Conversation turn-taking is the coordinated process through which people recognize opportunities to speak, project when another person may finish, prepare contributions, and manage transitions among speakers. It combines linguistic comprehension, production planning, timing, attention, and social action. In ordinary dialogue, listeners often begin preparing an answer while a question is still unfolding, because a fluent response may require substantially more planning time than the brief silence available at a turn boundary. The central scientific question is how people achieve this coordination despite limited processing capacity and uncertain knowledge of another person's intentions. Levinson and Torreira (2015); Bögels (2020).
The most durable finding is that conversation exhibits orderly yet flexible transitions. Speakers orient to recognizable possible completion points, and participants use meaning, grammar, prosody, sequential context, gaze, and embodied behavior to estimate when to respond. Short gaps are frequent, but gaps and overlaps are neither inherently pathological nor universally meaningful in the same way. Their interpretation depends on the action being performed, the relationship among participants, the language community, and the local sequence. Sacks, Schegloff, and Jefferson (1974); Stivers et al. (2009).
What Is Turn-Taking in Conversation?
A conversational turn is an interactionally recognizable contribution during which a participant holds or claims a speaking opportunity. Turn-taking refers to the locally organized distribution of such opportunities. A turn can consist of a word, a phrase, a sentence, a story, a question, or a nonverbal contribution within a broader interactional sequence. Its end is defined through participants' orientation to possible completion, rather than simply through a punctuation mark or the point at which sound ceases. Sacks, Schegloff, and Jefferson (1974).
Turn, utterance, floor, and speaking opportunity
An utterance is a stretch of produced language, while a conversational turn is a socially organized contribution. The floor describes the publicly recognizable speaking activity and the allocation of participation rights. A person can produce several grammatical sentences in a single turn, and a listener can produce a brief acknowledgment without competing for an extended turn. A speaker's pause may occur within a continuing turn, and simultaneous speech can involve cooperation rather than competition. These distinctions matter when researchers annotate recordings: a voice-activity algorithm can locate sound but cannot determine conversational action from sound alone.
Turn-constructional units and transition-relevance places
Conversation analysis identifies turn-constructional units, often abbreviated TCUs, as recognizable units from which turns can be built. Such units may be lexical, phrasal, clausal, or sentential. A transition-relevance place, or TRP, is a point where a turn could legitimately be complete and a speaker change could occur. Participants project possible completion using multiple cues, then update expectations if the speaker continues. This framework explains how someone can anticipate an end before hearing it and also why many projected completions do not yield an actual handoff. Sacks, Schegloff, and Jefferson (1974).
How the next speaker is selected
The classical model distinguishes current-speaker selection of a next participant from self-selection. An addressed question, a name, or directed gaze may make one person the relevant next speaker. If no one has been specifically selected, a participant may begin speaking at an appropriate opportunity. If nobody claims the next turn, the current speaker may continue. These are recurrent organizational possibilities rather than mechanical rules that dictate every conversational event. Multi-party interactions add changing participation frameworks and multiple plausible respondents. Sacks, Schegloff, and Jefferson (1974); Holler and Kendrick (2015).
Why Conversation Timing Is a Cognitive Puzzle
Producing speech usually involves selecting a communicative intention, retrieving words, organizing syntax and phonology, and preparing articulation. The time required for these operations can exceed the short intervals typically observed between conversational turns. Consequently, a strictly sequential account in which a person comprehends an entire utterance, then begins response planning, then speaks would frequently predict longer gaps than listeners actually produce. The timing puzzle motivates accounts that allow at least partial overlap between comprehension and production. Levinson and Torreira (2015).
Gaps, overlaps, pauses, and latency
A gap is an interval between the offset of one turn and the onset of the next. An overlap is a period in which two or more speakers are audible simultaneously. A pause can occur inside one speaker's ongoing turn. Response latency is often measured from a designated turn-ending event to the next response onset, but different annotation conventions change the observed distribution. Corpus studies caution against treating a single mean as a universal biological constant or against assuming that all silence longer than an arbitrary threshold communicates hesitation. Heldner and Edlund (2010).
How fast are speaker transitions?
A prominent cross-linguistic study of responses to yes-or-no questions found a broadly similar preference for relatively short transitions across sampled languages, alongside systematic cultural variation in average offsets. Often-cited descriptions of conversational gaps on the order of a few hundred milliseconds summarize selected distributions and interaction types; they do not describe every conversation or every community. Timing differs by whether the next action is a simple acknowledgment, a complex explanation, an agreement, a refusal, or a response requiring memory search. Stivers et al. (2009); Roberts, Torreira, and Levinson (2015).
Why speed alone does not explain coordination
Participants evaluate whether a response is relevant, informative, and fitted to the ongoing activity. A carefully delayed answer can be interactionally appropriate, and an early response can be disruptive if it starts before the other speaker has made the needed point. Shorter reaction time therefore is not a universal measure of conversational competence. The scientific target is coordinated timing relative to meaning and sequential opportunity, not simply the elimination of silence. Roberts, Torreira, and Levinson (2015).
A Mechanistic Sequence: From Hearing a Turn to Answering
A listener continuously extracts acoustic and linguistic information, constructs a provisional interpretation, identifies what action is emerging, estimates possible turn completion, selects a suitable response, and initiates speech. These operations can overlap and revisit earlier decisions. Different theories disagree about the exact degree of overlap, the point at which response selection begins, and whether anticipatory processing is primarily linguistic or social. Levinson and Torreira (2015); Bögels, Magyari, and Levinson (2015).
Incremental comprehension
Conversation is processed incrementally. Incoming words constrain possible grammatical structures, referents, and communicative intentions before the final word is heard. When a question begins with an identifiable request, its likely answer may become clear well before the speaker completes the sentence. Such early interpretation gives the responder time to prepare lexical material or a response plan. Incrementality also creates opportunities for premature interpretation; a late negation, qualification, or syntactic continuation can require the listener to revise a planned answer.
Recognition of social action
A listener must determine whether the current turn is asking, inviting, offering, complaining, informing, or pursuing some other social action. Recognizing an action is related to comprehending sentence meaning, but the two tasks are distinguishable. The same words may constitute an offer in one sequence and a joke in another. A turn beginning with “Would you like…” can project a question and possible acceptance or refusal, while the social meaning of the eventual reply depends on the shared situation. This is where language processing meets the sequential organization of action. Sacks, Schegloff, and Jefferson (1974); Roberts, Torreira, and Levinson (2015).
Projection of possible completion
After identifying likely action and grammatical structure, the listener estimates a window within which the turn may finish. This estimate depends on upcoming words, syntactic closure, intonation, duration patterns, and sometimes visual behavior. Projection is probabilistic: a speaker may append an explanation, repair a phrase, or extend a turn after an apparent completion. An accurate system must therefore combine readiness with restraint and adapt rapidly to changes. de Ruiter, Mitterer, and Enfield (2006); Bögels and Torreira (2015).
Response planning before turn completion
Planning can begin once the responder has enough information to identify an appropriate answer, even when comprehension of the remaining words continues. Experimental electrophysiology and more naturalistic interview studies find neural activity consistent with response preparation during the incoming question. These findings support overlapping processing, while the precise interpretation of a neural signature requires caution: it is evidence about task-related activity rather than direct observation of a private thought. Bögels, Magyari, and Levinson (2015); Bögels (2020).
Articulatory launch and last-moment adjustment
Once a response has been formulated sufficiently, the speaker coordinates respiratory preparation and articulatory onset with the estimated transition window. If the question continues unexpectedly, the planned onset may be delayed or reformulated. If another participant begins speaking, the responder may abandon the launch. Timing therefore emerges from a sequence of partially independent perceptual, cognitive, and motor operations rather than a single internal clock. Torreira, Bögels, and Levinson (2015).
Predicting When Someone Will Finish Speaking
Prediction in turn-taking refers to estimating the temporal and structural completion of a current contribution. The listener can anticipate upcoming words, infer grammatical closure, recognize a prosodic phrase boundary, or infer that a socially recognizable action has reached its conclusion. These components may agree, but they also sometimes conflict. The ability to use a cue in a controlled experiment does not prove that it is necessary or sufficient in every natural conversation.
Lexical and syntactic cues
The developing words and grammatical structure provide powerful information about possible completion. In a classic experimental manipulation, listeners could anticipate turn endings when recognizable linguistic content remained available even after intonational contour was flattened; their performance deteriorated when recognizable words were removed while intonation remained. The result established the importance of lexicosyntactic content under that task's conditions. It did not establish that intonation is irrelevant in real interaction. de Ruiter, Mitterer, and Enfield (2006).
Prosody, intonation, and phrase boundaries
Prosody includes pitch movement, timing, stress, rhythm, and phrase-final lengthening. Later experiments demonstrated that listeners use intonational phrase boundaries to project turn endings and that manipulating such boundaries can alter when participants expect a transition. The evidence therefore favors integration of linguistic structure and prosody rather than a single universal cue hierarchy. The contribution of a particular contour varies across languages, utterance types, and communicative functions. Bögels and Torreira (2015).
Anticipating forthcoming words
A projected turn ending may depend on predicting what words are likely to come next, not merely noticing that a syntactic unit is currently well formed. Gating experiments present participants with successively longer fragments of utterances and ask when they expect a turn to end. Performance can improve as contextual and lexical constraints narrow the possible continuations. This supports a predictive process, although the method cannot by itself determine whether listeners explicitly represent every anticipated word or use a coarser expectation about structure and action. Magyari and de Ruiter (2012).
Visual and discourse context
Face-to-face interaction supplies gaze, gesture, head movement, posture, and visible articulatory cues. In a 2024 experimental study, access to visual bodily signals and preceding conversational context improved turn-end anticipation under tested conditions. Visual signals interact with speech and context; a gesture that projects continuation in one setting can accompany completion in another. These findings strengthen multimodal accounts while leaving open the magnitude and generality of each cue's causal contribution. ter Bekke et al. (2024).
Comprehension and Speech Production at the Same Time
Coordinating a fast answer forces the cognitive system to allocate resources between understanding the current speaker and preparing the next contribution. This problem is central to psycholinguistics because comprehension and production are often studied separately in laboratories while ordinary dialogue requires them to interact. Levinson and Torreira (2015).
Evidence from electrophysiology
In a controlled question-answering experiment, Bögels, Magyari, and Levinson identified neural signatures associated with the opportunity to begin planning an answer before the incoming question ended. This provided direct experimental support for early response preparation. A subsequent study in less scripted interviews reported related effects under more natural conversational conditions. Together these findings make early planning a well-supported mechanism, although they do not fix a single timing point that applies to all utterances or participants. Bögels, Magyari, and Levinson (2015); Bögels (2020).
The cost of planning early
Planning and listening draw on partly overlapping cognitive resources. An experiment explicitly contrasting fast responses with anticipatory processing found a trade-off consistent with diminished processing of some later question information when responders were already preparing to speak. This qualifies the idea that earlier planning is unconditionally superior. In settings where late information is important, maintaining comprehension may be more useful than reducing the response gap. Bögels, Casillas, and Levinson (2018).
Incremental versus late-planning theories
An incremental-planning account permits response preparation as soon as enough information becomes available. A late-planning account emphasizes protecting incoming comprehension until the relevant utterance has been processed. A hybrid account treats timing as adaptive, depending on answer predictability, communicative stakes, processing demands, and the expected duration of the remaining turn. Evidence for some early planning does not imply that every utterance is planned to completion before its predecessor ends. Models must distinguish preparing a general action, selecting the content of an answer, and fully encoding its phonological form.
Attention, memory, and lexical retrieval
Conversation uses attention to monitor the speaker and relevant surroundings, working memory to maintain unfinished propositions, and lexical retrieval to prepare an answer. These domain-general resources contribute to variation in timing, but the organization of turn exchange is a distinct interactional phenomenon. A delayed response may reflect complex retrieval demands or careful evaluation rather than a generic attention problem. Roberts, Torreira, and Levinson (2015).
Social Sequence and the Meaning of Timing
A response is interpretable because it occurs after an identifiable prior action. Questions make answers relevant; invitations make acceptances or refusals relevant; stories make particular acknowledgments relevant. The sequential position of a response helps determine how silence or overlap will be understood. Conversation analysis provides evidence about these regularities through detailed examination of recorded naturally occurring interaction. Sacks, Schegloff, and Jefferson (1974).
Adjacency pairs and conditional relevance
An adjacency pair consists of linked social actions, often performed by different speakers, such as greeting and greeting, question and answer, or offer and acceptance. The first action establishes a normative expectation that a second action is due. Participants can exploit this expectation by answering immediately, delaying, repairing, or withholding a response. The pairing is a structure of social accountability, not a rigid linguistic template: an apparent question can receive an explanation, a counter-question, or a sequence of clarification turns.
Agreements, refusals, and delayed responses
Timing can help participants interpret what kind of response is emerging. A prompt acceptance and a delayed refusal may have different sequential designs, but timing never uniquely identifies agreement or disagreement. Delays may arise from lexical search, uncertainty, environmental interruption, and politeness considerations. The meaning of a delay must be established in the local sequence, not inferred from milliseconds alone. Corpus work shows that sequential organization contributes to turn-transition patterns alongside processing demands. Roberts, Torreira, and Levinson (2015).
Backchannels and continuers
Short acknowledgments such as “mm-hmm,” “right,” or “yeah” may display attention, understanding, or encouragement while another person maintains an extended turn. They may occur during another speaker's talk and therefore constitute overlapping vocal activity without a competing claim to the floor. Distinguishing continuers from new substantive turns is crucial for acoustic corpus analysis, automated segmentation, and interpretation of conversational synchrony. These practices show why turn-taking is an organized activity rather than alternating monologues.
Overlap, interruption, and collaborative completion
Simultaneous speech may involve a competitive entry, an anticipated completion, enthusiastic affiliation, shared laughter, or a brief acknowledgment. Calling every overlap an interruption collapses these distinct events. Researchers need context, turn design, participant responses, and sequential consequences to identify when an overlapping contribution was treated as disruptive. In many conversations, cooperation also involves completing each other's projected expressions. Sacks, Schegloff, and Jefferson (1974).
Repair and Misunderstanding at Turn Boundaries
Conversational repair is the organized set of practices people use when hearing, speaking, or understanding becomes problematic. Repair can be initiated by the person who produced a problematic expression or by a recipient who signals difficulty. It can occur within a turn, at a transition, or several turns later. Turn coordination and repair are closely connected, because a clarification request can temporarily interrupt the progress of a larger activity. Schegloff, Jefferson, and Sacks (1977).
Other-initiated repair
A recipient may say “Who?”, repeat a word with questioning intonation, or formulate a candidate understanding. Such actions locate trouble at different levels of specificity. A request for repetition indicates a different problem from a question about the intended referent. The timing and design of these actions can show how much of the prior turn was understood and what further information is required.
Self-repair and opportunities to continue
Speakers frequently correct or replace a word before handing over the floor. A self-repair can suspend a projected completion, while a recipient may hold back a prepared response to allow the speaker to finish reformulating. Early foundational work described an organizational preference for self-correction in recurrent repair environments. This is a claim about conversational practices and their distribution, not a rule about whose correction is always more accurate. Schegloff, Jefferson, and Sacks (1977).
Why repairs belong in the mechanism of coordination
A theory that predicts only clean alternation cannot explain real conversation. Participants must continuously monitor understanding and selectively suspend, restart, or redirect a planned contribution. Repair illustrates how turn allocation is embedded in the joint maintenance of intelligibility. The specialized mechanisms of clarification and repair are a neighboring independent topic; the present article addresses them insofar as they alter timing and floor transitions.
Eye Gaze, Gesture, Breathing, and Other Embodied Signals
Face-to-face dialogue distributes coordination across acoustic, linguistic, visual, and motor channels. The timing of a gaze shift or inhalation may help reveal when a participant expects a transition, yet the meaning of each signal depends on the activity. A purely audio-based account can miss important parts of this joint system. Holler and Kendrick (2015); ter Bekke et al. (2024).
Gaze as an index of next-speaker expectation
In recorded three-person conversations, unaddressed participants often shifted their gaze from the current speaker to the projected next speaker before the current turn ended. Close analysis also showed that these gaze movements were aligned with recognizable possible completion points. The findings demonstrate sensitivity to emerging sequential structure, but gaze alone cannot determine a person's conscious expectation or identify the next speaker in every interaction. Holler and Kendrick (2015).
Respiratory preparation
Speaking requires coordinated breathing and articulation. Measurements of breathing during question-answer sequences suggest that some respiratory preparations are timed with the unfolding conversational turn. Such findings connect language planning to bodily preparation, but inhalation is multifunctional: breathing may be governed by physical need, emotion, and the broader prosodic structure of speech. It is therefore one possible contributing signal, rather than an infallible predictor of turn-taking intention. Torreira, Bögels, and Levinson (2015).
Gesture and visible action
A speaker may extend an open-hand gesture while continuing an explanation or release a gesture as a contribution reaches completion. Visual action can make a projected transition more or less plausible, especially when the acoustic signal is ambiguous. Experimental evidence indicates that such visible behavior and preceding discourse jointly help listeners anticipate endings. The contribution of individual gesture types, however, requires language-specific and task-specific investigation. ter Bekke et al. (2024).
Telephone and text-based interaction
Telephone conversation preserves voice and sequential context while removing much visual information. Turn-taking remains possible because linguistic structure and prosody provide sufficient cues for many transitions. Text chat changes the available timing and feedback entirely: messages can be drafted privately, delivered asynchronously, and read after delay. Findings from spoken conversation cannot be transferred directly to message timestamps without accounting for these technological affordances.
Turn-Taking Across Languages and Cultures
The organization of speaking opportunities has been documented in many linguistic communities, but the distribution of response latencies, overlap practices, and social expectations varies. A scientifically responsible account recognizes both recurring interactional solutions and local conventions. The existence of cross-linguistic similarities does not establish that every culture values the same amount of silence or that a single preferred gap length should govern social judgment. Stivers et al. (2009).
What comparative research establishes
Stivers and colleagues compared the timing of responses to yes-or-no questions in ten languages. The investigation found cross-cultural similarities in turn-transition organization and measurable language-specific variation in timing. This is stronger evidence than isolated anecdotes about national conversational styles because the design targeted a comparable interactional environment. Nevertheless, the sampled activity was a specific question-response sequence, so its timing should not be generalized to all narrative, argumentative, institutional, or intimate dialogue. Stivers et al. (2009).
Linguistic diversity and cue structure
Languages differ in word order, morphology, and how grammatical information becomes available across a sentence. These differences can change when a listener can infer a likely social action or project completion. Prosodic structure and turn-final particles may also supply local cues. A theory based exclusively on English sentence endings risks overlooking these alternative mechanisms. Cross-linguistic research is therefore essential for distinguishing language-specific implementation from potentially widespread interactional organization.
Interactional norms and individual variability
A relatively long silence can have different interpretations across activities, speakers, and communities. What one participant experiences as respectful space for a considered answer another may treat as a missed opportunity. A scientific description distinguishes measured behavior, participants' demonstrated interpretations, and researchers' external evaluations. It does not rank communities by the speed with which they move from one speaker to the next.
The Development of Conversational Coordination
Early caregiver-infant interaction includes patterned alternations of vocalization, gaze, and bodily engagement. These exchanges supply opportunities to learn contingency and responsiveness before children can produce complex language. Mature spoken turn-taking develops gradually as the child acquires linguistic comprehension, speech planning, and sensitivity to social action. The 2025 review by Cosper and Pika synthesizes evidence that turn-taking development involves multiple partially separable abilities, and that adult-like timing should not be confused with the beginning of reciprocal interaction. Cosper and Pika (2025).
Prelinguistic exchanges
Adults often treat infants' sounds and gestures as contributions and respond in ways that organize an exchange. Researchers study the timing of these vocal and visual events to examine how recurrent contingencies develop. Such proto-conversational interactions are important precursors but remain different from adult exchanges built around fully specified questions, grammatical projections, and complex social commitments. Early alternation provides a developmental foundation rather than proof that infants already command the adult turn-taking system.
Changes during infancy
Longitudinal studies show that the timing of caregiver-infant exchanges changes substantially during the first year. Infant response patterns depend on the kind of signal measured, the conversational partner's behavior, and developmental stage. Simple comparisons with an adult average can therefore be misleading: a delayed infant vocalization may reflect emerging motor control or attention rather than failure to participate socially. Hilbrink, Gattis, and Levinson (2015).
Preschool and school-age language demands
Young children may understand that a question requires an answer while still needing extra time to retrieve words and plan a response. Studies of early language acquisition indicate that response timing is affected by the competition between comprehension and production, with adult-like timing emerging gradually. As questions become longer or answers more complex, the child must sustain comprehension while deciding what to say. Developmental comparisons should match action type and answer complexity before attributing differences to general conversational competence. Casillas, Bobb, and Clark (2016).
Learning from responsive interaction
Contingent exchanges can provide feedback about when a contribution is relevant and how an interlocutor's actions change the developing sequence. Caregivers model waiting, acknowledging, responding, and repairing. The relationship between such experience and later language outcomes is scientifically interesting, but a single correlation between parental responsiveness and performance cannot establish a specific causal developmental mechanism. Intervention claims require controlled evidence and appropriate developmental measures.
Individual Differences, Context, and Accessibility
Conversational timing varies with hearing conditions, language familiarity, fatigue, cognitive load, speech fluency, social expectations, and opportunities to prepare. These influences change the demands of coordination. A participant's behavior in one setting does not establish a stable personal trait, and atypical timing by itself is insufficient for identifying any disorder.
Language proficiency and multilingual interaction
When speaking a less familiar language, lexical retrieval and syntactic planning may take longer, especially for responses demanding precise wording. Listeners may also require more time to infer turn boundaries from unfamiliar prosodic or grammatical cues. The relevant comparison controls for task difficulty and interactional circumstances; multilingual experience does not imply a uniform conversational disadvantage or advantage. Cross-language differences in timing should be analyzed separately from broad claims about executive function.
Noise, hearing, and communication access
A noisy environment can delay recognition of a question or force additional monitoring of the speech signal. Participants may use visible speech and contextual expectations to compensate, and they may request repetition when the signal is uncertain. Accessibility practices, including allowing a slightly longer transition and ensuring that the speaker is visible when needed, can support participation. A conversational pause caused by reduced audibility is not evidence of deficient understanding.
Neurodevelopmental and clinical boundaries
Research on conversational coordination can inform hypotheses about individual variation in social communication, speech production, hearing, and language comprehension. Diagnostic categories such as developmental language disorder, autism spectrum disorder, aphasia, stuttering, and apraxia have distinct definitions and assessment requirements. No single gap duration, overlap count, gaze pattern, or turn-taking score functions as a clinical diagnosis. Condition-specific assessment and treatment belong to qualified clinical practice and the relevant dedicated clinical materials. This article describes the underlying psycholinguistic system.
Interactional accommodations
A person who uses augmentative and alternative communication, a speech-generating device, an interpreter, or delayed text-to-speech output may need a longer launch interval. Conversational partners can adapt by acknowledging the selected speaker, allowing enough response time, and avoiding premature completion of the other's contribution. The appropriate benchmark is successful and respectful joint participation, accounting for the modality and technology involved.
How Scientists Study Conversational Turn-Taking
Turn-taking research combines qualitative analysis of interactional sequences with quantitative measurement of acoustic timing and controlled experimental investigation. Each method answers a different question. Natural recordings reveal what participants demonstrably do; experiments identify the causal contribution of manipulated cues; neural measurements constrain hypotheses about processing time; computational models test whether a proposed mechanism can reproduce observed patterns. Strong explanatory claims integrate evidence across these approaches.
Conversation analysis and detailed transcription
Conversation analysis examines how participants organize actions through successive turns. Researchers repeatedly inspect recorded encounters and produce transcripts that may represent pauses, overlapping speech, cutoffs, elongated sounds, laughter, and audible changes in delivery. The interpretive goal is to establish the local significance of a practice using what participants do next. This yields detailed evidence about sequence organization, but the selection of illustrative episodes cannot by itself estimate a population-level prevalence. Sacks, Schegloff, and Jefferson (1974).
Acoustic corpora and temporal measurement
Large collections of recorded dialogue enable measurement of gaps, overlaps, turn lengths, response types, and conversational settings. Voice onset and offset may be annotated manually or detected algorithmically; either approach needs clear criteria for silence, breath noise, acknowledgments, and crosstalk. Skewed latency distributions require appropriate summaries and statistical models. Studies should disclose segmentation choices and report sensitivity to plausible alternative definitions. Heldner and Edlund (2010).
Gating, button-press, and projection tasks
Researchers can stop a recorded utterance at successive moments or ask participants to predict its endpoint. Experimental manipulations of available words, intonation, or conversational context reveal which information helps listeners project completion. However, button pressing is not identical to speaking an answer. The measured response includes task comprehension, decisions about the experiment, and motor execution. Laboratory findings should therefore be interpreted alongside live conversation data. de Ruiter, Mitterer, and Enfield (2006); Magyari and de Ruiter (2012).
EEG and event-related potentials
Electroencephalography captures brain activity with high temporal resolution and can reveal physiological changes around the point at which a response becomes identifiable. Studies of turn planning use these methods to test whether preparation begins before the incoming question ends. Yet an event-related potential is a statistical measure of neural activity, and its functional interpretation depends on the experimental contrast and alternative explanations. A timing-related neural effect does not uniquely identify a single cognitive algorithm. Bögels, Magyari, and Levinson (2015); Bögels (2020).
Eye tracking and multimodal recordings
Eye trackers measure gaze allocation during conversation, while synchronized video enables coding of gestures, head movements, posture, and embodied participation. A key advantage is observing how prospective speakers and observers attend to changing participation roles. A limitation is that a gaze direction can serve several functions and its interpretation must be grounded in interactional context. Multimodal measurements benefit from blinded coding, transparent exclusion criteria, and reliability checks. Holler and Kendrick (2015).
Experimental controls and ecological validity
A tightly scripted question can isolate one predicted turn-ending cue but may change participants' expectations and response opportunities. Free conversation preserves ecological complexity but introduces confounds in topic, social relationship, utterance difficulty, and speaker familiarity. Research designs can combine both approaches. Results should be described with the exact population and interactional setting rather than presented as universal properties of dialogue.
Leading Explanatory Models
A useful theory explains when a listener can project a completion, when preparation begins, why speaking starts at a particular moment, and why the same cue functions differently across actions. Existing theoretical traditions illuminate different parts of this problem; no single model currently resolves every linguistic, neural, social, and embodied level.
The interactional allocation model
The foundational conversation-analytic account explains turns through units, possible completion points, and local procedures for selection and self-selection. It is particularly effective for describing how participants collectively maintain an organized floor despite moment-to-moment uncertainty. It does not itself specify a complete neural implementation of word recognition or articulatory preparation. Its central contribution is a formal description of the interactional problem that a cognitive mechanism must solve. Sacks, Schegloff, and Jefferson (1974).
Predictive and incremental processing accounts
Psycholinguistic accounts emphasize incremental comprehension, anticipation of completion, and partially overlapping production planning. They explain how a responder can exploit available semantic information before a turn ends. Specific models vary in whether they postulate prediction of individual words, structural completion windows, or an emerging social action. Evidence from neural recording and turn-end estimation constrains the possibilities, while production planning remains flexible across situations. Levinson and Torreira (2015); Bögels (2020).
Interactive alignment
Interactive alignment theory proposes that dialogue partners tend to converge in aspects of their linguistic representations through interaction. Shared lexical choices, structural patterns, and conceptual coordination may reduce some processing demands and support mutual understanding. Alignment can help explain why dialogue sometimes proceeds efficiently, but it is not a complete account of precise handoff timing. A pair may align strongly in language while still encountering delays when an answer requires complex reasoning. Pickering and Garrod (2004); Garrod and Pickering (2004).
Social-sequential and cognitive-load explanations
An extended gap may result from greater lexical planning time, a socially delicate action, or both. Corpus analyses investigate how lexical and syntactic processing variables relate to transitions while also testing sequence organization and the type of response being delivered. An adequate integrated model allows social action to affect response selection and planning as well as the interpretation of the resulting latency. Roberts, Torreira, and Levinson (2015).
Integrated multimodal coordination
Multimodal models treat language, gaze, gesture, prosody, and bodily preparation as interacting sources of information. Recent experiments strengthen the view that contextual and visual signals can contribute to projection. Their relative weights likely vary according to signal quality, interaction type, language, and participant expectations. An integration model must specify testable predictions about cue conflict rather than treating the mere presence of several signals as an explanation. ter Bekke et al. (2024).
Established Findings, Open Questions, and Methodological Disputes
Findings supported by converging evidence
Conversation exhibits recurrent possible completion points and local organization of speaker selection. Listeners use linguistic structure and other available cues to anticipate transitions. Early response preparation sometimes overlaps with incoming comprehension, and multimodal signals can influence estimates of turn completion. These claims rest on complementary observational, experimental, and neurophysiological evidence. Their supported scope remains the conditions, populations, and interaction types actually examined. Sacks, Schegloff, and Jefferson (1974); Bögels, Magyari, and Levinson (2015); ter Bekke et al. (2024).
Questions still under investigation
The field continues to examine how semantic predictability, syntax, prosody, shared context, and social action are integrated online; how planning and comprehension compete for resources; and how these mechanisms vary with development, languages, and communication modalities. It is also unsettled how broadly a laboratory index of response planning generalizes to spontaneous social behavior. Diverse corpora and explicit replication are needed to estimate effect sizes and boundary conditions.
The lexicosyntax versus prosody debate
A prominent experiment emphasized lexical and syntactic information in projecting completion, while subsequent work demonstrated effects of intonational phrase boundaries. The most productive scientific reading is that these studies used different manipulations and interactional tasks, exposing distinct contributions of the available cues. The appropriate question concerns their conditions of use, redundancy, and relative weight rather than declaring one entire signal class irrelevant. de Ruiter, Mitterer, and Enfield (2006); Bögels and Torreira (2015).
Limits of timing metrics
A mean gap calculated across a corpus can obscure long tails, within-speaker variability, and differences among social actions. Researchers may disagree on whether a vocal acknowledgment is a turn, whether a breath belongs to the preceding turn, and where a prosodic completion occurs. Such choices can change both the measurement and the interpretation. Reporting operational definitions is part of factual precision, particularly when comparing cultures or clinical groups. Heldner and Edlund (2010).
Causation and ecological generalization
Correlations between gaze, neural activity, and speech onset can reveal coordination but cannot establish that one measured signal caused another. Experimental cue removal or manipulation provides stronger causal leverage, although the artificial task may differ from conversation in everyday settings. Research quality improves when predictions, analysis decisions, exclusion criteria, and population definitions are made explicit and multiple methods converge.
Practical Consequences for Real Conversations
Scientific understanding of turn-taking supports better communication design, teaching, accessibility, and interpretation of conversational behavior. Practical recommendations require attention to the social activity and the individual's needs. They should be proportionate to the evidence rather than advertised as universal scripts for effective conversation.
Listening without premature interruption
Listeners can use signs of possible completion while remaining open to the speaker's continuation. Allowing space for an unfolding answer is particularly useful when a question demands explanation, when the speaker is searching for words, or when a device adds production delay. The goal is to coordinate a suitable response with the actual completion of the other's contribution, rather than forcing the shortest possible gap.
Asking questions that facilitate answering
Questions that make the requested information recognizable early can permit earlier response planning. Requests containing complex qualifications may require greater listening time. In important conversations, concise formulation and an opportunity for clarification can improve understanding, although the relationship between these practices and specific outcomes depends on the situation.
Group dialogue and meetings
In multi-party interaction, a next speaker can be selected through name, gaze, institutional roles, or explicit facilitation. Meeting moderators can reduce ambiguous handoffs by clarifying who has been invited to answer, especially when participants join remotely or use interpretation and accessibility tools. Deliberate turn-allocation procedures are practical organizational adaptations, separate from the spontaneous self-selection practices of informal conversation.
Audio and video communication systems
Transmission latency, buffering, delayed playback, and missing gaze cues can disrupt expected transition windows. A speaker may interpret a technical delay as an interactional silence and start talking just as a remote partner responds. Designers can address such problems by minimizing latency, providing feedback about speaking activity, and supporting accessible turn selection. The exact engineering intervention requires system-level testing rather than extrapolation from a human gap average.
Human Dialogue and Artificial Conversational Systems
Conversational AI interfaces must detect when an input is complete, decide whether to wait, and time a response. These design problems resemble measurable features of human turn-taking, including voice activity detection, turn-end prediction, interruption handling, and multimodal coordination. A 2025 review surveys engineering approaches to turn-taking in conversational systems, illustrating active design work rather than establishing cognitive equivalence between a human speaker and a machine model. Patamia et al. (2025).
Voice activity detection versus conversational completion
Voice activity detection identifies the presence or absence of vocal signal. It can mistake a within-turn hesitation for the end of a contribution. A conversationally sensitive system also needs linguistic and contextual information about possible completion, plus interactional policies for waiting, clarification, and interruption. The gap between acoustic silence detection and social-action interpretation is a central engineering limitation.
What human comparisons can and cannot establish
If an artificial system produces fast, well-timed responses, that performance demonstrates success on the tested interactional tasks. It does not establish that the system uses the same neurocognitive representations as human interlocutors, experiences subjective awareness, or understands language in an identical way. Human processing theories can guide system evaluation, while the status of artificial subjectivity is a separate philosophical and scientific question. The Artificial Era supplies a broader historical context for examining multiple kinds of language-using agents, each requiring evidence appropriate to its own mechanisms.
Frequently Asked Questions About Conversation Turn-Taking
What exactly is turn-taking in conversation?
Turn-taking is the jointly organized distribution of opportunities to contribute to dialogue. Participants interpret developing turns, anticipate possible completion, allocate the next opportunity, and coordinate the onset of the following contribution. It includes both linguistic processing and sensitivity to what action is being performed.
Why do people answer so quickly after a question?
Listeners can often identify a question's likely action and relevant answer before the question is finished. Early response planning allows some production work to occur while the current turn is still being heard. Experiments show that this strategy can be cognitively demanding, so speed also varies with question complexity and the importance of later information. Bögels, Magyari, and Levinson (2015); Bögels, Casillas, and Levinson (2018).
Is there an ideal number of milliseconds between speakers?
No single duration is appropriate for every conversational action. The often-discussed short transition intervals describe statistical patterns in particular corpora. A brief acknowledgment, a considered refusal, and a complicated explanation have different temporal demands. The quality of a transition depends on its fit with the ongoing interaction, not merely its duration. Heldner and Edlund (2010).
How do listeners know when a speaker is finished?
Listeners integrate words, syntax, expected continuation, intonational phrase structure, social context, and sometimes gaze or gesture. Their estimates remain provisional because a speaker may extend an apparently complete turn. Controlled studies show that several cue classes can contribute under different conditions. Bögels and Torreira (2015); ter Bekke et al. (2024).
Is speaking at the same time always an interruption?
Simultaneous speech may be supportive or competitive. Brief acknowledgments, shared laughter, projected completions, and competing entries differ in their conversational functions. A recording's acoustic overlap must be interpreted with the surrounding sequence to determine how participants treated it.
Do children learn turn-taking before they can speak in sentences?
Infants participate in reciprocal vocal and bodily exchanges before they possess adult grammar. More advanced coordination develops gradually alongside comprehension and production. Developmental studies distinguish these early contingencies from the precise planning and social-action interpretation needed for mature dialogue. Cosper and Pika (2025); Casillas, Bobb, and Clark (2016).
Does a long response delay mean that someone has a communication disorder?
A single long delay does not establish a disorder. Question difficulty, distraction, language familiarity, signal quality, fatigue, and individual communication modality can all affect timing. Clinical assessment requires condition-specific criteria and context, not an isolated conversational latency.
Is turn-taking the same as conversational repair or common ground?
These processes interact but address distinct scientific questions. Turn-taking concerns speaking opportunities and transition timing. Repair concerns resolving difficulties in speaking, hearing, or understanding. Common ground concerns shared information and assumptions that make contributions interpretable. A single encounter uses all three, while each deserves independent investigation.
Synthesis: How Coordinated Dialogue Works
Conversation turn-taking is a joint system in which linguistic prediction, speech preparation, and social sequence continuously constrain one another. Speakers manage possible completion points; listeners anticipate how the turn is developing; prospective responders begin some planning before hearing the final word; and participants adjust their timing in response to prosody, visible action, comprehension demands, and the activities they are carrying out together. The evidence supports a flexible, adaptive architecture rather than a fixed pause interval or a single cue controlling the handoff. Levinson and Torreira (2015); Stivers et al. (2009); Bögels (2020).
The scientific value of this field lies in linking the observable organization of ordinary social activity to testable models of comprehension, prediction, attention, and production. Further progress depends on comparing languages and modalities, examining genuine interactions, separating action structure from processing demands, and using causal experiments to identify the contribution of particular cues. This framework explains why people are remarkably good at coordinating talk and why pauses, overlaps, and misunderstandings remain ordinary features of a system operating under time pressure.
Related Articles
• Cluster Language, Thought and Psycholinguistics — the scientific overview of cognitive language processing and related branches.
• Pragmatics in Psycholinguistics: Meaning Beyond Literal Words — how listeners infer communicative intentions beyond literal wording.
• Discourse Comprehension: How We Understand Connected Language — how context and connected propositions support understanding.
• Speech Production: How Words and Sentences Are Planned — planning processes that help determine response-launch timing.
• Predictive Language Processing: How the Mind Anticipates Words — incremental expectations that contribute to anticipating speech.
• Psycholinguistics: How Language Is Processed in the Mind — the broader science of language comprehension, acquisition, and production.
Conversational Repair: How People Resolve Misunderstandings — how speakers repair misunderstandings during turn-taking.
Sign Language Processing: Psycholinguistics of Visual Language — psycholinguistic evidence from visual-manual signed languages.
References
Bögels, S. (2020). Neural correlates of turn-taking in the wild: Response planning starts early in free interviews. Cognition, 203, 104347. https://doi.org/10.1016/j.cognition.2020.104347
Bögels, S., Casillas, M., and Levinson, S. C. (2018). Planning versus comprehension in turn-taking: Fast responders show reduced anticipatory processing of the question. Neuropsychologia. https://doi.org/10.1016/j.neuropsychologia.2017.12.028
Bögels, S., Magyari, L., and Levinson, S. C. (2015). Neural signatures of response planning occur midway through an incoming question in conversation. Scientific Reports, 5, 12881. https://doi.org/10.1038/srep12881
Bögels, S., and Torreira, F. (2015). Listeners use intonational phrase boundaries to project turn ends in spoken interaction. Journal of Phonetics, 52, 46–57. https://doi.org/10.1016/j.wocn.2015.04.004
Casillas, M., Bobb, S. C., and Clark, E. V. (2016). Turn-taking, timing, and planning in early language acquisition. Journal of Child Language, 43(6), 1310–1337. https://doi.org/10.1017/S0305000915000689
Cosper, S. H., and Pika, S. (2025). Human turn-taking development: A multi-faceted review of turn-taking comprehension and production in the first years of life. Psychonomic Bulletin and Review, 32, 2669–2695. https://doi.org/10.3758/s13423-025-02749-8
de Ruiter, J. P., Mitterer, H., and Enfield, N. J. (2006). Projecting the end of a speaker's turn: A cognitive cornerstone of conversation. Language, 82(3), 515–535. https://doi.org/10.1353/lan.2006.0130
Garrod, S., and Pickering, M. J. (2004). Why is conversation so easy? Trends in Cognitive Sciences, 8(1), 8–11. https://doi.org/10.1016/j.tics.2003.10.016
Heldner, M., and Edlund, J. (2010). Pauses, gaps and overlaps in conversations. Journal of Phonetics, 38(4), 555–568. https://doi.org/10.1016/j.wocn.2010.08.002
Hilbrink, E. E., Gattis, M., and Levinson, S. C. (2015). Early developmental changes in the timing of turn-taking: A longitudinal study of mother–infant interaction. Frontiers in Psychology, 6, 1492. https://doi.org/10.3389/fpsyg.2015.01492
Holler, J., and Kendrick, K. H. (2015). Unaddressed participants' gaze in multi-person interaction: Optimizing recipiency. Frontiers in Psychology, 6, 98. https://doi.org/10.3389/fpsyg.2015.00098
Levinson, S. C., and Torreira, F. (2015). Timing in turn-taking and its implications for processing models of language. Frontiers in Psychology, 6, 731. https://doi.org/10.3389/fpsyg.2015.00731
Magyari, L., and de Ruiter, J. P. (2012). Prediction of turn-ends based on anticipation of upcoming words. Frontiers in Psychology, 3, 376. https://doi.org/10.3389/fpsyg.2012.00376
Patamia, R. A., Dinh, H. P. T., Liu, M., and Cosgun, A. (2025). Turn-taking modelling in conversational systems: A review of recent advances. Technologies, 13(12), 591. https://doi.org/10.3390/technologies13120591
Pickering, M. J., and Garrod, S. (2004). Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences, 27(2), 169–190. https://doi.org/10.1017/S0140525X04000056
Roberts, S. G., Torreira, F., and Levinson, S. C. (2015). The effects of processing and sequence organization on the timing of turn taking: A corpus study. Frontiers in Psychology, 6, 509. https://doi.org/10.3389/fpsyg.2015.00509
Sacks, H., Schegloff, E. A., and Jefferson, G. (1974). A simplest systematics for the organization of turn-taking for conversation. Language, 50(4), 696–735. https://doi.org/10.2307/412243
Schegloff, E. A., Jefferson, G., and Sacks, H. (1977). The preference for self-correction in the organization of repair in conversation. Language, 53(2), 361–382. https://doi.org/10.2307/413107
Stivers, T., Enfield, N. J., Brown, P., Englert, C., Hayashi, M., Heinemann, T., Hoymann, G., Rossano, F., de Ruiter, J. P., Yoon, K.-E., and Levinson, S. C. (2009). Universals and cultural variation in turn-taking in conversation. Proceedings of the National Academy of Sciences, 106(26), 10587–10592. https://doi.org/10.1073/pnas.0903616106
ter Bekke, M., Levinson, S. C., van Otterdijk, L., Kühn, M., and Holler, J. (2024). Visual bodily signals and conversational context benefit the anticipation of turn ends. Cognition, 248, 105806. https://doi.org/10.1016/j.cognition.2024.105806
Torreira, F., Bögels, S., and Levinson, S. C. (2015). Breathing for answering: The time course of response planning in conversation. Frontiers in Psychology, 6, 284. https://doi.org/10.3389/fpsyg.2015.00284