operant conditioning

Pain behaviour: what is it and what do we do about it?


I’m re-reading Fordyce’s classic Behavioral Methods for Chronic Pain and Illness and once again I’m struck by how many of the concepts he introduced and systematically investigated are either mis-interpreted and ignored in our current approaches to helping people with persistent pain. Today I’ll explore just a tiny portion of what Fordyce described.

Pain behaviour refers to all the observable actions we do in relation to experiencing pain (NB some people include thoughts as well, but for today I’ll just focus on observable actions). There are roughly two groups of actions: those involuntary ones that we can call nocifensive responses that include reflex withdrawal underpinned by spinal reflexes but including brainstem circuits (see Barik, Hunter Thompson, Seltzer, Ghitani & Chesler, 2018); and those that are developed and shaped by learning (operant conditioning as well as social learning).

When I write about learning, I often have comments about this suggesting people have a choice about what they do, and that this learning must involve conscious awareness – the upshot of these comments is the idea that if we just tell someone that they’re doing something, information alone will be sufficient to change how often they’re doing it. Well, I don’t know about you, but if you’ve ever chewed your nails, changed your diet, decided to go on a social media diet, or do more exercise, you’ll know that there’s an enormous gap between knowing about and being able to follow through. So let me review some of the processes involved in learning and pain behaviour.

Pain behaviour probably has evolutionary significance. What we do when we’re sore acts as a signal to others, whether those actions are voluntary or involuntary. For example, while limping off-loads weight from the sore limb, it can also function to let other people know there’s something wrong. Groaning or sighing also lets people around us know that we’re not OK. Remembering that we’re a social species, being able to let others know that we need help – or not to do what we just did – means we’re more likely to receive attention, and also to warn others about potential danger. Of course, by eliciting help, we’re kinda obligated to help others when they do the same, which may be why when we see someone demonstrating prolonged pain behaviours we tend to feel annoyed: we might be asking ourselves “If they’re not going to reciprocate, why would I help? Dem’s the rules”

Now pain behaviour is also subject to learning principles. In other words, the specific behaviours we do develop in form and frequency depending on context. The underlying analysis goes like this: an antecedent is present (maybe it’s a particular person, location, or occasion), the behaviour occurs, then something in the environment/context occurs – and it’s this “something in the environment/context” that influences whether the behaviour is repeated, and/or the frequency of that behaviour. The easiest example of this is when you watch a three-year-old playing just a little distance from Mum and Dad. When she trips and falls, she’ll probably get up and brush herself down – and then you’ll see her look for Mum or Dad, and if they’re close enough, she’ll probably let out a bit of wail. In the context of Mum and Dad and her falling over, she’s learned that if she cries she’s likely to get a cuddle or some attention, and this is nice. In the absence of Mum and Dad, if she trips she’s less likely to cry because she’s not likely to get that cuddle. Clever huh?

So if that kind of learning occurs from the time we’re little, it’s easy to see how rapidly this pattern of behaving can become habitual, and when it’s habitual it’s unlikely to be something the child is aware she’s doing. Crying, or seeking attention, when we’re sore is something we’ve learned to do from an early age and while the form of that attention likely differs as we mature, the underlying mechanisms still apply (please don’t scream the place down when you go get your Covid vaccination! It’s OK for babies to cry, but not quite so socially appropriate for grown-ups to cry!).

How does the form of that behaviour change? It’s called “shaping” and it is something that occurs naturally through social learning, and it can also occur in a planned way. Take the example of the three-year-old falling and crying: crying is probably OK outdoors where there’s plenty of room and not too much attention being paid to the interactions between parents and child. Take that same behaviour indoors, perhaps in a supermarket or worse – a quiet waiting room – and it’s likely the parents will shush the child more quickly, and be a little more firm about any ongoing wailing. The context is different, the parents respond differently, and the child learns that it’s not OK to cry loudly where there are other people who might not approve. Over time children learn that in different contexts, different ways to attract attention are required. Clearly there’s more technical language we can use to describe this process, but for our purposes this is enough.

Why do we care about this?

Pain behaviour is normal. It’s something we all do. Mostly it functions in a positive way. We signal to others that we need help, we protect the sore body part, and gradually we recover and resume normal life. In some contexts, though, the tendency to continue doing pain behaviours outlives its welcome. In persistent pain it’s particularly problematic, but it’s also problematic in acute pain situations.

Let’s take the example of the dreaded pain rating scale. The 0 = no pain to 10 = most severe pain I can imagine scale. In the context of an emergency room, being asked to rate pain is a quick and very practical way for clinicians to decide how severe the presumed injury/tissue damage is, whether the person needs analgesia, and whether they’re responding to it. Give a number less than 3 or 4 and you’re probably not going to get a lot of pain relief. Give a number closer to 10, and you’ll get something. Give a number greater than 10 and you may get raised eyebrows. In an experiment by Herta Flor (Flor, Knost & Birbaumer, 2002), participants were given an electric shock and asked to rate their pain intensity (also nociceptive detection threshold (aka pain threshold) and pain tolerance). After they’d rated their pain over several trials, they were given one of two conditions: one in which they were given smiley faces and money when their rating was higher than their average rating for the previous trials, and one in which they were given a sad smiley when their rating was lower than their average. Flor and colleagues found that those people who had been given positive smiley faces for higher pain ratings rated their pain intensity significantly higher than those who had been given neutral or negative smileys.

This experiment doesn’t reflect changes in pain intensity. And this is a critical point to note! The stimuli were the same across both groups. What changed was the response offered to participants after they rated their pain. In other words, behaviour associated with experiencing pain and the resultant rewards given for higher ratings was reinforced.

This experiment, along with a large number of others, is one reason why I don’t like pain intensity measures being taken at every treatment session. Pain intensity ratings are behaviours subject to the contingencies that all behaviour is subject to – people learn what to do, and they do it. And they’re unaware of this process.

We often rely on pain intensity ratings in both experimental studies and clinical practice. Unfortunately, while a numeric rating scale or visual analogue scale are quick and dirty, they’re not like a pain thermometer. We just don’t have an objective measure of pain intensity. And we forget this.

Where am I going with this?

A couple of points. I don’t think we can always influence a person’s experience of their pain. This means that we’re often needing to influence what they do about it – because prolonged distress and disability is not good for anyone. Given the social nature of our species, and the involuntary nature of our response to another person’s distress, we’re inclined to try to reduce distress by offering comfort. Nothing wrong with that except where it gets in the way of the person beginning to do things for themselves. As clinicians we need to reinforce actions a person does to increase their capabilities. We also need to limit our reinforcement of illness behaviour, and we need to do this with the consent of the person – being open about why we’re doing this. Remember people learn this stuff without knowing they’re learning it! This means that as clinicians we must stop judging people and what they do in response to pain. Pain behaviour is learned over a loooong time, and it’s reinforced in so many places. People don’t do pain behaviour on purpose. So we can’t judge people as being “non-copers” or having “exaggerated illness behaviour” – we can just gently show the person what happens, why it happens, and what the effect of that pattern of behaving is having on their life.

The second point is that we can’t treat pain ratings as Truth with a T, and think that we’re getting a pure measure of pain intensity – because rating pain on a scale is a behaviour, and it’s influenced in exactly the same way as all behaviours are. This doesn’t mean ignoring someone’s pain intensity – it just means we need to listen to what the person is trying to communicate.

Barik, A., Thompson, J. H., Seltzer, M., Ghitani, N., & Chesler, A. T. (2018). A Brainstem-Spinal Circuit Controlling Nocifensive Behavior. Neuron, 100(6), 1491-1503 e1493. doi: 10.1016/j.neuron.2018.10.037

Flor, Herta, Knost, Bärbel, & Birbaumer, Niels. (2002). The role of operant conditioning in chronic pain: an experimental investigation. Pain, 95(1), 111-118. doi: https://doi.org/10.1016/S0304-3959(01)00385-2

Around and around and around we go – or…back to the beginning again


ResearchBlogging.orgIn the early days of non-medical chronic pain management, operant models of behavioural learning were used by Prof W Fordyce.  At the time, this was a revolutionary approach to helping people with chronic pain reduce their disability, distress and depression by altering the reinforcement schedules operating on pain behaviour.  It was such an innovation because it moved the focus from pain intensity to disability associated with pain, and in doing so, it offered hope to the many people who were otherwise thought unlikely to ever “get better”.

Operant models gradually fell out of favour as cognitive behavioural approaches for managing pain became popular over the 1990’s and later.  Refinements to behavioural approaches have been incorporated into ACT (Acceptance and Commitment Therapy) and graded exposure treatments for pain-related anxiety and avoidance – but this hasn’t been extended to operant conditioning.

What exactly is operant conditioning? The simplest way to think of it is to remember “operant” conditioning “operates” on the environment with behaviour maintained by what occurs afterward – so we need to think about rewards, punishments, and positive and negative conditioning.  In other words, the chance I’ll reach for a certain brand of chocolate bar will go UP if I receive some sort of reward for doing so, while the chance of my watching the Rugby World Cup will go DOWN if the people around me groan in unison whenever a game comes on the TV.  But this depends on whether I want the reward offered by the chocolate manufacturer – or I’m bothered by people groaning when I have the TV on!

In a really interesting study by researchers from Eastern Michigan University and University of Washington (Seattle), three hypotheses were tested: fear-avoidance, operant, and pain models, and their ability to predict activity levels.

To recap, the so-called fear-avoidant (pain-related anxiety and avoidance) model involves anticipatory anxiety, or specific fear of movement/reinjury and the resultant avoidance of activity and is usually associated with a classical conditioning process.  A person moves and simultaneously experiences pain or the fear of pain/reinjury, and therefore movement becomes associated with pain.  It doesn’t take long before the person avoids movement because it hurts, and because of the meaning of the hurt.

The operant model involves the person’s behaviour “operating on” the environment (usually involving people’s responses to the behaviour) which serves to increase or reduce the likelihood the person will do that particular behaviour again.  Think of someone coming home with a headache, looking poorly, maybe lying down, and the partner offering comfort – chances are, if the person liked this response, he or she will do the same thing next time a headache comes to visit!

In the pain model, it’s believed that the experience of pain itself dictates the degree of disability an individual has.  So, the person may have a neurobiological vulnerability leading to a greater sensitivity to painful input, and an augmented experience of pain.

In this experiment, 20 patients with chronic low back pain attending a tertiary pain management centre were asked to complete a range of questionnaires, and then were tested with a pressure algometer to determine the pressure at which they experienced pain.  This is thought to help identify the degree of “pain sensitivity” people may have and in this experiment was used to test the pain model.

Then patients were asked to wear an accelerometer to determine their activity level over five days – except when in water (showering, bathing, swimming).

Working some awesome statistical analyses (that, as usual, I won’t go into! Read the paper itself for more details), the researchers found that “the operant model emerged as the only model accounting for a statistically significant amount of the variance in physical activity”.  While the other two models were also relevant, and did contribute to the level of activity people carried out, family and/or spouse responses that were less solicitous, less punishing, and more distracting predicted higher activity levels.

The researchers concluded that“significant others’ responses account for a significant amount of the variance in physical activity, confirming that reinforcement of behavior is a powerful mechanism for behavior change”.  In a sense, this is not news – the response of family and others has always been thought to be important.  What is news is that when the other two models were tested concurrently, the factor that had the most impact on activity level was that of family and partners.

Here’s a question: how often do we

(a) assess the responses of these important people in behavioural terms?  While we might ask about the impact of pain on relationships, I think we need to examine what people do when the person is experiencing pain, and we also need to know how the person demonstrates or displays their pain – what the significant other is responding to, and

(b) specifically address how the person with pain can manage the responses of other people to their pain behaviour? Do we tell the person to communicate effectively about their pain to help them let other people know what they would like them to do when they’re sore?  It can be a very delicate process to negotiate how to ask the partner not to take over the activities the person with pain is beginning to do again! and finally

(c) include the family and partner in our treatments, goal setting and pain “education” so they are more fully informed about what their family member is trying to do, and so they can know what their best options for offering support might be?

It’s almost like deja vu – operant conditioning is coming back in from the cold.  I’d love to know your take on how we can do this within our usual practice. 

Alschuler, K., Hoodin, F., Murphy, S., Rice, J., & Geisser, M. (2011). Factors contributing to physical activity in a chronic low back pain clinical sample: A comprehensive analysis using continuous ambulatory monitoring PAIN DOI: 10.1016/j.pain.2011.07.017

Pain-related anxiety and avoidance: a practical application of theory and research to clinical practice


In 1995 I started work at the Burwood Pain Management Centre.  It wasn’t my first foray into pain management, but it was the first time I had worked in a fully integrated interdiscipinary team environment.  It was also significant because of the use of the words ‘fear-avoidance’, ‘guarding’ and ‘anticipatory anxiety’.

What the team had observed was that there were a specific group of patients who were not just worried about experiencing pain, but were also showing the signs of increased physiological arousal, avoidance of specific activities, and firm beliefs about harm, reinjury, or doing further damage.

At the time we used an operant conditioning model (based on Fordyce, 1971 & 1976), along with graded reactivation using a physical conditioning model, to help these people learn to be more functional.  We had some successes – but a number of less successful responses too.

Towards the end of the 1990’s, a new model started to emerge.  It was the ‘fear-avoidance’ model proposed firstly by Lethem et al (Lethem et al. 1983), then elaborated on by Vlaeyen and colleagues (1999, 2000).

This model proposed that people demonstrating avoidance of activities may be developing a ‘phobic’ response to their situation: based on catastrophic interpretations of their pain, beliefs that their body will be dangerously harmed by movements, increased physiological arousal in response to these beliefs, avoidance of situations the same as or similar to those that provoke their beliefs, and subsequent reward of this avoidance by negative reinforcement (ie, by avoiding a negative experience, they increase the avoidance behaviour). (more…)