Fittingness: Rational success in concept formation

Epistemic status:

This post offers a philosophical argument for distinguishing truth and fittingness as parallel success concepts for two phases of rational inquiry. I have graduate degrees in philosophy, with relevant expertise in logic, type theory, epistemology, phenomenology, ontology, and philosophy of science. I am confident (0.9) about truth the central claims, and the fittingness of central concepts. At the end I include a brief and much more speculative section (0.5) on how these ideas might related to pattern recognition in the brain.


The concept of truth is widely held to be a central concept of success for rational inquiry. It captures the notion of accuracy of beliefs in modeling reality. In this post, I suggest that truth needs a companion concept of fittingness. I will argue, that as truth is the essential success concept for rational belief, so fittingness is needed as a success concept for rational concept formation (equivalently: ontology building or type assignment). Both truth and fittingness capture an idea of accuracy in world modeling, but they operate at different type theoretic levels. Roughly, fittingness concerns the accuracy or appropriateness of how we divide reality by attaching concepts to it, by typing it, how we cut reality at the seams with language. Truth captures the accuracy of statements (propositions) within a given ontology. In order for questions about the truth of a proposition to be meaningful fruitful and salient, the concepts in which the proposition is expressed must be sufficiently fitting. [Slider’s comment below makes the point that truth conditions can be meaningful even when the concepts employed are not fitting. He gives the example of the concept grue from philosophy. This example also illustrates that even when a statement is truth evaluable it may not be fruitful or salient for the purpose of rational inquiry.]

Fittingness in Inquiry

There are two phases of rational inquiry. These two phases do not occur in strict linear sequence, but rather as a cycle, frequently looping back and forth. In one phase I’m engaging with some domain (or a novel aspect of a domain) that is confusing, anomalous, or hard to make sense of. I don’t have good pre-existing concepts to describe or articulate what’s happening here. I’m not even sure what questions to ask, or how I should identify salient features of the domain.

Despite my ignorance, I can engage with this domain by looking at examples, by interacting and observing what happens. As I do this I may be struck by an association, metaphor, or a loose analogy with something I already understand (usually something in a seemingly unrelated domain). I notice this association and “try it on for size”. I attempt to map it to the current domain more closely to turn the intuitive association into something more rigorous. Sometimes, this mapping clicks into place. I keep observing my interactions with the domain, and the new analogy holds closely. Moreover, it might become a fruitful guide to my inquiry, suggesting new avenues of exploration, new questions to ask, etc. In time, if it proves to be particularly fruitful, coherent, and stable across the domain, it may solidify into a feature of the domain.

Once features (which might also be called properties, objects, relations, agents, or processes depending on context) start to form, we can move into the second phase of inquiry. We can formulate precise statements about these features and what might or might not hold between them. At this point, truth becomes relevant as the success concept for evaluating such statements. A question such as “Do objects characterized by property A enter into a process B?” has a truth value. It can be evaluated based on observation and argument. In contrast, the question of whether the domain is properly characterized by property A and process B, is not a question of truth, it is a question of fittingness.

Just as true has an opposite false, and it can (for some domains) admit of degrees called probability, we can similarly say that fitting has an opposite of distorting or ill-fitting, and there are degrees of fit.

Spandrels: An Example of Fittingness in Biology

A scientific example to illustrate these phases. Stephen Jay Gould and Richard Lewontin introduced the term spandrel into evolutionary biology based on an analogy with an architectural feature of old churches.

Stephen Jay Gould; Richard Lewontin (1979), “The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme”, Proc. Roy. Soc. Lond. B, 205 (1161): 581–598.


The spandrel was meant to capture the idea that a trait of organism could evolve as a structural byproduct or side effect of another trait that which evolved due to natural selection. Analogously, architectural spandrels arise from the necessary space left at the top of an archway. The analogy is fitting because with the spandrels due to arches, architects eventually made use of them for decorative flourishes, much as species with spandrel traits might eventually evolve uses for these traits that confer selective advantage. The conceptual mapping between architectural spandrels and biological spandrels is tight and multi-layered. It was not be obvious before it was suggested, but once one sees the connection it clicks into place.

When an analogical mapping is fitting, when the fit is tight and consistent, and when it illuminates the structure of a domain it quickly morphs into a literal feature. Within a few years of Gould and Lewontin’s original paper, the term spandrel was replaced by the more biological sounding term “exaptation”. This became a new standard concept in the conceptual repertoire of evolutionary biology. It is no longer a mere analogy, but a literal feature. Biologists can use this new concept to ask precise scientific questions such as “Did bilateral symmetry arise in animals as an exaptation or an adaptation?” Such questions can be evaluated in terms of truth or falsity. Insofar as such statements are taken seriously in actual scientific research, They presuppose conceptual fluency with the relevant concepts they presuppose the fittingness of the concepts used in their formulation [updated based on Slider’s comment]. We cannot fruitfully engage with truth evaluation until we’ve already embraced the fittingness of the concepts used to pose the question.

Can fittingness be reduced to truth?

To anticipate a potential object to the true-fitting distinction. I acknowledge that questions of conceptual fit can be turned into questions of truth by asking whether the statement, “X is fitting for domain D” is true. However, this is a meta-statement about the relation of our concepts to the domain, it is not a truth-evaluable statement within our target domain. So the cost of converting fittingness questions to truth questions is to shift the relevant domain of inquiry into a meta-domain. Ascending to such a meta-domain does not help us evaluate fit. We must still rely our ground-level judgements of fit as we engage/​interact with the domain while trying on different concepts for size.

Why is this Distinction Important?

Intractable Disagreement

I think that often when disagreements persist between well meaning intelligent people, it is because of a mismatch in the taken-for-granted concepts and word meanings we use to try to discuss the issues at hand. We can try to resolve such mismatches by offering explicit definitions and disambiguation of terms, but these measures do not adequately address the questions of whether a given term, definition, or concept is relevant, useful, natural, illuminating, etc.

When disagreements resist straightforward approaches based on evidence and argument, when they seem to turn on different views of what is salient to talk about, this is a good clue that we need to shift from a truth-focused inquiry to a fittingness-focused inquiry.

Sub-Optimal Concepts

The culture of rationalism (and science generally) may have an unacknowledged bias towards truth-focused inquiry over fittingness-focused inquiry. Inquiring into fittingness can feel messy, childish, too subjective, certainly not what a respectable researcher should be focused on. Rationalists like clean lines and sharp distinctions, not loose analogies, metaphors, and associations. However, this means we might end up rushing the first phase of inquiry, leaving us stuck with ill-fitting concepts. When we try to formulate truth claims in terms of these ill-fitting concepts, we may give the illusion of clarity and precision, while actually obfuscating the domain we wish to understand.

The concept of fittingness can help us avoid this tendency by allowing us to explicitly signal when we need to shift modes between the two phases of inquiry. Rather that this seeming like some illegitimate move in the discussion, it can become a well accepted and non-mysterious oscillation between two well-defined phases of inquiry. In short, I am arguing the the concept of fittingness is fitting for the domain of rational inquiry.


In general, questions of fittingness concern the correct formation of the concepts, ontologies, or types (in a type theoretic sense) that we use to structure our statements about a domain. In contrast, questions of truth concern the evaluation of statements made within an already formed scheme/​ontology/​typing of the domain.

What does fittingness-focused inquiry look like? It is about orienting to a domain, examining examples, connecting the dots, drawing analogies, telling stories. We get our hands dirty with the domain while looking for conceptual connections. This can lead us to an experience a spontaneous shift in our perception of how the domain is organized as it is interpreted through a new conceptual lens. Not just any shift will do; for it to be fitting, the shift must be one that transitions our experience of the domain to become more intelligible or rationally structured. The domain becomes less confusing, more navigable, more naturally divisible, and more compositional.

Speculation: Fittingness and Pattern Recognition in the Brain

It seems likely that on a neurological level, inquiry that leads to fitting concepts is grounded in the brain’s basic pattern recognition abilities. This is an automatic process that learns patterns in experience without pre-existing abstract concepts. While the brain can learn innumerable patterns, our pallet of linguistically articulable concepts is much more limited. Since we can’t name every micro-pattern we are able to learn, we must find an economical way to express the most important patterns, ideally by re-purposing and extended existing concepts through analogy. A judgement of fittingness is the perception of a higher-order structural match between the pattern of a concept and the pattern we’ve learned to recognize non-linguistically in the domain.

Note: this previous paragraph is a just a hypothesis of how fittingness may relate to the brain. The main philosophical arguments of this post do not depend on the correctness or even plausibility of this hypothesis.