The causal conundrum
Who is to blame for the over assumption of causation on new scientific findings?
We live in a world obsessed with cause and effect, but not everything has a causal relationship. A correlational study was conducted in Maine that showed a 99% correlational relationship between divorce and per capita margarine consumption. However, the above study does not have a causal relationship, but simply a correlational one. Correlational studies, unlike experimental studies, do not manipulate a variable and have random assignment. Correlational studies simply measures the extent to which two variables relate to one another. The existence of correlational studies allows for the science community to study factors that would be impractical, unethical, or impossible to measure otherwise. It is imperative to note that correlation does not equal causation, and therefore correlational studies should be careful to imply otherwise as it can be misleading. So, if you’re living in Maine, have no fear, you can still eat as much margarine as you wish.
If the science community agrees that correlation does not equal causation, why do correlational studies still get misinterpreted as having causal outcomes? One aspect that leads to misinterpretation is a lack of understanding by journalists who are reporting on new findings. Additionally, titles and abstracts are often written in a misleading manner that implies a causal relationship even when one is not present. Researchers often do this because it allows their paper to have a more ‘flashy’ and intriguing title.
Schellenberg (2019) sought to test how often causation is inferred in psychology and neuroscience papers that explored musical training. Why? A large majority of the non-experimental research that exists on musical training often implies, if not states, a causal relationship. Therefore, Schellenberg sought to conduct a literature review to test how big of a problem this really is. Additionally, he sought to explore if the problem of inferring causation is more common is psychology or neuroscience. A total of 114 articles were compiled by one research assistant on the basis that they had a correlational design, which “tested for associations between musical training and nonmusical ability, brain function or brain structure.” After compiling all 114 articles, the research assistant designated all articles as either a psychology or neuroscience article. This designation was based on whether or not a neuroscientific method such as EEG, MRI, etc. was used, or if the article’s title contained words such as brain, neuroscience, or neuropsychology. Of the 114 articles, 62% were written by neuroscientists.
Surprisingly, a large majority of the non-experimental research that exists on musical training often implies, if not states, a causal relationship. After articles were collected and designated, the important step of the research came: determining whether a causal relationship could be inferred from the title or abstract. A second research assistant was assigned to designate the articles as either inferring causation or not inferring causation. In order for this research assistant to remain impartial, they were not told that discipline was a factor in the overall study. All titles that contained causal verbs such as “enhance,” “strengthen,” and “limit,” were coded as inferring causation. However, articles with a causal verb and quantifying term such as “may,” “perhaps,” and “possibly,” were coded as not inferring causation. Additionally, for any article that contained titles that were unclear, the research assistant was instructed to reference the abstract. Furthermore, all articles which designated musical training as an outcome variable, as opposed to a predictor variable, were also coded as not inferring causation. In order to ensure accurate results, a third research assistant was also assigned to code the articles as inferring or not inferring a causal relationship. This addition of a second coder allowed for inter-rater reliability, the comparison of the ratings of different coders. The two raters had an initial agreement of 76%, and after meeting to discuss their differences the agreement rate jumped to 96%.
The results showed that in ~61% of articles written by neuroscientists inferred causation was present. However, in articles that were written by psychologists, inferred causation was present only ~49% of the time. One possible explanation for why inferred causation is so common in music training literature is the emphasis music training research often places on cognitive abilities and brain imaging.
The research shows there is a clear problem of inferred causation across both neuroscience and psychology research.This research is important because it makes both academics and journalists more aware for the future. Journalists are often the ones to disperse new scientific findings to the masses, but they rarely read entire scientific articles, instead they are more likely to reference just a papers title and abstract. Although, those working in both psychology and neuroscience academia cannot control what journalists write, they can control the titles and abstracts of their papers.
Additionally, scientists as a whole can put in greater effort to ensure that press releases about their articles, which the majority of news articles are based off of, contain accurate information. Additionally, research found the presence of scientific jargon in summaries written for the public exceed an acceptable amount, i.e. to much jargon was present in the summaries for a lay man’s understanding. Thus, the blame for misinterpretation lies both with how the scientists are writing and with how the journalists are interpreting. Therefore, it is important the scientists are more conscious of what and how they write, and it is also equally important for journalists to ensure they are accurately reporting on new findings. As doing so will hopefully help ensure that accurate information about new scientific findings is portrayed to the masses.
Have you ever read a misleading article about new scientific findings? Comment the link below!
The above blog post is based on the findings from the following article: Schellenberg, E. G. (2019). Correlation = Causation? Music Training, Psychology, and Neuroscience. Psychology of Aesthetics, Creativity, and the Arts.
