The Hawthorne Effect - Case Notes

SCI Home | SCI Faculty Page | Cover Sheet | Reference List | Users Notes

Author - Tim Brazill - Summer 1998

Background - The Hawthorne Effect refers to the tendency of human subjects to modify their behavior (from what might otherwise be expected) when they know (or believe) that they are part of an experiment (or are simply under observation). The name of the effect stems from the locale of the original research that first brought the problem to light: The Hawthorne Works (Chicago) branch of the Western Electric Co. The management of this plant began research in 1924 to discover ways to improve employee productivity (and later, morale). The first segment of the research (which lasted until 1927) focused on the effects of illumination in the work place (Roethlisberger & Dickson, 1941, 14-18). They discovered that changes in illumination, generally (though somewhat unpredictably) produced increases in production, regardless of whether the light was increased or decreased. They initially attributed these results to a variety of uncontrolled factors, and thus sought to refine their techniques. They replicated the study in another department, and this time introduced a control group. Both the treatment and control groups showed the same increase once again. The researcher's realized that they thus could not attribute the increase for the treatment group to the changes in illumination, but did not yet realize the source.

This brings us to their next step (and the principal focus of the case study), which was to further control the environment by isolating a very small (n=5) group of workers for an intensive period of study. The workers were recruited for their cooperative nature (ibid., 21-22), and then were told of the importance of the experiment (though they were urged to work at their "natural rate", ibid. 32-33). You should note once again that this is not ideal research protocol any longer.  Anyway, they then subjected these five women (two of which were replaced after the first seven months) to two years of various test periods, each incorporating a variety of rest periods, breaks, and changes in schedule. As is clear from the results (ibid. 76), their output showed the same general increase over the course of the study. The researchers proposed and tested five new hypotheses (ibid., 87-89) to explain the results, each of proved somewhat unsatisfactory. Finally, after intensive interviews with the women it became clear that what they were responding to was the fun, excitement, and/or unusualness of the experimental situation itself. Thus, an interesting facet of human behavior (see also: social facilitation), and a major potential source of bias in human behavior research was discovered serendipitously.

Important Notes:

• This case is essentially an object lesson in what NOT to do when experimenting with (or obtrusively observing) human beings. This point should be driven home at the end of the case (using discussion of their findings on Group Worksheet #3).
• In setting the background for the case, I would NOT discuss the earlier illumination studies, as it gives away the solution (the use of control groups, even though Roethlisberger and Dickson didn't realize it).
• It might be helpful to point out how the workers were paid (see pages 12-14), and that the pay rate change alluded to in Period 3 was meant to make sure they were NOT getting any more than what they had gotten in the regular shop floor (see pp. 34-36).
• It might also be helpful to describe their normal working conditions as specified in Period II (page 33).  Note that they were provided with a noon-time 45 minute meal each day.  The "lunches" referred to in Table III in the initial reading are actually what we would call snacks.
• Some students will want to claim that the increases might just be a result of increasing skill over the study period, but the women had been working at or near the shop average output for two years or so (on average) prior to the experiment. See pace 23.
• While students are completing worksheet #1
  • You might want to try to tease out of them the reasons for the strict control of the environment (to rule out the effects of extraneous variables that might intrude on the cause-effect relationship we're trying to establish).
  • The importance of the first two periods is, of course, to establish a baseline of behavior against which you can judge the effects of the testing periods.  You're setting them up here, of course, because what is also needed is a control group.
  • The second question is a little vague. "Increasing or decreasing" compared to what? The baseline or the previous period? I leave that up to the user to clarify as you see fit.
  • I'd discuss their predictions as a class, prior to moving on.
• While students are completing worksheet #2
  • Students might point out that the sample is small, and so the study is flawed from the beginning. This might be true if the researchers were trying to generalize from this ONE study to a much larger population. You might bring up the importance of replication. More to the point though, try to keep them on the track of just seeking to explain the behavior of these 5 individuals.
  • Hopefully, their predictions will NOT match well, and thus they will be surprised by the results, and their curiosity piqued.
  • I'd also talk about their potential explanations before giving them the punchline.
• Group Worksheet #3
  • You might want to have them work on this at home or as a second class-day assignment (brainstorming without the Carey text might be fun, but I don't know if they will come up with the control group idea on their own).
  • Some solutions include
    • more unobtrusive observation or field experiments (but this lessens control)
    • deliberate deception or the use of a cover story (but this raises ethical concerns)
    • asking the subjects to be "faithful" (but note, Roethlisberger and Dickson did this and it didn't work)
    • the use of a control group that experiences all the same attention but none of the independent variable manipulations (this is the preferred method).
  • Note the inclusion of the related idea of the "placebo effect"
  • And you can wrap it all up with the role of serendipity in science