This Or That: How Useful Are Dichotomies Really?

Last week I posted an article about prevention vs promotion thinking, and how most people will largely fall into one category or the other, which you can read here. A lot of people really resonate with this kind of thinking, because it makes everything simple. The simpler things are the better we like to feel we understand things, especially when it comes to human motivation and behavior. But is this kind of dichotomous thinking always a good thing?

A dichotomy is when two things are so clearly opposed to one another they are considered opposites. In a dichotomy those two things cannot be similar to one another, and rarely interact with their respective opposing side. It's very simple to explain dichotomy and dichotomous thinking in pairs of opposites: black or white, light or dark, masculine or feminine are just a few key examples of dichotomies. It’s the idea that you're either one or the other – not both.

Why is this kind of thinking so ever present then? It's easy enough to explain, but why does it exist in our brains in the first place? Well dichotomies are very efficient at categorizing information, which is what our brains like to do. When we categorize information, we can better understand the world around us and make better decisions on how we choose to interact with that world. We also are taught many abstract concepts when we're younger in terms of dichotomies. Because it's so simple for our brains to use and understand, you can teach a child certain emotions, thoughts, or feelings are that we can't always easily explain using just words. Sad is defined as the absence of happy and vice versa, wet is defined as the absence of dry, dark is defined as the absence of light.

So is this type of thinking always a good thing? Well the obvious answer is no and the intuitive reason is because things are often more complicated than just a "one or the other" scenario. Happiness and sadness exist on a spectrum, much like light and all its different hues and shades. The simplistic way of categorizing this information doesn't always capture the whole picture.

MacCallum, Zhang, Preacher, and Rucker in their 2002 study published in Psychological Methods entitled On the Practice of Dichotomization of Quantitative Variables test the benefits and detriments of using dichotomous thinking in scientific research. What MacCallum and colleagues did was they took existing quantitative studies and ran the statistical analysis once when the X values were distributed according to the data, and once when the X values were split into a dichotomy, providing high and low values at the median. They found that a "comparison of the results of analysis of the association between X and Y before and after dichotomization of X shows a distinct loss of effect size and loss of statistical significance". This means that when you take complex variables and try and simplify them into two categories, you often lose a lot of the really important and meaningful stuff in the process.

They conducted similar tests across numerous different cases and found comparable results. When dichotomization is used, it takes away from the significance of the results because they're no longer capturing the whole picture, but a much more simplified version of it. They found that the biggest culprits of dichotomization were those using self-rated psychological scales. Things like levels of depression, anxiety, self-esteem, and narcissism were taken from a scale where you rate yourself from 1 - 5 for example, to "depressed or not depressed", "anxious or not anxious", etc.

You can see how this could be problematic. People with mild or moderate depression/anxiety are all of the sudden placed into one category or the other. MacCallum and colleagues go in detail about the defenses of dichotomization of variables and a few of them have some weight. The idea that results are "more impressive" and statistically significant when variables are split into a dichotomy is one researchers will use a lot. You may be thinking this is terrible and why would they do this it's so unethical! Well yes it definitely asserts something as true that may not be completely, but remember, these researchers need funding and publications. A lot of times journals won't publish studies unless they're statistically significant, because who wants to read a study where the null hypothesis isn't rejected and no statistically significant effect happens? This is potentially a problem with academia at large and not necessarily the individual researchers (although I’m sure individual culpability exists).

There are many other defenses of dichotomization, such as dichotomized variables are more reliable, dichotomies bring light to underlying categories of individuals that otherwise wouldn't be represented, and dichotomization resulting in higher levels of correlation, all of which MacCallum and colleagues dissect and debunk. The two instances in which they conclude that dichotomization of variables is justified is when there are clearly two very different categories that are distinct from one another, and when the distribution of numerical data is extremely skewed. The example they give for the second situation is if participants are asked how many cigarettes they smoke per day, and a very large portion of participants answer "zero" and the rest give a variation of values that are nonzero. This would potentially justify splitting the categories into "smokers" and "nonsmokers", which would be useful for understanding the data. Although even then they say that subsequent research should recognize this dichotomy and test for some of the intricacies of that middle group, which consists of very occasional smokers, that gets lost in the process.

They ultimately conclude that dichotomization is rarely justified in quantitative research, as so much is lost in the process. This rings true for other aspects of life as well. One of the biggest dichotomies in our country right now is democrats vs. republicans. This is useful for trying to categorize and understand large groups of people as a whole but loses so much of the nuance between them. Think about it, whatever political affiliation you align yourself with, would you want to be likened to the absolute most extreme person in your party? But because it's so easy for us to dichotomize, we very quickly and often lose the intricacies and nuances that make a person unique and begin making assumptions about them. This is true of gender, race, religion, and so many other ideologies. Things are complicated. But our brains don't like it when things are complicated so we make it simple.

So, am I going back on the post I made last week? Maybe a little, in which case call it growth. But really, I'm providing more information reminding you to take things in dichotomies with a grain of salt. Remember dichotomies are really efficient and useful for understanding complex topics. Just be mindful as they can also be extremely damaging when we only think in terms of one or the other.