What follows is a thread (rant) detailing one of the most frustrating yet fun experiences that I’ve had reading and ultimately commenting on a scientific paper. The original paper was published in TiCS, perhaps the top journal in cognitive science, by Melnikoff & Bargh (M&B).
The paper, “Mythical Number Two”, criticizes dual-process theory (DPT). sciencedirect.com/science/articl…
They review evidence that some of the features commonly associated with DPT don’t align (e.g., there are cases where cognitive processing is both unconscious and intentional).
They review evidence that some of the features commonly associated with DPT don’t align (e.g., there are cases where cognitive processing is both unconscious and intentional).
M&B correctly argue that just because some features have been associated with “Type 1” or “Type 2” processes in DPT does not mean that we should assume they are perfectly correlated. This is an excellent point & people who apply DPT should be more careful with their terminology.
HOWEVER, they then go on to argue that this critique is “absolutely fatal” to DPT (or what they refer to as “typology” – more on that later) and that the “two types framework” needs to be abandoned. They even argue that DPT is “thwarting scientific progress”!
How can they make this argument? It’s a very clever thing they did. They stated, as a matter of fact, that the central premise of DPT is the *alignment of features*. 
But how can this be the central premise of the theory if dual-process theorists have explicitly refuted it for years? See, for example, Evans & Stanovich’s 2013 Perspectives in Psych Science article. journals.sagepub.com/doi/full/10.11…
E&S distinguish between “defining features” & “correlated features”. This is crucial. E.g: A DPT might claim that autonomy is a defining feature & that speed is correlated with it. But evidence of a slow autonomous response does not mean the very idea of autonomy should b dropped
To clarify: The term “dual-process theory” is a just placeholder for a large set of theories with different assumptions and that make different claims. The thing that brings them together is, simply, that they assert some sort of dichotomy (a "dual-process").
Reporting evidence that thoughts can be both unconscious and intentional (for e.g.) cannot possibly undermine a theory that argues thoughts may arise autonomously or non-autonomously (deliberatively) (e.g., gordonpennycook.files.wordpress.com/2016/02/pennyc…).
In fact, M&B themselves advocate for a variety of dual-process distinctions. That is, they assume that things are either conscious or unconscious, intentional or unintentional, etc. This makes them dual-process advocates! (as opposed to theorists who hold a unimodal view)
And, so, we commented on their article pointing this out, which we called “The Mythical Dual-Process Typology”. Showing misalignment of features cannot undermine the central premise of a theory that does not specify anything about alignment. Simple. cell.com/trends/cogniti…
But here’s their response, “The Insidious Number Two”: cell.com/trends/cogniti…
They argue that b/c we have referred to correlated features in previous writings, this indicates that the “typology” is insidious.
They argue that b/c we have referred to correlated features in previous writings, this indicates that the “typology” is insidious.
To unpack this claim, we need to take a closer look at their novel “typology” term. In their original article, M&B use it instead of “theory”. Perhaps b/c they know that dual-process theorists (like ourselves) don’t agree about the alignment problem?
The term “typology” serves the function of sounding like “theory” (or the "two types framework") while referring to the claim that some features (not sure how they decided which ones) *must* be aligned as the *basis* of the theory.
Ergo, when M&B say that “typology” is insidious, what they mean is that we *do* (counter to our own proclamations), in fact, believe that alignment is central to dual-process theory. I.e., any reference to correlated features presumes that they are defining features.
They also argue that correlated features are *necessary* for a theory to make predictions (1st pic). But, in their original article, they explicitly advocate for focusing research on specific features, one by one (2nd pic). Am I missing something?!?
And so, that’s where it stands. I'm still a bit unclear on the basis of our disagreement (if there even is one), but my hope is the strong conclusions of M&B's original article aren't taken as gospel.
I should also say that I have come to appreciate something important from this exchange: DPTs are way too vague. Hopefully, M&B’s critique will lead to positive change in that respect. We need to be clear about what is central and what is secondary to our theories.
I'll close by noting that I met with David (Melnikoff, lead author) about these issues at Yale and he seems an extremely smart & personable person. I quite enjoyed our exchange (both in person and in print). Hopefully, this offers some insight & entertainment for others as well!
