As with Steveson's data, our class data was very interesting and seemed to disprove a lot of our prior hypotheses about how this study would run. The first thing that was very striking was just how different the success rates were based on the target person. The fact that only 25% of Antonio-bound folders made it while 80% of Susan's made it, convincingly disproves my hypothesis from Part I that the target characteristics would not matter that much in whether or not the folder would reach the final target. The results seem to show that Susan is much easier to reach than Antonio and at a shorter average length (3.25 vs. 4.5). This leads in my mind to two possible conclusions as to why this is the case. The first may be that Susan is much more prestigeous or well-known in the Penn community and thus according to Monge has a higher in-degree within the community. The other is that the Susan Yoon group was significantly better at selecting their first target than Antonio's group. When looking at the data, there are a few dimensions that seemed to significantly differ when regarding characteristics of the first alter. One interesting difference is that the Susan Yoon group was much more likely to pass the folder first to a member of the same sex (8 out of 10) than the Antonio group (3 out of 8). Another difference seemed to be that (6 out of 8) people in Antonio's group passed it first to fellow students (while 1 passed it directly to faculty) while only 5 out of 10 of Susan's group passed it first to another student (and 4 handed it to faculty). As far as the first number is concerned, the literature seems to make no mention that choosing a second alter of the same sex increases final success. Milgram simply says that exchanges are much more likely to occur between member of the same sex and Killworth's paper on accuracy simply states that all network members only choose the correct alter (in order to reach the target on the shortest path) 50% of the time, without mentioning if same-sex passages were typically more accurate and/or successful. Thus, this doesn't seem to explain why so many more folders reached Susan. The second fact is interesting in that 3 out of 8 completed chains in the Susan group were handed first to faculty and then proceeded to Susan strictly through faculty. This is very similar to the Stevenson study in which 88% of the time, a faculty member, grad student or staff member who received the folder was able to complete the chain to the administator. However, in my case (with Antonio) I was the only one in my group to pass it to a faculty member and my chain died after that faculty member. Thus, there are no guarantees that passing it to a faculty member will necessarily be more likely to lead to a completed chain.
This data seems to suggest that the most likely reason for the major difference in completion rate had to do with the characteristics of the target. Unlike the Stevenson study, our study involved reaching a facutly member or staff member rather than just an administator. Thus it makes sense that our average chain lengths were longer than their 1.25, considering that administration is typically more central to the university as a whole. In our case Susan (a faculty member) seemed much easier to reach than Antonio (a staff person). While the sample size is very small, the data show that faculty members, students, and staff people were all represented as the final link when the target was a faculty member, yet only students represented the final link when the target was a staff person. A possible explanation is that faculty (being more presitigious) have a much higher in-degree than staff people and are much better known throughout the university community. This would make sense considering faculty teach (and thus have contact with students) and work together in departments that are often interconnected (which gives them contact with staff and faculty). Staff members on the other hand (like Antonio) are much more likely to only be connected to those in their department, and further only with students who specifically work in their department. This explanation is very convincing and plausible considering that Antonio has actually been here one more year than Susan, and common sense would dictate that he'd be able to make more contacts within an extra year. Clearly though, in this case affiliation of the target seems to be a much more important determinant of success than the amount of time the target has been a part of the community.
My personal data makes much more sense when I consider the fact that affiliation of the target is so important. I figured that passing my folder straight to a faculty member would make it much more likely for the chain to reach Antonio, than first passing it to a person of any other affiliation. However, my chain literally stopped after Diana (despite the fact that I thought the chain would only take a week and a half and go through four links!). Unforutnately, I was left with a dearth of information, considering that the postcard Diana sent in was not postmarked and thus, I have no idea as to when she got around to doing the task. However, assuming that she is trustworthy and didn't just leave the folder in a pile and forget for a few weeks (which is a possibility), it seems that once she passed it on the next respondent was either just very lazy or indifferent or just completely clueless about where to turn next to find a relatively isolated staff person. This was not the case at all in Susan's group where once the folder reached a faculty member it reached the target 75% of the time and there was only one case like mine where the chain stopped after the first exchange. Furthermore, unlike the classic Milgrim studies only one of the penultimate alters was responsible for delivering more than 1 folder to the target (June C. who only delivered 2). Together all of these data suggest very convincingly that I was completely wrong, and that affiliation has a major impact not only on the length of the chains but also on whether or not they were completed. No matter the characteristics of who you pass it to, it seems much easier to reach a typical faculty member than a typical staff member.
Another interesting finding is that while the completion rate of Antonio's group was very comparable to Stevenson's data (25% completion for Antonio vs. 27% in the Stevenson study), a much higher percentage of folders made it to Susan (80%) than we see in any of the literature. The highest completion rate we see in Stevenson's literature review is that of Lundberg, who found that 57% of folders reached the target in an organizational study with most studies in the larger society averaging between 20 and 30% completion. Furthermore, Stevenson predicted that only 50% would reach the target, with the target working in the same building as many of the participants! Thus, it's very surprising that so many people were able to reach Susan in a university community of more than 20,000. This is very hard to explain, even if I am correct in asserting that her affiliation as a faculty member leads to a much greater chance of the folder reaching her. A possible reason for this is that there was some very close connections between the departments of many of the starting students and the Graduate School of Education. Unlike Stevenson's study our folders were not handed out to starting participants at random and instead we had a large proportion of our starting participants sharing the same department and school (in fact only one participant was not a communication major and all were in the school of arts and sciences). Thus, due to our starting characteristics the study may have been biased considering that comm students, staff, and faculty are much more likely to have connections at the school of education (which houses many undergraduate classes and likely shares some common study interests with the communication deparment) than they are to a medical institute that is not as closely related to the department and that many students and faculty have never even entered. Another potential bias that differs from Stevenson's study is that the students in this class who started the chains, already knew the basic structure and "tricks" to these studies, and were furthermore probably more motivated to follow up with their first contact than a random student would be. I, personally was not as interested in making sure to remind my contact to complete the task as I wanted to see what would happen naturally and if the folder could really progress on its own. Clearly, in my example my folder seemed to follow the same dead-end path as most of the folders in the Stevenson study. An interesting follow-up to this study would be to ask how many people in each group communicated with their 2nd alter in order to remind them about this project or check on the status of the folder. Then we could see if this bias has a major effect or not. In any case something must be accounting for this tremendously high success rate with Susan Yoon and these seem to be the most likely causes based on my prior experiences and the data.
Furthermore, in the Antonio group we see that the two chains that were completed were started by 5th year students. Stevenson's study showed that while there was a clear isolation effect for freshmen in the sample, seniors starting the chain were less likely to complete it than juniors and sophomores, and most chains that were passed from student to student went to upper-classmen before reaching the target. Our data don't seem to match up, especially considering that these two participants not only passed down (which nobody in the Stevenson survey did) but both passed it down first to sophomores! In one of the chains three sophomores were involved before the folder reached an upper-classman. Even though we can expect that 5th year students may pass down (considering most kids graduate in 4) it is very intriguing that they passed so much lower and furthermore that it worked. We also see in the Susan group that more students passed the folder on to younger students than those in their same year and none passed it higher! Perhaps again this is due to the bias that most of the class consists of seniors and juniors and hence it's much harder to pass it higher. However, the fact that more people passed it lower and we had a better overall success rate than the Stevenson study really challenges the notion that upper-classmen are more involved in these transfers and more vital to their final success than under-classmen.
Finally, it's important to point out what this study says about tie strength and who we trust and expect will be able to best help us with this task. As I stated in part 1, Granovetter would say that it's important to utilize weak ties because they are the best outlets to diverse resources and social circles. Burt however, would note that while structural holes are important, trust is also very key in asking for help in an endeavor. In my case, I was hoping that trust would win out over a weak tie, and unfortunately with no postmark on the postcard I received, I have no way of knowing if Diana was actually trustworthy or if the next alter simply was clueless. When considering all of the data it seemed that most people also preferred to hand off the folder first to a moderate-very strong tie (13 of a possible 16). Six of these chains were completed (with two data points outstanding) while only 2 delivered first to very weak or weak ties were completed. This is evidence that strong ties can be very helpful in beginning a task like this and are usually quite trustworthy in helping you out. An important potential weakness with this result is that in much of the literature there is simply a dichotomy of strong and weak rather than a gradual five-point scale. With a scale like this, it's very hard to distinguish what a "weak tie" is in the Granovetter sense and thus we see that simple categories can have a major impact on our interpretations of data. Regardless though, this shows that there is some ambiguity about the best way to go about choosing a second alter and for that matter choosing the next person at any part of the chain.
In conclusion, the take home point from our University experiment seems to be that target characteristics really do matter. While our sample size was very small and clearly biased in many ways, these data offer some very intriguing results that are worthy of further investigation.
Part 1: http://www.mysocialnetwork.net/blog/481/r32/2006/09/small_university_part_ia_big_d.html