Operant vs Respondent
One thing that can help to understand as we approach animal training is the differences between Operant and Respondent behaviors.
Respondent behaviors are autonomic reflexes, that are responses to a stimulus. Classical Conditioning is sometimes called, Respondent Conditioning for exactly this reason. There are some things in life that we are born, instinctively knowing how we feel about. Food, water, sex feel good - pain, discomfort, lacking a necessity (like food/water) feel bad. From birth without any conditioning we know how these things feel. But when any stimulus is repeatedly paired with one of these things - it begins to take on that meaning. When a click is repeatedly paired with food, it takes on the appetitive value of the food. Like Pavlov's dogs and the sound of the bell, it elicited the respondent behavior of salivating, because the bell predicted food. While if something like a wiggling finger is paired with the sharp pain of a halter tug, the wiggling finger takes on the meaning of the pain. It becomes a conditioned aversive and elicits the same respondent behaviors that pain elicits. Respondent behaviors however can only be conditioned, not trained. Let me explain...
Operant behaviors are the trained behaviors, these are active, thought out behaviors we chose to do. Operant behaviors can be operantly conditioned, through the four learning quadrants. Positive Reinforcement and Negative Reinforcement can make these behaviors happen more. While Positive Punishment and Negative Punishment can make these behaviors happen less.
Punishment and Reinforcement however do not influence whether or not a respondent behavior will happen more or less!
Respondent behaviors are not under voluntary control, so they can't be encouraged to happen more or less. However they can be elicited and connected with our operant learning...
So when we Positively Reinforce a behavior (let's use "back up" as an example) we are adding a primary appetitive (food in this example). We not only Operantly Condition the "back up" behavior by positively reinforcing it. We ALSO Respondent/classically condition the behavior, the cue, ourselves (and all the antecedents) by pairing this scenario with appetitives!! So ourselves, our cue, the environment, and the behavior become respondently conditioned as appetitive. While only the behavior was operantly conditioned to happen more. But next time we are in this scenario the horse remembers it felt good and can elicit the similar appetitive respondent behaviors, like the feelings of happiness and anticipation of food and good things!
However when we use aversives in our training, either as punishment or reinforcement, we will be respondently conditioning our cue, behavior, and all the antecedents as an aversive. While we may negatively reinforce the behavior, making the operant behavior happen more - we may actually be aversively respondent conditioning ourselves, our cue, the behavior, and our whole environment. Thus eliciting the respondent behaviors that are paired with aversives.