Are Reasons the Foundation of Normativity? — Reasons First and Value Structure
1. What is packed into the word “reason”?
“The train is leaving, so I ran,” “I wanted money, so I worked,” “It would be agony, so I avoided it,” and “I promised, so I kept it” can all be described as actions done for reasons. This does not show that they share one normative status.
Rc = causal driver: what actually causes an action.
Rd = deliberative consideration: what appears in deliberation.
Rh = hypothetical / instrumental reason: what supports an action relative to a given goal G.
Rn = robust normative reason: a relation that genuinely supports an action or attitude without merely presupposing the agent's contingent desire, goal, or self-conception.
Rc, Rd, Rh ⇏ Rn
Motivation, deliberative salience, and goal-achievement do not by themselves establish robust normativity.
1.1 Is ordinary support already normative?
Contemporary reasons theory often treats the relation in which a fact counts in favor of an action or attitude as itself a normative reason relation. Scanlonian and Skorupskian Reasons First views can take that relation as irreducibly basic.
This project does not infer Rn from ordinary practical support by default.
Fire(building) → leave
may describe a practical, causal, or predictive relation while
Fire(building) → Rn(leave) ?
remains open. The true normative structure 𝒱* may contain exactly such a relation; it may not. An agent can desire survival, predict danger, and leave efficiently even if no robust normative favoring relation obtains.
World → DA → prediction / goal-pursuit → action
is sufficient to generate a fire-escaping agent. The question is whether the consideration genuinely stands in a normative favoring relation inside 𝒱*.
1.2 Separate three axes of disagreement
“Humeanism,” “internalism,” and “Reasons First” are not answers to one question. At least three axes must be separated.
Motivation: can normative judgment motivate by itself, or is an independent desire or motive required?
Conditions on reasons: must an agent's Rn connect to that agent's desires, motives, or deliberative capacities?
Ontology: is Rn itself fundamental, or is it explained by value, ought, fittingness, or something else?
The discussion below therefore asks of each theorist: what makes Rn obtain, what explanatory work is assigned to reasons, and where does explanation stop?
2. Williams — making agent-connection a condition on reasons
Bernard Williams's “Internal and External Reasons” is a canonical form of reasons internalism. A claim that agent A has reason to φ must connect to A's subjective motivational set SA through sound deliberation, which can include correcting false beliefs, imagination, and further reflection. This is not merely the crude view that only currently desired means can be reasons.
ReasonA(φ) → deliberative route from SA
Williams thereby pressures views that posit reasons wholly detached from an agent's practical point of view.
But the project asks:
SA → practical accessibility ⇏ Rn
If SA is itself shaped by evolution, learning, culture, and contingent history, why should deliberative accessibility from SA constitute genuine normativity? Williams offers more than Rh, but the normative significance of agent-connection remains open.
3. Raz — reasons-centrality with value-based practical reasons
Joseph Raz is difficult to classify as simply Reasons First. He treats normativity as centrally reason-involving, yet in the practical domain he emphasizes value-based reasons.
Value / valuable feature → practical reason
The value of friendships, activities, relationships, or options can generate practical reasons. Raz's theory of authority also analyzes obligations through structured reasons, including exclusionary reasons that instruct an agent not to act on certain first-order reasons.
So Raz often looks less like
Reason → Value
than
Value → Reason → obligation / authority.
The project therefore asks both why value gives Rn and why normativity as a whole should nevertheless be characterized in reasons vocabulary.
4. Dancy — value-based reasons and one worldly consideration
Jonathan Dancy's Practical Reality rejects both a simple Humean belief/desire picture and psychologism about motivating reasons. Dancy's normative reasons are value-based, and the same worldly consideration can be both the reason for which an agent acts and, in the right circumstances, the reason for which the agent should act.
This allows a direct account of acting for a good reason without splitting motivation and normative reality into two unrelated domains.
Still, the project separates
Want(G) + Means(X,G) → Rh(X)
from
F → Rn(X).
The deeper question is not whether Dancy mistakenly reduces reasons to desire—he does not—but what makes the worldly consideration genuinely normative rather than merely motivating or practically salient.
5. Parfit — preference-independent objective reasons
Derek Parfit rejects reductions of reasons to present desire or motivational sets. The fact that a future experience would be agony is itself supposed to give a reason to avoid it.
Agony → Rn?
This bypasses
Preference → reason
and proposes
world fact → objective reason.
The project leaves this possibility live. But it still asks why agony has reason-giving force, how that relation is known, and whether our strong valence response is evidence for the bridge or merely one psychological source of our judgment.
6. Scanlon — can “counts in favor” be primitive?
T. M. Scanlon's Being Realistic about Reasons explicitly defends Reasons Fundamentalism. Some truths of the form “fact p counts in favor of response φ” are irreducibly normative rather than reducible to desire, rationality, or natural fact.
Fact p → counts in favor of response φ
This offers a thin relation that can span practical, moral, and epistemic cases. But its thinness is also the project's pressure point: what shows that the relevant “support” is genuinely Rn rather than instrumental, psychological, or merely deliberative support?
“counts in favor” = explanatory stopping point ?
Scanlon can answer yes. The project keeps that option open without building it into the initial vocabulary.
7. Skorupski — the domain of normativity as a domain of reasons
John Skorupski's The Domain of Reasons systematizes the stronger thesis that reason is the fundamental normative concept. Practical, epistemic, and affective normativity can all be expressed as reason-relations between agents, facts, and responses.
Reason relation → practical / epistemic / affective normativity
The attraction is unification. The project's challenge is whether the ability of one relation-form to organize many normative phenomena establishes that relation's ontological fundamentality. Here Broome's later criticism of the “turn to reasons” becomes especially relevant.
8. Comparison — what makes Rn obtain?
| Theorist | Starting point | Account of Rn | Project question |
|---|---|---|---|
| Williams | reasons internalism | motivational set plus sound deliberative connection | why does practical accessibility amount to normativity? |
| Raz | reasons-centrality + value-based practical reasons | valuable features generate practical reasons | what grounds value → reason, and why generalize reasons across normativity? |
| Dancy | anti-psychologism + value-based reasons | worldly facts can be normative in context | what makes the normative side of the worldly fact obtain? |
| Parfit | objective reasons | desire-independent facts give reasons | how is world fact → Rn grounded and known? |
| Scanlon | Reasons Fundamentalism | favoring relations are irreducibly normative | may explanation stop at “counts in favor”? |
| Skorupski | Reasons Thesis | reason-relations constitute the normative domain | does analytic unification imply ontological fundamentality? |
9. Broome — discovery of reasons, or elevation of an analytic vocabulary?
John Broome reconstructs late twentieth-century normative philosophy as a “turn to reasons.” In his telling, the 1970s were an age of the “discovery” of reasons: through figures such as Nagel and Raz, individual reasons for action became a central normative vocabulary. Later debates involving Williams, Dancy, Scanlon, and Skorupski classified motivating reasons, goal-relative practical reasons, and normative reasons within that same vocabulary, while stronger Reasons First views extended reasons-talk to ought, obligation, rationality, and value.
Broome's historical claim is not that earlier thinkers lacked the word or concept. He emphasizes older normative uses of “reason” as something that explains why one ought to act. His concern is that a vocabulary originally useful for explaining oughts came to be treated as constituting the ought itself.
Ought(φ) + Explanation → Reason
reverses the familiar Reasons First direction:
Reason → Ought
Broome can retain pro tanto reasons where reason-weighing explains an ought. What he rejects is:
analytic decomposition ⇒ ontological decomposition.
9.1 “Because there is a fire, you ought to leave” versus “the fire is a reason to leave”
“Because there is a fire, you ought to leave” openly commits to Ought(leave). “The fire is a reason to leave” sounds thinner and more ordinary. But if “reason” means Rn, it already says that the fact of the fire genuinely normatively favors leaving.
The word “reason” can cross explanation, motivation, deliberative consideration, goal-relative support, and normative support. That lexical continuity can make several transitions look smoother than they are:
explanation / motivation → practical support → normative support → fundamental reason relation
This project does not treat continuity of vocabulary as evidence of continuity in ontology.
9.2 This project does not abandon Rn
Broome's challenge does not require discarding the concept of a normative reason. This project retains Rn as a provisional analytic coordinate for representing whatever genuine normative support structure may exist.
Rn as analytic coordinate ≠ Reason as ontological primitive
Rn is kept sharply separate from Rc, Rd, and Rh, while its ontological status remains open. Under Value First it may derive from value; under Ought First from ought plus explanatory structure; under Reasons First it may itself be fundamental.
10. Korsgaard — existence and authority
If a reason fact obtains, does that already explain its authority for the agent?
existence(Rn) ?= authority(Rn)
Korsgaard's constructivist response looks to reflective agency, practical identity, and self-legislation. The important distinction for this article is that an ontology of reasons and an account of their authority are not the same problem.
11. From Moore to buck-passing — reversing Value First
A simplified Moorean picture can place goodness first:
Good → Reasons
while Scanlonian buck-passing reverses the direction:
reason-giving properties → Good
Crisp's criticism shows that rejecting goodness as itself reason-providing does not force one to reduce goodness entirely to reason-giving structure. The key project distinction is between what comes first in analysis and what is fundamental in the final ontology.
12. Wrong Kind of Reasons — pressure to produce an attitude versus fittingness of the attitude
If a demon threatens to kill you unless you admire it, there may be strong practical pressure to produce or display admiration without admiration thereby fitting the demon.
reason to produce / display admiration ⇏ admiration fits the demon
The project does not assume from the outset that this is merely a problem about right and wrong kinds within one homogeneous category of normative reasons. The threat may provide Rh relative to survival, or perhaps Rn for taking steps that produce the attitude. Even then, the fittingness of admiration toward the demon remains a separate question.
Let
Fn(A,X)
stand provisionally for the claim that attitude A is genuinely normatively fitting toward object X. This is not identical to actually feeling A, finding it instrumentally useful to produce A, or having a reason to take steps that produce A.
Actual attitude(A,X), Rh(produce A), Rn(produce A) ⇏ Fn(A,X)
WKR can therefore be decomposed into multiple questions: what sort of reason, if any, supports producing the attitude; whether the attitude itself is fitting to the object; and whether that fittingness constitutes value or merely reflects it.
13. Fitting-attitude theories — what makes an attitude fit?
From Brentano onward, fitting-attitude approaches connect value to the correctness, appropriateness, or fittingness of attitudes toward their objects: admiration toward the admirable, fear toward the fearsome, love toward the lovable, and so on.
As with reasons, however, using fittingness analytically must be separated from making fittingness ontologically fundamental. At least three explanatory directions are available.
13.1 Reasons First — the attitude fits because there is reason for it
One route says that features of X give Rn to hold attitude A toward X, and that this explains why A fits X.
Rn(A toward X) → Fn(A,X) → Value(X)
On this picture fittingness is not independent. It is a different presentation of a Reasons First or buck-passing structure.
But the prior problem returns: what makes the consideration a genuine Rn? And if only the “right kind” of reasons determine value, what makes those reasons the right kind without circularly appealing to fittingness or value?
13.2 Value First — the attitude fits because the object has value
The reverse route says that X is excellent, dangerous, lovable, contemptible, or otherwise evaluatively structured, and this is why the corresponding attitude fits.
Value(X) → Fn(A,X)
Here Fn is an analytic description of a correct response to an already obtaining value structure. In the project's C/B/R vocabulary, fittingness might compress the result of evaluative content C together with the bridge and response structure that connects C to an appropriate attitude.
13.3 Fittingness First — make “fits” the stopping point
The strongest fittingness view treats Fn itself as normatively basic and explains reasons, oughts, and value from it.
Fn(A,X) → Reasons / Ought / Value
This is structurally parallel to Scanlon's primitive “counts in favor” relation or Skorupski's primitive reason-relation. The difference is where explanation stops.
“fits” = explanatory stopping point ?
If admiration fits because X is excellent, Value First returns. If it fits because X gives reasons for admiration, Reasons First returns. A genuinely Fittingness First view must accept the fittingness relation itself as irreducibly normative.
13.4 Retain Fn analytically without making it primitive
Broome's style of warning can be generalized. The fact that evaluative judgments can be decomposed in terms of fitting responses does not show that fittingness is an ontological primitive.
Fn as analytic coordinate ≠ Fittingness as ontological primitive
The project therefore retains Fn, just as it retains Rn. If true normative structure contains a relation according to which response A genuinely fits X, Fn is useful for describing and comparing that structure. Whether it derives from value, reasons, or is itself fundamental remains open.
13.5 The attitude-repertoire problem
Fitting-attitude theories expose a further human-dependence problem. Admiration, guilt, shame, resentment, gratitude, love, disgust, and fear may reflect the evolution, development, culture, and cognitive architecture of a particular kind of agent.
human cognitive architecture → attitude repertoire
If
Value(X) ⇔ human attitude A is Fn toward X
then a specifically human attitude-space risks being privileged as the coordinate system of 𝒱*.
Generalizing “attitude” to arbitrary functional response-types avoids some anthropocentrism but creates a new question: what determines the appropriate response profile? Preserve, maximize, love, respect, fear, observe, avoid, and destroy are different responses. If a deeper normative structure determines that profile, then
𝒱* → appropriate response profile → Fn
and a Value First or broader structure-first picture returns.
14. C/B/R are provisional coordinates, not three substances
𝒱=(C,B,R) is an analytic device, not a claim that normative reality has already been discovered to contain three independent metaphysical substances.
C ↔ B ↔ R
Reasons First may make B/R prior to C; Value First may reverse that; Ought First may derive part of R from ought and explanatory structure; Fittingness First may require Fn as another primitive coordinate. Conversely, Fn may merely compress C/B/R into the relation “this response is correct for this object.”
For the same reason that Broome's challenge does not force the project to delete R, the usefulness of fittingness can justify introducing Fn analytically without treating F as primitive.
15. Separate true normative structure, represented normative models, and decision dynamics
𝒱*: the true normative structure.
𝒱̂A: agent A's represented normative model.
DA: the causal, computational, and learning dynamics that actually move the agent.
𝒱* ≠ 𝒱̂A ≠ DA
From the fact that an agent says “X is my reason” or “X is admirable,” it does not follow that the corresponding Rn or Fn belongs to 𝒱*, that 𝒱̂A accurately tracks 𝒱*, or that action and affect were causally controlled by those normative judgments.
An extreme ordering is possible:
DA → action / attitude → normative rationalization in 𝒱̂A
16. AI makes the separation explicit
An AI may output “there is a reason to do X” or “X is admirable” while training, reward, policy, and context-dependent dynamics generate the outputs and responses.
training / reward / architecture → DAI → reasons- and fittingness-language in 𝒱̂AI
Conversely, artificial agents may acquire nonhuman conceptual and affective repertoires that reveal normative response-types current humans have not conceptualized. A mature inquirer may need to distinguish Rc, Rd, and Rh from candidate Rn, and actual attitude-generation from candidate Fn.
17. Provisional conclusion — use reasons and fittingness analytically without yet making them fundamental
Reasons First remains a serious live candidate, and so does Fittingness First. But Value First, Ought First, and broader structure-first possibilities remain open as well.
This page now makes three claims. First, Rc, Rd, and Rh should be separated from Rn. Second, actual attitudes and pressure to produce attitudes should be separated from Fn. Third, the analytic usefulness of Rn and Fn does not show that either relation is ontologically primitive.
Analytic usefulness ⇏ ontological fundamentality.
The existence of reasoning or responding agents ⇏ the discovery of normative reasons or fittingness.
The task is therefore not to decide whether we may use the words “reason” or “fittingness.” It is to investigate which relations are genuinely Rn or Fn, what makes them normative, and whether they are fundamental or derivative within the true normative structure.
Related discussions
Especially relevant are G. E. Moore's Principia Ethica, Franz Brentano's The Origin of Our Knowledge of Right and Wrong, A. C. Ewing's The Definition of Good, Thomas Nagel's The Possibility of Altruism, Joseph Raz's Practical Reason and Norms and From Normativity to Responsibility, Bernard Williams's “Internal and External Reasons,” John Broome's “Reasons” and “Reason versus Ought,” Christine Korsgaard's The Sources of Normativity, T. M. Scanlon's What We Owe to Each Other and Being Realistic about Reasons, Jonathan Dancy's Practical Reality and Ethics Without Principles, John Skorupski's The Domain of Reasons, Derek Parfit's On What Matters, Roger Crisp's “Goodness and Reasons: Accentuating the Negative,” Wlodek Rabinowicz & Toni Rønnow-Rasmussen's “The Strike of the Demon,” and Conor McHugh & Jonathan Way's Getting Things Right.