Darbe Schlosser, MA
Founder, Think & Move Well
A practice-based, hypothesis-generating report from a Parkinson's movement-coaching practice.
Cueing works for freezing of gait. The difficulty is what happens when the cue goes away.
The strongest trial in this literature taught people to use cues at home, measured real gains, and then measured those gains deteriorating once the cueing stopped. Its authors concluded that what the field needs is permanent cueing devices and continuing treatment.
That a self-generated cue can work is no longer in question. Work from Washington University in St. Louis compared self-generated against external cues directly, and self-generated cues came out ahead — in people who freeze as well as those who do not.
The question this article takes up is a different one: whether a person can reach that cue when reaching it is hard. A cue produced on request, in a laboratory, walking a marked course, is not the same act as producing one mid-episode in a supermarket aisle, eight months after anyone last coached it. This practice treats that retrieval as the thing being trained — not the cue itself, but reliable access to it under load. The components below are how it is built: a countable target set before moving, and the voice synchronized to each heel strike, rehearsed in calm conditions precisely so they are reachable in difficult ones.
This article reports no outcomes. Nothing here measures whether self-generated cueing reduces freezing, and no comparison with device-based cueing was made. What follows is a description of what is taught, how often, and a set of falsifiable predictions.
The evidence types below are labeled wherever they are used: published research · Think & Move Well methodology · practice observation · proposed mechanism (hypotheses, explicitly not findings).
Published research.
Freezing of gait is episodic, disabling, and only partly responsive to medication. Three findings frame the problem this coaching addresses.
Cueing helps, modestly, and then stops helping. In a randomized trial of home-based cueing training, gait improved by a small but significant margin: gait speed rose about 5 cm/s and step length about 4 cm, and among freezers specifically, freezing severity fell by roughly 5.5%. Balance confidence improved. But there was no carry-over to daily activities, quality of life, or fall rate, and at six-week follow-up the gait and freezing measures had all deteriorated significantly. The authors' own conclusion was that the decline underscores the need for permanent cueing devices and follow-up treatment (tmw-1295).
Freezing involves the loss of automaticity, and the compensation is expensive. Freezers show impairment in both executing and acquiring automatic movement, alongside deficits in inhibition and set-shifting. The proposed model is that a movement can be handled either by an automatic route or by a controlled, effortful one; non-freezers compensate for lost automaticity by shifting to cognitive control, but in freezers both routes are impaired, so that compensation overloads cognitive resources and the breakdown is the freezing episode itself (tmw-1296).
Anxiety is a cause, not just a consequence. When anxiety was experimentally raised rather than merely observed, freezers spent significantly more of each trial frozen and had more episodes, and step-to-step variability rose under threat. The authors conclude that anxiety causes freezing rather than simply accompanying it (tmw-1294).
Published research.
That external cueing produces measurable, if modest, gait benefit in Parkinson's, including a reduction in freezing severity among freezers (tmw-1295). That those benefits fade once cueing is withdrawn, with some limited uncued carry-over suggesting a small amount of motor learning is retained (tmw-1295). That freezing is associated with de-automatization and executive-control deficits (tmw-1296). That dual-task load and turning provoke freezing (tmw-1300). That anxiety causally increases it (tmw-1294). That freezing is best characterized as an episodic phenomenon rather than a continuous deficit (tmw-1301), and that stride-to-stride variability is dysregulated in people who freeze (tmw-1298).
And, importantly for what follows, that self-generated cues have already been tested against external ones. A body of work from Washington University in St. Louis has compared the two directly. Internal cueing improved gait velocity, cadence and stride length and reduced variability, while external cueing produced minimal improvement and a decline in gait stability (tmw-1350). Singing at a tempo matched to comfortable walking pace improved gait variability without cost to velocity, cadence or stride length — and, in the same study, a generic verbal dual task slowed and destabilized gait while the singing condition did not (tmw-1351). Covert mental singing reduced variability more than listening to music, and more than singing aloud (tmw-1352).
Most directly of all: people with Parkinson's with and without freezing of gait were both able to use self-generated cues, and while external cues increased gait variability, self-generated cues did not (tmw-1349). Cue response is not universal, however. In a comparable paradigm, 23 of 35 participants responded and 12 did not (tmw-1353).
Published research. Placed before the method deliberately.
Correction, 2026-08-02. An earlier version of this article stated that no published study tested this approach and that no trial had compared self-generated against externally supplied cueing for freezing of gait. Both statements were wrong, and the work described in the preceding section had been missed by our searches. The section below states what is genuinely unresolved. We are grateful to have the error pointed out and have left this note rather than quietly amending the text.
Durability after training stops has not been tested for self-generated cues. The comparisons above measured gait while the cue was in use. RESCUE measured what happens after external cueing is withdrawn and found the gains deteriorated (tmw-1295). No equivalent withdrawal-and- follow-up study exists for a self-generated cue. This, not novelty, is the open question, and it is the entire premise of what follows.
Access under real-world stress has not been tested. The published work establishes that people can use a self-generated cue in a laboratory walking task, on request. It does not establish that someone can retrieve one in a supermarket aisle, mid-episode, under threat — which is when it would matter, and which anxiety findings suggest is the hardest case (tmw-1294).
Coaching someone to construct and train their own cue has not been studied as a taught skill. In the published studies the cue is supplied to the participant: a song, a tempo, an instruction to sing. Whether the capacity to generate and deploy one can be trained, and what that training consists of, is not addressed.
Non-response is real and uncharacterized. Roughly a third of participants did not respond to cueing in a comparable paradigm (tmw-1353). Nothing identifies who they are in advance.
One published finding cuts against the specific method described here. Covert mental singing outperformed singing aloud (tmw-1352), whereas the coaching described below couples an audible voice to heel strike. Our rationale is proprioceptive-vocal coupling rather than rhythm alone, but that is a rationale, not a result, and the published comparison currently favours the covert version.
The mechanism literature still contains an argument against this approach. If freezers have impairment in both the automatic and the controlled route, and if the cognitive compensation is itself what overloads and produces the episode (tmw-1296), then asking someone to generate and hold their own cue adds load to the very system that is failing. Dual-task load provokes freezing (tmw-1300). This objection is weakened but not removed by the finding that a structured vocal cue did not destabilize gait where a generic verbal dual task did (tmw-1351).
Think & Move Well methodology.
The approach described here is one component of a broader coaching system, and does not represent the program's full movement, cueing, or exercise-design methodology. A companion article, Counting as Coaching, describes a different component of the same system: how a memorized digit sequence structures movement practice.
Its organizing idea is that the cue must be reachable, and that reachability is a trained skill rather than a property of the cue. Conventional cueing supplies timing from outside the person: a metronome, a line on the floor, a therapist's voice. A device left at home helps nobody in a supermarket aisle — but neither does an internal cue the person cannot summon once their attention is captured by the freeze itself. The coaching described here treats construction and ownership as the means: a cue you built yourself is one you can rebuild under pressure. What is trained is the rebuilding.
Three components recur, and they are taught as things to do before and during a freeze rather than as exercises performed separately from it.
A countable target, set before moving. The client scans the environment, selects a destination, and commits to reaching it in a stated number of steps. The number converts an open-ended and often frightening transition into a bounded task with a defined endpoint.
The voice synchronized to heel strike. Rather than following an external beat, the client produces their own vocal rhythm and synchronizes it to each heel strike. Attention is deliberately anchored there: during ambulation, heel strike is the timing event that cues the vocal output. The voice is not trained as an independent rhythm that happens to coincide with the step — it is synchronized to each heel strike, so the timing signal is generated internally and anchored to proprioceptive feedback rather than to an external sound. The stated intent is to promote temporal consistency and symmetry in gait. That is the design rationale. No effect on either was measured.
Trained retrieval under increasing difficulty. The first two components are what is retrieved; this is the training of the retrieval itself, and it is the component this article treats as decisive. Strategies are first rehearsed when the person is calm and not frozen, then practiced under conditions that progressively resemble the ones in which they will be needed — divided attention, unfamiliar environments, the medication off-state. The intent is that the strategy remains reachable as conditions degrade. Whether it does is the paper's central untested claim, and no outcome is reported here.
What the method actually trains. The components above describe what is produced. They do not, on their own, distinguish this from asking someone to count their steps, which is common practice. The difference is in the requirements the task imposes, each of which is separately coachable and separately failable: intentional control of attention, so the cue is deliberately placed rather than merely present; accurate targeting and timing of the voice, so it lands on the intended event rather than near it; synchronization of vocal output to movement, described above; respect for the inverse relationship between spatial and temporal parameters, because a faster count and a longer step trade against each other; and consequently pacing the count so that sufficient spatial displacement can occur before the foot lands, rather than allowing the count to drive the step short. A client counting quickly and stepping shortly is performing the task incorrectly even though every element is present. The sixth requirement — maintaining all of this as conditions get harder — is the retrieval training described above. These requirements are stated here as method. Their full development, and the evidence for each, is the subject of a companion article in preparation.
What distinguishes this from a device is not the rhythm, and not only the ownership — others have already shown a self-generated cue can outperform an external one while both are in use. It is the claim that retrieval can be trained, so the client can produce the cue in a doorway, with nothing in their hands, on a day nobody is coaching them. That is the design intent. It is not a claim that it works.
Practice observation. Uncontrolled, and not an outcome.
A note on how to read the numbers first. Clients differ widely in how many sessions they attend, so a client share and a session share are not interchangeable: a component reaching half of clients may appear in a fifth of sessions. Percentages in this article are client shares throughout, with session counts given alongside them.
Freezing is discussed with 228 of the 299 clients in this corpus (76%), across 596 of the 1,189 de-identified sessions. Self-generated cueing runs throughout that rather than sitting alongside it: freezing and a self-generated cue appear in the same passage for 117 clients (39%), across 174 sessions and 268 passages. Passage-level co-occurrence was used deliberately rather than session-level, because two topics appearing forty minutes apart in the same call is not evidence that they are linked.
The individual components are similarly widespread. A countable step target appears with 142 clients (47%), across 203 sessions. Voice coupling appears with 161 clients (54%), across 265 sessions.
For ambulation tasks, voice coupling was operationalized as synchronized vocalization to each heel strike, and heel strike is therefore the retrieval marker used to measure it. This is an operational definition rather than an adjudicated classification: sessions discussing heel strike for other reasons are not distinguished from those teaching the coupling.
One methodological note that shaped these figures. The phrases the literature would use, "self-cueing" and "internal cue," appear with only 25 clients (8%), in 26 sessions. Measured by terminology, this method would look marginal. Measured by what actually happens in sessions, it reaches roughly half of clients. The counts above measure the mechanism, not the label, and a reader should treat them as retrieval counts from stated queries rather than as adjudicated classifications.
Coaches ask clients directly whether the strategies are being used outside sessions, and clients report using them selectively, typically when they get stuck rather than as a constant practice. Coaches also check whether the added cognitive task is interfering with walking rather than assuming it is not. None of this is outcome data. It is what the coaching records contain.
Proposed mechanism. None has been tested here; each is stated with what would falsify it.
Hypothesis 1: Ownership changes durability, not only immediate performance. A self-generated advantage during use has already been reported (tmw-1350, tmw-1349). The untested prediction is that a further difference appears after withdrawal, because there is nothing to withdraw. Falsified if self-cued gait deteriorates on the same timescale as device-cued gait after training stops.
Hypothesis 2: A bounded target reduces the anxiety component. Given that anxiety causally increases freezing (tmw-1294), converting an open transition into a countable task with a visible endpoint may reduce threat rather than add load. Falsified if measured anxiety or step-length variability is unchanged or higher when a step target is used.
Hypothesis 3: Anchoring to proprioceptive feedback differs from following a sound. Coupling the voice to a felt event in the gait cycle may recruit different resources than matching an external beat. Falsified if the two produce equivalent gait and equivalent dual-task cost.
Hypothesis 4: The load objection is structure- and dose-dependent rather than fatal. The mechanism literature predicts that adding a cognitive task should worsen freezing (tmw-1296, tmw-1300). There is already evidence against the strong form: a generic verbal dual task slowed and destabilized gait while a rhythmically coupled singing task did not (tmw-1351). The remaining proposal is that a rehearsed, highly practiced self-cue becomes cheap to run, whereas a novel one is expensive. Falsified if dual-task cost fails to fall with practice, or if freezing frequency rises when the strategy is used.
Hypothesis 5: Retrieval under load is trainable, and is what the training is for. The prediction is not about the cue's effect while in use, which is already reported (tmw-1350, tmw-1349). It is that the probability and fidelity of retrieving a self-generated cue — initiating it unprompted, at the right moment, with timing intact — degrades under environmental and cognitive load, and that this degradation reduces with graded practice. Falsified if retrieval latency, initiation rate or timing fidelity under induced load fail to improve with training, or if trained retrieval does not transfer to conditions that were not practiced.
The comparison this calls for is a randomized trial of self-generated cueing against device-based cueing, with the primary endpoint measured after withdrawal rather than during use. Both arms would train for a matched period; both would be assessed cued and uncued at baseline, at the end of training, and at follow-up far enough out to capture the fade the RESCUE trial documented at six weeks.
Primary outcome: freezing frequency and duration uncued at follow-up. A second arm should test Hypothesis 5 directly, since durability and retrievability are separable: measure unprompted initiation rate, retrieval latency and timing fidelity under graded load — divided attention, unfamiliar environment, time pressure — at baseline, after training, and at follow-up, including in at least one condition never practiced. Secondary: dual-task cost during self-cueing and whether it falls with practice, which tests Hypothesis 4 directly; state anxiety and step-length variability during transitions, which tests Hypothesis 2; and adherence, reported for all randomized participants rather than completers only.
A negative result would be informative and is entirely plausible. If self-cued gait fades the same way device-cued gait does, the ownership premise is wrong and the honest conclusion is that permanent cueing support is what people need, exactly as the RESCUE authors argued.
There is no control group, no randomization, no blinding, no standardized instrument, and no outcome measurement of any kind. Coaching records were made for coaching, not research.
Retrieval counts are not diagnoses. A passage containing both freezing and a self-generated cue has not been adjudicated by a second reader, and the co-occurrence window was chosen by us. An early version of these searches counted video-call freezing alongside gait freezing, which we removed once identified; other false positives may remain.
Sessions were selected because they contained the teaching, which is a form of selection bias. This article can describe what is taught, not how often it fails to take, and it says nothing about clients who stopped attending.
All observations come from the practice that developed the method, interpreted by its developer. Client speech is paraphrased rather than quoted, because the underlying sessions have not individually completed de-identification review.
Commentary from Think & Move Well. This section is opinion, not a finding.
The cueing literature has largely asked which cue works best. We think the more useful question for someone living with freezing is who holds it. A metronome in a drawer and a laser cane in a cupboard are both, functionally, cues that were never there when it mattered. We are aware that the mechanism literature can be read as predicting our approach should fail, and we would rather state that plainly than argue around it. We are also aware that we initially claimed more originality than the record supports: others tested self-generated cueing before us, in freezers, and found in its favour. What we can defend today is specific: the approach exists, it is defined, it is taught to roughly half of the clients in this practice, and it produces predictions specific enough to be wrong. Whether it holds up is a question for someone with a control group.
Think & Move Well provides movement coaching and education. It does not provide medical advice, diagnosis or treatment. Nothing here is a recommendation to start, stop or change any medication, therapy or treatment. Freezing of gait carries a real risk of falling. Strategies that add a mental task while walking can increase difficulty in the moment and should be introduced gradually, in a safe environment, and with appropriate supervision. Consult your neurologist, movement disorder specialist or physical therapist before changing what you do.
Corrections policy. We correct errors. If you find one in this article, including in how we have represented any cited study, email darbeschlosser@motorvationusa.com with the section and what is wrong. We will review it, correct the page if the error is confirmed, and add a dated note recording what changed.