Psy 121 Announcements re: First Writing Assignment
Dell will not be on campus this afternoon (Thurs) or tomorrow (Fri). She’ll be available sporadically to answer e-mails over the week-end, but would prefer talking with you in person. Her office hours next week are:
Monday, 4-5:30
Tuesday 4-5:30
Wednesday 1-2, 3:20-5:00
Thursday, 1-2, 3:20-4:30
There are sign-up sheets available on the bulletin board to the right of her office door (Psy 130) if you want to reserve a time slot.
There will be a group Question & Answer session on Monday, Sept 17, 6-7 p.m., Psy 105.
THE SMOKER HYPOTHESIS
A suggestion has been made that the Lips Group liking ratings might be higher than the ratings for the Teeth and Knees Groups because of a conditioned positive affective response to the lips pen position for smokers. Here’s an expanded exploration of why this potentially interesting explanation should be abandoned in this case.
First, I hesitated to challenge the suggestion on theoretical grounds in class, because I appreciate the kind of thinking it reflects. However, the literature on conditioned drug effects demonstrates that stimuli/actions associated with drug self-administration usually produce conditioned withdrawal-like responses when presented without the drug itself (see pp. 208-209 in GRG and the references provided therein).
Second, it is critical to recognize that a smoking-based conditioned positive affect associated with the lips pen position could not explain our failure to obtain the predicted difference (based on the facial feedback hypothesis) between the teeth and knees groups’ liking ratings.
Finally, even if the previous literature demonstrated a reliable conditioned positive affective response to (something like) the lips pen position, this conditioned response could only explain the obtained effect in this experiment if certain very unlikely situations held. These are outlined below.
One version of the suggestion that smoking status mattered in this experiment argues that IF positive affect is produced by (only) the lips pen position in smokers, smokers’ ratings would inflate the mean ratings of only the lips group. I agree with this assertion. But this is a good example of why we do statistics. In short, if each of the groups contains some smokers and some non-smokers (expected on the basis of random assignment to groups of a population consisting of both), then the selective effect of the lips position on smokers would be expected to increase not only the mean but also the variance in the lips group. This is because in the lips group, some participants (the smokers) will show evidence of conditioned positive affect and others (the non-smokers) won’t, whereas none of the participants in the other two groups will evidence conditioned positive affect. However, the variance we obtained for the liking ratings in the lips condition was not larger than that for the other two conditions. (The numbers in parentheses after each mean in the distributed tables are the standard deviations, a measure of variance.) It is also worth noting that, if the variance had been increased, it would have been more difficult to obtain statistical significance in the analysis of the groups’ means.
After one of the Friday labs, it was suggested that perhaps the smokers in the class ended up predominantly in the Lips group. This suggestion was based on the observation that the experimental packets were not fully intermixed, combined with the speculation that smokers might sit in particular parts of the room (i.e., that seating patterns of smokers/nonsmokers in the room also were not random). In other words, there might not have been a random assignment of participants to groups, such that the Lips Group consisted entirely of smokers (or nearly so).
The probability that the two semi-nonrandom events described above overlapped sufficiently to produce the Lips effect (i.e., an elevated mean with a standard deviation equal to that of the other two groups, with group sizes 29-37) is extremely small. Nothing is impossible, and I certainly agree it would have been better if the packet types had been better interleaved. But to propose that the seating patterns were blocked in just the right way to match the way the packets were blocked moves beyond likely coincidences.
Finally, to clarify the conditions under which a variable is a confound. IF there was a nonrandom assignment of smokers/non-smokers to the three groups, so that the Lips group consisted of smokers and the other two groups each contained few smokers (assuming about 50% of the total population was smokers), then smoking status was confounded with the manipulation of pen position. But such a nonrandom assignment of participants to groups defines a poorly-designed experiment from which one would not want to conclude much of anything. In particular, in order to demonstrate that there is, in fact, a conditioned positive affect for the lips pen position, one should have an appropriate control (Knees) condition consisting of smokers and/or a second Lips group consisting of non-smokers. Without at least one these, even in the unlikely event that there was a smoking status confound, it’s impossible to conclude that the elevated liking ratings in the Lips group resulted from their status as smokers.
I hope that these comments help to put “the smoker effect” in context. I applaud the attitude that “you can’t know for sure unless you’ve collected the appropriate data.” However, for a variable (e.g., smoking status) to merit much energy, there must be a reasonable likelihood that it: (1) is confounded with the manipulated independent variable and (2) can account for the pattern of obtained effects. There are good reasons to doubt that either of these situations were present in our experiment. I strongly urge you not to spend much time exploring this variable in your writing assignment.