garbage can model of organizational theorizing
I’ve noted a seemingly popular model of organizational theorizing of late: the garbage can model. The garbage can model of theorizing consists of efforts to (randomly) match and unite potential candidate theories with potential correlations in one’s data. What I have seen scholars do, is, first mine their data for various possible correlations. Then scholars sift through various theories for possible explanations for these correlations in their data. And, then, voila!, thats the theory. Yes, that’s it. Not too different from how Cohen et al conceptualized an organization’s efforts to match potential problems (in this case: correlations in data) with extant solutions (in this case: 70s theories).
In all, I don’t know how prevalent the garbage can model is, it’s of course hard to tell post hoc, once a paper is published. I suppose one’s shameless prostitution to various theories might show up in a person’s research program, or, lack thereof. Theorizing, of course, is a messy process. But, somehow the garbage can model is unsatisfactory, to say the least. It’s so opportunistic, so void of perspective, so void of cookie-pushing, so phenomena-driven. So lame.
And, of course, I would personally never succumb to anything like this.
tf
April 28, 2008 at 8:06 am
I’ve been talking about this trend for years. We are not problem- or theory-driven but rather data- and method-driven. Theories just sit on the shelf as post-hoc explanations for ’significant’ (p<0.05) correlations. I like to call it what it is…pseudo-science.
We all know why it occurs. Pressure for publications, no time to gather primary data, reviewers who are impressed by large sample sizes and statistical methods they don’t understand. The good news is that the scholaticism movement in the middle ages dragged on for centuries so if we keep quiet about it then it may be quite a while before someone points out that the emperor has no clothes.
Steve Phelan
April 28, 2008 at 7:49 pm
I like to think that there are two ways of doing theory-testing using statistical samples:
1. Honest applications of the hypothetico-deductive approach, where the researcher theorizes before looking at the data. We can never follow the hypothetico-deductive (HD) method exactly, probably not even approximately, but one thing we can do is take the theory-before-data principle seriously in our research practice. Some of us do.
2. “Purely rhetorical” applications of HD, where one’s research is reconstructed post hoc to reflect the normative methodological idealization. HD in this case is primarily an external referent of prestige.
Option (2) is pretty much a special case of loose coupling: put HD in the display window and no-one will challenge you with what’s going on backstage.
The problem is that we can’t tell by looking at a manuscript which strategy the author is using, hence, I have no idea what the percentages are. Also, there’s a fine line that separates (1) from (2).
If there’s an emperor who’s naked out there, gosh darn it, we have no way of telling who it might be.
Mikko Ketokivi
May 6, 2008 at 11:13 am