Your face is gonna freeze like that (Part 3)…….(aka: How many Steves?)

The mental manipulations that we described in Part 1 are actually a form of psychological priming. Priming occurs when the way in which a person responds to an event (stimulus) is influenced by a previous stimulus. Perhaps most intriguing is the fact that much of this influence takes place outside of our conscious awareness. In part 2,  we presented our working hypothesis that priming may work with dogs; specifically priming dogs to feel happy and relaxed will enhance learning during touch-then-treat training.

So, can a dog’s emotional responses and behaviors be primed? Like most people who live with and love dogs, my intuitive response is “Of course they can“! But, intuitions aside, if we are being good canine scientists, how might we collect actual data that will either support or refute this hypothesis?

Suppose that our friend Terry decides to test the theory on her dog Steve.

single Border Collie

STEVE

The Steve Study:  Terry sets up a study protocol of daily training sessions with Steve. In each session, Steve is primed  using a favorite food treat. Terry decides to use 10 repetitions of reinforcing eye contact, something Steve readily and regularly offers, as her priming stimulus. She follows this portion of the session with a series of touch-then-treat exercises in which Steve’s front paws, mouth, and ears are each handled gently (see below). Terry repeats the entire routine once per day for a period of three weeks.

Touching Ear Touching feet Touching teeth

Data Collected: For each training session, Terry rates Steve’s response to the three types of handling (ears, paws, mouth) using a five-point rating scale that ranges from “does not accept at all” (score of 1) to “accepts completely” (score of 5). This type of scale, called a “Likert scale” is commonly used in studies that rate subjective experiences. Terry scores Steve’s responses before the experiment begins (pre-test) and daily (after each session) until the end of the study. Here are Steve’s mean (average) weekly scores, before and after testing, for each handling exercise:

HANDLING

PRE-TEST

WEEK 1

WEEK 2

WEEK 3

EARS

4.1

4.8

4.8

5.0

PAWS

1.4

3.5

4.2

5.0

MOUTH

2.1

3.2

4.4

4.4

Results and conclusions: Steve’s scores show that he was quite tolerant of having his ears handled at the start of the study (mean score 4.1), but that he did not tolerate handling of his paws (mean score 1.4) or of his mouth (mean score 2.1) at the start of Terry’s study. All three areas showed improvement over the 3-week training period. Steve completely accepted Terry’s handling of his ears and paws and showed a great deal of improvement in allowing Terry to work with his mouth by the end of the three-week study. Terry concluded that “priming is an effective training tool that leads to  improvement in a dog’s acceptance of handling exercises

Is Terry justified in this conclusion? Does Steve’s response tell you if priming enhanced classical conditioning (touch-then-treat training) in dogs? In a word, NOPE.

There are several problems with using a Steve Study (or a Muffin study, or a Rover study or a Cooper study). Some of these may be immediately obvious to you; others perhaps less so. Here are four reasons that Steve’s response (though very nice for Steve and Terry) should not be used to make conclusions about our hypothesis:

1. Confounding factors: Several other factors may have influenced Steve’s behavior. The passage of time alone (especially in an adolescent dog) may have caused him to more readily accept handling. Second, Steve may have started to tolerate handling regardless of any type of training intervention, simply in response to the daily scheduled interactions with Terry. Third, his response may have been due to touch-then-treat alone, regardless of priming. These uncontrolled factors are why all studies need to include a control group.

2. Experimenter bias: Terry was not blinded to the treatment and had expectations that it would be effective. This leads to a common cognitive error called confirmation bias; more about this in a later blog). In this case Terry’s bias is also caused by a placebo effect and would lead her to err on the side of seeing improvement in a subjective measure where it may not have existed.

3. Steve ain’t stable: No, this is not a derogatory statement about Steve. What I mean is that Steve, like all biological creatures, varies in his day-to-day behaviors. That variation, which occurs both within an animal and between animals, must be accounted for when trying to determine if an animal is actually responding to a treatment or if the results that we are seeing are purely by chance and are caused by normal day-to-day fluctuations. In this case, fluctuations in behavior or learning, but this applies to all biological processes.

4. One Steve = Anecdote (not a study): Last but not least, the fact that Steve varies day-to-day in his behavior is further complicated by the fact that ALL dogs vary from one another (we all know this, of course). In this particular case, we would expect that all dogs vary in the degree to which they respond to priming (if priming works). The trick with research is….here is the punch line……separating the normal variation that occurs within and among dogs from the variation that may be caused by our treatment (in this case, priming) is why we need to study groups of dogs and why we always need one or more of those groups to be a control group This is also the reason that we need statistics – to keep us from making incorrect conclusions that occur because Steve just happens to be an unusually smart dog (which of course, he is), because of normal differences among dogs, or because Steve just happens to be having a good day.

So, instead of just testing our hypothesis on just Steve, we actually need to test it on a sample of the population (i.e. multiple Steves):

many BCs

MULTIPLE STEVES (sample size: n = 10)

OOPS! WAIT you say! These are all Border Collies! Don’t we need a sample that is representative of all dogs in the population? Yep – we sure do……

Group-of-dogs-resized-600

STUDY SAMPLE (n = 9)

This looks a bit better. Let’s test our hypothesis on a study group, a sample that of dogs that is representative of the population of dogs that we would like to make conclusions about. Let’s start with the group of happy fellows above, but keep adding until we have about 75 young adult dogs available for our study. What are the treatment and control groups that are needed for this study? To thoroughly control this study, we need three groups:

Treatment (Experimental) Group: Also called the “test group”, this is the group of dogs that receive the experimental treatment; in this case this is a series of Priming + Touch-then-Treat training (25 dogs or n = 25).

Positive Control: Positive control groups are used when we need a control that is expected to have a positive result, allowing the researcher to show that the protocol was  capable of producing results. In our study, this would be a group of dogs who are not primed, but are trained to accept handling using Touch-then-Treat only. (n = 25)

Negative Control: Negative control groups are used to make  sure that no confounding variables affect results and to factor in any likely sources of  bias. A negative control can also be a way of setting a baseline. No training. (n = 25)

Study Protocol: Dogs are randomly assigned to one of the three groups. (We might also “block” dogs across treatments, which matches dogs by sex, age, breed or other sources of variance that we want to control; more about this in another blog). We use the same study protocol, with one exception. The person who scores each dog after their daily training (or no training session) with Terry, is not Terry. Rather, we will use a scorer who is “blinded” to the study treatments and is present only to score each dog’s response to handling, with no knowledge of the treatment.

In the final part of this series, we will look at data from the Multiple Steves study and what conclusions we might make from our hypothetical study. For now though, you may be asking……WHY IS THIS APPROACH IMPORTANT? Well, let me tell you…….gotta first climb…..

Up on my Soapbox…..

soapbox   

        Testing new ideas using the scientific method protects us from making Steve Study mistakes. Sure, you may see improvements in your dog’s coat, vitality, agility performance, or health when you switch him from a cooked to a raw diet, when you decide to go grain-free, when you train him to balance his front feet on a ball, or when you use the ointment for his ear infection that Joe next door (who knows a lot about dogs) concocted and gave to you. However, without adequate study that includes groups of subjects, control groups and (gasp!) statistical analysis, you cannot know if something that has not yet been studied is actually doing what you think it is doing. While not infallible, the scientific method  is constructed to prevent or minimize bias, to test sample groups that represent a population, and to prevent us from coming to conclusions based only upon our biases, our intuitions and our dear beloved, and very smart dog, Steve.

to be continued[1]