The fellow was treating a patient with a complex cervicogenic headache — three months of unilateral temporal pain, intermittent dizziness, a history of two prior cervical spine episodes that had resolved with manual therapy. The evaluation findings were consistent with upper cervical segmental dysfunction. The manual therapy approach was technically precise. By the fourth visit, the headache frequency had decreased by roughly half. By the sixth visit, it had plateaued.
The fellow adjusted. Added thoracic mobilization. Progressed the motor control exercises. Revisited the home program. The adjustments were logical. Each one addressed a real impairment. But the headache frequency held at its plateau, and the dizziness — which the fellow had been treating as a secondary cervicogenic symptom — had not changed at all.
The mentor was sitting in the corner of the treatment area, reviewing documentation on another case. Watching.
The mentor could see the pattern forming. The dizziness did not correlate with the cervical interventions. It was present on some mornings regardless of the previous day’s treatment. The patient had mentioned a medication change six weeks ago — a new antihypertensive — in a sentence the fellow had noted on the intake form and not revisited. The vestibular screening at evaluation was negative, but it tested for peripheral vestibular dysfunction. It had not probed for medication-related dizziness or for the kind of postural blood pressure variation that a new antihypertensive could produce.
The mentor’s instinct was to walk over and point it out. Name the medication. Suggest a blood pressure check. Redirect the clinical reasoning toward the pharmacological contributor that the mechanical framework was not capturing. In the residency, that would have been the right move. At the residency level, the mentor’s role is to make the invisible visible — to correct, to redirect, to fill the gap between what the clinician sees and what the clinical picture contains. A mentor who corrects efficiently produces a clinician who steadily accumulates knowledge.
At the fellowship level, pointing it out is often the wrong move.
Not because the fellow does not need the information. The way the fellow acquires information determines the kind of clinician the fellow becomes.
The mentor who tells the fellow what they missed produces a clinician who knows one more thing. The mentor who asks the right question at the right moment produces a clinician who thinks differently (Jensen et al., 2000). The first adds information. The second changes the reasoning process. The distinction is not about pedagogical style. It is about what the training is actually for.
In the residency, the developmental target is routine expertise — efficient pattern recognition, organized knowledge structures, reliable application of evidence-based frameworks within established clinical presentations. The mentor at the residency level models reasoning, names decision points, and makes the implicit explicit. This is appropriate and necessary. The resident does not yet have the organized framework necessary for self-correction. The scaffolding must be present and visible.
In the fellowship, the developmental target is different. The fellow already has the framework. The fellow can manage the straightforward case independently and the moderately complex case with indirect supervision. The fellowship is developing something the framework alone cannot produce — the capacity to recognize when the framework is insufficient and to generate a different approach. That capacity is not built by being told what is wrong. It is built by sitting with the discomfort of an incomplete explanation until the discomfort itself becomes a clinical signal.
The NAIOMT Fellowship EPA Guide describes a progression from “mentor explicitly names the clinical reasoning steps and decisions being made” through increasingly independent levels of practice to “mentor provides feedback based on the fellow’s written summary.” The progression is a deliberate reduction of scaffolding — not because the scaffolding was unnecessary, but because the scaffolding has done its work. The fellow must now reason without it. Every time the mentor points out what the fellow missed, the scaffolding goes back up.
So the mentor waited. And then asked: “The headache is responding. The dizziness is not. What does that tell you about your hypothesis?”
The fellow paused. Began to reason out loud. If the headache was responding to the cervical interventions, the cervicogenic component was likely real and being addressed. If the dizziness was not responding, either it was a separate mechanism or it had a contributor that the current framework was not addressing. The fellow looked at the chart. Saw the medication change. Began to consider whether the dizziness had ever been cervicogenic at all, or whether the assumption that it was had shaped six visits of treatment without being tested.
The fellow did not need the information. The fellow needed a reason to look for it.
Something else happens in this process — something the literature on mentoring has only recently begun to examine. The mentor who regularly asks fellows to articulate their reasoning begins to articulate their own more precisely. The mentor who pushes fellows to identify their assumptions begins to catch their own. The mentoring process is a form of deliberate practice in adaptive reasoning for the mentor as much as for the fellow.
This is counterintuitive. The assumption has always been that mentoring flows in one direction — from expertise to development, from the experienced clinician to the developing one. But a 2018 study on online clinical mentoring found that mentors reported improved clinical decision-making as a result of the mentoring process itself, not despite it (Westervelt et al., 2018). The act of formulating the right question — the one that dismantles an assumption without replacing it — requires the mentor to reason about the fellow’s reasoning, which is a level of cognitive work that sharpens the mentor’s own clinical thinking.
A scoping review of mentor training programs in medicine found that mentor preparation remains inconsistent and poorly evaluated (Sheri et al., 2018). Another review of mentorship in residency found that most programs rely on satisfaction surveys at a single time point as their primary evaluation strategy (Joe et al., 2023). The profession invests in the structure of mentoring — the hours, the requirements, the scheduling — without investing in the quality of the mentoring relationship itself. After thirty years of supervision research, there is no robust empirical evidence demonstrating that supervision directly improves patient outcomes — the supervisor remains what one review called “the largely unknown party in the supervision experience” (Watkins, 2012).
Medicare reimbursement, adjusted for inflation, is roughly half of what it was thirty years ago (MedPAC, 2023). The economic pressure that shapes every clinical minute also shapes every mentoring minute. The mentor who waits rather than tells is spending time that the productivity model does not account for. The mentor who formulates a question rather than providing an answer is doing cognitive work that the system does not recognize. And the bidirectional development that quality mentoring produces — the sharpening of the mentor’s own reasoning through the act of cultivating someone else’s — is invisible to every metric the profession currently tracks.
Fewer than 12% of DPT graduates pursue residency training (Osborne et al., 2019). The fellowship pathway narrows further still. The mentoring that distinguishes correction from provocation — the kind that changes reasoning rather than adding facts — exists inside a structure that serves a fraction of the profession. The rest learn by accumulation, by trial, by the slow accretion of experience that may or may not organize itself into the adaptive capacity the patient in the treatment room actually needs.
The fellow finished the session with a different plan. Kept the cervical manual therapy — the headache was responding, and the cervicogenic component was real. But added a seated-to-standing blood pressure check. Asked the patient to track the timing of the dizziness relative to the new medication. Scheduled a conversation with the referring provider about the antihypertensive. The dizziness was no longer being treated as a cervical problem. It was being investigated as its own question.
The mentor watched from across the room. The wrong move — pointing it out, correcting efficiently, adding one more piece of knowledge — would have been faster. It would have solved this case. It would not have changed the clinician.
