Saturday, May 5, 2012

Aversive images and Evaluative conditioning in food choices

     This is the third installment of my weekly blog post; and I am running through the longest article, up to this point.  The article is here for anyone with access.  The title of the article is "Using aversive images to enhance healthy food choices and implicit attitudes: An experimental test of evaluative conditioning" and I'm excited to talk about the real-world applications that this study portrays.  Let me define some terms, just to help clarify the points of the study.  Evaluative conditioning deals with how we can come to like or dislike something through an association (  So basically, evaluators wanted participants to form an association between energy-dense snack foods and aversive images.  The evaluators also wanted to see how this evaluative conditioning would affect implicit attitudes - the positive or negative thoughts, feelings, or actions towards objects which arise due to past experiences which one is either unaware of or which one cannot attribute to an identified previous experience (  [Later on I will talk about the problem I see with this definition of implicit attitude and the experiment].  They would assess this using the Implicit Association Test (IAT) which required participants to choose between fruits or snacks to take home with them.
    The experiment was conducted with 134 participants (101 females) who were recruited from King's College London staff and students.  Participants were randomly assigned to either a control or intervention group which determined what set of pictures they would be viewing.  Participants were asked to watch a slideshow of 100 images (five snack images shown 20 times in a random order).  Each trial lasted for 2.5s, with the snack image being shown for one second, followed by one of five aversive bodily images (intervention condition) or just a blank screen (control condition).  The aversive images were carefully chosen through pilot participants and came to include two images of obesity (in men and women), two images of arterial disease (with the text "artery disease" written on them) and one of heart surgery.
    Two outcomes were measured.  The primary outcome measure (an index of behavior) involved participants choosing to take fruit or snacks home with them and secondly, a voucher to spend at either a fruit or confectionery stand at the nearby train station.  The participants' choices were scored on a 5 point scale from -2 (two choices of snacks) to 0 (no preferences either way) to +2 (two choices of fruit).  The secondary outcome measure dealt with the IAT.  The assumption with the IAT is that responses to categories that are highly associated in memory are assumed to be faster when they share the same response key, relative to the response to categories not associated in memory.  Four categories were laid out, as two targets (fruit, snacks) and two attributes (pleasant, unpleasant).  So, in a first task, participants assigned a target with an attribute, so they would assign "fruit" and "pleasant" to the left computer key ("E") and "snacks" and "unpleasant" to the right key ("I").  In the second task, the targets were switched, so the left key was for snacks-pleasant and the right key for fruits-unpleasant.  Then, category headings were displayed throughout, in the left and right hand corners of the screen and participants had to press the correct button that corresponded to them.  Errors in sorting were pointed out, and an accurate response was necessary to move on to the next presentation.  The IAT scores were measured as the average response time between the two versions of the critical combined task.  In other words, if responses were faster for when fruit-pleasant were on the same key (than with fruit-unpleasant), then that must have meant that an association had been made in the mind of the participant and a more positive score would be assigned indicating a more positive attitude to fruit, and a more negative attitude to snacks.  In addition, explicit attitudes of the participants were also measured.  Participants were asked to express their ratings to "For me, eating fruits/snacks is" along 5 semantic differential scales (not at all healthy - healthy; bad - not at all bad; not at all enjoyable - enjoyable; not at all unpleasant - unpleasant; good - not at all good) and their responses were scored to gauge their overall preference for fruit.
     The results of the experiment showed that with the primary outcome, participants in the intervention condition preferred to choose fruit over snacks more often than those in the controlled condition.  This was predictable, since those in the control group were simply seeing biscuits, cookies, brownies, etc. with no association or link to them being healthy or unhealthy.  Those in the intervention group had made an association between the aversive images and the snacks and decided not to choose them when given the chance.  With the secondary outcome (test of implicit attitudes) it was shown that there were higher IAT scores  than in the control group, indicating a relatively stronger implicit preference for fruit over snacks.  In addition, both the implicit and explicit attitudes explained the variance in food choice that was observed.  Results were always compared to a baseline in the pre-intervention to see how participants' choices had changed.
       One potential problem I see with these results is that there was an overwhelming majority of females in the study who may be more likely to change their attitudes and select healthier snacks, because of an underlying preference to eat healthier. Also, since this study was conducted at a College, it is not representative of the general population and may include participants who are generally more health-conscious and more likely to change their behaviors and eating habits as opposed to older individuals who may be more fixed in their eating preferences and behaviors.  Another problem with the study may be the lack of controlling for the aversive images.  Maybe it is the aversive stimuli that are causing the participants to choose one way or the other.  Ideally, a control group should be established with participants exposed to just the aversive stimuli and see how they respond, or even a control group where fruits are shown with the aversive images to see if participants would prefer to choose the energy-dense snacks over the fruits that were associated with aversive images.  These would, potentially, make the experiment a lot more valid and convincing, but there are always costs of money and time associated with recruiting more participants and sorting through the data from all the various experimental conditions.  Another point to note in this experiment (not necessarily a problem) was that there was an inability to change the implicit attitudes of those who already held negative attitudes towards the snacks, which could be explained by ceiling effects.  As explained in my psychology class, ceiling effects arise when responding to a certain stimulus has already reached its maximum potential, and you are not going to get any increase in responding, as if you had hit a ceiling.
     It is interesting to note that in this experiment, the conditioned stimulus (CS) is the food, and the unconditioned stimulus (US) is the aversive images of health consequences (obesity, arterial disease, etc.) This is reversed, where normally we would see food as the US in other experiments because it is enjoyable and biologically essential, and elicits an unconditioned response.  The results of the experiment did not say a lot about how explicit attitudes were changed in the participants, but it is noted that implicit attitudes, rather than explicit attitudes have been shown to be changed through the evaluative conditioning when there is a clear, observable link between the CS and the US.  In this case there was that strong association between the CS and the US, which is why we saw a large change in implicit attitude in those participants who had stronger implicit preferences for snacks over fruit at baseline.
     When I first read through this experiment, I immediately connected it to the real-world applications of advertising and marketing.  Imagine billboards with images of sodas and aversive images of obese adults on them, along with a message saying "GOT MILK?" on the bottom?  The possibilities are endless with conditioning people to associate two images and change implicit attitudes to the foods/drinks that they previously enjoyed so much.  The study itself talks about aversive images used on tobacco packaging (the skull and bones), signaling how this procedure is already underway and will potentially become more widespread as more research is conducted and experiments begin to dive deeper into the underlying mechanisms of behavior change and the best methods of intervention.  I believe that implicit attitudes are not going to be changed overnight with a simple experiment linking foods and aversive stimuli; because, even in this study, people are clearly going to remember that they saw hundreds of aversive images next to the snacks that they were later asked to choose from (it is not like they don't remember their past experience, and they suddenly feel a change in their behavior).  I believe that implicit attitudes are much harder to change in most people who already have established food preferences and behaviors.  This is why I believe the research should have been conducted on more adults.  Potentially, the best way to prevent people from choosing energy-dense snacks over fruits is to instill a mind-set from an early age.  If people are exposed to aversive stimuli and develop a fear response to certain foods and the possibility of developing obesity and arterial disease, then they will be less likely to choose these foods over healthier alternatives.  This brings up ethical issues, however, about manipulating younger children and mentally affecting them (with aversive images), when they are most impressionable.  A whole range of convincing research needs to be conducted in order for the full effects of these advertisements to be known and eventually proliferated.

1 comment:

  1. Your best blog post yet! Although you incorrectly state "Those in the intervention group had made an association between the aversive images and the snacks and decided not to choose them when given the chance." you do question this statement in the following paragraph. It would have been more accurate to say that the authors of the paper interpret their results to indicate associative learning, and then you could comment that their design fails to rule out pseudoconditioning.

    Nevertheless, a very well written blog post, and I enjoyed your analysis and discussion. You are spot on that kids learn to crave energy-dense (but nutrient bereft) snacks from an early age and choose them over fruits and other healthy foods, such as fish and shellfish, meats, veggies, and nuts. I like your suggestion that evaluative conditioning could be a useful technique to instill in children a fear of or aversion to junk food. There are solid epidemiological and scientific data extolling the disease-inducing properties of junk foods (industrial processed foods) over foods closer to their natural state, so I think the ethics issue can be diffused. The ethics issue can be diffused even further when we realize that the kids are already subjected to massive amounts of evaluative conditioning on the part of the advertisers linking pictures of children's favorite cartoon characters, superheros, tv shows, etc. on the packaging of the junk foods. There's a lot of evaluative conditioning going on to the detriment of the kids and something must be done to counter that!

    I see more use among health care industry of messages about how healthy vegetables and fruits are (though, unfortunately, the same messages aren't being spread about meats, especially from pastured animals), but this evaluative conditioning may not be sufficient to overcome kids' (or adults) preferences for junk foods. We need some aversive conditioning to devalue the junk foods in addition.