A client leaves session having genuinely understood the intervention. They agreed with the rationale, could restate it back accurately, seemed engaged. The following week, the homework was not done — not from resistance, usually, but because the week happened and the intervention simply did not survive it.
This is one of the most consistent and least discussed frictions in talk-based practice. Session-level insight is comparatively easy to produce. Between-session behavior change is not, and the gap between the two is large enough that homework non-adherence rates in the literature are frequently reported in the 20 to 60 percent range depending on modality and population, even among clients who report agreeing with the intervention.
For practitioners, the useful question is not why clients "resist." It is why understanding and behavior operate on different tracks — and what the adherence literature says narrows that gap.
Two systems, not one failure of will
The general finding across cognitive-behavioral homework research is that comprehension of a task and completion of a task are only loosely correlated. A client can understand a thought-record exercise perfectly and still fail to complete it at home, not because the understanding was shallow, but because completing a task outside the session draws on a different set of processes — cue-triggered initiation, competing demands, environmental friction — that comprehension alone does not touch.
The clinical error is treating non-adherence as evidence the client didn't really understand or didn't really want it. The more consistent finding across the literature is that understanding and doing are supported by different mechanisms, and strengthening one does not reliably strengthen the other.
This maps onto a broader distinction worth making explicit with clients themselves: insight and internalized change are not the same event, and clients who can articulate an idea with total clarity often report that it still does not feel true or automatic — a pattern also visible in clients who know something logically without feeling it.
What the adherence literature says actually moves the needle
A few findings recur across the homework-adherence and implementation research worth building into practice design.
Specificity of the task predicts completion better than the client's stated motivation. Vague homework ("notice your thoughts this week") shows lower completion than homework with a specified cue, action, and moment ("when you sit down at your desk Monday morning, write one thought before opening email"). This is consistent with implementation-intention research more broadly — pre-specifying the situational trigger for an action reliably increases follow-through compared with a general intention to act.
In-session rehearsal of the homework, even briefly, predicts higher completion than assigning it and moving on. Practicing the first instance of the task inside the session appears to reduce the activation cost of doing it alone later.
Immediate self-monitoring outperforms delayed recall. Clients asked to record a target behavior at the moment it occurs show meaningfully higher and more accurate adherence than clients asked to reconstruct the week from memory at the next session — largely because retrospective recall of one's own behavior over days is unreliable, and because the friction of "I'll remember it later" is where most homework quietly dies.
Why this matters beyond the homework itself
The deeper implication is that between-session change is less a test of a client's commitment and more a design problem: how is the evidence of the target behavior being captured, and how close to real time is that capture happening?
Clients frequently self-report as "not doing the work" when what actually happened is that they did fragments of the work and lost track of them by the next session — the fragments were real but never recorded, so neither the client nor the practitioner has anything to point to. This produces a discouraging narrative ("I never follow through") built on an incomplete record rather than an accurate one, and it compounds the internalization gap it was meant to close.
Where a low-friction capture layer fits
This is the specific problem MyDopa's daily-capture design addresses, and it maps directly onto the adherence findings above: immediate, low-friction, specifically-cued recording of a target behavior, in the moment, rather than reconstructed later from memory.
Used as an adjunct between sessions, it gives clients a place to log the homework instance the moment it happens — closing the gap between doing the thing and losing track of having done it. It also gives the practitioner an accurate record to work from at the next session, rather than a client's best reconstruction of a week that has already partly evaporated. It does not replace the clinical relationship or the intervention design; it addresses the specific mechanical failure point the adherence literature keeps identifying — that the record of change is usually the missing piece, not the client's willingness to change.
A practical adjustment worth testing
Practitioners who want to test this directly can start small: take one existing homework task, specify the exact cue and moment for it the way implementation-intention research suggests, and ask the client to record the instance immediately rather than reporting from memory next session. The adherence literature suggests this single change — specificity plus immediate capture — accounts for more variance in follow-through than most of what gets attributed to client motivation.