Table 1

Guide to Understanding the Article

Section

 Tips, Comments, and Problem Areas

Abstract

Note that there isn’t one—It wasn’t required in 1969. Your paper, however, should have an Abstract.

Introduction

2^nd paragraph: Coleman and James assumed that group size has nothing to do with people leaving or joining a group. Thus, according to their model, about  10% of people leave a group within a certain time, that percentage will be the same regardless of whether the group size is 10 (in which case, the group would lose 1 member) or 10,000 (in which case, the group would lose 1,000 members). Furthermore, according to their model, the size of the group doesn’t have an affect on people’s willingness to join the group. According to their model, a person is just as likely to join a 2-person group as a 20-person group. This assumption—that group size has nothing to do with people joining a group—is tested in the Milgram, Bickman, and Berkowitz study.

Next to last paragraph: Note that in 1969, sexist, noninclusive language was allowed. However, such language is not allowed today. So, don’t use “he” to refer to a person who may be a “she.”

Method

Note that analysis of their data would have been easier if they had used equally spaced levels of crowd size (1, 4, 7, 10, 13, 16) or proportionally spaced levels (1, 2, 4, 8, 16, 32).   

Data Analysis

Today, this section would usually be in the Results section. It would have been nice to know that (a) the judges made their ratings independently and (b) that the judges completely—or almost completely—agreed with each other.

Results

You can see that a straight line (linear trend) fits the data points for those who stop fairly well. You can see that, for those who look up, there is alos a  linear component. In addition, however, there is a curved component (in this case, an upside down “u”-shaped [quadratic]) component, suggesting that adding members to a small group has a larger impact than adding  members to a large group.

As we mentioned before, doing the statistical tests to determine which components were significant was difficult because the authors did not choose the right levels of group size. If you want to test for the linear components, you can follow the instructions in “Using Table E-4 to Compute Trend Analyses” in Research Design Explained using the following weights: –5, -4, -3, -1, 4, and 9, and using 148 as the weighting factor for the linear effect. Note that the sum will be the proportion times 5. Thus, for the one-person crowd for those who stop, the condition sum would be .20 (5 X .04). That sum would be multiplied by –5 to get -1 (-5 X .20) = -1. After multiplying each condition’s sum by its weight, you would add up all those sums and then divide by 5 (number of observations in each condition) X 148 (the weighting factor for the linear effect). You would then divide the result by MS Within (0.004). The F you get should be within 10 of the F  of 101.7 listed in the article.

Discussion

Note that the authors seem to suggest doing a study in which you manipulate two variables: (1)group size and (2) interest of the scene. To our knowledge, such a study has not yet been done.