26/04/2026
Contaminating the Decision: Why Stepping In After a Change of Behavior Costs You More Than the Find
A change of behavior is not a source. It's the visible signature of an unfinished computation.
When the dog throws a COB, a head snap, a check-back, a deceleration, a re-cast, what we're watching is the dog updating a probability estimate inside a turbulent odor field. The plume is intermittent. The boundary layer is doing things the dog can feel that we will never see. Pressure gradients are pulling odor in directions that don't match where the source actually sits. The dog is mid-Bayesian-update, weighting current sensory data against the prior it built two seconds ago. Then the handler steps in.
What "stepping in too soon" actually does
It looks helpful. It feels like teamwork. Mechanically, it's an interruption of an ongoing operant chain at the worst possible moment, the moment immediately following a prediction error, when the dog's nervous system is most plastic and most ready to encode whatever happens next.
What gets encoded? Not the odor problem. The handler.
The dog learns that COB → handler movement → reorientation → reward proximity. The contingency the dog forms is between their behavior and your behavior, not between their behavior and the odor. You've inserted yourself into the their decision making. You are now part of the cue stack. And every time the problem gets hard going forward, the dog will check that cue stack before working the air.
This is how dogs become handler-dependent without anyone meaning for it to happen. No one taught it deliberately. It was taught one premature step at a time.
The fluid dynamics problem the dog is actually solving
When odor leaves a source, it doesn't travel in the straight lines our diagrams pretend. It moves through a turbulent medium governed by the Péclet number, the ratio of advection to diffusion. In most operational environments, advection dominates. That means odor arrives in intermittent packets, separated by gaps, with concentration gradients that lie about source direction more often than they tell the truth.
The dog's job in that moment of COB is to integrate across multiple packets, weight the recent ones more heavily, and decide whether the current hypothesis ("source is that way") still holds or needs revision. That integration takes time. Sometimes it takes a cast. Sometimes it takes a step back and a recommitment. Sometimes it takes a head shift that looks to us like nothing.
It is not nothing. It is the dog finishing the math.
When stepping in is correct
Stepping in is a tool. Like any tool, the question isn't whether to use it, it's when.
Step in when the dog has clearly exhausted the available information and is now searching the same space with no new data coming in. Looping without updating. Casting without committing. The signal is repetition without resolution, the dog is no longer integrating new packets, just rehearsing the old ones.
Step in when the environmental geometry is genuinely beyond what the dog can solve from their current position. A sealed container blocking odor flow. A pressure pattern pulling the plume into a space the dog cannot legally or physically access. The dog's nose is correct; the architecture is the obstacle.
Step in when fatigue or stress is degrading the dog's processing, not the odor picture. You're not solving the problem for them. You're resetting the conditions under which they can solve it.
Do not step in to confirm a COB. Do not step in because the silence is uncomfortable. Do not step in because a clock in your head says it's been long enough. Those are handler problems disguised as dog problems.
The discipline
Reading the dog is the easy half. Trusting what you read, long enough that the dog gets to finish their own sentence, is the half that takes years.
The dog is doing physics you cannot do. Your job is to protect the conditions under which they can keep doing it, and to recognize the narrow window where your intervention adds information rather than replacing it.
Everything else is contamination.
No matter how you train, know why.
