Lexical ambiguity occurs when one spoken or written word form can support more than one interpretation. The English word “bank” may identify a financial institution or the land beside a river; “paper” can denote a material, a newspaper, a scholarly article, or an examination. Readers and listeners ordinarily resolve these possibilities with remarkable speed. They combine information about the word, its more familiar meanings, its grammatical role, the surrounding sentence, the situation, and the conversation's history. Psycholinguistics studies how this selection unfolds across milliseconds, how wrong interpretations are revised, and how experience changes which meanings become available. Rodd (2020); Eddington and Tokowicz (2015).
Consider “The fisherman sat on the bank” and “The customer went to the bank.” The form bank is identical; the surrounding information establishes different intended meanings. Yet the explanation is more demanding than saying that context supplies a dictionary definition. Some alternative interpretations can receive measurable activation before the system settles on the appropriate one; the timing and strength of that activation depend on meaning frequency, contextual constraint, the form of ambiguity, and the experimental task. The principal scientific question is how lexical knowledge and contextual evidence interact during real-time comprehension. Swinney (1979)90355-4); Duffy, Morris, and Rayner (1988)90066-6).
What lexical ambiguity means in psycholinguistics
A lexical item is ambiguous when a single lexical form is associated with distinguishable interpretations. The form may be a sequence of sounds recognized during listening or letters recognized during reading. A listener must identify the form while also assembling a representation of what the utterance means. Lexical access refers to the processes that make stored word-related information available; word-meaning selection concerns how an interpretation becomes favored, sustained, and integrated into the sentence. These processes interact, and their boundaries remain important theoretical questions. Rodd (2020).
Ambiguity is ordinary linguistic organization. A word can carry a family of closely related senses, independent meanings, or an interpretation that extends from one domain to another. A dictionary may divide this continuum into entries and numbered senses, but such editorial divisions do not map directly onto the mental representations that language users employ. Researchers therefore specify whether a study concerns different lexical meanings, related senses, repeated forms with different pronunciation, or contextual variations that are not always counted as separate senses. Eddington and Tokowicz (2015).
Homonymy, polysemy, homographs, and homophones
Homonymy concerns a shared form associated with relatively unrelated meanings. The financial and riverside meanings of bank are a conventional English example. Polysemy concerns a form with semantically related senses: the mouth of a person and the mouth of a river share a structured resemblance, whereas “paper” can refer to a substance or to a document made from it. Semantic relatedness is graded, culturally dependent, and sensitive to the way examples are selected. Some words sit between obvious homonymy and obvious polysemy. Experiments that ignore this gradient risk grouping distinct processing phenomena under one label. Klepousniotou, Titone, and Romero (2008); Eddington and Tokowicz (2015).
A homograph shares a written form across interpretations. A homophone shares pronunciation, as in flower and flour, even when spelling distinguishes the lexical items. Some homographs have different pronunciations, as in lead meaning to guide versus lead the metal; their ambiguity is strongest in writing, where pronunciation may not yet be supplied. The particular channel therefore matters: print may leave information unspecified that speech has already resolved, while speech can merge forms kept apart by spelling. Homographs, homophones, homonyms, and polysemous words overlap as classifications rather than forming a single exhaustive taxonomy.
Lexical ambiguity versus syntactic ambiguity
Lexical ambiguity concerns which meaning or sense of a word is intended. Syntactic ambiguity concerns how a sentence's words combine into a grammatical structure. “I saw the man with the telescope” permits different attachments for “with the telescope,” even when the meanings of the individual words remain fixed. A sentence can contain both kinds of uncertainty, and lexical interpretations may constrain grammatical parsing. The familiar garden-path effect refers to revision of an initially favored structural analysis; a lexical misinterpretation can also require revision, but it need not involve a structural garden path. Leinenger, Myslín, Rayner, and Levy (2017).
How context selects the intended word meaning
Context is the collection of evidence relevant to interpreting an expression. It includes the immediate words, sentence grammar, earlier discourse, topic, the physical situation, a speaker's probable intention, and the listener's knowledge and expectations. It can act before an ambiguous word appears by raising the availability of a likely interpretation, at the word by modifying the balance among candidates, and afterward by prompting integration or reinterpretation. Different studies detect these effects at different moments, so a claim that context “works early” or “works late” requires a specified task and measure. Vitello and Rodd (2015); Rodd (2020).
Sentence-level semantic constraints
In “She deposited her paycheck at the bank,” the words deposited and paycheck establish strong support for the financial meaning. In “The canoe drifted toward the bank,” canoe and drifted support the riverside meaning. These cues operate through learned semantic relationships and through a developing model of the event described. Sentence context may reduce the difficulty of selecting a contextually appropriate meaning, but some theories allow residual activation of alternatives even under strong bias. The strength and timing of this reduction must be established experimentally. Swinney (1979)90355-4); Duffy, Morris, and Rayner (1988)90066-6).
Contextual support has degrees. A general topic such as finance favors the financial bank; a specific syntactic and semantic phrase such as “opened a savings account at the bank” supplies stronger converging evidence. Context is sometimes measured with completion norms, asking independent participants to continue a sentence, or with ratings of how strongly a passage favors a particular meaning. A researcher cannot simply label a sentence “strong” by intuition and assume that every participant will experience the same bias.
Meaning dominance and relative frequency
Ambiguous words are often unbalanced. One interpretation is encountered much more frequently than another in a particular language community, making it dominant; the rarer interpretation is subordinate. An approximately balanced ambiguous word has meanings of similar availability. These terms describe distributions of experience and experimental responses, not permanent properties that are identical for all speakers. Dominance can be estimated through free association, sentence completions, or other norms, and the method influences the estimate. Duffy, Morris, and Rayner (1988)90066-6); Eddington and Tokowicz (2015).
A dominant meaning may receive a substantial initial advantage, even when context supports a subordinate alternative. In classic eye-tracking research, reading an ambiguous word with a highly dominant meaning could take longer when preceding context required its subordinate meaning: the subordinate-bias effect. The phenomenon helped motivate models in which context changes the order or relative strength of accessible meanings without necessarily excluding the dominant meaning from initial processing. Neither every ambiguous word nor every sentence produces that cost; contextual constraint, experimental matching, and reader experience all matter. Duffy, Morris, and Rayner (1988)90066-6); Dopkins, Morris, and Rayner (1992)90023-Q).
Grammatical, discourse, and situational cues
Grammar can narrow interpretation because the possible syntactic category and argument structure of a word restrict how it can fit the sentence. A preceding paragraph may also define a topic strongly enough that the relevant interpretation remains accessible across several sentences. In spoken conversation, an image, gesture, shared task, or speaker's pointing can make one meaning pragmatically appropriate. These cues contribute evidence rather than operating as an infallible semantic switch: a listener may still briefly commit to another meaning, especially when an utterance is misleading or a familiar sense is much more accessible. Rodd (2020).
The same cue can be used differently as a sentence unfolds. A word may first be understood under the current topic, then interpreted under a later qualifying phrase, and finally integrated with information from a wider conversation. The process is incremental: comprehension develops while the input arrives. This continuous construction makes ordinary language efficient, while also creating the possibility of temporary misinterpretation and revision.
Do multiple meanings become active at once?
This is one of the defining questions of lexical ambiguity research. A strictly selective account proposes that sufficiently informative context permits access primarily to the intended meaning. An exhaustive or multiple-access account proposes that alternative meanings are initially activated despite contextual bias, with selection developing subsequently. Intermediate models allow contextual cues to reorder or weight access. The theoretical alternatives concern unobservable processing mechanisms, so researchers infer them from carefully timed behavioral and neural measures. Swinney (1979)90355-4); Sereno (1995).
What classic cross-modal priming established
In Swinney's influential 1979 experiments, participants listened to sentences containing an ambiguous spoken word while also making rapid decisions about visually presented letter strings. Immediately at the offset of the ambiguous word, recognition was facilitated for test words associated with either of its meanings, including when the preceding sentence favored one interpretation. At a later test point, facilitation remained for the contextually appropriate meaning. These findings supplied influential evidence for early availability of multiple meanings followed by context-sensitive selection. The result is tied to the method's temporal probes; it does not prove that every conceivable sense of every word is fully activated in every context. Swinney (1979)90355-4).
The cross-modal paradigm provides a narrow temporal window by presenting a probe at a controlled delay relative to speech. Its strengths are sensitivity to meaning-related activation and the ability to compare immediate with delayed access. Its limitations include the secondary decision task, properties of the probe, variability in lexical associations, and the possibility that laboratory attention differs from everyday listening. Later research varied stimulus balance, context and probe methods, producing a more differentiated account than any simple universal activation rule. Eddington and Tokowicz (2015).
What eye movements reveal during normal reading
Eye tracking records where readers look and how long they fixate words while understanding connected text. Because fixations occur during the reading process itself, they can reveal local difficulty without asking for an overt meaning judgment on every trial. Duffy, Morris, and Rayner showed that eye-movement patterns varied with meaning balance and with whether the disambiguating clause appeared before or after the ambiguous word. Dopkins, Morris, and Rayner subsequently compared competing interpretations of such findings. These studies established that context and meaning frequency influence the time course of reading, while leaving room for debate about the precise internal order of activation and selection. Duffy, Morris, and Rayner (1988)90066-6); Dopkins, Morris, and Rayner (1992)90023-Q).
First-fixation duration, gaze duration, regressions, and later rereading are distinct measures. An increase at the ambiguous word can reflect difficulty in initial recognition or meaning selection; increased rereading downstream may reflect integration failure and reinterpretation. The temporal relationship between an ambiguous word and the clause that clarifies it is essential for interpreting these signals. A longer fixation cannot be converted directly into a statement about a single neural stage.
Why early multiple activation and context sensitivity coexist
The evidence supports a processing system in which alternative interpretations can be available transiently while contextual information continuously shapes their relative activation and ultimate selection. The strength of that conclusion differs across stimulus types, meaning frequencies, methods, and stages of comprehension. Early activation does not imply prolonged indecision; a weakly activated alternative can become irrelevant within a brief interval. Strong contextual influence does not imply that the alternative never becomes available. Rodd (2020); Vitello and Rodd (2015).
Major models of lexical ambiguity resolution
The competing models emerged partly because experimental tasks sample different portions of the process. A model of access explains what meanings become available when the word is encountered; a model of selection explains how one interpretation gains priority; a model of integration explains how that interpretation fits a sentence or discourse; a learning model explains how these tendencies change with experience. A comprehensive account must specify their interactions and generate testable predictions. Rodd (2020).
Exhaustive and autonomous access accounts
In a classical exhaustive-access account, encountering an ambiguous word activates more than one stored meaning, with contextual constraints operating subsequently. The attraction of this proposal is its straightforward explanation of early priming for multiple meanings in some experiments. Its strong version is challenged whenever prior context, distributional frequency, and meaning similarity appear to modify the earliest detectable processing. “Automatic access” describes limited dependence on deliberate choice within particular tasks; it does not establish that activation is immune to all surrounding information. Swinney (1979)90355-4); Eddington and Tokowicz (2015).
Selective and context-sensitive access accounts
A context-sensitive model gives prior evidence a role in initial activation. If a sentence very strongly favors the riverside meaning of bank, that interpretation can gain priority at or near first contact with the form. Strong forms of the account predict little measurable initial influence from meanings inconsistent with context; weaker forms predict a continuous change in relative activation. The experimental challenge is to distinguish true absence of an alternative from activation below the detection threshold of a particular method. Contextual strength must also be measured independently of the outcome being explained. Sereno (1995); Duffy, Morris, and Rayner (1988)90066-6).
Reordered access and the subordinate-bias effect
The reordered-access model assumes that context and frequency jointly affect which interpretation becomes available or influential first. When a dominant meaning is highly practiced, it can compete with a context-supported subordinate interpretation even if context moves the latter forward in the processing sequence. This helps explain patterns in eye movements, especially subordinate-bias conditions. Follow-up studies have debated whether exceptionally strong contexts can remove the effect entirely. A 2017 set of eye-tracking experiments found that inserting more neutral material between an ambiguous word and later disambiguation did not produce the predicted increase in difficulty; the authors interpreted their findings as compatible with reordered-access and probabilistic approaches rather than the particular resource-constrained model they tested. Duffy, Morris, and Rayner (1988)90066-6); Leinenger, Myslín, Rayner, and Levy (2017).
Distributed, attractor, and settling accounts
Distributed models treat meanings as patterns across many interacting semantic dimensions rather than entirely separate dictionary entries. A word form starts a dynamic activation process; contextual evidence and the network's learned structure guide the system toward a more stable interpretation. Multiple related senses can share representational features, while more distant meanings may compete for distinct states. Rodd, Gaskell, and Marslen-Wilson used computational modeling to show how distributed semantic representations can reproduce contrasting ambiguity effects in word recognition. The models demonstrate mechanistic possibilities and make quantitative predictions; simulation fit alone does not identify a unique neural implementation. Rodd, Gaskell, and Marslen-Wilson (2004).
Rodd's later ambiguity-centered account emphasizes a high-dimensional semantic space, multiple linguistic and paralinguistic cues, and continuous learning that shapes the stability and availability of meanings. This perspective places differences between readers and listeners inside the explanation: two individuals can know the same word yet possess different experience-dependent meaning biases. It also makes disambiguation a central property of everyday word recognition rather than an exceptional repair mechanism. The proposal synthesizes multiple research traditions; its precise architecture remains a theoretical interpretation open to further tests. Rodd (2020).
Probabilistic interpretation and prediction
Probabilistic approaches describe how multiple sources of evidence change the relative likelihood of candidate meanings. A highly frequent interpretation can begin with a stronger prior, while local words and the situation supply evidence favoring another interpretation. Such a description captures the graded influence of frequency and context and clarifies why a less common meaning can ultimately win. It does not by itself settle whether the brain explicitly predicts a particular word or meaning before it arrives, or whether facilitation reflects rapid integration after input. Distinguishing prediction, access, competition, and integration requires designs that test their separate temporal signatures. Leinenger, Myslín, Rayner, and Levy (2017); Kutas and Federmeier (2011).
Why related senses behave differently from unrelated meanings
A word with several related senses can support an interconnected region of semantic space, whereas a homonym may connect a form to relatively distant semantic representations. Behavioral studies have identified situations in which polysemous words are processed faster in lexical decision, while unrelated ambiguity imposes competition costs in other tasks. The effects depend on exactly how semantic similarity, meaning number, and sense frequency are defined and measured. The useful generalization is that the structure of meaning matters, not merely the number of dictionary definitions. Rodd, Gaskell, and Marslen-Wilson (2002); Eddington and Tokowicz (2015).
Klepousniotou, Titone, and Romero examined how comprehension varies with overlap among senses, demonstrating the importance of internal semantic similarity rather than treating polysemy as a single homogeneous category. In a distributed account, overlapping senses can support one another on features they share; widely separated meanings create a different landscape of activation. Yet whether such facilitation appears in lexical decisions, reading passages, speech, or explicit interpretation judgments depends on the demands of the task. The evidence calls for graded description and careful stimulus norming. Klepousniotou, Titone, and Romero (2008).
When ambiguity can help recognition
The phrase “ambiguity advantage” names findings in which certain ambiguous words are recognized more quickly than suitably matched unambiguous words in tasks such as visual lexical decision. It does not imply that ambiguous sentences are always easier to understand. The advantage can arise from rich semantic connections or distributed activation of related senses, while sentence-level disambiguation can still impose a cost when interpretations compete. Rodd, Gaskell, and Marslen-Wilson's 2002 lexical-decision experiments challenged explanations that attributed the advantage simply to the presence of multiple unrelated meanings. Rodd, Gaskell, and Marslen-Wilson (2002).
Why dictionary counts are insufficient
Lexicographic entries reflect descriptive choices about how to subdivide senses. A participant's semantic similarity judgments and the usage patterns represented in their experience may organize those senses differently. Studies must distinguish homonymy, polysemy, sense relatedness, word frequency, familiarity, concreteness, length, spelling neighborhood, and contextual diversity. Without such controls, an apparent ambiguity effect may partly reflect another property of the materials. Eddington and Tokowicz (2015).
What happens when the first meaning is wrong?
Incremental comprehension sometimes produces a plausible but incorrect interpretation. If a listener hears “The visitor waited near the bank” without enough prior information, a familiar meaning may gain priority. The continuation “until the floodwater receded” shifts the intended interpretation toward the riverside. The system must now incorporate the new evidence, reduce commitment to the mistaken interpretation, and revise its understanding of the developing event. The timing of clarification changes which operations are needed: early bias can steer access, while delayed disambiguation may require reinterpretation of already constructed meaning. Vitello and Rodd (2015); MacGregor et al. (2020).
Selection, suppression, and reinterpretation
Meaning selection refers to the growing preference for an interpretation consistent with evidence. Suppression is a family of theoretical explanations for declining influence of irrelevant meanings; an experiment showing reduced priming for one sense does not uniquely prove a dedicated inhibitory mechanism. Reinterpretation involves changing the semantic representation already constructed for part of a sentence. Selection and revision can overlap, and researchers distinguish them by positioning disambiguating words at different points in the sentence. Neuroimaging implicates frontal and temporal language-related networks in these demands, but the assignment of specific regions to one psychological operation remains debated. Vitello, Warren, Devlin, and Rodd (2014); Vitello and Rodd (2015).
The cost of delayed clarification
Later context can expose the cost of maintaining an incorrect interpretation or switching away from it. A reader may slow on the word that makes the previous meaning impossible, reread earlier material, and then proceed with a revised model. This outcome is related to, but separable from, the early cost observed on the ambiguous word itself. The 2017 study by Leinenger and colleagues found no evidence for the specific prediction that a longer neutral stretch between ambiguity and clarification would necessarily increase the difficulty of later disambiguation. This constrains models that posit a general “digging-in” process under those experimental conditions; it does not eliminate all costs of semantic revision. Leinenger, Myslín, Rayner, and Levy (2017).
How researchers measure lexical ambiguity
Lexical ambiguity is a test case for cognitive science because a word's visible or audible form can be held constant while the intended interpretation changes. Experimenters manipulate meaning dominance, prior context, position of clarification, relatedness among senses, modality, or exposure history, and then measure responses. A convincing experiment also matches nonambiguous comparison items and documents the properties of target words so that a frequency or familiarity difference is not mistaken for ambiguity. Eddington and Tokowicz (2015).
Lexical decision and semantic priming
A lexical decision requires the participant to decide whether a string is a word. Ambiguous words can show different reaction times from unambiguous words, but the task requires word identification rather than full interpretation of a sentence. Semantic priming uses a related probe to estimate the availability of a meaning: recognizing “money” more readily after bank than after an unrelated control can indicate activation of the financial interpretation. Cross-modal designs combine spoken sentences with visual probes at precisely controlled delays. An association effect reflects activation under those task conditions; it is not a direct readout of a fully conscious interpretation. Swinney (1979)90355-4); Rodd, Gaskell, and Marslen-Wilson (2002).
Eye tracking, self-paced reading, and comprehension questions
Eye movements permit continuous measurement of fixation durations and rereading. Self-paced reading measures how long participants spend advancing through segments and can locate delayed processing costs, although it alters natural reading behavior. Comprehension questions assess whether readers ultimately selected the intended interpretation. Because successful final answers can coexist with transient wrong interpretations, combining early and late measures produces a more informative picture than relying on accuracy alone. Duffy, Morris, and Rayner (1988)90066-6); Leinenger, Myslín, Rayner, and Levy (2017).
EEG, ERPs, MEG, and the N400
Electroencephalography and magnetoencephalography measure rapidly changing signals associated with brain activity during language processing. The N400 is a widely studied event-related potential component sensitive to semantic processing and contextual fit across many kinds of meaningful stimuli. It is useful when asking how expected and unexpected meanings are processed, but an N400 effect cannot uniquely identify lexical prediction, selection, suppression, or conscious understanding. The component is shaped by multiple processes and experimental conditions. Kutas and Federmeier (2011).
MacGregor and colleagues combined MEG and EEG during spoken-sentence comprehension to distinguish responses to an ambiguous word from later responses when a subordinate meaning became clear. The authors reported an enhanced response to ambiguous words roughly 400–800 milliseconds after their acoustic offset in the tested conditions, along with later responses linked to reinterpretation. Their work illustrates that ambiguity resolution may involve temporally distinct operations; timing from one paradigm is not a universal latency for every sentence or speaker. MacGregor et al. (2020).
Functional MRI and causal inference
Functional MRI identifies distributed regions where blood-oxygen-level-dependent signals covary with conditions demanding greater semantic selection or reinterpretation. Reviews associate ambiguity-related sentence processing with left inferior frontal and temporal regions, among others. Such findings support a network account, but the spatial association is not proof that one region uniquely “stores” a meaning or performs a single mental operation. Distinguishing retrieval, competition resolution, reanalysis, and broader task demands requires converging designs and carefully specified causal claims. Vitello and Rodd (2015); Rodd, Vitello, Woollams, and Adank (2015).
Brain mechanisms: semantic representations and control
A scientific account of lexical ambiguity involves interactions between representations of word meaning and processes that guide selection within a sentence. Temporal regions are widely implicated in aspects of semantic processing, while left inferior frontal regions are repeatedly recruited when context places greater demands on controlled retrieval, selection, or reinterpretation. This functional characterization describes group-level experimental patterns. It does not imply that the same word meaning is confined to one small location or that a frontal signal can be equated with a single abstract mechanism. Vitello and Rodd (2015).
In an fMRI study that manipulated the position of disambiguating information in spoken sentences, researchers observed responses in the left inferior frontal gyrus both to initially ambiguous words and to subsequent disambiguating information. Other temporal responses differed in their sensitivity to later clarification. The results support a division of computational demands across interacting systems, while several accounts of the precise frontal contribution remain viable. Later combined MEG/EEG work adds information about the temporal sequence of these processes. Rodd et al. (2012); MacGregor et al. (2020).
What neuroscience does and does not establish
Neural measurements provide convergent evidence that ambiguity can increase semantic and control demands and that delayed disambiguation can elicit additional processing. They cannot settle all disputes between autonomous, reordered, interactive, or distributed-access theories on their own. The same network can participate in several cognitive operations, and an observed difference between conditions may reflect difficulty rather than a dedicated ambiguity module. The strongest accounts connect behavioral timing, stimulus properties, brain activity, and independently specified computational predictions. Vitello and Rodd (2015).
Lexical ambiguity changes with experience
A person's meaning preferences reflect exposure to particular domains, communities, and recent conversations. A scientist who frequently encounters “cell” in biology, a musician who encounters “key” in harmony, and a computer programmer who encounters “port” in networking may retrieve specialized interpretations more readily than someone whose experience differs. These are illustrations of how learned frequencies can matter; a specific individual's bias must be measured rather than assumed from occupation. Rodd's model treats learning as a continual influence on the organization and accessibility of word meanings. Rodd (2020).
Experiments on word-meaning priming demonstrate that hearing an ambiguous word in a sentence favoring one meaning can bias later interpretation of that word. Rodd and colleagues reported persistence after approximately 20 minutes and evidence that the effect was not simply tied to the identity of the original speaker. These results support learning or retuning of meaning availability after contextual exposure. They do not establish that all meanings are equally malleable or that a single sentence permanently rewrites the lexicon. Rodd, Lopez Cutrin, Kirsch, Millar, and Davis (2013).
Individual differences and vocabulary knowledge
Readers differ in language history, vocabulary, exposure to particular senses, reading habits, and familiarity with specialized terms. Such differences can change both the speed of recognizing a word and the probability of selecting a particular interpretation. An experiment based on average meaning dominance may obscure systematic variation between individuals. A modern account therefore treats meaning frequency as an estimate derived from a population and recognizes that each mental lexicon has a partially individual structure. Rodd (2020).
Spoken language and written language
Speech unfolds over time and can carry prosody and acoustic cues; print provides a stable sequence of letters that readers can revisit. Pronunciation may resolve some homographs in speech, whereas overlapping spoken forms can create ambiguities not present in spelling. These modality differences affect experimental interpretation. The general need to combine lexical information with context spans modalities, but evidence from one channel must not be converted into a claim that timing and cue use are identical in the other. Eddington and Tokowicz (2015); MacGregor et al. (2020).
Bilingual experience and cross-language ambiguity
A person who knows two or more languages may encounter word forms and meanings that overlap across them. A familiar written form may correspond to different words in different languages, and a translation equivalent may represent a different range of senses. This can increase the number of plausible candidate interpretations in some circumstances. The outcome depends on language proficiency, task language, exposure, spelling and phonological similarity, and experimental setting. Cross-language homographs and translation ambiguity should be distinguished from ordinary within-language polysemy; neither phenomenon establishes a general cognitive advantage or disadvantage of bilingualism. This topic connects to bilingual language processing but retains its own narrower empirical questions.
Why ambiguity matters for conversation, reading, and learning
In conversation, people ordinarily convey their intended meanings because lexical cues converge with the ongoing topic, the environment, and mutual knowledge. Misunderstandings occur when speakers and listeners assign different interpretations to the same expression or when essential disambiguating cues arrive late. A useful explanation targets the word and the missing context: “By bank, I mean the side of the river.” This practical observation follows from the basic structure of ambiguity resolution, rather than from a clinical intervention claim.
In reading, an ambiguous word can be understood without noticeable delay, especially when the context supports a dominant or accessible interpretation. Balanced meanings, rare intended senses, and late clarifications can impose measurable costs in some designs. Educational and technical writing can reduce avoidable ambiguity by naming referents, introducing field-specific definitions, and positioning key context near a potentially unfamiliar term. These are communication-design implications, not evidence that every ambiguous expression should be eliminated: metaphor, humor, literature, and ordinary flexible language also depend on multiple meanings. Duffy, Morris, and Rayner (1988)90066-6); Rodd (2020).
Meaning selection in humor and creative language
Puns intentionally preserve access to more than one interpretation or invite a sudden reinterpretation. Their effect depends on shared lexical knowledge and a context that supports a second reading. A sentence can remain easy to recognize at the word level while requiring a change in the larger situation model. Such language provides a natural illustration of coexisting possibilities, although laboratory results about priming and fixation times cannot by themselves explain the aesthetic or social experience of humor.
Implications for dictionaries and language teaching
For learners, teaching a word as a single translation can underrepresent its usage. Examples across multiple genuine contexts may help develop flexible connections between form, sense, and situation. The available experimental evidence on ambiguity processing supports the importance of contextual learning and experience-dependent meaning preferences. Specific instructional techniques require their own educational outcome studies; it would be unwarranted to infer from a short-term priming effect that one teaching method must outperform all alternatives. Rodd, Lopez Cutrin, Kirsch, Millar, and Davis (2013); Rodd (2020).
Boundaries with related psychological and clinical topics
The science of lexical ambiguity asks how a linguistic form maps to meanings and how context guides real-time interpretation. General semantic knowledge and its long-term organization belong to a broader study of memory. Lexical retrieval failures, including tip-of-the-tongue experiences, concern access to known word forms and should be distinguished from choosing among several meanings of a word that has already been recognized. Clinical language disorders can affect semantic processing and discourse comprehension, but the presence of ambiguity in everyday language is a normal linguistic property and does not constitute a symptom or diagnosis. Rodd (2020).
Dyslexia primarily concerns difficulties in accurate or fluent word recognition, decoding, and spelling; it must be evaluated using its own clinical and educational evidence base. Aphasia, developmental language disorder, and other recognized communication conditions involve different diagnostic frameworks. Research findings in healthy adult readers cannot be used as a screening test for these conditions. The same separation applies to interpersonal communication skills: a conversation may fail because the parties disagree about goals or social meanings, even when neither word was lexically ambiguous. This article addresses the scientific mechanism of word-meaning selection.
Lexical ambiguity and computational language models
Computational systems can represent contextual relationships among words and select different outputs for the same form under different prompts. Distributional models and large language models can therefore serve as tools for formalizing hypotheses about context sensitivity, sense similarity, and learning from linguistic exposure. Comparing model behavior with human priming, eye movements, or neural responses may expose useful similarities and divergences. The comparison requires matched tasks, clearly defined metrics, and attention to differences between training distributions and the participant's experience.
Successful disambiguation in generated text establishes that a system can produce contextually appropriate linguistic behavior in that setting. It does not establish that the system uses human lexical representations, undergoes human neural processing, or possesses subjective experience. Computational fit, functional competence, cognitive mechanism, and philosophical accounts of understanding are separate levels of inquiry. Lexical ambiguity is a particularly informative test case because it makes the relation among form, possible meanings, and contextual evidence explicit. Rodd, Gaskell, and Marslen-Wilson (2004); Rodd (2020).
What the evidence establishes and what remains debated
The established empirical foundation is that many word forms support more than one interpretation; the accessibility of those interpretations is influenced by semantic relatedness, usage frequency, preceding context, and the point at which disambiguating information appears. Eye-tracking, priming, lexical-decision, and neural studies provide converging demonstrations of these influences. Early activation of alternative meanings has been observed in classic paradigms, and readers can show additional processing demands when a subordinate meaning must be selected. Swinney (1979)90355-4); Duffy, Morris, and Rayner (1988)90066-6); Eddington and Tokowicz (2015).
The contested territory includes whether particular contexts preclude initial access to alternative meanings, how access and integration are separated in time, which semantic features represent polysemous senses, and how frontal and temporal activity should be assigned to specific computations. Some findings are task-specific, and different studies define “early,” “selective,” and “multiple access” differently. Distributed and probabilistic theories provide coherent interpretations of many results, while their architectures remain subject to experimental testing. The defensible conclusion is a dynamically context-sensitive system with measurable competition in some circumstances, not one universal mechanism proven for all words and situations. Vitello and Rodd (2015); Rodd (2020).
Methodological priorities for future experiments
Researchers can strengthen inference by norming the frequency of each meaning, separately measuring contextual constraint, matching lexical frequency and orthographic variables, comparing homonymy with graded polysemy, and using convergent early and late outcome measures. Preregistered distinctions between competing models are especially valuable when theoretical predictions overlap. Individual-level language histories, spoken versus written modalities, and ecological differences between isolated decisions and discourse comprehension also deserve explicit attention. These methodological proposals follow from the known dependencies documented in the review literature. Eddington and Tokowicz (2015); Rodd (2020).
Frequently asked questions
What is an example of lexical ambiguity?
The word bank can mean a financial institution or the side of a river. In “She withdrew cash from the bank,” the financial meaning is supported. In “The heron stood on the bank,” the riverside meaning fits. The word form is shared, while context changes the favored interpretation. This is a convenient example of relatively unrelated meanings, although actual interpretation speed depends on readers' experience and the full sentence.
How does context resolve multiple word meanings?
Context provides information that shifts the relative accessibility of meanings and determines which interpretation fits the current sentence and situation. The process can begin before the ambiguous word arrives and continue after it, particularly when later words require revision. Frequency and semantic relatedness modulate the effect. Rodd (2020).
Are all meanings activated whenever we read an ambiguous word?
Several influential experiments detected early activation of more than one meaning, but the result does not warrant a universal rule for all meanings and all contexts. Stimulus balance, timing, contextual strength, sense relatedness, and task determine what an experiment can reveal. Some models predict graded and short-lived activation that can be difficult to detect with a given measure. Swinney (1979)90355-4); Eddington and Tokowicz (2015).
What is the difference between homonymy and polysemy?
Homonymy concerns relatively unrelated meanings sharing a form, while polysemy concerns several related senses of a form. The distinction is useful in experiments because relatedness can change processing, but semantic similarity is graded and some cases resist sharp classification. Klepousniotou, Titone, and Romero (2008).
Why does an uncommon meaning sometimes take longer to understand?
A highly familiar meaning may have an initial processing advantage. If a sentence requires a subordinate interpretation, the reader or listener may need to overcome competition or revise a prior interpretation. The cost depends on how much context is available and when it arrives. Duffy, Morris, and Rayner (1988)90066-6).
Does lexical ambiguity mean a reader has a language disorder?
No. Multiple meanings are a routine property of natural language. A temporary interpretation error in an ambiguous sentence is a normal cognitive event. Clinical concerns require independent assessment of the person's broader language functioning and appropriate diagnostic criteria; experimental ambiguity effects are not a diagnostic test.
Can recent experience change which meaning comes to mind?
Yes, under studied conditions. Experiments show that exposure to an ambiguous word in a meaning-biasing sentence can affect later interpretation, including after a delay. The duration, generalization, and strength of the change depend on the method and exposure history. Rodd, Lopez Cutrin, Kirsch, Millar, and Davis (2013).
Related Articles
Cluster Language, Thought and Psycholinguistics — the published scientific chapter connecting the mechanisms of comprehension, word recognition, semantics, and language use.
Semantic Memory: Facts, Meaning, and Knowledge — the wider organization of conceptual and semantic knowledge in long-term memory.
Tip-of-the-Tongue Phenomenon: Why Words Feel Just Out of Reach — a related but distinct problem of retrieving an available word form.
Perceptual Set: How Expectations Influence What We Perceive — a cross-domain account of contextual expectations in perception.
Word Recognition: How the Mind Identifies Spoken and Written Words — the recognition processes that connect language input with lexical knowledge.
References
Dopkins, S., Morris, R. K., and Rayner, K. (1992). Lexical ambiguity and eye fixations in reading: A test of competing models of lexical ambiguity resolution. Journal of Memory and Language, 31(4), 461–476. https://doi.org/10.1016/0749-596X(92)90023-Q90023-Q).
Duffy, S. A., Morris, R. K., and Rayner, K. (1988). Lexical ambiguity and fixation times in reading. Journal of Memory and Language, 27(4), 429–446. https://doi.org/10.1016/0749-596X(88)90066-690066-6).
Eddington, C. M., and Tokowicz, N. (2015). How meaning similarity influences ambiguous word processing: The current state of the literature. Psychonomic Bulletin and Review, 22, 13–37. https://doi.org/10.3758/s13423-014-0665-7.
Klepousniotou, E., Titone, D., and Romero, C. (2008). Making sense of word senses: The comprehension of polysemy depends on sense overlap. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(6), 1534–1543. https://doi.org/10.1037/a0013012.
Kutas, M., and Federmeier, K. D. (2011). Thirty years and counting: Finding meaning in the N400 component of the event-related brain potential (ERP). Annual Review of Psychology, 62, 621–647. https://doi.org/10.1146/annurev.psych.093008.131123.
Leinenger, M., Myslín, M., Rayner, K., and Levy, R. (2017). Do resource constraints affect lexical processing? Evidence from eye movements. Journal of Memory and Language, 93, 82–103. https://doi.org/10.1016/j.jml.2016.09.002.
MacGregor, L. J., Rodd, J. M., Gilbert, R. A., Hauk, O., Sohoglu, E., and Davis, M. H. (2020). The neural time course of semantic ambiguity resolution in speech comprehension. Journal of Cognitive Neuroscience, 32(3), 403–425. https://doi.org/10.1162/jocn_a_01493.
Rodd, J. M. (2020). Settling into semantic space: An ambiguity-focused account of word-meaning access. Perspectives on Psychological Science, 15(2), 411–427. https://doi.org/10.1177/1745691619885860.
Rodd, J. M., Gaskell, M. G., and Marslen-Wilson, W. D. (2002). Making sense of semantic ambiguity: Semantic competition in lexical access. Journal of Memory and Language, 46(2), 245–266. https://doi.org/10.1006/jmla.2001.2810.
Rodd, J. M., Gaskell, M. G., and Marslen-Wilson, W. D. (2004). Modelling the effects of semantic ambiguity in word recognition. Cognitive Science, 28(1), 89–104. https://doi.org/10.1016/j.cogsci.2003.08.002.
Rodd, J. M., Johnsrude, I. S., and Davis, M. H. (2012). Dissociating frontotemporal contributions to semantic ambiguity resolution in spoken sentences. Cerebral Cortex, 22(8), 1761–1773. https://doi.org/10.1093/cercor/bhr252.
Rodd, J. M., Lopez Cutrin, B., Kirsch, H., Millar, A., and Davis, M. H. (2013). Long-term priming of the meanings of ambiguous words. Journal of Memory and Language, 68(2), 180–198. https://doi.org/10.1016/j.jml.2012.08.002.
Rodd, J. M., Vitello, S., Woollams, A. M., and Adank, P. (2015). Localising semantic and syntactic processing in spoken and written language comprehension: An activation likelihood estimation meta-analysis. Brain and Language. https://doi.org/10.1016/j.bandl.2014.11.012.
Sereno, S. C. (1995). Resolution of lexical ambiguity: Evidence from an eye movement priming paradigm. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(3), 582–595. https://pubmed.ncbi.nlm.nih.gov/7602263/.
Swinney, D. A. (1979). Lexical access during sentence comprehension: (Re)consideration of context effects. Journal of Verbal Learning and Verbal Behavior, 18(6), 645–659. https://doi.org/10.1016/S0022-5371(79)90355-490355-4).
Vitello, S., and Rodd, J. M. (2015). Resolving semantic ambiguities in sentences: Cognitive processes and brain mechanisms. Language and Linguistics Compass, 9(10), 391–405. https://doi.org/10.1111/lnc3.12160.
Vitello, S., Warren, J. E., Devlin, J. T., and Rodd, J. M. (2014). Roles of frontal and temporal regions in reinterpreting semantically ambiguous sentences. Frontiers in Human Neuroscience, 8, 530. https://doi.org/10.3389/fnhum.2014.00530.