Holism

Chapter 2 endnote 21, from How Emotions are Made: The Secret Life of the Brain by Lisa Feldman Barrett.
Some context is:

Holism explains why bread baked in my home in Boston is never as tasty as bread baked at my friend Ann’s house in Berkeley, California. The Berkeley loaf has a superior flavor because of the different yeasts floating naturally in the air and the elevation above sea level. These additional variables can dramatically impact the end product, and expert bakers know this.

Holism is the idea that parts of a system cannot be studied separately from the entire system. Holism is necessary for studying emergent phenomena, including emotion.

Science has many examples of holism or the lack thereof. The idea that genes have a norm of reaction (i.e., they have differential effects in different environmental contexts) is an example of how genetics must be studied holistically. I am particularly fond of the cautionary tale described in The Triple Helix, a book by the noted evolutionary biologist Richard Lewontin. Scientists were studying fruit flies, performing highly controlled experiments to determine which genes controlled certain physical characteristics. A single gene with a particular mutation might produce, say, deformed curly wings instead of desirable straight wings under precise lab conditions that controlled for temperature and moisture. If so, the scientists would conclude quite naturally that the given gene mutation always produces curly wings. It all seemed straightforward until the scientists left the lab, looked at wild fruit flies, and found that the same mutations did not produce the curly wings seen in the lab. In the lab, the scientists studied one gene while trying to hold all else constant, when in reality they were studying the gene in one particular context. Those curly wings emerged within a particular configuration of gene mutation, lab temperature and humidity, and so on. The outside world, however, is a complex system with many contexts, so the emergent characteristics were different.

Holism is particularly important for understanding brain function, although scientists rarely consider this. Neuroscientists tend to treat brain networks as if they are completely separable, like Lego blocks or parts of a machine. But networks do not operate in isolation of one another; they are anatomically and functionally linked.  Therefore it is ill advised to study them isolation.  Each network serves as the internal context for the others. Therefore, brain networks should be studied holistically.

Failure to recognize the importance of holism is, in part, responsible for the so-called “replication crisis” in science. Sometimes experiments don’t yield the same results because of small changes, such as the time of day the study was conducted, other contextual factors in the lab, or small changes to the experimental methods that scientists incorrectly assume are inert. The context is an important part of the experiment, but scientists often don’t realize this and fail to include information about context in the method section of their papers. This problem stems from essentialism. When an experiment fails to replicate, this does not necessarily mean that the original experiment was “wrong;” instead, the findings were more likely under certain conditions that the scientists were not aware of and/or did not report. When other scientists tried to recreate the experiment, they were working with incomplete information and the replication failed. A good example can be found in chapter 3 of How Emotions are Made, that posed, Western facial stereotypes for emotion (pouting in sadness, scowling in anger, etc.) are identified by people from remote, non-Western cultures only under certain conditions.

To practice holism, scientists need to model the context rather than merely try to control it. (Contextual influences can be random, or stochastic. Randomness is a normal part of nature. Essentialist thinking also prevents people from understanding the importance of stochastic processes. A holistic approach can be used to model stochastic influences as well).