SPOILER: Punishment is not necessary for effective behavioural modification outcomes, and behaviour modification using extinction with punishment carries a significant probability of undesirable side-effects.
If you are a bit of a nerd like me and are interested in the nitty gritty of the evidence, read on.
What is differential reinforcement? Differential reinforcement (DR) is a behaviour modification (BM) technique which involves selectively reinforcing target behaviours to increase their probably of occurring. Variations include DR with extinction, and DR without extinction. DR with extinction is a behaviour modification (BM) technique which involves selectively reinforcing desired behaviours while withholding reinforcement or punishing undesired behaviours. Some FF trainers use the former method of extinction. DR without extinction involves reinforcing both unwanted and target behaviours manipulating variables in the schedule of reinforcement. Those variables might include reinforcement value, using multiple concurrent reinforcers, duration of reinforcer, and value of the reinforcer. DR of alternative behaviours (DRA) aims to replace an unwanted behaviour with a new target behaviour, while DR of other behaviours (DRO) aims to reinforce the abstinence of the behaviour, i.e. rewarding any socially accepted alternative behaviour to the undesirable behaviour. DR is the basis of BM for both balanced and force free (FF) trainers. One of the criticisms of FF training is that one cannot achieve effective BM outcomes without an extinction by punishment component. As a FF trainer, my own experience and anecdotal evidence refute this assertion. As a scientist, I need more than anecdotes.
What does the literature tell us about the efficacy of DR with and without extinction? To be honest, the literature is sketchy and results between studies vary. There is inconsistency in methodology, many DR studies don't stand up to method scrutiny, and none of the experiments directly compare DR with and without extinction.
Fortunately, some very clever people (1) analysed the variables and outcomes of 109 experiments across 32 papers using the SCARF model to analyse the outcomes and quality of the research. SCARF analysis helps to score the variables for equitable comparison.
Study participants had varying degrees of intellectual disabilities, and displayed unwanted behaviour ranging from food refusal, to serious harm to others. The literature review and SCARF analysis did not distinguish between DR with extinction by punishment and withholding when reporting outcomes, i.e. outcomes of DR with extinction, whether by punishment or withholding were reported together. In order to assertain if there was significant differences in fallout risk between the two extinction strategies, I had to review the methods of the studies individually.
In lay terms, the research shows:
DR experiments using extinction by punishment reported high rates of successful extinction of the undesirable behaviour and an increase in the occurrence of the target behaviour. However, there was a significant risk of extinction burst and emergence of new aggressive behaviours, and in some cases, the new behaviours were more serious than the initial ones. The emergence of new behaviours does not affect the quantitative analysis of the experiment's data, i.e. if there was successful extinction of the undesirable behaviour, it was recorded as a success, regardless of the resulting unintended fall-out.
Punishments must be consistent and reliable, or BM could "result in sustained countertherapeutic responding", meaning, DR using punishment is not effective unless extinction is consistently and reliably delivered.
Experiments using extinction by withholding also reported high rates of successful extinction of the undesirable behaviour and an increase in the occurrence of the target behaviour.
An important variable to the success of reinforcing the target behaviour with extinction was the timing of the delivery of the reinforcer of the target behaviour.
This is where things get very interesting.
DR without extinction, where both undesirable and target behaviours are reinforced, had varying degrees of success across experiments, and was found to be most effective in decreasing instances of unwanted behaviour and increasing instances of target behaviour when more advantageous schedules of reinforcement were used to reinforce the target behaviour. In other words, effective BM was achieved by reinforcing both the undesirable behaviours and target behaviours at differing schedules and degrees of reinforcement. A 100% success rate was reported in experiments using concurrent reinforcement .of target behaviours.
"When creating socially valid interventions, practitioners should consider manipulating parameters associated with concurrent schedules of reinforcement as this may lead to reduction in problem behavior, increased rates of appropriate behavior, and potentially reduce risks associated with extinction." (1)
"Results [...] showed that behavior was often sensitive to [reinforcement] manipulations of duration, quality, and delay in isolation, but the largest and most consistent behavior change was observed when several dimensions of reinforcement were combined to favor appropriate behavior."(2) (On DR without extinction.)
The literature not only tells us that punishment is unnecessary for effective behavioural modification, it tells us that effective BM outcomes can be achieved using DR without extinction when the concurrent schedules of reinforcement are appropriate.
But these experiments are human experiments, not canine experiments, you might say. This is true, however, Applied Behavioural Analysis is the core of DR which is predominantly used to treat people with intellectual disabilities and animals, and is the basis of operant conditioning.
This research does not consider any other variables besides BM outcomes. Fallout, whether behavioural, social, emotional, or psychological, falls outisde of the scope of the research. And still, FF produces the most favourable outcomes.
Success of BM without the risk of fallout does require some skill, including delivery timing, choosing the right reinforcer, and the dog's preferences. Any trainer worth their salt should have the aforementioned skills necessary to produce effective outcomes without the use of force and punishment. Why then is it so hard for some to let go of punishment?
Comments