NOTICE: Posting schedule is irregular. I hope to get back to a regular schedule as the day-job allows.

Friday, June 15, 2012

The GUIDE: Operant Conditioning. [Full link to blog for email clients.]

In discussion with a colleague this week, I was struck by the differences in how we measure memory in lab animals, and in humans.  In the animal research lab, we tend to use computerized tools that present information, and give the animal a chance to respond within a restricted time frame.  Since we can't give verbal instructions to the lab subjects, we use these techniques to exert a certain degree of control over what information the animal's brain processes at a given time. With humans, the tests of memory and cognition are much more likely to be written or verbal tests, taking advantage of human communication skills to instruct and control the test parameters.

So, how exactly do we train animal subjects (and no smart remarks about training graduate students!)? 

Domestication and/or training of any animal is a learned process.  If you have ever watched circus animals, movie dogs or the "Walrus, Otter, Sea Lion" show at Sea World you have seen the results of what is called "operant conditioning."  Most folks should recall reading or learning about Pavlov's experiments in classical conditioning.  Pavlov rang a bell simultaneous with the smell/taste of food.  The dog salivated at the food stimulus, and eventually learned to salivate at the sound of the bell alone.   

In behavioral terms, we talk about an "unconditioned stimulus" (UCS) – that is, a stimulus that produces the desired response without learning.  In the Pavlov example, smell of food is the UCS and salivation is the response.  Then there is the "Conditioned Stimulus" (CS) – this is a stimulus (bell) that we *want* to produce the same response as the UCS.  So we present UCS and CS together for a *lot* of repetitions – usually with the CS occurring before the UCS.  Eventually the conditioned response (salivation) starts to occur after the bell, but before the smell of food.  Once that happens, we call it a "conditioned response" (CR). Once our experiment begins to look like this:  CS – CR – then UCS, we have conditioned the animal and can even omit the smell of food (UCS) and still get salivation (CR) just by ringing the bell (CS).  Voila - Pavlov's classic experiment in behavioral conditioning, also known as "classical conditioning." 
Operant conditioning differs in that we actually teach the animals a chain of UCS stimuli, and those stimuli require the animal to *do* something.  In the Sea World show example, the sea lion is conditioned to know that ringing a bell means a food reward will be given.  This is classical conditioning, but then we add a twist – now the sea lion gets a chance to explore the bell – if it rings the bell itself, it gets a fish!  This is the first essential step in operant training, the animal makes its own response and receives a reward.  To make more complex behaviors, we just train in one preceding step at a time – before the sea lion can ring the bell, it must climb some steps.  So in this step we have two operant phases: (1) climb steps, in order to be able to (2) ring bell, to receive the reward.  It's just like training dogs, horses, or even people.  Once enough steps are "chained" together, our sea lion becomes the star act and appears to be behaving just like a human – but it's only conditioning at work.

However, many laboratory tests require that animals not be conditioned to the task.  We accomplish this by changing the information (although not the behavioral procedures) required for a given task.  A "delayed-match-to-sample" task requires an animal to see or sense a stimulus ("Sample"), then find that same stimulus again ("Match") after a delay.  The stimulus and response behaviors are operantly conditioned, but the actual information in the form of the Sample varies from trial to trial.  Again, we automate the process because we can produce many hundreds or thousands of stimuli, repeat the trial process hundreds of times per day, and provide very precise timing of sample, delay and match onset in order to examine precisely what the brain is doing when the information is presented.  

Nevertheless, in a laboratory, the animals become conditioned to many factors:  not just the tasks they run, but also which humans feed them, take them out for the experiments, provide veterinary care, and what time of day these processes occur.  Lab animals become accustomed to regular routine, and we rely on that to reduce the random external factors that confuse or obscure what we want to record.  

Frankly, I'm a bit prejudiced toward automated, computer controlled procedures, and I tried to convince my colleague that such would be needed for any human studies if we are ever to successfully translate all of the brain science we've learned in animals to practical prosthetics for the brain.

So, until next time, train and protect your brain, but stay away from ringing bells!

No comments:

Post a Comment

Please add comment - no links, spammers will be banned.