One factor that complicates many of the debates surrounding applied behavior analysis is the fact that its technical language groups things together based mainly on their effects on specific (and usually observable) behaviors. While this makes it excellent for its intended purpose (analyzing behavior), it also makes said language extremely poor for use in an ethical debate.
To illustrate, let's take the hypothetical example of a boy -- let's call him "Chuck" -- who likes to act out in class. Let's also take three hypothetical teachers, "Jane", "Matt", and "Alletta", each of whom decided to take a different approach to making him behave.
Jane decided to assign Chuck additional homework on days when he misbehaved, gently informing him of her reasons whenever she did so. After a bit, Chuck realized that acting out in her class only led to more work and started paying attention more.
The second teacher, Matt, decided on a simpler course of action. He hooked a cattle prod to a remote control and strapped it to Chuck's back during class. Whenever Chuck acted out (or attempted to remove the cattle prod), Matt activated the device. After a few jolts, Chuck realized that both actions only resulted in terrible pain and stopped acting out, paying attention in a manner that only the truly terrified can.
The third teacher, Alletta, was an advocate of corporal punishment, but had promised Chuck's father that she wouldn't hurt him. As such, she instead called Chuck's best friend, Bill, to the front of the class and spent a few minutes torturing him whenever Chuck misbehaved, making sure that Chuck knew why she was doing so. Not wanting to see his friend hurt, Chuck quickly fell in line.
In the language of behavior analysis, all three of these teachers implemented the same basic strategy: punishment by contingent presentation of an aversive stimulus. All three strategies were punishment because they directly targetted and decreased an undesired behavior class (acting out), they were contingent because the strategy was only implemented when the behavior occurred, they were presentation of a stimulus because all three strategies involved giving Chuck something (be it extra homework, a painful shock, or a traumatizing emotional experience) and those stimuli were aversive because Chuck moved to avoid them (1).
This is not, however, to say that all three of those strategies were equally ethical. Very few people (outside school-age children, anyway) would have problems with what Jane did. Matt's actions were a somewhat exaggerated version of a highly controversial technique used in certain behavioral institutions (2). Alletta's actions, however, haven't been considered even remotely acceptable since the 1700s and even then, they were confined to punishing certain royals.
For those historians reading this, yes, I did name Chuck and the people in the third example after certain individuals. And while Matt's method may have prevented Ollie's eventual hissy fit, just imagine Jimmy's reaction!
To get to the point, however, I also don't mean that these strategies would all have the same effects. While all of them decreased the target behavior, Jane's strategy would have had the fewest unpleasant side effects.
The strategy closest to what commonly comes to mind when the term "aversive" is mentioned to people familiar with today's ABA establishment, however, is Matt's.
If this were an isolated case of misused or misunderstood terminology, it would be a fairly minor problem. Unfortunately, that isn't the case. The language of behavior analysis is -- perhaps unsurprisingly -- designed for analyzing behavior. As such, methods are categorized based on their effects on (usually specific and observable) behaviors and not their effects on the person who is behaving.
Ethical considerations, on the other hand, are almost always based on benefit and harm to people, something which the language of behavior analysis is decidedly poor at describing.
If this was widely understood, it wouldn't be a problem. Unfortunately, many people who use the techniques produced by ABA don't understand the distinction. The debates surrounding their use are often muddied by this.
As a general rule, people with postgraduate educations in behavior analysis understand its technical language and use it to talk about ABA's techniques. Everyone else... doesn't, although there are exceptions. Things get more complicated when terms used in ABA have or acquire different meanings outside of that language.
The debates surrounding aversives are, quite literally, a textbook example of this.
(1) According to Cooper, et al. (2007), the most influential textbook on behavior analysis used today, an aversive stimulus is defined as "a stimulus change or condition that functions (a) to evoke a behavior that has terminated it in the past; (b) as a punisher when presented following behavior, and/or (c) as a reinforcer when withdrawn following behavior." In this case, it's (b) that applies.
(2) While modern accounts usually don't involve literal cattle prods, instead involving a variety of other shock devices, there are accounts of literal cattle prods -- and improvised devices made from them -- being used as late as the early to mid-'90s.
Cooper, J., Heron, T., & Heward, W. (2007). Applied Behavior Analysis (2nd Ed.) Upper Saddle River, NJ: Pearson Education, Inc.