Collins, H. M. 1992. Changing Order: Replication and Induction in Scientific Practice. Chicago and London: University of Chicago Press.
Review by Michael Beach
The original publication of this book was in 1985. I read the updated 1992 edition. The focus on this book is an examination of how we perceive order and our need to replicate. In science, replication is important in particular because when one makes a factual claim, it must be based on evidence and any empirical evidence must be replicable. Collins shares three chapters of specific examples to make his point. He discusses the TEA-Laser, detecting gravitational radiation, and paranormal experiments.
Order and perception are subjectively designed. Whenever there we examine a large number of things we try to categorize. In science, categorization is necessary to make sense of differences. The problem with categorizing is that things don’t really exist in discrete groups of things, but rather as a continuum of individual things. Scientists attempt to define categories by defining attributes. Whenever one creates experiments that either attempt to define a category, or attempts to make conclusions about subjectively defined categories, it becomes difficult to take the next step. That is, it is difficult to draw general conclusions about specific empirical outcomes. This is the problem with inductive reasoning. As soon as one attempts to apply a finding in a specific situation to larger groups, the generalized conclusion will inevitably have to include exceptions.
The major concern Collins points out about replication is that each group defined and included in an experiment will influence outcomes differently. Generally, empirical work includes undocumented steps. “My concern is not how we could be certain in principle about induced regularities but about how we actually come to be certain about regularities in practice” (Collins 1992, 6). For example, some processes are taught from one lab worker to another through tacit practice. Explicit documented procedures carry a project so far, but there are different ways of doing lab tasks. Practices vary from lab to lab and from practitioner to practitioner. This means that replication includes variation, and variation leads to different exceptions when attempting to analyze and generalize findings.