On the Nature of Reason
A Philosophically Rigorous and Theory-of-Mind Analysis
I. Framing the Ontology of Reason
Reason is not a thing but a capacity—an emergent faculty residing within certain minds capable of abstraction, normativity, and meta-cognition. To treat it ontologically demands distinguishing it from cognition broadly construed. Where cognition refers to any process of acquiring and manipulating information—be it associative, perceptual, or procedural—reason specifically refers to the norm-governed, inferential, and self-reflective manipulation of propositional content. The domain of reason, then, is propositional attitudes and their normative interplay: believing, inferring, doubting, justifying.
Reason is not simply about internal mental events, but about their logical structure and evaluative status. That is, not all cognition is reasoning; much is automatic, unconscious, and subrational. Reasoning begins where the agent becomes responsive to justification—to what Wilfrid Sellars termed the “space of reasons,” as opposed to the “realm of causes.”
What Is Reason?
Reason is not merely a function of logic or the mechanical application of rules. It is the capacity to draw inferences, to evaluate propositions, and most significantly, to seek truth through methodical coherence and correspondence with reality. It is a normative faculty: reason does not merely describe how we think but prescribes how we ought to think. In this sense, it is more than calculation or heuristic processing. It is the systematic alignment of belief with evidence, tempered by reflection and capable of self-correction.
Thanks for reading Craig’s Substack! This post is public so feel free to share it.
Epistemological Function of Reason
In epistemology, reason is a foundational source of justification. Rationalism claims that certain truths can be known a priori—independently of experience—through the sheer power of reason (e.g., Descartes' cogito, Kant’s synthetic a priori). In contrast, empiricists like Locke and Hume regard reason as secondary to experience, a servant that processes sensory inputs. Yet even Hume, in denying the rational justification for causality, implicitly affirmed that reason is the only tribunal to which such denials could be submitted.
The key lies in how reason interacts with belief: it operates both deductively (as in formal logic) and inductively (as in empirical generalisation). The standard model in analytic epistemology defines a justified belief as one formed through a reliable process—reason being paramount among them.
Philosophical Anthropology: Humans and Reason
Human beings are often defined as rational animals (Aristotle’s ζῷον λόγον ἔχον). What distinguishes humanity from other entities is not consciousness alone, but rationality—logos—the capacity for structured thought, reflective deliberation, and normative constraint.
Kant insisted that reason is what gives moral law to the self. It is autonomous in its higher function: to legislate maxims. Unlike instinct or learned association, reason can operate against desire. It introduces the distinction between what is and what ought to be, which no animal instinct can generate.
Reason also entails second-order thinking: the capacity to reflect upon one’s thoughts and revise them. This reflexivity is not observed in non-human animals beyond rudimentary problem-solving. The recursive capacity for metacognition (thinking about thinking) is a hallmark of reason, not just intelligence.
Reason in Non-Human Entities
Artificial systems, from Turing machines to neural nets, mimic elements of reasoning. Yet they do not possess reasons in the normative sense. A machine processes data. It does not evaluate whether a belief is justified, only whether an output is syntactically or probabilistically correct given inputs.
In animals, tool use and social learning show intelligence and adaptability. But even advanced species such as corvids and apes lack the propositional structure of reasoning. They respond to stimuli and patterns but do not engage in self-reflective justification or abstract conceptual analysis.
II. The Epistemic Framework: Justification and the Normativity of Reason
From the epistemologist’s perspective, reason’s role is justificatory. To reason is not merely to move from one mental state to another, but to do so under the governance of norms of correctness. The central philosophical insight here is that reasoning involves not just transitions between beliefs, but entitlements to beliefs. This is the critical difference between sub-personal cognitive processing and full-blown reasoning.
Davidson’s triangulation theory insists that propositional thought and reasoning only emerge within a public space of mutual interpretability, where beliefs are attributed against a background of shared norms. In this sense, reason is not an individualistic process—it requires a background of language, community, and normativity.
Brandom develops this further in his inferentialist semantics: to have a concept is to be able to use it in a game of giving and asking for reasons. Inference is not mechanical derivation; it is rule-governed participation in a discursive practice. Thus, the very structure of reasoning is social and normative.
Moreover, reasoning is intentional. The agent takes one belief to follow from another, or sees an inconsistency and acts to resolve it. This intention-directed movement distinguishes reasoning from mere pattern-recognition or stimulus-response behaviour.
III. Theory of Mind and Metacognition
A central component of the capacity to reason is a robust theory of mind—understood not just as the capacity to model other minds, but also one’s own. To reason, the agent must have what Peter Carruthers calls “metarepresentational thought”—the ability to represent mental representations as such, and not merely the world directly. This underpins the recursive self-correction involved in reasoning. When one recognises a belief as unjustified, one does not simply discard it; one alters the grounds of belief.
This recursive loop—wherein thoughts are themselves the object of thought—is essential. A child may form beliefs, but unless she can reflect on whether her beliefs are coherent, contradictory, or unfounded, she does not yet reason. Metacognition provides the feedback loop that transforms associative inference into reasoned justification.
Crucially, this recursive self-reference demands a first-person perspective. Unlike machines or animals, humans can regard themselves as agents who possess beliefs, evaluate them, and alter them. This involves not only computational recursion but what Thomas Nagel described as the subjective character of experience—a point machines fundamentally lack. It is not just that I believe something, but that I know that I believe, and that I can evaluate whether I should believe it.
IV. Reason as a Norm-Governed Practice
Reason operates under norms of validity, soundness, and coherence. Importantly, these norms are not reducible to causal regularities or evolutionary advantages. They are deontic in character: they prescribe how one ought to think, not how one does think. This distinction is pivotal in separating reason from biological adaptation. A neural net may converge on a correct output, but it does not know that the conclusion is justified. The knowledge of justification is a second-order belief: “I believe that p because q, and I take q to be a reason.”
This is what John McDowell means when he says that the deliverances of reason are “rationally intelligible”—they can be brought into the space of reasons, explicated, defended, challenged. It is this structure that permits critical reflection, moral responsibility, and rational disagreement. One cannot meaningfully argue with a thermostat.
V. Comparative Cognition: Reasoning and Non-Human Minds
To address whether non-human minds—animal or artificial—can reason, we must clarify the criteria. The minimal conditions include:-
Possession of propositional attitudes: Beliefs, desires, intentions—not mere associations or dispositions to act.
-
Recognition of logical relationships: Entailment, contradiction, conditional dependence.
-
Responsiveness to norms: A capacity to alter beliefs because of their epistemic status, not simply due to reinforcement histories.
-
Metacognitive awareness: The ability to assess one's own reasoning and revise it in light of that assessment.
Very few non-human systems meet these conditions. Apes and corvids exhibit tool use, social learning, and even rudimentary planning. But they do not seem to restructure their beliefs in light of propositional evaluation or engage in discursive justification. A chimp may stop trusting a conspecific, but it does not revise its beliefs on the basis that one belief contradicts another. Nor does it engage in rule-governed inference to new beliefs by recognising the validity of a logical form.
As for machines, they execute formally valid operations, but this is not tantamount to reasoning in the philosophical sense. The distinction here is between syntax and semantics. Machines manipulate symbols, but they do not understand them. This is Searle’s Chinese Room argument: computation is not cognition. Until a machine can know that it knows and revise its beliefs for reasons it understands as such, it is not reasoning.
Criterion for Defining Reason
The ability to reason, therefore, can be defined by the following capacities:-
Inference: Drawing conclusions from premises.
-
Justification: Evaluating beliefs according to standards of truth.
-
Normativity: A sense of obligation to logic and coherence.
-
Abstraction: Manipulating symbols beyond immediate perception.
-
Self-reflection: Modifying one’s reasoning processes upon examination.
These together form a structure far more intricate than computation or instinctual adaptation. Reason, then, is the architecture of thought that seeks to understand not merely how but why.
VI. The Phenomenology of Reason
Another neglected but essential element is the phenomenology of reasoning—the lived experience of being a rational agent. This includes not only the “aha” moment of insight but also the felt pressure of contradiction, the satisfaction of coherence, the unease of a hasty inference. Reasoning is not a dispassionate chain of syllogisms; it is a mode of self-awareness. It lives in the tension between certainty and doubt, order and error.
Here we might draw on phenomenologists like Husserl or Merleau-Ponty, who saw rationality as embedded in the intentional arc of experience. Reason is not detached from embodiment; it is enacted through language, gesture, and interaction. It is a mode of dwelling in meaning, not a static set of axioms.
VII. Reason, Freedom, and Responsibility
Finally, reasoning is intimately tied to freedom. To reason is to choose among beliefs, not based on compulsion or appetite, but on what one judges to be true. This is why Kant tied reason to moral law: a being that can reason is a being that can be bound by principles it recognises as authoritative. In contrast to merely causal creatures, the rational agent is autonomous—a lawgiver unto itself, though not lawless.
This autonomy implies responsibility. If one reasons badly, one is blameworthy. One ought to have seen the contradiction, noticed the fallacy, questioned the premise. Such judgments only make sense in a framework where reason is not just causal efficacy but normative fidelity. The rational agent is not simply a locus of mental events; it is a moral subject.
Conclusion
Reason is not computation, not intelligence, not information processing. It is the norm-governed, metacognitive, justificatory, self-reflective capacity of agents to operate within the space of reasons. It presupposes intentionality, normativity, and subjectivity. It is irreducibly first-personal and ethically loaded. While animals and machines may simulate fragments of reasoning, they do not dwell in its full structure.
To reason is not to think, but to think well, and to know what it means to think badly. That capacity—fragile, contested, profoundly human—is the ground of our dignity, our fallibility, and our freedom.
Final Contrast
The difference is existential. A bacterium can adapt. A crow can solve. A machine can predict. But only a human being can ask whether the prediction is true and why it matters. Only a human can defy a logical conclusion because it violates a higher-order belief or moral maxim. That revolt against the merely given is the signature of reason.