A “systematic error” in a mental health database has led to the retraction of a 2017 paper on how people with psychosis process facial expressions.
Joana Grave, a PhD student at the University of Aveiro, in Portugal, and her colleagues published their article, “The effects of perceptual load in processing emotional facial expression in psychotic disorders,” in Psychiatry Research, an Elsevier title.
According to the abstract of the paper:
Psychotic … [p]atients have difficulties in identifying facial expressions and appear to be highly sensitive to the presence of emotional distractors. Yet, no study has investigated whether perceptual load modulates the interference of emotional distractors. Our goal was to test whether psychotic patients were more sensitive to irrelevant emotional stimulus, even when the task demands a high amount of attentional resources. Twenty-two participants with schizophrenia or schizoaffective disorder and twenty-two healthy controls, performed a target letter discrimination task with emotional task-irrelevant stimulus (angry, happy and neutral facial expressions). Target-letters were presented among distrator-letters, which could be similar (low perceptual load) or different (high load); participants should discriminate the target-letter and ignore the facial expression. Results showed that patients were more prone to distraction by task-irrelevant stimulus, especially under high load, suggesting difficulties in attention control. Moreover, in psychotic patients, happy faces caused higher interference with the task, whereas neutral and angry faces resulted in less interference. These findings could provide innovative approaches regarding attentional deficits on social contexts in patients with schizophrenia spectrum disorders.
But Grave’s team said that, sometime later, they realized that the facial expression data on which they’d based their analysis had been scrambled. Per the retraction notice:
This article has been retracted at the request of the authors and Editor-in-Chief.
The Authors have recently discovered an error in the database that affects the results and, consequently, the discussion, conclusions and abstract of this article on the effects of perceptual load and task-irrelevant facial expression during a target-letter discrimination task in psychotic disorders.
As the notice explains, the researchers discovered the mistake:
during the analysis of new data using the same experimental task, but collected from a distinct clinical sample, and involves a systematic error in the database: pictures of one actor were coded as depicting the same valence (angry, happy, neutral) during the programming of the experimental task, regardless of the real facial expression portraited by the picture in each trial. As a result, the information provided by the ‘Emotion’ variable in the database, created by the E-Prime software for each participant during the performance of the task, was not correct. Because the Authors used the ‘Emotion’ variable to perform the statistical analysis, the results of lower accuracy for individuals diagnosed with psychotic disorders in trials where the task-irrelevant stimuli are happy facial expressions – as revealed by a significant interaction between group and facial expression (p = .016) and between group, perceptual load and facial expression (p = .027) – are not correct, thus partially affecting the discussion, conclusions and abstract of the manuscript.
The authors added that they were able to correct the error and reanalyze the data, and that they will try to republish their work:
The error had no influence on the design and procedure of the experimental task, as the order of the pictures was fully randomized per participant.
And they close with a bit of advice to other researchers:
The Authors apologize for the inconvenience caused by these mistakes and agree with the retraction. The Authors advise all researchers to pay special attention when scripting/coding the stimuli to be used in the experimental tasks, and to confirm all scripting/coding with independent researchers, whenever possible, to prevent mistakes (Strand, 2021).
The notice ends with an apology from the journal for failing to detect the error during the peer review process.
Writing in response to praise from another researcher on Twitter, Grave said:
She added:
Grave said she found inspiration in the story of Julia Strand, with whom Retraction Watch readers may be familiar.
Like Retraction Watch? You can make a one-time tax-deductible contribution or a monthly tax-deductible donation to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
I hope to be as brave if one day I find a mistake in my data.
If you work with enough people, you will eventually be involved in a retraction. Happened four times to me. There is no shame provided you were not acting dishonestly. There really is no problem if all you do is 1) run statistics & 2) read pathology, like I did. What amazes me is how grossly dishonest persons get caught & still continue to do scientific work.
So honesty is everything. If you did not cook the data or fake things, you will suffer no ill consequences.
I do admire the authors for doing the right thing in correcting a database error. But the more significant errors were (1) assuming that this simplistic experimental task in any way mimics the real task load of anybody engaging in a real social interaction, and (2) collecting data from a tiny sample of individuals in each arm of the experiment. In general, fixing analysis errors will not solve the problem of bad measurement, which seems to plague some research disciplines. Like the old joke, in the retracted study the food was terrible, and the servings were so small!
Good luck finding 100+ people with schizophrenia for a research study. Sometimes 20 is the best you can do.
Sure, research on neuropsychiatric disease is hard. But the incidence of schizophrenia and schizotypal syndromes is very high, ~1% of all humans globally. In WEIRD countries, the incidence may be higher, and a large proportion of those people are in medical care of some kind and readily recruited for study. It’s not that hard.
Even if one can’t round up hundreds of people for such a study, getting a noisy signal from a bad measurement of a small sample is still a bad noisy estimate of the effect. No one should take the estimate of the effects in that paper seriously even after correcting the database error. As they say, papers like this fill a much-needed gap.
I just hope the academia is more tolerant towards retraction. If the retraction is related to data manipulation, deliberate fabrication, or plagiarism, this is not acceptable.
However, if some mistakes are indeed due to some sort of innocent mistakes, why don’t give the researchers a chance to revise.
Which is why the biorxiv model should be the future of academic publishing. Postpublication peer review and revisions should be the standard, not just in cases of significant mistakes.
Correction: I believe you mean *Julia* Strand.
Fixed, thanks!
Are you kidding, Dr. Oranski? Scientists have panic of L. Schneider. I heard recently a researcher telling that she was not afraid of retracting or correcting a paper but of him. So she was not going to contact the EIC because she was scared of LS. There are more and more scientists thinking he’s damaging science, not improving it.