blog

Guest post by Alex Holcombe, Nicholas Brown, Patrick Goodbourn, Alexander Etz and Sebastian Geukes

A study published by Christopher Thorstenson, Adam Pazda, and Andrew Elliot in Psychological Science in August 2015 suggested that sadness impairs color perception, based on two experiments they conducted. Almost as soon as it was published, many people pointed out problems with the study, on Twitter and on blogs. But there are also broader problems in this research area that we would like to see addressed. Our current article “Does sadness impair color perception? Addressing the flaws in published research” examines both the details of the Thorstenson et al. study and looks at the bigger picture.

Shortly after the Thorstenson et al. manuscript was published, five of us got together over e-mail to write a letter to Psychological Science calling for retraction. Before submitting, we contacted the first author, Christopher Thorstenson, and after a few exchanges of e-mails, he told us that he and his colleagues were going to retract the paper. Three months after publication, in the midst of a growing chorus of criticism on social media, the article was duly retracted.

In their retraction notice, the authors described some problems with the data in their second experiment. They suggested that after revising their second experiment to address the problems that they noted, they would seek to re-publish their original Experiment 1 with a revised Experiment 2. When we saw the retraction notice, we noticed that only a few of the problems with the experiments were mentioned. In our view, the retraction notice did not describe all the major problems with the study, and some of these issues are not uncommon in this research area.

There has been reason to be concerned about the design and implementation of studies not unlike those of Thorstenson et al. for several years; we would like to see a general improvement in this research area. For example, a long-running research program has investigated the effect of exposure to colors such as red on test and athletic performance. But like the Thorstenson et al. study, most of these studies have compared only one example involving the color of interest to a more neutral color. In addition, in the studies of red on test performance, exposure to just one particular red color (i.e., a single combination of hue, saturation, and brightness values) is typically compared to another color. Such a study design makes it impossible to know whether the results on performance are really about red generally, or just a particular red compared to a particular neutral color.

To describe remedies for this limitation and others, we revised our original letter into a full-length article, published here at F1000 Research to explain in more detail the additional problems not mentioned by Thorstenson et al. in their retraction. We hope that our article will help researchers in this area to improve their methods.