Spearman’s rank always gives a value between -1 & 1, -1 being perfect negative correlation, and 1 being perfect positive correlation. Therefore my results gave a moderate positive correlation. This can be shown also on the graph. This graph has a pattern of points, meaning there is some correlation (no correlation is shown by an arrangement of no pattern). This pattern resembles a line from bottom left to top right, showing positive correlation, where as one value goes up, so does its corresponding value. Negative correlation would be shown the other way, going from top left to bottom right, where as one value goes up, the other goes down (and vice versa).
This graph, however, shows much less of a correlation. The previous graph showed the average values, which clustered due to averaging (using such a small sample of ratings meant they were more varied than they would be on a larger scale). This graph above, on the other hand, shows the rankings, which meant the small differences between the average values, shown clearly on the other graph, were irrelevant. Instead this graph only took into account how each score rated in terms of 1st, 2nd etc. This graph is a more accurate representation of the Spearman’s rank value I obtained earlier, as it shows some degree of positive correlation, but not enough to assume this was not just be chance.
To calculate my Spearman’s rank value’s importance, I then conducted a significance test. This is to see how likely this value is due to chance. I used a significance level of 5%, which is to test whether this value would be likely to occur 95% of the time if the experiment was repeated (therefore testing whether this value is unlikely due to chance). With my test being 1 tailed, and using 10 pairs of data, the critical value is 0.5636. This was found using a maths formulae book referenced later in my work. My result (p) is 0.51515151… (17/33). 17/33<0.5636 As p My full table of results is in the appendix, together with my full calculations.
The results of this experiment have been shown to give no support to previous hypotheses, such as those by Goffman and Walster et al. My value of Spearman’s rank was less than that of the critical value for my sample size and hypothesis; although only by around 0.05 (my value was 91.4% of the critical value). Although statistically this is reasonably likely to be by chance, it appears to be close enough to the critical value to warrant further repeats of this test.
It also appears on closer investigation that my final result was greatly influenced by the difference in only two couples: couple A and couple J. These could be anomalies, and this could be investigated through statistical methods such as standard deviation at another time. The difference in my results and those from previous studies could be because of many reasons e.g. small sample size, different age in participants or changes to original study. However, as there are no statistics easily available from these previous studies, it is hard to compare them with mine empirically.
One limitation of my study is that I only used 20 participants. This means my results could unreliable as they may not be representative enough for my whole target population. My target population is also so small that it is difficult to compare to previous studies. Another problem with my study was that me participants were in the age bracket of 16 – 19, while the people in the photos I used were up to age 30. Therefore it was difficult for my participants to rate the attractiveness of the people in the photos when some of them were nearly twice their age (other studies have shown that the older someone is, the less likely it is that they will be seen as attractive). A further limitation is that conducting the study in Exeter was restrictive, as it has very few ethnic minorities compared to many bigger cities. Using participants just from college also meant a more middle-class viewpoint.
Improvement As mentioned in limitations, I could improve on my research by using different age ranges. However, using photos of people aged 16 – 19 would have the possibility of needing parental consent, as well as having couples who are less likely to be in a serious relationship. Instead I could improve my study by asking people (either students or not) who are 18-30 to rate people who would be in the same age bracket as them – therefore eliminating the limitation of the previos study.
Further Research One way in which I could alter my study would be to investigate whether the matching hypothesis is true for same-sex best friends as well as couples, or for gay couples. It has been suggested that while we pick potential partners on physical attractiveness, we pick potential platonic friends on their social qualities. However, it would be interesting to see whether friends judge each other on ‘what is fitting’ physically too, as Brown proposed we do with partners.
If I was using just my results alone, I would not have any evidence to support the Matching hypothesis theory. However, as my results were reasonably close to the significance level, I would not dismiss the theory, as it would warrant further investigation. Taking into account the previous studies aforementioned in my introduction, I believe the Matching Hypothesis can be supported, to a certain extent. As my study was only on a small scale I believe previous evidence to be more reliable on the subject.
Walster, E., Aronson, V., Abrahams, D. & Rottman, L. (1966). The importance of physical attractiveness in dating behaviour. Journal of Personality and Social Psychology, 4, 508-516. Walster, E., & Walster, G.W. (1969). A new look at love. Reading, MA: Addison Wesley. Murstein, B.I. (1972) Physical attractiveness and marital choice. Journal of Personality and Social Psychology, 22 (1), 8-12. Brown, R. (1986) Social Psychology (2nd Ed.) New York: Free Press.