There are a lot of ways to divide up psychology. For present purposes, I’m only interested in the one I just made up, which is something like this: One type of psychological research looks at a mundane mental process and tries to classify its components and articulate its mechanisms; the other looks at a behavioral phenomenon of interest and tries to pick apart why it happens.
As phrased, this isn’t a very crisp distinction, so a demonstration by induction may have to serve. Let’s first attempt to clarify via onomastics: Call the first type of psychology Process Psychology (PrP) and the second Phenomenon Psychology (PhP). Examples of PrP include the study of cognitive architectures (e.g. ACT-R, the Emacs of psychology), mental representations (e.g. what I think of as the Fodor-Kosslyn debate over the linguistic nature of representation—I hedge because I don’t know whether Fodor and Kosslyn have ever actually engaged on this), attitudes (I don’t have any specific e.g.s, but the voluminous research on the relationship between attitudes and behavior is more or less what I’m talking about), persuasion (e.g. the two really influential theories of persuasion that I’m too lazy to Google), and, for neuroscience’s sake, memory formation (e.g. Gary Lynch’s crazy molecular work on long-term synaptic potentiation). Examples of PhP include the research on various heuristics and biases in judgment and decision making research (roundly ridiculed, or at least keenly questioned, by my cohabitators Anuj Shah and Danny Oppenheimer—subscription to PSYCHOLOGICAL BULLETIN required, sorry), Stanley Milgram’s immensely disturbing work on compliance, Leon Festinger and Joel Cooper’s more subtly disturbing work on cognitive dissonance, studies of various mental illnesses, and—for the sake of a dubious distinction—Eric Kandel’s famous work on LTP.
PrP is usually drier, harder to understand, and of much less obvious relevance than PhP. I think this is because PrP tends to be driven by much vaguer questions: How is the mind put together? How is knowledge represented in it? It’s hard to figure out a good way to decide on a starting point for this type of research, and your choice of starting point constrains your subsequent inferences, so models can vary wildly. Whereas you always have a starting point with PhP, namely your phenomenon. If you discover that people will shock other people based on orders from some putz in a lab coat, it’s easy to think of parameters that might govern that phenomenon, facts that might explain it, and so on.
However, whereas PhP tends to produce parameterizations and explanations, PrP tends to produce widely integrative, generative models. (The models, of course, generate predictions, which can then become their own little eddies of PhP.) Isn’t this the kind of psychology we’re ultimately aiming for? I know this objective bothers some people and strikes others as unknowable, but really—instead of discovering boundary conditions on phenomena post hoc, don’t we gain much more credibility as students of the mind if we have a model that predicts those phenomena and their boundary conditions—as well as other phenomena, and the degree and nature of their relationship?
I don’t have any strong interest in reifying this still-murky distinction more than it deserves, and I don’t mean to suggest that integrative and generative psychological models are more accurate than parameterizations of phenomena—I imagine the opposite, on average, is true. But the phenomenon-by-phenomenon approach to psychology seems like an uphill battle, and one more subject to overfitting. The advantage of an integrative model is that it’s constrained by all extant data pertaining to its domain; whereas a model of a highly specific phenomenon is constrained only by data on that phenomenon.
Of course, your priority might just be to understand a given phenomenon, not cognition in general. If you’re interested in predicting Senate elections, you might not care how the variables that determine the decision fit into models of cognition more generally—and the domain-specific accuracy you sacrifice might be worth more than the broader understanding you stand to gain.