Science Into Practice: When a Trained Behavior Looks Like a Reflex

Watch a highly experienced detection dog work a vehicle.

The dog moves through the search pattern, hits the wheel well, and the final response is there — immediate, clean, committed. No deliberation. No visible hesitation. It looks automatic. It looks like the odor triggered the behavior the way a tap on the knee triggers a leg kick.

It looks reflexive. Fluent in every situation. Sort of like blinking.

How you explain a behavior determines how you train it, how you maintain it, and how you troubleshoot it.

You know I'm big on terminology. So here we go. Some language housekeeping — because this matters.

Most people learned the phrase "classical conditioning."

Behavior analysis replaced it with a more accurate term: respondent conditioning. The behavior responds to a preceding stimulus that elicits it. The name describes the mechanism. That's a more useful framework, and it's the language you'll see in contemporary behavior science.

Skinner, B.F. (1938). The Behavior of Organisms. Appleton-Century-Crofts. — Skinner introduced the respondent/operant distinction as a functional classification based on what controls behavior.

So what is respondent conditioning, and how is it different from operant?

Respondent conditioning is what Pavlov described. A neutral stimulus gets paired repeatedly with one that already produces a biological response — and eventually the neutral stimulus produces that response on its own. The dog salivates when it hears the bell, not just when food appears. The key feature is that the organism is a passive participant. The behavior is elicited — pulled out by a preceding stimulus. It's involuntary.

Consequences don't change it.

You cannot reinforce or extinguish a salivation response, a blink due to an eye puff, or a knee jerk reaction to a knock on your knee. It doesn't work that way.

Operant conditioning is different at the root level. The organism emits behavior and the consequences of that behavior determine whether it happens again, more often, or less often. The behavior is voluntary. The behavior is sensitive to what comes after it.

This is the distinction that matters: elicited versus emitted. Controlled by what comes before versus controlled by what comes after.

Cooper, J.O., Heron, T.E., & Heward, W.L. (2020). Applied Behavior Analysis (3rd ed.). Pearson. — The definitive text on the functional distinction between respondent and operant behavior classes.

Pavlov, I.P. (1927). Conditioned Reflexes. Oxford University Press.

Now here's where it gets complicated in practice.

When an operant behavior becomes extremely fluent — trained to the point where there is no latency, and the behavior is fast, constant, and precise — it starts to look respondent. The behavior appears immediately after the antecedent, as if elicited. And trainers, handlers, and even experienced practitioners start describing it as classically conditioned.

It isn't.

A highly fluent operant is still operant. The controlling variable hasn't changed. What's changed is the dog's reinforcement history with that behavior in that context. The behavior is fast and consistent because it has been reinforced extensively under clear stimulus conditions. That looks like a reflex. It isn't one.

This matters enormously in detection work.

When a dog's final response looks completely automatic — when the odor seems to pull the behavior out of the dog — it's easy to conclude that the behavior is classically conditioned (aka reflexive). That the consequences no longer matter. We seem to think that reinforcement does not maintain the behavior at this point.

If the behavior is operant — and it is — then it remains sensitive to consequences.

Reinforce it, and it stays strong. Let the reinforcement lapse, and the behavior will change. The quality will shift. The dog's propensity to respond correctly will change. You may not see it immediately. You will eventually.

Intermittent schedules of reinforcement are still reinforcement. And they will maintain the behavior and even make it stronger when applied correctly.

Lindsay, S.R. (2000). Handbook of Applied Dog Behavior and Training, Vol. 1: Adaptation and Learning. Iowa State University Press. — Addresses the distinction between conditioned reflexes and trained operant behaviors in applied canine contexts.

Here's a diagnostic test worth keeping in your back pocket.

If you want to know whether a behavior is truly respondent or a fluent operant, ask: does changing the consequence change the behavior?

With a truly respondent behavior — a salivation response, a reflexive response like a blink — consequences don't touch it.

With a fluent operant, consequences matter. Put the behavior on extinction and it will change. Change the reinforcer and the behavior will reflect that, eventually. Shift the context and you may see the behavior weaken in ways that feel surprising — but shouldn't, because context is part of the discriminative stimulus picture that controls operant behavior.

This is also why the Clever Hans research on detection dogs is so important. The handler's subtle cues can influence whether the dog offers a final response. That influence exists because the final response is operant. You cannot inadvertently cue a salivation response. You can absolutely influence an alert.

Fluency is a training goal. We want our dogs to be fast, consistent, and precise. But fluency can obscure our ability to read what's actually happening. A beautiful, automatic-looking behavior can mask deterioration in the underlying operant association. By the time it shows up as a problem in the field, the lack of reinforcement has been building for a while.

If you're not tracking your dog's performance over time, you're missing the early signs. Fluent behavior can hide a lot. The only way to know if the operant control is holding is to track the correct data that allows you to maintain fluency.

Be curious.

Robin Greubel | K9ensus Foundation