Liu Hua is an independent researcher who has lived with stuttering for nearly three decades. Since 2019 he has worked with over sixty adults who stutter, focusing on the motor mechanisms underlying speech fluency and the social application of speech. In 2019 he developed the Reconstruction Protocol, and in 2025 he proposed the concept of a motor speech homeostatic repair mechanism, the first of three theoretical papers he has completed on the subject. His work integrates motor learning theory, neurobiology, and clinical observation, and the framework and accompanying materials are openly documented on Zenodo as a contribution to the stuttering community. He writes from the dual perspective of someone who has lived with stuttering, and who has spent seven years observing how others change.
|
Abstract
Stuttering is often described by its visible shades—severity, frequency, type of dysfluency. Beneath these surface variations lies a deeper architecture: a motor speech homeostatic repair mechanism that protects whatever speaking pattern has been repeated most often, whether fluent or dysfluent. This mechanism is not unique to speech: it is how the body protects practiced movement in general. Drawing on seven years of clinical work with adults who stutter, this paper proposes that speech is not merely pronunciation but the application of pronunciation within living social contexts. I describe how a protected dysfluent program can be replaced, and what changes in the person afterwards. Understanding this hidden dimension reveals why two people with identical visible stuttering may inhabit entirely different internal shades.
Statement of sources
Claims in this paper come from three distinct sources, which I mark throughout: established findings in the motor control and stuttering literature, cited in brackets; my own clinical series of over sixty adults who stutter (2019–2026), archived in the dataset cited below; and the terms I propose, which are introduced explicitly as proposals.
Introduction: Many Shades, One Mechanism
The 2026 ISAD theme, “Shades of Stuttering—What’s Yours?” invites us to honor the diversity of stuttering experiences. Some stutter openly and visibly; others hide it so effectively that even close friends never notice. I lived with stuttering for nearly thirty years and tried breathing techniques, slowed speech, and counseling; some helped temporarily, none lasted. I eventually realized that I had been looking at the wrong layer of the problem. This paper describes the architecture I found beneath it.
The Motor Speech Homeostatic Mechanism
Speech is not the only movement that becomes habitual. Walking, handwriting, driving and swimming all settle into patterns repeated without conscious direction. When an action is repeated countless times, the brain transfers control from the conscious cortex to subcortical structures such as the basal ganglia and cerebellum [1,2]. The action becomes automatic, requires less energy, and feels comfortable. Whatever has been repeated most often becomes the protected default.
For people who stutter, that default is stuttered speech. The body maintains speech motor patterns within a stable range, much as it maintains temperature and blood pressure—an analogy, not an identical mechanism [3]. Any deviation, whether from therapy, stress, or deliberate effort, registers as a disturbance and is gradually repaired back toward the default, during sleep, rest, and daily micro-adjustments.
This repair mechanism is not itself pathological. It keeps anyone’s speech stable, and it protects fluent speakers too: after heavy exercise, emotional upset or external pressure, speech can waver, and within a day or two it returns to its range on its own. Stuttering is not a missing mechanism but a wrong turn the mechanism was never meant to defend. Unlike these temporary deviations, stuttering is a stable pattern the mechanism actively restores [4,5]. This may explain why relapse is so common after fluency-shaping therapy: the therapy creates fluency during sessions, without changing the protected default, so once conscious effort relaxes, the old pattern tends to return.
The Umpire Trap
In my clinical series, people who stutter often appear trapped in a paradox of self-assessment: what feels “right” or “comfortable” is calibrated to the pattern they have spoken for years. Comfort is not neutral evidence. A pattern performed for years is also cheaper to produce, and the body registers lower effort as “correct”—the laryngeal and articulatory adjustments of stuttered speech have become the least costly way of speaking this person knows. A new way of speaking, even when objectively more fluent, therefore feels wrong—not because it is wrong, but because it is unfamiliar and, at first, more effortful [6]. Their subjective feedback, on which most therapies depend, is filtered through a mechanism committed to preserving the old pattern. I call this the umpire trap: the final authority over correctness has been shaped by the very condition it seeks to escape.
The Invisible Shades
If stuttering is maintained by a protected motor program, then visible shades are only surface expressions of a deeper reality. In the same series, one man’s stuttering appeared mild—he rarely blocked or repeated. Yet his internal experience was debilitating: he had restricted his active vocabulary to those words he could produce without stuttering, becoming a master of circumlocution. To the world he was fluent; inside he was severely restricted. Severity ratings based on observable behavior may therefore mislead us about the true nature of a person’s stuttering [7,8].
Disassembly, Reassembly, and a New Baseline
If stuttering is a protected motor program, the challenge is not to modify it but to replace it. Disassembly returns the sounds of speech to their most basic, independent form: in people who stutter, fluent and dysfluent elements are bundled into the same motor packages, and these must be separated, as a machine is stripped to its parts before rebuilding [9]. Repairing a single sound does not help here, for the same reason that repairing one broken part does not restore a machine; what is needed is a return to the original settings, from which the whole program is built again. Reassembly then builds new connections free of the old dysfluent associations [10].
One implication is that this process may bypass the homeostatic repair mechanism. If the old pattern is not practiced while a new pattern is built and practiced intensively, the body may gradually shift its protection to the new pattern. What was once repaired toward dysfluency is then repaired toward fluency.
Speech Is More Than Pronunciation
Speech is not simply producing sounds; it is using them with particular people. Speakers adjust, by default, to the person in front of them—an older relative, a child, a friend, a stranger—though they differ in how quickly they can switch. For people who stutter, this layer of adjustment has been shaped by years of negative experience: the difficulty is not only whether to speak at all, but speaking appropriately, at the moment and in the form the situation calls for [11,12]. The body has learned not only how to stutter but also when to stutter, and when to avoid speaking altogether.
In the wider body of work behind this paper, I have developed a four-part model of how speech works in real contexts—receiving information, forming a thought, selecting the words, and producing the sound—and found that most training addresses only the last part. One implication is the familiar observation that reading aloud, or speaking to oneself, often proceeds without difficulty, while speech in real dialogue does not [11]. Self-directed practice can work; live conversation is where the difficulty appears.
This layer also explains a familiar failure. A speaker who rehearses a coming encounter for days may perform worse than if they had not rehearsed, because prepared speech is a monologue, while real dialogue is a closed loop in which each utterance is drawn out by the other person’s response. The prepared form receives no answer, and therefore cannot be sustained.
Addressing that layer without changing the motor program provides coping strategies but leaves the root mechanism intact. Both must change together.
The Adaptation Journey
In the same clinical series, three patterns recurred after a new motor program began to take shape. This is not simply practicing more; it is relearning how to be a speaker in the world.
First, expression can surge. Because the new program carries no old baggage, people temporarily say things they could never say before, across situations that previously defeated them. The surge is usually followed by a settling period, as a new baseline emerges that reflects the person’s real capacity rather than the capacity imposed by stuttering. Some experience this as a loss and miss the peak; the peak was transitional, the floor is the gain.
Second, some chase “better” speech—faster, more impressive—until they exceed what their cognitive organization can support, producing fragmented thoughts and filler words. The remedy is to step back and rest and let deliberately slowed speech restore balance.
Third, identity does not automatically follow the motor change. Some, having lost the struggle that shaped them, unconsciously look for another struggle to take its place. This transition is rarely discussed but deeply real.
Why Some Therapies Work and Others Do Not
The framework explains mixed therapy results. Behavioral techniques help when the motor program stays relatively flexible; when it is deeply entrenched, they are insufficient, because the homeostatic mechanism repairs the deviation as soon as conscious effort relaxes [13,14]. Purely psychological approaches reach the layer of social use but leave the motor root untouched [15,16]. Effective intervention must match the depth of the problem.
Conclusion
The motor speech homeostatic framework adds a dimension that visible categories miss: beneath every pattern lies an invisible motor architecture protected by the body’s own stability systems, and beneath every effort to speak lies the work of speaking with other people. Understanding this does not make stuttering disappear, but it turns it from a mysterious curse into an intelligible process.
References
- Doyon, J., & Benali, H. (2005). Reorganization and plasticity in the adult brain during learning of motor skills. Current Opinion in Neurobiology, 15(2), 161–167.
- Watkins, K. E., Smith, S. M., Davis, S., & Howell, P. (2008). Structural and functional abnormalities of the motor system in developmental stuttering. Brain, 131(1), 50–59.
- Guenther, F. H. (2016). Neural Control of Speech. MIT Press.
- Boberg, E., & Kully, D. (1994). Long-term results of an intensive treatment program for adults and adolescents who stutter. Journal of Speech and Hearing Research, 37(5), 1050–1059.
- Bernstein Ratner, N. (2005). Evidence-based practice in stuttering: Some questions to consider. Journal of Fluency Disorders, 30(3), 163–188.
- Bloodstein, O., Bernstein Ratner, N., & Brundage, S. B. (2021). A Handbook on Stuttering (7th ed.). Plural Publishing.
- Sheehan, J. G. (1970). Stuttering: Research and Therapy. Harper & Row.
- Van Riper, C. (1982). The Nature of Stuttering (2nd ed.). Prentice-Hall.
- Schmidt, R. A., & Lee, T. D. (2020). Motor Learning and Performance: From Principles to Application (6th ed.). Human Kinetics.
- Kleim, J. A., & Jones, T. A. (2008). Principles of experience-dependent neural plasticity. Journal of Speech, Language, and Hearing Research, 51(1), S225–S239.
- World Health Organization. (2001). International Classification of Functioning, Disability and Health (ICF). WHO.
- Yaruss, J. S., & Quesal, R. W. (2004). Stuttering and the International Classification of Functioning, Disability, and Health (ICF): An update. Journal of Communication Disorders, 37(1), 35–52.
- Bothe, A. K., Davidow, J. H., Bramlett, R. E., & Ingham, R. J. (2006). Stuttering treatment research 1970–2005: I. Systematic review incorporating trial quality assessment of behavioral, cognitive, and related approaches. American Journal of Speech-Language Pathology, 15(4), 321–341.
- Craig, A. (1998). Relapse following treatment for stuttering: a critical review and correlative data. Journal of Fluency Disorders, 23(1), 1–30.
- Menzies, R. G., O’Brian, S., Onslow, M., Packman, A., St Clare, T., & Block, S. (2008). An experimental clinical trial of a cognitive-behavior therapy package for chronic stuttering. Journal of Speech, Language, and Hearing Research, 51(6), 1451–1464.
- Corcoran, J. A., & Stewart, M. (1998). Stories of stuttering: A qualitative analysis of interview narratives. Journal of Fluency Disorders, 23(4), 247–264.
- Liu, H. (2026). Reconstruction Protocol for Stuttering: A motor speech homeostatic repair mechanism-based intervention (2019–2026) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21823997
Hi Liu Hua, I found your discussion post to be interesting. Thank you for sharing your findings. I am currently learning about the same thing you researched. I was wondering if you could further discuss the motor program and how you have seen this framework work for you personally? Your explanation would greatly help deepen my understanding. Thank you very much!
Hi, thank you — glad it connected. I’ll take your two questions in turn.
On the motor program. The mechanism is not specific to speech. Anything the body builds from repeated muscle movement settles into a form it then protects — walking cadence, handwriting, a signature. Motor learning calls this a generalized motor program; the idea itself is well established, and my contribution is only to apply it to speech. In speech, that protective mechanism is what makes stuttering durable: the pattern repeated most often becomes the default, and the body repairs deviations back toward it. It is also why an ordinary technique does not hold — it creates a deviation without changing the default.
The awkward part is the direction of the work. Most training uses repetition to build a program. Here the program already exists and has to be changed, and changing an entrenched motor program is genuinely hard — almost anyone finds it very difficult to change their own handwriting, which is precisely why a signature works as evidence. The difficulty is not in the movement; it is in the protection. In my own case what made the change possible was disassembly first — separating the fluent and dysfluent elements back to their most basic, independent form — and only then rebuilding. Repairing a single sound leaves the bundle intact.
On what this demands in practice. On your second question — I’d rather describe what the framework demands in execution than my own case, because the difficulty is not in the concept. It is in what happens when you try to act on it.
First, protection pushes back. Reaching in to change the core produces a strong, physical sense of wrongness. The closest everyday comparison is being made to change your handwriting: the discomfort and the resistance are real, and they are the mechanism working, not evidence that the attempt is failing.
Second, there is no positive feedback along the way. There is no stage at which you train for a while and see a partial result. The result arrives at the end, or not at all — which means the person has no instrument for judging progress except how it feels, and that feeling is calibrated to the old pattern. My impression is that most attempts die here rather than at the motor problem.
Third, the old program does not stop running. Two hours of daily practice sit against eight hours of ordinary living in the old pattern. How the two change relative to each other, and how to switch between them deliberately, has to be planned, not discovered.
Fourth, once a new program exists, both are available. Getting the old one to withdraw in an orderly way is a separate problem from building the new one.
And in a real sense the whole thing is relearning speech. The reconstruction runs the same course as normal development, only compressed — from childhood through to adulthood — so the process has stages rather than being one continuous improvement. One stage matters more than the rest. When training reaches the equivalent of the early teens, pronunciation has become essentially mature and complete, and this is where an execution problem appears: the person can now speak well enough to feel finished, but what he has is a young teenager’s capacity. It is developed, and it still cannot carry an adult’s conversational life. The predictable consequence is that the new ability gets used too early — the excitement of finally speaking pushes it straight into situations that demand a maturity it does not yet have. So the person has to hold that excitement back on purpose and wait for stability and strength to come up before widening the range of use. Learning to adjust to different listeners is part of what is still outstanding at that stage.
On a motor-program account, then, resolution has two tracks that have to run together: the pronunciation problem and the adaptation problem. Most of what I’ve seen addresses one and calls it the whole.
Hi Liu Hua, I am currently completing my masters program and learned a lot from this article. Your proposed motor speech homeostatic mechanism seems related to neuroplasticity and the idea that “practice makes permanent” or reinforces established motor patterns. From a clinical perspective, how can SLPs balance repeated practice to establish a new speech motor pattern with spontaneous conversational practice, especially given your point that rehearsed speech may not generalize to real-world dialogue?
Hi — my working language is Chinese, so I needed to translate your comment into Chinese, work through it and revise it there, and then translate it back. I am therefore not certain that I have understood your question accurately. What follows is my answer as I understood it; if I have read it wrong, please say so directly.
Your question, as I understand it, is this: if rehearsed speech does not carry over into real conversation, what makes me think that a new motor pattern established through heavy repetition will be able to?
Let me start with the neuroplasticity point. Part of it is the same: repetition is what decides which pattern the body protects. But the mechanism itself is passive, and it has a ceiling. Repetition within a given window strengthens the learning; past that ceiling the effect reverses — the pattern reverts, or blurs. This is clearest during the adaptation period: conversations with strangers are generally held to seven or eight a day, three or four in the morning and three or four in the afternoon, about ten minutes apart. Beyond that number, the adaptation does not hold.
This is not specific to stuttering. Look at a single word long enough and it stops looking like a word — you can no longer recognise it. Past a certain point, repetition reverses the effect. Motor learning points the same way: distributed practice outperforms massed practice, and the interval is itself part of the effect. So the ceiling does not look like an isolated case.
So the amount of repetition is not what matters. The form of the repetition is.
I think you have pointed at something important — and at something that has to be settled before training even starts. No fixed form of communication can be allowed to become the norm during training: reading aloud, memorising a script, repeating after a model, and so on. The whole training therefore takes place in the form of ordinary back-and-forth speech — free, unpredictable, not even returning to the same topics.
And so there is no transfer problem here. From the first day, what is being practised is the situation it will be used in. The question of how to carry it into real conversation only arises when the training happened somewhere else.
Training and rebuilding speech on the platform of ordinary conversation also requires the whole process to be one-to-one. By the time training ends, the person can already speak and converse normally with the trainer. What follows is the adaptation phase: extending from speaking with the trainer to speaking with a range of different people. The capacity to repair has the same ceiling — past a certain point, repair takes longer, and staying in that stretched state damages the repair mechanism in turn. That is what makes the adaptation phase meaningful: it is not speech training, it is adjusting the mechanism.
Hi there! Thank you for sharing you research. As a student clinician, I’m curious what you’ve found helps most with carryover outside of the clinic?
Hi, thank you — it is a fair question. The answer I would give is not a technique but an arrangement: let the training itself take the form it is going to be used in. Then there is nothing to carry over, because you never left the target situation.
In my work there is no clinic-versus-outside gap to cross. The training is not conducted in a fixed form — no reading aloud, no memorising scripts, no repeating after a model. It is ordinary conversational speech: free, unpredictable, not even returning to the same topics. So from the first day, what is being practised is already the situation it will be used in, and there is nothing left over to transfer.
What comes after training is not more speech training either. It is a widening of range — from speaking with the trainer to speaking with an increasing variety of people. That phase has its own ceiling: conversations with strangers are held to about seven or eight a day, spaced apart, and past that number the adaptation does not hold.
So what I would say is less about how to carry it outside the clinic, and more about where the work is done in the first place.
A follow-up for anyone reading the questions here.
Two of the points raised — whether this mechanism is simply neuroplasticity and “practice makes permanent”, and how repeated practice is actually meant to carry over into real conversation — are set out at length in a preprint just published.
A follow-up for anyone reading the questions here.
Two of the points raised — whether this mechanism is simply neuroplasticity and “practice makes permanent”, and how repeated practice is actually meant to carry over into real conversation — are set out at length in a preprint just published (in Chinese, with a bilingual abstract).
It covers three things: what the repair mechanism itself is; the ceiling on repetition, and the ceiling on the capacity to repair; and what the adaptation period looks like in practice.
Liu, H. (2026). The Self-Repair Mechanism of Speech Stability: A Passive Process, Its Ceiling, and What It Looks Like in Stuttering [语言稳定性的自我修复机制——一套被动形态及其上限，基于口吃案例的观察]. Zenodo. https://doi.org/10.5281/zenodo.23391847
Hi Liu Hua, as a future SLP who is currently in graduate school, I found it interesting how you shared that a person’s stuttering may appear mild based on their speech, even though they could experience more challenges with communication in everyday situations. I was wondering, how can SLPs better identify these less noticeable challenges during an evaluation, especially when a client shows only mild signs of stuttering?
The reply I sent you has been posted below, you can take a look.
Hi, thank you — what you describe is real: a case that looks mild is not necessarily mild, and it is very easily hidden by the single reading of “sounds fluent”.
I would separate the problem into two layers.
1. There are two layers to what is hidden
One is a specific articulation problem; the other is a problem of language use.
On the face of it, the use layer is the important and difficult one — but in actual practice it largely stops being a problem. What genuinely has to be gone through one by one is the articulation layer.
2. Why the use layer can wait: define “everyday speech” first
Many people who stutter set themselves the same goal — to speak the way everyone else does. But “everyday speech” gets defined very differently from person to person. Some mean ordinary conversation in daily life; others take it to mean that particular settings and particular content must also be fluent — a speech, a performance.
Those two things have to be separated.
Everyday life is the set of situations the person can reproduce on their own initiative. Anything they cannot reproduce on their own initiative is not everyday life.
Once that line is drawn, “resolving stuttering” and “improving speaking ability” become two different goals.
And everything outside the line? It does not need to be worked on. The mind is an internal mapping of the external world — when the external world changes, the mind changes with it. That is the psychological leap: once the conditions are in place it resolves itself, and nobody has to go in and move it. A person who is afraid of water stops being afraid once they can swim. (See my separate paper, Understanding the Psychological Leap: Addressing the Psychology of Stuttering [正确认识心理”跃迁”——处理口吃心理的一个思路], DOI: 10.5281/zenodo.23336866.)
So whatever “hidden use problem” remains after the line is drawn is, by definition, already outside it — those are situations that cannot be reproduced, and they are not what is being addressed here.
3. What actually has to be looked at is the articulation layer — and only two things: range, and fluency
① Range of articulation
When you talk with the person, watch whether they are communicating through repetitive forms of expression and repetitive sounds — working with the same small set of words and constructions over and over.
Then step outside the content of what they are saying and ask in a divergent way, to find out whether they can actually produce each sound (A, B, C … one by one).
There is a trap here, and it is about language itself.
In English, one word carries one meaning, and a single profession comes with a dictionary of its own — the logic is not like Chinese, where words are assembled from characters. So in ordinary speech, anything beyond a person’s knowledge, anything they do not know, simply never comes up.
The other way round: if they can say it, that word has appeared in their everyday life — they have heard it, and they know it.
So when you test range, use the words their life actually contains. Test with dictionary entries or specialist vocabulary and what you measure is not articulation but vocabulary.
② Fluency of articulation
Articulation is the result; underneath it is muscular coordination — lungs, vocal folds, oral cavity, pharynx, nasal cavity. Repeat a movement often enough and the coordination improves, and “fluency” appears in the articulation.
Fluency is not something you either have or don’t have; it is a level. Stuttered speech has a fluency of its own — it simply sits at a different level.
So: when you find someone whose articulation sounds normal but shows no fluency, that means their articulation is the problem.
Hi Liu Hua, I enjoyed reading your article! I read that in your biography you write from the dual perspective of an independent researcher and someone who lived with stutter. I was wondering how did your own lived experience with stuttering shape your initial interest in this field of research?
Hi, thank you — this one deserves a careful answer.
It started by accident. I stuttered from childhood, and it began because a neighbour deliberately taught me to do it. The memory that stayed with me is an argument I had as a child: walking and talking at the same time, over a distance of about twenty metres, I could not get the second word out. There was a remedy going around where I grew up — hit the child on the head suddenly on a rainy day and the stutter goes away. So as a child I was hit on the head fairly often.
In high school I had two close friends. One of them, the mischievous one, started imitating the way I spoke, on purpose. After two days of imitating me, his own speech started going wrong — that is what made him stop.
I went away to university, where I had to learn Mandarin; the local dialect was not allowed, and after that I kept speaking Mandarin. For a while I believed my speech had become normal — until, some years later, I heard a recording of myself. The stutter was still there, only much milder. I asked the people around me. They had known all along; they were simply used to it, and nobody had said anything. And I had gone on believing I was fine.
Those two things taught me early to distrust two sources: my own sense of how I sound, which can be wrong; and other people’s ears, which will not volunteer what they hear. That is why I later insisted on recording everything and letting other people watch.
As for how I ended up working on this — in 2018 I had to choose a direction to work in. Two options were in front of me: the adult-products industry, or stuttering. I was too embarrassed for the first, so I chose the second. My thinking was simple: at least I would get my own stutter resolved. I had read almost nothing on the subject, so I had no idea how hard it was; if anything, the university episode had left me thinking it would be easy. The plan was simple too — three months, and then I would stop.
Halfway through, I nearly let it go several times, because by then my own stutter was resolved. But everything already done would have become a sunk cost, so I kept going. I am still going.
So if you ask how my own history shaped this work, the answer has two halves.
One half is that I spent nearly thirty years inside it. That is my qualification for attempting this.
The other half is that I came to the literature very late. When I started, the field’s default assumptions were not yet in my head — so the first thing I did was not to test what others had said, but to examine the problem through my own experience.
Thirty years inside, and only then a view from outside. That is probably the only advantage I have over people trained in the field.
Hi Liu Hau, I loved reading your article. You emphasize that speech is shaped not only by motor patterns but also by the social context in which communication occurs, including the person we are speaking to and our past experiences. How might clinicians incorporate real-world conversational interactions into therapy so that changes in a person’s motor speech pattern can actually carry over into everyday communication?
Hi, thank you — I want to answer this one carefully.
Answering it in theory would be simple. But in practice you run into a great many things that are hard to handle. So I would rather answer from what I have actually dealt with, so that you can see where the sticking points are and avoid some wasted effort.
First: define “everyday communication” before anything else.
Almost everyone’s goal is stated the same way — “to speak like a normal person”. But what counts as “everyday” varies enormously. For some it means speech in ordinary daily life; for others it means speaking well in specific settings and on specific content as well — in a speech, in a performance.
What I do is draw the line this way: situations the person can reproduce on their own initiative count as everyday communication; situations they cannot, do not.
That single cut also separates two goals: resolving the stutter, and improving speaking ability.
Without that boundary, the work has no end point. “Speaking well in a speech, too” is a target that can be pushed up indefinitely. “Resolving the stutter” is not.
Second: the medium of training cannot be a fixed form.
I don’t know whether you have come across this saying — muscle built in a gym is stiff muscle; muscle that is alive is found on people who work on building sites.
Speech is the same. The point of training speech is ordinary use. However beautiful the articulation, if it is stiff it is of no use.
So the training cannot be carried out in a fixed form — reading aloud, memorising a script, repeating after a model. It may look very good. It cannot be used.
The training has to happen inside the shape of ordinary conversation: the trainer speaks normally, and the person does the work of disassembly and reassembly on their own articulation as it goes.
In the room it looks like this. The trainer says a whole sentence; the person answers slowly, one word at a time. As it becomes more practised, it moves up to whole sentences at a time.
But I want to be clear about why this stretch is hard: it is hard on two people at once.
On the person’s side, it means taking their own articulation apart and putting it back together one word at a time — unnatural, and uncomfortable. On the trainer’s side, it means listening to an adult speak like that for the whole session.
There are people on both sides who cannot stand it. That is very rarely written down.
So this is not a matter of “having a method”. It is a stretch that both people have to get through.
(On why reading aloud cannot do this work, see: How to Practise Speaking — Why Reading Aloud and Talking to Yourself Do Not Help, and What Actually Goes Wrong in Conversation. DOI: 10.5281/zenodo.23417825)
Third: make use of the psychological leap.
The inner state is a projection of a person’s actual situation. Change the situation, and the inner state re-matches on its own, and quickly.
So the order is not: deal with “I am afraid of speaking to a certain kind of person”, and then build the ability. It is: build the ability, walk through the real situations, and that layer recedes by itself.
One thing happens often during training: the person and the trainer fall out mid-conversation — each certain they are right, neither accepting the other’s view. Very often this is not a disagreement in understanding. It is a difference in where each of them is standing.
A trainer has to recognise this. Used well, the psychological leap saves a great deal of
I’m replying to you below. There might be a little problem with the webpage on my end, so I wanted to reply under your comment.
Hi Liu Hua, thank you so much for sharing your insights from your clinical work! I especially liked how you mentioned that an adult whose stuttering was mild with more significant emotional impacts because as clinicians we often forget that patients can each experience their stutter differently. What was something else from your research that surprised you?
Hi, thank you — after long enough in this work, surprises become the ordinary, so this one is not easy to answer. If I have to pick one:
The relationship between repetition and outcome is not a straight line.
We all assume, without thinking about it, that more practice is better. It is also the most common default in stuttering work: when the speech will not come out, add force, add hours. What I have seen instead is this — within a given span of time, past a certain amount, the effect goes the other way: the pattern reverts, or blurs.
The capacity to repair has the same ceiling. Past a certain point, repair takes longer; and staying in that stretched state long enough damages the repair mechanism in turn.
So “more practice” is right inside a range, and reverses outside it.
For clinical work this means: if you meet someone who is getting worse the more they practise, it may not be a lack of effort — it may be that they are past the ceiling. And where that ceiling sits differs from person to person; there is no single standard, it can only be found one case at a time.
(I have written this up in a preprint: DOI 10.5281/zenodo.23391847)
Hi Liu Hua,
Thank you for writing such an insightful article. I was particularly interested in the idea of the “umpire trap” in therapy. In your opinion, what are some things clinicians can do to help clients provide more objective feedback? What are some ways to achieve this?
Hi, thank you — you picked out that point, and you asked it concretely.
One premise first: the umpire trap is not a problem of “judging inaccurately”. It has two components — the capacity to judge, and the willingness to bear the consequence.
A tool can only address the first. So the three things below only cover that half.
1. Give the person a mirror from outside: record it.
This is the most direct one, and it is how I was corrected myself — I believed my speech was normal until I heard a recording of it; then I asked the people around me, and they had known all along.
The important part is watching it afterwards, not watching it live. Watching live interrupts the speech, and it turns attention onto “how am I doing” — which takes you straight back into the umpire trap.
2. Write the criteria down first, so the standard cannot drift with the feeling.
“Was that better or worse than last time” — during the training period there is almost no answer to that question, because the person has no instrument. So do not ask “how do you feel it went”. Ask about things that can be observed:
After something goes wrong, does it come back on its own if they rest?
How long has normal speech been sustained? (I use three to six months.)
Is any device or medication needed?
Put the standard first, and the feeling can only speak from inside it.
3. Turn “normal” into a range.
Many people get stuck on “I still feel there is something not quite right”. But normal is not 100%; it is 98 or 99, approaching 100 without reaching it. Inside that range, it counts as normal.
This is not reassurance — it turns “I am not completely recovered” from a conclusion into a statement that can be judged. Once there is a range, the question “am I there yet” has an answer.
One last thing, which is what I most want to say:
All three give the person an external reference point — and none of them makes them objective.
You can hand someone a mirror, but you cannot look into it for them. And whether they are willing to use that mirror is the second component — a tool cannot reach it.
Hi Liu Hua,
Thank you for sharing your article! It was really interesting! You discuss how the brain may return to old stuttering patterns even after fluency imrpoves in therapy. How can clinicians help clients maintain new speech patterns outside therapy, and especially in stressful speaking situations.
Hi, thank you — I will answer this in two parts.
First, a qualification: in my work the phrase “outside therapy” does not really hold.
Because the training is itself conducted inside ordinary conversation — no reading aloud, no memorised scripts, no repeating after a model; it is done in real dialogue. So there is no gap between “learned in the clinic” and “used outside it”; there was nothing that needed transferring in the first place. (I have answered this point elsewhere, so I will not repeat it here.)
The second part is the point: in situations under pressure, the problem is not “maintaining” — it is quantity.
The adaptation period is not “the more contact, the more adapted you become”. It has a ceiling: conversations with strangers are generally held to seven or eight a day, spaced about ten minutes apart. Past that, the adaptation does not hold.
What is special about a stressful situation is this — pressure is not one quantity; it is several quantities stacked together: number of people (one-to-one / one-to-many), interlocutor (familiar / unfamiliar), content (familiar / unfamiliar), standing (equal / unequal).
So when you add pressure, add only one thing at a time. Add all four at once and it is equivalent to a day’s worth of dozens — that is not adaptation, that is overloading.
And more dangerous than quantity is this: the most common failure under pressure is not speaking badly. It is wanting to speak well.
Under tension a person instinctively starts managing their own speech. And the direction of that management is calibrated to the old pattern. So the usual outcome under pressure is not “he failed to maintain it” — it is that he walked his own speech back.
But there is a further layer, and it matters more than any of the above: the old set cannot be removed.
The old set is, in essence, a degradation of speech function in a particular direction. It is the same repair mechanism, and its object of protection has not finished changing over — during the transition the old one is still pulling, and the new one has only just taken hold. So the aim was never “to get rid of the old one”. It is to let it degrade to the point where it no longer interferes with normal speech.
An analogy: someone stops speaking their home dialect in adulthood; decades later they can still switch back to it, only less fluently. The dialect has not disappeared — it has receded to an unused position. That is exactly what I aim for with the old speech pattern: it is still there, but it does not interfere.
This changes what counts as failure clinically:
Slipping back into the old pattern under pressure is not failure. It only shows that the old one is still there, and has not yet receded far enough. The real criterion is not “does he slip back” — it is “after slipping back, can he come back on his own”.
Thanks for sharing your research. I’m always intrigued by the use of the word NORMAL. What is normal? Is it fluency? Or is normal the way most people speak? As no one has 100% fluency, no one speaks without afiller word, stumbles over a word, has a block, etc. So, doesn’t that make stuttering normal, as we too sometimes speak fluently and sometimes not? How do we, and can we even, define “normal”? And do we even want to be like everyone else, or worse, perfect? I know it’s a rethorical question, but it’s an interesting one. 😉
Keep talking
Anita
Hi Anita, thank you — this is the question I get asked most often, and the one I most want to answer.
First: you are right.
The standard I use for “normal” was never 100% to begin with. What I write is 98, 99 — approaching 100 without reaching it. Inside that range, it counts as normal. Because nobody is at 100%. That number does not exist.
But there is one place in your question where I would change the wording.
Whether there is a block is not the dividing line.
People who don’t stutter also stumble once inside a sentence, also say “um”, also start again halfway through. The difference is not whether the thing happens. It is what happens after it:
Someone who doesn’t stutter: blocks for a moment, then goes on. The block is over.
Someone who stutters: blocks for a moment, and then starts handling it — pushing harder, swapping in another word, going around it, stopping and starting again. And the handling itself makes the next one worse.
So I don’t treat the block as the problem. I treat the handling as the problem.
Second, on “do we even want to be like everyone else”.
My criteria don’t include “sounds like most people”. I use a different ruler:
After something goes wrong, does it come back on its own if he rests?
That doesn’t need the concept of “normal”, and it doesn’t need comparison with anyone. It uses him, against his own next time.
And as for “perfect” — I don’t ask for perfect. What I ask is that it doesn’t have to be sustained by input.
If someone has to practise five hours a day to buy a window of normal speech, that looks good in the data — but it is another kind of dependence.
So I don’t ask “does he sound like most people”. I ask “what does it take to keep it”.
One last thing: I don’t think your question is rhetorical — I think it is the question. If “normal” cannot be defined, then the judgement has to be replaced with something else.
What I replaced it with is: whether it comes back on its own.
(The five criteria are set out in full here: 10.5281/zenodo.23335647)