Introduction
Filler words such as "um," "like," "you know," and "so," along with other repeated speech patterns such as habitual hedging phrases or a fixed rising inflection at the end of statements, are commonly framed as errors of attention or willpower that a speaker simply has not tried hard enough to correct. Wikipedia's coverage of speech disfluency has generally treated these patterns as linguistic or clinical phenomena without connecting them to the substantial psychological literature on habit formation and behavior change, a body of research spanning cognitive and social psychology, behavioral therapy, and speech-language pathology. This article reviews that literature and considers what it implies about why certain speech patterns persist, why efforts to eliminate them through suppression alone tend to fail, and what timeframes the research actually supports for changing them.
A distinction is maintained throughout between two kinds of claims. The first kind is established habit-formation research: findings from controlled studies and diary methods on habitual behavior in general, cited to their original sources. The second kind is an extension of that research to speech patterns specifically. Direct empirical study of filler words or other verbal disfluencies as habits, in the technical psychological sense, is limited. Where this article applies general habit theory to speech, that application is a reasonable structural inference rather than a directly tested finding, and it is flagged as such.
Speech Patterns as Habitual Behavior Rather Than Deliberate Choices
In behavioral psychology, a habit is typically defined not by how undesirable or repetitive a behavior looks from the outside, but by its underlying mechanism: a habit is a learned association between a stable context or cue and a response, such that encountering the cue activates the response with little or no conscious deliberation [3][5]. This is distinct from a merely frequent behavior. A behavior performed often but still under conscious control, such as choosing a restaurant based on evaluating options each time, is not a habit in this technical sense. A habit is one where the cue itself, rather than a decision, is doing most of the work of producing the behavior [5].
Diary studies by Wood, Quinn, and Kashy asked participants to report their ongoing thoughts and behaviors at intervals throughout the day. Behavior classified as habitual, meaning performed almost daily in stable contexts, was associated with participants' thoughts running to matters unrelated to the behavior itself, consistent with the behavior not requiring active conscious guidance. Non-habitual behavior, performed less often or in shifting contexts, was associated with thoughts that tracked the behavior more closely, consistent with a greater need for conscious oversight. Habitual behavior was also associated with lower reported stress than non-habitual behavior [1].
Applied to speech, this framing offers a plausible account of why filler words and similar patterns are experienced as involuntary rather than as a series of small choices. Public speaking, meetings, interviews, and other contexts that reliably recur for a given speaker function as stable cues. A verbal pattern that reliably co-occurs with those cues, such as inserting "um" at a pause point or hedging with "I mean" before a claim, fits the general structural definition of a habit: a learned link between a recurring context and an automatic response. This is an extension of the general habit literature to a domain, spoken language production, that the cited studies did not themselves examine. No study identified in preparing this article specifically measured filler-word production using the diary or reaction-time methods that habit researchers use to establish automaticity. The claim being made here is therefore one of structural fit with established habit criteria, not a reported empirical result about filler words.
The Habit Loop in Behavioral Psychology
The popular description of habits as a three-part "cue-routine-reward" loop originates largely from journalist Charles Duhigg's 2012 book The Power of Habit, a popular-press synthesis rather than a peer-reviewed source [19]. The underlying research it draws on describes a similar structure in more precise terms: habits are context-response associations stored in memory that form through repeated pairing of a stable cue with a behavior that has been rewarded or reinforced in that context in the past [5]. Repetition under consistent conditions is what builds the association; the reward's main role, in this account, is during the learning phase, after which the behavior can persist even without deliberate anticipation of a reward, cued directly by context [3][5].
A practical implication of this cue-based model is that habits are more effectively changed by altering or removing the cue than by resisting the response once the cue has already occurred. Verplanken and Wood, reviewing intervention research on consumer habits such as travel mode choice and food purchasing, found that interventions disrupting the stability of the context, for example through a change in routine or environment, were more effective at breaking existing habits than interventions aimed only at changing attitudes or providing information, because habits are governed more by contextual cuing than by conscious belief [4]. They proposed that effective habit-change interventions generally need to either interrupt the cue-response link directly ("downstream" strategies) or restructure the environment so the old cue no longer reliably occurs ("upstream" strategies), rather than relying on persuasion alone [4].
Extended to speech, this suggests that a filler-word pattern cued by, for instance, the specific tension of an unscripted pause in a formal presentation may be less responsive to general resolve to "stop saying um" than to changes that alter what happens at the cue itself, such as substituting a different response (a pause, a breath) at the moment the trigger occurs. This is again an inference from the general cue-based model rather than a tested finding about speech specifically.
Progressive Exposure and Graduated Behavior Change
A separate strand of behavioral research addresses how gradual, structured exposure to a triggering situation changes behavior, distinct from but complementary to habit-cuing research. The foundational theoretical account, developed in the context of fear and anxiety, holds that repeated, controlled exposure to a feared or avoided stimulus allows new, corrective associations to form and compete with the old response, provided the exposure is structured so the old response can be experienced and gradually modified rather than avoided [18]. This principle underlies graduated exposure hierarchies used broadly in behavior therapy, in which a person is exposed to increasingly challenging versions of a trigger rather than the most difficult version at once.
Speech-language pathology has applied structured, graduated behavioral techniques to fluency disorders for decades, independent of but conceptually related to the general exposure literature. Azrin and Nunn's habit-reversal procedure, originally developed for nervous habits and tics, was directly adapted for stuttering: a controlled trial by Azrin, Nunn, and Frantz compared a regulated-breathing method, described by the authors as built from general habit-reversal components, against an abbreviated desensitization procedure for reducing stuttering episodes, reporting a substantial reduction in stuttering with the regulated-breathing method [15]. A broader systematic review of stuttering treatment research from 1970 to 2005 found that, among approaches meeting basic trial-quality criteria, treatments combining self-management strategies, response-contingent techniques, and structured practice were associated with the strongest outcomes for both children and adults who stutter [16]. More recently, a controlled trial testing a cognitive-behavior therapy package, which included graduated exposure to feared speaking situations, alongside standard speech restructuring found that the addition of the exposure-based CBT component produced significant, sustained improvement in participants' anxiety and avoidance around speaking situations, though it did not by itself improve fluency beyond what speech restructuring alone achieved [17]. That result is a useful caution: graduated exposure in this literature is well supported as a treatment for the anxiety and avoidance that often accompany a speech pattern, but the evidence for it directly reshaping the disfluency itself is more limited and treatment-specific.
Applying a progressive exposure framework to filler words or other repeated speech patterns outside of clinical fluency disorders (that is, to informal disfluency rather than diagnosed stuttering) is again an extension beyond what the cited studies tested directly. The structural logic, that gradually increasing exposure to the contexts that cue a pattern, while practicing an alternative response, allows a new association to compete with the old one, is consistent with both the exposure literature and the cue-based habit literature. It has not been isolated and tested as a treatment for filler-word frequency specifically in the sources reviewed here.
Why Suppression and Willpower-Based Approaches Tend to Fail
A substantial body of research in cognitive psychology addresses what happens when people attempt to directly suppress a thought or a behavior through effortful control. Wegner's ironic process theory proposes that intentional mental suppression requires an ongoing monitoring process that checks whether the unwanted thought or impulse is present, and that this monitoring process can itself increase the accessibility of the very thing being suppressed, particularly when attentional resources are limited or under load, producing a rebound or paradoxical increase in the suppressed content [11]. A meta-analysis of controlled studies on thought suppression by Abramowitz, Tolin, and Street found consistent support across the reviewed literature for this paradoxical pattern, especially for suppression attempts sustained over longer periods and under higher cognitive load [12].
This research was not conducted on filler words or speech patterns, and the mechanism it describes concerns unwanted thoughts and, in related work, motor and behavioral impulses more broadly, not verbal disfluency in particular. As an extension, it offers a plausible explanation for a commonly reported experience: an instruction to "just stop saying um," which asks a speaker to actively monitor for and suppress the word in real time while also managing the content of what is being said, may increase the cognitive load under which the ironic-process effect is most pronounced, plausibly making the filler word more, rather than less, likely to occur under pressure. This remains an inference from adjacent research rather than a directly demonstrated result for speech.
Why Awareness-Based and Cue-Based Approaches Tend to Succeed
Consistent with the cue-based model of habit and the difficulties documented in the suppression literature, behavioral interventions for repetitive, habit-like behaviors have generally been built around identifying the cue and building an alternative response, rather than around suppression of the target behavior itself. In Azrin and Nunn's original habit-reversal procedure, the first formal component is awareness training, in which a person learns to accurately detect the earliest signs of the habitual behavior and the situations that reliably precede it, before any attempt is made to change the behavior [13]. Miltenberger, Fuqua, and Woods's review of habit-reversal research, which explicitly discusses application to stuttering among other habits and tics, describes awareness training together with a "competing response" component, in which the person practices a specific alternative behavior physically incompatible with the habit, cued by the same trigger that previously cued the habitual response, as the core mechanism through which habit reversal produces change [14].
A related line of research on implementation intentions, developed independently of the habit-reversal tradition, found that forming a specific, pre-planned "if-then" link between an anticipated situational cue and an intended response substantially increased the likelihood that people followed through on a range of goals, compared with a general intention to change without a specified cue-response plan [10]. Gardner's review of how "habit" is used across the health-behavior literature similarly concludes that because habitual behavior is triggered by cues rather than motivation, effective behavior-change strategies generally need to work at the level of the cue and the response linked to it, rather than relying only on increasing a person's motivation or willpower to resist [6].
Together, this literature offers a reasoned account of why cue-identification and awareness-first approaches are better supported for changing habit-like behavior than suppression-only approaches: they work with the automatic, cue-triggered nature of the target behavior rather than against it, avoiding the ironic monitoring burden associated with direct suppression. Its application to filler words specifically follows the same logic as the general model, identifying the cues that reliably precede a given filler word and building a specific alternative response to those cues, but has not been isolated as a tested intervention for filler-word reduction in the literature reviewed for this article.
Realistic Timelines for Habit Change
A widely circulated claim holds that a habit can be formed or broken in a fixed period, most commonly cited as 21 days, and sometimes as 66 days. The 21-day figure traces to an informal observation by a mid-twentieth-century plastic surgeon about how long patients seemed to take to adjust psychologically to changes in appearance, not to a controlled study of habit formation, and it has circulated in popular self-help writing far beyond its original, non-experimental context. The 66-day figure is a more recent and more accurate popularization, but it is frequently repeated as though it were a fixed, universal rule, which the underlying research does not support.
The study most directly responsible for the 66-day figure is Lally, van Jaarsveld, Potts, and Wardle's real-world investigation of habit formation. Ninety-six volunteers each chose a single health-related behavior (an eating, drinking, or activity behavior) to perform daily for 12 weeks in response to a consistent chosen cue, such as a specific meal or time of day, and reported their sense of how automatic the behavior felt each day using a self-report habit index. Automaticity increased along an asymptotic curve, rising quickly at first and then leveling off, and the researchers modeled the point at which each individual's curve reached its plateau. The average time to reach that plateau was 66 days, but individual times ranged from 18 days to 254 days depending on the person and the specific behavior chosen, with simpler behaviors generally reaching automaticity faster than more complex ones. Missing a single opportunity to perform the behavior did not meaningfully disrupt the overall trajectory toward automaticity [8][9].
This finding does not establish a fixed universal timeline for habit formation, and the study's authors did not claim one. It establishes that, for a specific set of self-selected, once-daily health behaviors performed in response to a consistent cue over a 12-week window, automaticity followed a predictable curve with a wide range of individual completion times and an average around 66 days [8]. A follow-up practical review by Lally and Gardner, drawing on this and related work, discusses strategies such as selecting a stable, specific cue and prioritizing consistency of repetition over intensity, while cautioning that the amount of practice needed to reach automaticity varies considerably by behavior and by individual, and that no single number reliably predicts how long a given habit will take to form [7].
No study identified in preparing this article has directly measured how long it takes a speech pattern such as a filler word to change under a specific intervention using the same automaticity-tracking methodology Lally and colleagues used for health behaviors. Extending their finding to speech pattern change, the most defensible inference is not a specific day count but the shape of the underlying finding: meaningful behavior change of this kind tends to follow a gradual curve rather than a sudden shift, tends to vary substantially across individuals and across the specific behavior involved, and is not reliably disrupted by occasional lapses. Applying the specific 66-day average to speech pattern change as though it were a directly established number would overstate what the research supports.
Applying Habit-Formation Research to Speech Pattern Modification
Drawing the strands above together: filler words and related speech patterns share the structural features that define a habit in the behavioral literature, namely a learned, context-triggered response that operates with reduced conscious oversight [1][3][5]. Cue-based intervention models, awareness training, and graduated practice all have independent empirical support as approaches to habit-like and disfluency-related behavior change, including direct evidence from stuttering treatment research showing that self-management and response-contingent, exposure-informed techniques outperform less structured approaches [4][10][13][14][16][17]. Suppression-based approaches to unwanted verbal or behavioral output are undercut by well-replicated findings on ironic mental control [11][12]. And the realistic timeline literature counsels against fixed-day claims in favor of a gradual, individually variable, cue-dependent process of change [8][9].
None of this constitutes direct experimental proof that filler-word frequency specifically follows the same curve, responds to the same interventions, or takes the same range of time to change as the health behaviors, tics, and consumer habits that the cited studies examined. It is, instead, a coherent and well-supported extension: speech patterns meet the structural definition of a habit used throughout this literature, and the general principles of cue-based, awareness-first, gradually practiced behavior change are the best-supported account available for how such patterns are likely to change, pending direct research on verbal disfluency using the same methods.
FAQs
Does research support the popular claim that habits take exactly 21 days to form?
No controlled study of habit formation supports a fixed 21-day timeline. The widely cited real-world study on this question, Lally and colleagues' 2010 investigation, found that participants took an average of 66 days to reach automatic performance of a daily health behavior, with individual times ranging from 18 to 254 days depending on the person and the behavior involved. The finding is a wide, individually variable range with an average around 66 days under that study's specific conditions, not a fixed number applicable to all habits.
Why does deliberately trying to suppress a habitual behavior sometimes make it worse?
Research on ironic mental control proposes that intentionally suppressing a thought or impulse requires an ongoing monitoring process that checks for the very thing being suppressed, and that this monitoring process can itself increase the accessibility of the unwanted content, particularly when attention is divided or under load. A meta-analysis of controlled thought-suppression studies found consistent support for this paradoxical pattern. This is offered as one plausible explanation, drawn from adjacent research, for why direct effort to eliminate a habitual behavior such as a filler word often proves less effective than approaches built around identifying the behavior's triggers.
What is the difference between a habit and simply a frequent behavior, in psychological research?
In the behavioral literature, a habit is defined by its mechanism rather than by how often it occurs. A habit is a learned association between a stable context or cue and a response, such that the cue triggers the response with little conscious deliberation. A behavior performed often but still requiring active decision-making each time it occurs, rather than being triggered automatically by a consistent cue, does not meet this definition, even if it happens frequently.