Poisoned Cues

with Jesus Rosales-Ruiz

The idea of poisoned cues first requires a good understanding of what a cue is and isn't. A Cue is defined as “a discriminative stimulus (SD) established through positive differential reinforcement” this means the learner knows that cue/antecedent=behavior X=positive reinforcement. Remember our ABCs? Antecedent-Behavior-Consequence (loop). A “cue” is one that results in a positive (addition of appetitive) Consequence. A cue eventually becomes a conditioned positive reinforcer, just like a click. Remember Classical Conditioning when X=Y? When a neutral stimulus is paired repeatedly with an appetitive stimulus is becomes a classically conditioned appetitive. So when Cue=Behavior=R+, Cue=R+! As simple as that! We can also test this – the animal will work for the ability to get the cue that results in R+, remember this is how behavior chains work!

But what about R- trained behaviors? These we call a Command, so we can differentiate in common language. A Command is “A discriminative stimulus (SD) established through negative reinforcement”. So Command=Behavior=R- this can include when the command itself is no longer aversive just a predictor of an aversive. For example saying “walk on” giving a squeeze, kick, spank until you get the behavior then R-. Now the behavior is trained thoroughly enough that you get the behavior immediately after the “walk on” verbal command. Just like the cue became a conditioned appetitive – the command becomes a conditioned aversive. So command=behavior=R-, so command=R-. so command=aversive. This can be proven/tested by seeing how the command can be used to punish a behavior (reduce it's frequency) – and in this case the removal of this command can be used to reinforce a behavior (like a release).

The Poisoned Cue is the combination of a command and a cue “A discriminative stimulus (SD) established through the combination of negative and positive reinforcement”. The aversive can be used in the shaping process or be part of the antecedents of the behavior or be part of the consequences. Whether you use an aversive to “correct” or punish or if you use it to coerce the desired behavior - if an aversive becomes any part of your ABC loop it can poison the entire scenario. In testing, the SD can be used as a positive reinforcer but with a severely diminished outcome. So Poisoned cue=behavior=R+ and R- so Poisoned cue=?? Either? Neither? Both? Poisoned cues create reluctance, latency, conflict behaviors, behavioral breakdown proceeding cue (unclean repetitions) and/or behavioral breakdowns after the cue.

Remember when we say a “cue” is poisoned, we do mean the cue and all other antecedents, not just the intentional cue we've chosen to put on the behavior. The smell in the room, the color of your hat, the lead rope, the halter, the way you're positioned, the tone of your voice – everything they see, smell, feel, taste, and hear can all be poisoned depending what they think predicts the aversive. Find out what has become poisoned – find your baseline, then add in variables, find when latency, distraction or conflict behavior starts. Is it the presentation of an object? The wearing of a tack piece? Or your verbal/visual cue? Remember even your food reward can be poisoned. Remember how we condition a neutral stimulus to become an appetitive stimulus by pairing? This is classically conditioning with an appetitive. We can also counter condition by turning an aversive stimuli into an appetitive with careful pairing, we often do this to help our learners overcome fearful stimuli. Unfortunately if we aren't careful with our timing and arrangement we can actually counter condition the appetitive to predict and aversive. If you only pull out your treats when something aversive is going to happen, that food is going to predict something bad is going to happen.

Once we've determined which stimuli are poisoned it's truly easiest to change them. This can be changing the color, shape, size, design, sound, feel, visual cue. Ideally we should just change it. If this isn't possible for whatever reason change it as best as possible and rebuild the behavior. While it may or may not be possible to “unpoison” a behavior it's much quicker and more reliable to just start over.

<- Back to Training Articles