359 effect sizes from 42 studies of glossed reading. Across the studies, readers with gloss support learned substantially more target vocabulary than readers without glosses on both immediate and delayed tests. The exact advantage varied by gloss design and learner/text characteristics, but the overall pattern strongly favors giving readers accessible meaning support rather than assuming unsupported context alone will do all the work. Yanagisawa, Webb & Uchihara, 2020
A newer meta-analysis focused specifically on English L2 incidental vocabulary learning through reading reached the same broad conclusion: glosses were associated with significantly better vocabulary outcomes, although effects varied considerably across study conditions. He, Shaari & Lay Shi, 2026
This does not mean you should open a dictionary for every unfamiliar word.
It means that checking a meaning is not a failure of reading strategy. It can be part of the learning process.
Direct comparisons do not produce one universal winner
Studies that directly compare inferencing and dictionary consultation are especially useful because they test the exact decision learners face.
The findings are mixed rather than contradictory.
Alahmadi and Foltz found significant learning under both inferencing and dictionary consultation, with no overall difference between the two strategies for their sample. Learners with larger existing vocabularies tended to benefit more under both conditions. Alahmadi & Foltz, 2020
Other experiments have found an advantage for dictionary-supported learning. Chen's study of Chinese university learners reported stronger incidental vocabulary acquisition when learners could use bilingualized dictionaries than when they read without dictionary access. Chen, 2012
Zou likewise reported stronger vocabulary learning from dictionary consultation than from inferencing in a reading task with intermediate learners. Zou, 2016
These results make more sense if we stop treating the two strategies as identical cognitive events.
Inferencing asks you to generate a meaning from imperfect evidence.
Lookup gives you external meaning information that you must connect back to the sentence.
Both can involve attention. Both can be shallow. Both can be done well or badly.
The best choice depends on what the reading moment requires.
Myth: the harder strategy always creates stronger memory
Myth: If you struggle to guess a word before checking it, you will necessarily remember it better than if you simply look it up.
Reality: Effort can be useful when it leads to productive retrieval or deeper processing, but difficulty by itself is not a learning mechanism. A confident wrong inference can leave you with the wrong meaning; an immediate lookup can also be forgotten seconds later if you never connect it to the sentence or encounter the word again.
Recent research has begun testing hybrid designs rather than forcing learners into pure guessing or pure glossing. For example, work combining retrieval prompts with later gloss feedback examines whether learners can first attempt meaning retrieval and then receive corrective support. That logic is promising because it preserves effort while preventing unresolved errors from remaining unresolved. Wang, Gao & Gui, 2025
The broader principle is simple:
Useful effort needs useful feedback.
Your existing vocabulary changes how well context works
Context does not operate independently of what you already know.
To infer one unknown word, you need enough of the surrounding language to interpret the clues. If every sentence contains several unfamiliar words, the evidence itself becomes unstable. You may be trying to solve one mystery using five other mysteries.
This is one reason lexical coverage matters so much in reading. If unfamiliar vocabulary is dense enough to keep breaking comprehension, the problem is no longer “Should I guess this one word?” It is that the text may be poorly matched to your current level. We explore that threshold problem in Do You Need to Understand Every Word When Reading English?.
The Alahmadi and Foltz study also found that larger vocabulary size was associated with greater learning in both the inferencing and dictionary conditions. Alahmadi & Foltz, 2020
That fits a practical intuition: the more surrounding language you know, the more useful evidence you have for interpreting the word you do not know.
Not every unknown word deserves a lookup
Stopping for every unknown item can destroy the reason you were reading in the first place.
If you are reading a novel and already understand that a character walked through some kind of decorative doorway, the exact architectural meaning of lintel may not matter. Looking it up could add knowledge, but it may contribute almost nothing to your understanding of the scene.
By contrast, if you are reading a news article and the entire argument turns on whether a policy was mandatory, recommended, or provisional, approximate understanding is not enough.
The decision should depend on the word's importance, not merely its unfamiliarity.
Look it up when one of these is true
- The word is central to the sentence or argument. If misunderstanding it changes the main point, verify it.
- Your inference is weak or has several plausible meanings. “Something negative” may be enough for flow, but not enough for learning.
- The word keeps appearing. Repetition is evidence that the word may be useful beyond this one sentence.
- You want to use the word yourself. Productive use requires more precision than rough reading comprehension.
- It belongs to a technical, academic, legal, medical, or other precision-sensitive context. Approximate guessing becomes riskier when distinctions carry real consequences.
Keep reading when the word is low-value and the sentence still works
Sometimes the strongest reading strategy is to do nothing.
Paribakht and Wesche observed that learners ignored many unfamiliar words during meaning-focused reading rather than attempting to infer every one. Paribakht & Wesche, 1999
That can be rational.
If the sentence remains understandable, the word is peripheral, and stopping would break the flow, continuing may be better than turning every page into a vocabulary worksheet.
Incidental vocabulary learning from meaning-focused exposure is real, but it is gradual. A 2023 meta-analysis covering reading, listening, viewing, and related input conditions found clear overall learning effects while also showing that only a proportion of encountered target words were learned and that outcomes varied with proficiency, materials, spacing, and other factors. Webb, Uchihara & Yanagisawa, 2023
Exposure creates opportunities. It does not guarantee that every unknown word becomes known after one encounter.
A better default: infer, decide, verify
Instead of choosing one permanent rule, use a three-step decision.
The three-step reading loop
- Infer. Before reaching for a dictionary, use the sentence, grammar, morphology, discourse clues, and your background knowledge. Ask what kind of meaning would make the sentence coherent.
- Decide. Is your rough interpretation enough for the purpose of this reading? Is the word important, repeated, useful, or uncertain enough to deserve attention?
- Verify when needed. Check a concise definition or gloss, then return immediately to the original sentence and reinterpret it with the confirmed meaning.
The final step matters. A lookup should not become an isolated flash of information in another tab. Reconnect the meaning to the exact context that made you notice the word.
If the word seems worth keeping, encounter it again later or test whether you can retrieve its meaning without the definition in front of you. That distinction between recognizing a word and being able to retrieve it is explored in Recognizing a Word Is Not the Same as Being Able to Recall It.
Do not confuse a correct guess with knowing the word
A successful inference may give you only the aspect of meaning required by one sentence.
You might correctly infer that lean means “not having much fat” in lean meat while missing its verb sense in lean against the wall, its adjective sense in a lean year, or common collocations that determine how native speakers actually use it.
Vocabulary knowledge is multidimensional: meaning, form, grammatical behavior, collocations, register, multiple senses, and the ability to retrieve the word all develop over time.
Context can start that process. A dictionary can clarify part of it. Neither single exposure completes it.
Guess first when you can. Look it up when the answer matters.
You do not need to choose between becoming a good contextual guesser and becoming a good dictionary user.
They solve different problems.
Inferencing protects reading flow and trains you to extract meaning from linguistic evidence. Lookup adds precision, prevents unresolved mistakes, and can strengthen the form–meaning connection. Research supports both strategies, and the strongest practical rule is conditional rather than ideological.
When the context is rich and the word is not important, make the best inference you can and keep moving.
When the word carries the argument, appears repeatedly, seems useful, or leaves you uncertain, check it.
And when you do check it, return to the sentence. The goal is not to collect definitions. It is to make the next encounter with that word easier, faster, and more accurate.
Sources and further reading
- Alahmadi & Foltz (2020) — Exploring the effect of lexical inferencing and dictionary consultation on undergraduate EFL students' vocabulary acquisition
- Yanagisawa, Webb & Uchihara (2020) — How Do Different Forms of Glossing Contribute to L2 Vocabulary Learning from Reading? A Meta-Regression Analysis
- He, Shaari & Lay Shi (2026) — The effects of glosses on English L2 incidental vocabulary learning through reading: a meta-analysis
- Webb, Uchihara & Yanagisawa (2023) — How effective is second language incidental vocabulary learning? A meta-analysis
- Paribakht & Wesche (1999) — Reading and “Incidental” L2 Vocabulary Acquisition: An Introspective Study of Lexical Inferencing
- Chen (2012) — Dictionary Use and Vocabulary Learning in the Context of Reading
- Zou (2016) — Comparing Dictionary-induced Vocabulary Learning and Inferencing in the Context of Reading
- Laufer (1990) — Guessing: No substitute for systematic learning of lexis
- Wang, Gao & Gui (2025) — Impact of retrieval prompts with posterior L1 glossing on novel words learning and processing during L2 reading