On the Semantics of the Ought-to-Do Author(s): Hector-Neri Castaeda Reviewed work(s): Source: Synthese, Vol. 21, No.

3/4, Semantics of Natural Language, I (Oct., 1970), pp. 449-468 Published by: Springer Stable URL: http://www.jstor.org/stable/20114739 . Accessed: 31/07/2012 21:14
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Synthese.

http://www.jstor.org

HECTOR-NERI

CASTA?EDA

ON THE SEMANTICS

OF THE OUGHT-TO-DO

i
Deontic modal and permissible concepts like ought, right, obligation, forbidden, have benefited from the philosophically exciting work in the semantics of

and concepts done by Kanger1, Hintikka2, Kripke3, Montague4 others. Their semantics illuminates both the topic and the contribution to deontic logic: the topic is what of the standard axiomatic approach used to call the Ought-to-be. On the other hand, the non philosophers

standard approach deontic systems of represented by early axiomatic ours deals with the Ought-to-do. Thus, rather than competing with the com to deontic standard approach logic, our non-standard approach our non it. This can, however, be seen only by providing plements of semantical foundations. standard approach with a minimum This is precisely what this essay attempts to do. We shall also provide a rationale some proto for our non-standard semantical system by formulating of the system and data that both guide the development philosophical serve as tests of adequacy sophical, not technical. There are, of course, for it. In fact, our concern is primarily philo

reasons for important general philosophical our appreciation of standard systems is developing divergent systems: (i) enhanced by contrasting them with non-standard ones, and (ii) our under of a set of concepts improves by seeing them in different lights. standing 1. Ought-To-Be The and Ought-To-Do

standard approach to deontic logic conceives of deontic expressions is obligatory that', 'it is permissible that', 'it is forbidden that', 'it is ('it wrong that', 'it is right that', 'it ought to be the case that', etc.) as ex pressing operators that have the same domain and range : the domain of propositions deontic (or states of affairs) and properties. Linguistically, in the standard approach, operators whose domain and expressions are, range are both the domain of sentential forms and sentences, i.e., the total

Synthese 21 (1970) 449-468. All Rights Reserved

Copyright ? 1970 by D. Reidel Publishing Company, Dordrecht-Holland

450
set of all well-formed D is a deontic

HECTOR-NERI

CASTA?EDA

formulas

/is. world

expression and/is the beautiful Semantically,

(wffs). Thus, on the standard approach if a wff, Df is slwff of the same general kind idea is that Z>/is true in a given possible

w if and only if/is on D) in some (or true (or false, depending world which is ideal with respect to W. Here a possible every) possible world W is ideal with respect to W if and only if all obligations belonging to W both are also obligations to W and are fulfilled in W'. belonging idea has been both explained in detail and put to important uses by in 'Deontic Logic and Its Philosophical Morals' Hintikka (presented in a of the symposium on deontic logic at the meeting of theWestern Division This

American A

Association Philosophical valuable distinction tremendously

in Cleveland, Hintikka

Ohio, in May, 1969). makes is that between

analysis of the truth and in is straightforward for 'It ought to be the case that/' is what makes our world have genuine, non-empty obligations sightful: but its falling short of an ideal in some respects, i.e., its having nothing conditions something realization must is (or would be) true in an ideal world; since the of certain ideals may prevent others from being realized, we consider not only one ideal world but a set of them, not necessarily false which of perfection: there may be alternative roads

and deontic consequence. logical consequence The intuitive idea behind this standard semantical

arranged in a linear sequence to perfection.

of deontic logic is the The primary contrast in the above conception contrast between what is and what ought to be. The idea of who is to so that the approach can handle is not considered, realize the obligation statements like "There ought to be no very nicely genuinely impersonal pain", meant merely to articulate something about the universe, which is not conceived as an agent but simply as the totality of all existents and all facts. This impersonal statement tells of what would be a necessary lack in every universe, and neither attributes responsibility for any action to some body nor demands any action from anybody : the statement is oriented to no agents: it has the structure of the statement "It is (would be) desirable that there were no pain". the sentence 'There ought to be no pain' may very well be Naturally, used in ordinary agential statement, language to formulate a personal, the sentence 'God e.g., "God ought to have created no pain". Conversely, ought to have created no pain' can express a non-agential statement

ON THE SEMANTICS OF THE OUGHT-TO-DO

451

some form of desirability or 'oughtness' to the state of affairs attributing or proposition statement may, "God created no pain". This non-agential be more naturally expressed by the sentence 'It ought to be the perhaps, case that God created no pain'. Ordinary sentences can easily be made to is that we different, though related statements. What matters statement "There ought to be no the preceding non-agential distinguish from the agential statement that attributes to God a duty or obliga pain" express tion, and articulates not merely an ideal creation, for blaming or praising God for his creation. In earlier papers I have adopted a non-standard view focuses but establishes a basis

That

view of deontic logic. on the agential ought statements, and is based on a of the common structure underlying all processes of legis generalization or rule-making, whether they go on continuously in high legislative lating chambers of legislative bodies of of nations, or in pompous meetings institutions, or in pedestrian discussions of procedures by the members of informal groups or clubs. A formal analysis of the general underlying

of rule-making, whatever its kind, is this : (i) a framework of nature governing the facts is recognized ; (ii) a set ? of facts and laws of or commands or endorsed or imperatives are adopted prescriptions, to overriding considerations in cases of conflicts) ; (iii) the set ? (subject limits the freedom of action, i.e., of decision and command, of the agents the commands in ? are 'addressed' or directed to. Thus, a piece of rule structure making confronts the real world, with all its facts and laws of nature, with other possible practical worlds in which the same facts and laws of nature hold, but different decisions and orders take place: the making of a set of rules is, at bottom, nothing more (on this analysis) than the adoption of a system S? of alternative practical worlds which share both the same facts or commands. A contingent prescrip and the same set ? of prescriptions tion or command (with respect to S?) is one that holds in some worlds but in other worlds of S?. What is obligatory with respect to the or enacting of ?, i.e., what is obligatory ^ iswhat is 'commanded' legislating in every world of is obligatory is is, thus, in a sense, what S?: What necessarily prescribed or commanded. The preceding is a crude statement of the intuition behind our approach to deontic logic. It is non-standard with respect to the ideal-world anal ysis, yet it, too, has long and respectable roots in the history of philos ophy. There is no need to remind anybody of the imperativist analysis of fails to hold

452

HECTOR-NERI

CASTA?EDA

or of traditional ought-statements must be clear that the preceding

deontic statements is that they and an agent and a demand that the agent do, or fail to do, the action in question. Many an ordinary statement that is apparently agentless demands an action; e.g., "Cars ought to have plates" demands that some agents put plates on cars; it is short for a statement to the effect involve an action that people in a relationship R of a certain kind etc.) put plates usership, managership, ports imperatives of the form "If you are R to that car, put plates on it". That statement contrasts very sharply with the agentless statement "There ought to be no pain" discussed above. In short, deontic statements divide (i) those that involve agents and actions and support im and (ii) those that involve states of affairs and are agentless and peratives, have by themselves nothing to do with imperatives. The former belong to what used to be called the Ought-to-do and the latter to the Ought-to-be. neatly into: Our non-standard approach, then, is not suited for the Ought-to-be; we hope, it is adequate for the Ought-to-do. 2. Proto-Philosophical We have Data but, on that car. The to a car (ownership, statement sup

analysis of ought-statements truth-conditions of ought-statements. an important feature of certain Now,

it analysis of laws. However, is not concerned with the paragraph themselves, but only with the analysis of the

command

as our topic by means of the Ought-to-do just demarkated imperatives. Thus, in order to gather data for any theory of the Ought-to as do we must engage in a phenomenological analysis of imperatives, of normative or deontic of course, as in a phenomenological well, analysis
statements.

2.1. Prescriptions. pieces of advice,

shall call commands, We orders, petitions, requests, entreaties, and suggestions mandates. Clearly mandates belong into families. One and the same imperative sentence, e.g., 'Karl, go home at 3 p.m.', may express an order, a command, apiece of advice, a petition, a request, etc. What all these have in common is a structure consisting of a reference to an entity named 'Karl' and the predicative This structure we shall call a prescription. action going-home-at-3-p.m. or statement; but in this simple or atomic case it a proposition It is like differs from the corresponding expressed with the sentences proposition 'Karl goes home at 3 p.m.' and 'Karl will go home at 3 p.m.' in the way

ON THE SEMANTICS OF THE OUGHT-TO-DO

453

subject and predicate are related, i.e., in the copulation of the subject and predicate. The prescriptive copulation, we shall say, makes of the subject an agent prescriptively considered, and, likewise itmakes of the action an action prescriptively considered. The crucial point is that an agent pre considered has a certain asymmetry with respect to the other scriptively members of a relation. A clue to this appears in the fact that an indicative sentence formulating or the performance (i.e., proposition-expressing) of an action has both a passive and an active counterpart; but there doing is no such counterpart for an imperative (i.e., prescription-expressing) sentence. For example: we can say either "Paul hit Mary" or "Mary was hit by Paul" asserting the same proposition. But we cannot choose to put the prescription embedded in "Paul, hit Mary" in the passive version "Mary, be hit by Paul". differ from their corresponding prescriptions propositions in the different copulation of their atomic constituent ultimately pre are, then, abstractions from families of mandates Prescriptions, scriptions. that disregard not only the modality of the mandates, (e.g., command, Complex order, suggestion, entreaty), but also the act of issuing and the elements of this act: the time, the place, and the issuer. We shall in our proto examination philosophical represent prescriptions by the clause resulting from imperative sentences by deleting the comma after the agent name; thus, in the above
go home'.

example structural

our prescription

is expressed

purely

by 'Karl

An mixed

important compounds

of prescriptions

prescriptions (1)

(or mandates).

feature of prescriptions and mandates is that are and propositions (or mandates) for instance: Consider,

Karl, do the following : if it rains, close the windows if and only if the awnings are not up, and if it hails, turn circulator A on if and only if circulator B is off.

in (1) there are two atomic prescriptions, Karl close the Evidently, windows and Karl turn circulator A on, linked by connectives to four and the whole of (1) is a mandate having at its core a mixed propositions, This is an extremely important point. Aside from complex prescription. the imperative prefix of (1) 'Karl, do the following', it is patently clear that the prescriptive clause 'if it rains ... off' is, though an incomplete the conditioning clauses of (1) sentence, a mixed compound. Palpably,

454
need not

HECTOR-NERI

CASTA?EDA

express states of things or of the weather: of Karl himself, e.g., performances (2)

they may

express

if you come late, don't close the Karl, do the following: if and only if you-raise-the-awnings, windows but do raise the awnings if it is raining. see both

the proposition "You [Karl] raise the awnings" and as constituents "Karl raise the awnings" the prescription of the same mandate. We also see again that ordinary English has a widely applicable mandate mandate
given any

Here we

operator that yields a canonical notation for the expression of a as analyzed into the imperative operator and a prescription:
clause C formulating a prescription P9 where '#/, 6a2,..., 'an'

are names
tence ral9

or terms referring
a29..., an9 do the

to the agents
following:

referred

to in C, then the sen

a mandate whose core

C"1 expresses

is prescription (3)

P. For example:

if it rains, Paul mow Paul, Mary and Ted, do the following: the lawn, and Ted put the sprinkler on; and if it does not rain, and Mary paint the window Ted open the windows, sills.

2.2. The Parallelism Between Imperative and Indicative Logic. Philosophers have a logic at all, have argued and continue to argue whether mandates can actually and whether mandates whether they have a two-valued logic, be inferred from other mandates. As far as we can see these disputes are to a large extent verbal disputes. The fact is that mandates do stand to other in formal relationships holding between (and and propositions mandates formal relationships is of and the study of mandate among) propositions, of ethics and of action. How we call great importance for the philosophy them is unimportant. Thus, for convenience, we shall here simply use the in an extended sense to cover the of logicians customary terminology mandate and prescriptive relationships. is two-valued or not is not a purely whether the logic of mandates Now, issue. We must, however, remember that in the case of propositions verbal we find cases which present truth-value gaps and that by some procedures we do manage pretty well with a two-valued proposi of regimentation tional of certain tough cases of that a parallel two-valued work well enough. But aside an impressive mass of

the mere appearance logic. Likewise, is not a proof and prescriptions mandates theory of prescriptive from merely defensive implications moves, we won't

can marshall

ON THE SEMANTICS OF THE OUGHT-TO-DO evidence that supports

455

prescriptional appear sound:

logic. To

the claim of parallelism between propositional and start with, inferences of the following forms

(Fl)
(F2)

X9 do A
Therefore, X, don't fail to do A. X, don't fail to do A Hence, X9 do A.

The validity of (Fl) and (F2) amounts to the rejection of three values for mandates ;clearly 'A' can represent any action whatever simple or complex. More evidence for the parallelism between propositional and prescrip tional logic comes from the fact that the following schemata implication seem to hold:

(F3) (F4)
(F5)

% %

do A' implies 'X, do A and do A9 do A and B9 implies 'X9do A9

*X, don't both do A and fail to do B9 implies 'X, don't do both of the following: one, fail to do B and some action C; two, do C and A.9 A 'X, do A9 and 6X, don't do both : and not-B9 imply 'X, do B.9 the following connection between schemata is also apparent :

(F6) Furthermore, (F7)

If 'X, do A9 implies 6X, do B99 then the corresponding pre is scription of the form 'X, don't do both A and not-i?' necessarily (logically) 'binding' on X.

(F1)-(F6) constitute strong evidence of the parallelism between imperative and indicative logic for the reason that (F3)-(F5) correspond to the axioms of, and (F6) to the rule of inference of, a classical two-valued system of propositional logic put forward by J. Barkley Rosser5:

Al A2
A3

p z) (p & p)
i.e., ~(p & ~(p &p)) ~p)

(p &q) 3 p
& ~ ~(rp)) ~(~(qr) & ~q) & ~ ~(~{qr) & ~ ~(rp))) i.e., ~(~(/? Modus ponens: from p and ~(p & ~q) infer q. between these axioms and (F3)-(F5) are: (i) that axioms i.e., ~((p&,q)& ~(p & ~q) 3

The differences

456
(Al-A3) whereas are wffs

HECTOR-NERI of Rosser's

CASTA?EDA

system and belong to the object language, are the corresponding so statements, meta-linguistic (F3)-(F5) of (F3)-(F5) in that we should consider the object-linguistic counterparts stead of these; (ii) that the double negations of A3 are missing in (F5). But

an implication into a 'valid' (F7) erases difference (i) by mapping prescription. Likewise if we allow the substitution of the equivalents estab lished by (Fl) and (F2), we have, then, the erasure of difference (ii). Hence schema system. Natu (F1)-(F7) give us a system virtually isomorphic to Rosser's 'X99 'A9, rally, this isomorphism holds only for the values of the variables Since 'X' can represent any sequence of agents 'j?', and 'C of (F1)-(F7). referred to in the atomic prescriptions good deal of generality. Nevertheless, constituting A and B, the result has a we grant that there is no effective way the range of the variables 6A\ 'B9, and 'C All we can say is of determining that the validity of schemata (F1)-(F7) does establish that for a class of to the two-valued actions the logic of mandates is isomorphic logic of

propositions. We must, of course, be prepared to encounter in the case of to the truth-value gaps well-studied in the mandates problems analogous case of propositions. we must also be ready to meet those problems But - as with all the weapons in the propositional armory already available well as with new ones that investigation may deliver. In short, we have excellent treatment of support for a two-valued to develop the two and it is of philosophical importance imperatives, even if merely to discover the valued point of view in detail. This is so, the Ought statements

precise junctures at which it really breaks down. 2.3. Deontic Statements. One of the most striking things about to-do is, as we pointed out in Section 1, the fact that deontic are intimately linked to imperatives. ordered, (i.e., commanded,

bidden, wrong, permissible, transforma above naturally yields deontic sentences by two grammatical tions: (a) replace the imperative mood and put in its place the subjunctive the comma before form, by dropping prescriptional to the result a deontic prefix, for example: prefix (Id) It is permitted that Karl the main verb;

ismandated To begin with, whatever for requested, etc.) is said to be obligatory, etc. For instance, the imperative sentence (1)

(b)

if it rains, close the do the following: windows if and only if the awnings are not up, and if it hails, turn circulator A on if and only if circulator B is off.

ON THE SEMANTICS OF THE OUGHT-TO-DO the structural operators those of connection revealed here is this:

457

Phenomenologically, D?.l. Deontic perhaps

of the Ought-to-do type (of course, the Ought-to-be type) are operators

not on

prescriptions. in the range of a deontic operator can be the prescriptions Furthermore, as in the case of (Id). Now, it seems that deontic atomic or compound, are themselves not prescriptions. statements of the Ought-to-do The type verb 'Ought' does not even have an imperative form, and expressions 'Be obligated to do A9 or 'Be forbidden to do A9 sound nonsensical. course, one can perform actions sense to advice or order "Make and Mary stay home mand the performance the mandate demands
obtain.

like Of

so that it is that bring about obligations, it the case that it is obligatory that John after 6 p.m." But here the imperative does not de of the act of being obligatory: there is no such act; the doing of an action that causes an obligation to are statements or propositions.

Indeed, Thus: D?.2.

deontic

statements

Deontic
or

operators

of the Ought-to-do

type yield propositions

statements.

An

immediate

consequence operators

of D?.l

and D?.2

is type are not iterative.

D?.3.

Deontic

of the Ought-to-do

of D?.l and D?.2 since This consequence provides a partial confirmation or (ii) stuttering iterations of deontic prefixes are: either (i) nonsensical, repetitions, or (iii) the iterated prefixes are not meant in the same sense, as 'It is obligatory
clearly shows.6

that it is obligatory

that everybody

pay his income tax'

D?.3 does not, however, preclude a deontic operator from applying to a as in "It is statement as a constituent having a deontic prescription obligatory that Paul do the following: do A, if it is obligatory that he do A." 2.4. Overriding Ought and Prima Facie Oughts. It is a commonplace that that there are conflicts of there are many types of deontic statements, is not duties, so that what we ought to do everything being considered what we ought to do given certain considerations.7 Thus, what we advise to do is not an action that we believe he ought to do, but one somebody

458
that we believe That is, While

HECTOR-NERI he ought to do above

CASTA?EDA considered.

else, everything

being

D?.4.

the overriding ought in "X ought to do A" may be said to imply or fully support the imperative "X, do A", the prima facie oughts in conflict do not imply or fully support their corresponding imperative.

the considerations may be which determine obligations, it Now, whatever a set clear that those considerations, is phenomenologically by determining of deontic statements of the form "It is obligatory the values of namely, a set of mandates, at their core. Thus, formally we can represent a preceding prescriptions set of deontically relevant considerations by the set of prescriptions they determine. if Formally, we shall identify different sets of considerations determine the same set of prescriptions. they necessarily 2.5. Ought-to-do and Can. Philosophers continue to debate whether Ought a set of prescriptions, they also determine that .4", determine also 6A9 in that form. Likewise, to wit, those which have the

seems to be that there are implies Can or not. The truth of the matter several uses of the deontic words in some of which Ought does, and in others Ought does not, imply Can, especially when the Can in question is not a logical Can. Here we choose to be concerned just with those deontic statements adopt: D?.5. "It is obligatory that X do A" "X to do A" is consistent. is a natural elucidation implies that the prescription of the Ought-to-do subtype that imply Can. In particular, we

of the idea that Ought implies Can, which that Xdo A" implies "It is possible for is often put thus: "It is obligatory Xtodo^". D?.5. There is one case that deserves that everybody obligatory implies neither "Everybody consider (4) special analysis. We all agree that "It is has a wife bring her to the meeting" who has a wife" nor "Someone has a wife". But

It is obligatory that someone her to the meeting. (4) imply "(There is) someone

(or other) who

has a wife bring

Does

[in the universe

of agents

in question]

ON THE SEMANTICS OF THE OUGHT-TO-DO

459

(who) has a wife"? The answers run in both directions. And, again, the is that the deontic words have many different senses truth of the matter or uses, so that for some statements expressed with the sentence (4) the implication holds, and for others it does not. However, we can provide a two such sets of senses or uses. In the senses for distinguishing or uses in which Ought implies Can, (4) does imply "Someone has a wife", for if nobody among the agents in question has a wife nobody can fulfill the philosophers' established by (4). Naturally, the obligation job is, in criterion fact, to elucidate the structure note prejudice: D?.6. thus, we must For of our conceptual that frameworks without

concepts, statements of the form "It is obligatory that p and A", imply the corresponding statement that p, where and 'A9 '/?' stands for a proposition for a prescription. of deontic 'Paradox'.* deontic It is normally statements : held that a principle

some families

2.6. The Good-Samaritan like the following (P) governs

If X9s doing A implies 7's doing B, then for X to do A (1) that it is obligatory, obligatory; for Y to do B, and (2) that it is wrong; X to do A. for Y to do B implies

implies

that

it is

that it is wrong,- for

to deontic in a standard approach of Now, logic that allows mixtures deontic and non-deontic statements, the antecedent of (P) is interpreted as the meta-statement that 'Z does (perform) A" implies "7 does (per form) B". And this has given rise to the so-called good-samaritan paradox. the case of Arthur, whose duty is to bandage a man, his em Consider ployer, whom he will kill a week hence. Since "Arthur bandages a man he will kill a week hence" does imply 'Arthur will kill a man a week hence", it is taken that by (P) it follows that Arthur has a duty to kill a man. If there is a paradox here, then solutions based on differences of whom is both the bandager agents and times are beside the mark: here Arthur and the killer, and the time of the killing is later than that of the bandaging, it can be earlier than the latter. but, obviously, However, interpreted in the present case there is really no paradox, to have an antecedent about the implication even between if (P) is two

460

HECTOR-NERI

CASTA?EDA

statements: Keeping suffices to dispel the scope distinctions performance here. The sentence 'Arthur has a duty to bandage a man he will paradox kill' really means:

(5a)
and not

(3x) (awill kill x & a has a duty to bandage x)9 (3x) (a has a duty to bandage x & kill x).

(5b)

Thus, it does not follow from (5a) by (P) (1) that Arthur has a duty to kill, since the part "a has a duty to bandage x" of (5a) implies nothing about killing. In general, several of the cases that have been proposed as variants of the good-samaritan 'paradox' can be analyzed as involving confusions on the scope of the deontic operators at issue, or of a definite description: case that cannot be analyzed away by scope but not all. A beautiful in "Good Samaritans, 'paradox' of the Knower, and Epistemic Obligations", pages 366ff .8 Contrary-to-Duty Imperatives, the case of a man, say Jones, whose job is to know what is done Consider wrong by other people in a certain office. Suppose that Smith did A, which distinctions is wrong by the rules of the office. Thus, "It is wrong,, that Smith do A" and "Jones ought,- to know that Smith (does) did A" are true. Since "Jones knows that Smith (does) did A" implies "Smith (does) did A"9 by (P) (1) interpreted as we have been doing, we have, then, "Smith to do (have done) A", which contradicts the hypothesis that it ought,is wrong,- for Smith to do A. Here the scope distinction does not help. For one thing, there is apparently no satisfactory analysis of knowledge so that we can take, in the model of (5a), some conjuncts of the analysis outside the scope of the deontic operator ought j. But suppose that we can analyze "Jones knows that p" as "/? and Jones believes that p and Jones has evidence for that/?". Then the scope analysis patterned after(5a),of "Jones ought,- to know that Smith did A" yields "Smith did A and Jones ought i to both believe that Smith did A and have evidence for this". But the fact is that a duty to know is not the same as the duty to believe and have evidence: surely one can have the latter without having the former. there are psychological attitudes that one must acquire, or acts that one must perform that imply that something that psychological happens to be wrong has occurred. Such cases give rise to troubles for (P). in those cases in which there is no purely The trouble is compounded In general, is ?qvist's

ON THE SEMANTICS OF THE OUGHT-TO-DO

461

psychological

pure psychological core psychological having incorrectly, done

is the content, that can be extracted, in the way believing core of knowing. For instance, there is no purely that can be really obligatory when one is, allegedly said to be obligated to repent, to lament, or to apologize for, some action A which it is wrong; to do.

different types of proposed solution consisted of distinguishing ?qvist's in 'Some duties. But his proposal was shown by Lawrence Powers, Deontic Logicians', pp. 384-388, not to be at all adequate. Of course, we do not object to distinguishing types of duties: we have done so in our of conflicts of duties. But we do not have examination phenomenological In fact, we have already to resort to this to solve the Knower paradox. of (P), that deontic operators apply to prescrip found, independently tions. Thus, we can recognize ositional one, namely: D?.7. in (P) a prescriptive principle, not a prop

If prescription A implies prescription B, then (1) "It is obligatory; that A" implies "It is obligatory; and (2) "It is wrong; that B" implies "It is wrong;

that B9"

that A". to all have:

Now, D?.7 provides an immediate, sharp and unified solution form we the forms of the good-samaritan 'paradox'. In ?qvist's (6) which

It is obligatory,- that Jones know that Smith did A, "Jones know that Smith did A". This contains the prescription

"Jones knows that Smith did implies neither the proposition prescription A" nor the prescription "Jones do A". Hence, from (6) by D?.7 we cannot derive that it is obligatory^ [or wrong;] for Smith to do (have done) A. we may properly and consistently Thus, accept that the system of rules governing the tasks of all the people in Jones' office, including duties in exactly the same sense, and even of the same type. 3. A Basic Language for We pass now to describe the Ought-To-Do. Jones, are

structure of a pure deontic the syntactical a large Actually we start by constructing language for the Ought-to-do. number of such syntactical structures, one for each prima facie obliga toriness and one for the overriding ought. These languages will be called = ?>;*, for i l, 2, 3,..., where Dt* is the language of the pure overriding ought. We shall call the union of these languages Z>*.

462
Primitive

HECTOR-NERI

CASTA?EDA

individual variables, predicate signs: individual constants; '~' and '&'; the the connectives constants; sign '__' to indi underlying cate prescriptivity; or obligatoriness;, the sign '0?' of oughtness; and '(' and ')'. Rules of formation: We the small letters '/?',V, use Quine's corners implicitly throughout. Let indicatives (i.e., expressions of let the capital letters *A\ 'P', 'C

'r', range over

or propositional functions); propositions or prescrip of prescriptions range over prescriptives (i.e., expressions tional functions); let '/?*' and '#*' range over both indicatives and pre scriptives; let 'Z' range over predicates and 'x' over individual variables, unless otherwise (a) Indicatives (1) specified. are sequences having one of the form:

xi9..., xn)9 where Z is an w-adic predicate and each Z(xi9..., is an individual constant or variable, and no x? is underlined ; X;

(2)

(-/>);

(3)
(4) (5)
(1)

(P&q);
(x)p; OtA;
are sequences of signs having one of the followingforms
xi9..., xn), where Z is an w-adic predicate, each xn if

(b) Prescriptives
Z(xi9...,

an

individual

variable

underlined

(to indicate

or constant, and the agency of x?);

at

least one x? is

(2)
(3) (4) (5)

{-Ay,
(P&A); (A Scp); (x)A.

are all the wffs of Ot*. We write The indicatives and the prescriptives 6Ax9 and 'Ax_9, in the case of a monadic predicate, for 6A(x)9 and *A(x)\ on parentheses We adopt the customary conventions and respectively. ' ' ' v ', 3 ', and = '. usual definitions of We tions: DDL DD2. DD3. RtA= WtA = ~Oi~A Ot~A ("It is right; that A") ("It is wrong; that A") & ~0;~??("lt is completely introduce the other deontic terms by means of the following defini

F?A=~O?A that A")

free or optional

ON THE SEMANTICS OF THE OUGHT-TO-DO We

463

can already see that each D* satisfies some of the proto-philo sophical data of Section 2 above: (i) the difference between an atomic and its corresponding indicative lies in the copula; (ii) the prescriptive to some singular expressions the role of ex copula assigns prescriptive agency; (iii) mixed indicative-prescriptive perative; (iv) deontic operators map prescriptives there are no iterations of deontic operators. pressing We shall call the non-quantified respectively; i.e., Df Dc* and Df are compounds into indicatives; im (v)

subsystem of Z>* and of each Z>f*, is D;* without the rules of formation

(a) (4) and (b) (5), and similarly for Dc\

4. The Axiomatic Our non-standard Systems Ot* syntactical. Its latest

representation9, Z>;* to an axiomatic rules of inference. Axioms:

approach has so far been mainly consists of the extension of each system Ot*, by adjunction

structure syntactical of the following axioms and all wffs having at

aside from the axioms for quantification, one of the following forms are axioms: least 01. 02a. 02b. 03. 04. p*,

if p* has the form of a truth-table tautology. Note: the overriding OtA 3 ~Ot~A ought ZD A takes as axiom schema 02b OxA instead of 02a, while the pri (ptA & OtB) 3 Oi(A & B) ma facie oughts take 02a in (p &OiA) 3 Oi(p & A) stead of 02b ; all other axiom schemata hold for all Oughts to-do that imply Can, as dis cussed in Section 2.5.

Rules M.P.

of inference : (modus ponens) : If \~ip* and h;/?* 3 q*, then \-tq*.

DR1. DR2. UG. We adopt

3 OtB. If YciA => B, then h;0^ 3 A, then Vtp 3 OtA. If Vcip If \-iP*9 \-i(x)p*. the usual definitions on brackets. of 'proof, 'theorem', etc. as well as the Here 'h;/?*' means that/?* is a theo part of Ox* will be called non-quantified

customary conventions rem of the calculus 0?*. The

464

HECTOR-NERI

CASTA?EDA

the axioms of quantification Of; more precisely, Of is Ot* without rules UG, (a) (4) and (b) (4). We shall write 'r-;cP*' to mean without p* is a theorem of Of.

and that

deontic of propositional axiomatization The preceding logic is con a transliteration of Quine's sistent and admits of decision procedures, e.g., cum propositional variables.10 It for uniform quantification procedure adequately Thl. Th2. MT.l. MT.2. MT.3. MT.4. 5. Models As represents the following principles.

\~icOiiA & B) 3 OtA & 0;P ^p&OiA h;cO;0&^) If a prescriptive Ax^is Ax is logically valid. logically valid, valid. entails entails the indicative Ax. the indicative ~ Ax. then the corresponding

is logically ~Ot(Ax^ & ~Ax) nor ~Oi(Ax) Neither O ?Ax) nor ~0?Ax) Neither O ?Ax) for the Systems 3 we continue Dt*.

to refer: (i) to arbitrary wffs of Dt*9 whether indicative or prescriptive with the asterisked letters '/?*', '#*', and '/**'; (ii) and and V, to arbitrary indicatives of Dt* with the plain letters '/?',V to arbitrary prescriptives of Dt* with the capital letters *A\ 'P\ and 'C. (iii) in Section

where 5.1. Of. A model M for a system Df is an order triple (W0, W,iy, a nonempty set of entities called possible deontic worlds, or just W_ is worlds, for short, W0 is a member of J^ and is called the real or designated function that assigns to each pair of a world, and 2 is a two-argument world the following Rl. R2. and a wff of Df just one element of the set {1,2}, in accordance of W. rules, where Wj and Wh are members L(p*> Wj) l(~p*,Wj) l(~P*,Wj) = l or 2, if p* = l9 = 2. if and is atomic, only i.e., p* with

is a wff of D?* by otherwise

formation rule (a) (1) or (b) (1).

if ?(p*9 Wj)=2;

R3.
R4. R5a.

= = /((/?* &#*), _/(4*5 T^) l; otherwise, IF,) 2. = If there is a world Wj such that /(/?, J^) l, then for every world Wh: ?(p9 Wh) = l. = 1, if and only if for every world Wj in HOiA, W0) ^different = l. fromW0:I(A9Wj)

= = /((/>*&?*), W}) 1, if and only if both /(/?*, W}) 1 and

ON THE SEMANTICS OF THE OUGHT-TO-DO R5b. = l, W0) = l. Wj) if and only if for every world

465 in W:

?(OtA, ?(A,

Wj

\= icp*, if and only if for every model p* is valid in Df, = for/and inM. And/?* has a model if and only if l9 W0) W0 M,I_(p*, for some model M, /(/?*, W0) = l, for /, and W0 inM. It is a simple thing to show that If hic/?*, then h icp*. the proof of the following : completeness If/?* is consistent, proceeds along the lines of all proofs of

We

define:

MT1. And

Henkin MT2.

/?* has a model.

Outline of proof. By standard procedures it can be shown that the set of wffs of Df is denumerable and that every consistent set can be extended to a maximal set. Take any maximal consistent set of wffs of consistent con Df that includes p*9 and call it W0. Take as W the set of maximal sistent sets Wj generated from W0 as follows: every indicative/? of W0 is in Wj, and for every indicative of the form OtA in W0, A is in Wj\ in the case A is also in W0 if we are dealing with OtA. We let /be the latter = 1 if and function I such that : (i) /(/?*, only if/?* belongs to Wj9 and WJ) = 2 if and does not belong to Wj. It is clear from (ii) /(/?*, Wj) only if/?* the construction that (W0, W, /> is a model for/?*. We have, therefore, from MT1 and MT2, by standard reasoning, that: MT3. t~icP*, if and only if t= ic/?*.

5.2. Ot*. The models for the full systems Ot* are ordered quintuples W, D9 \j/9/>, where W0 and W are as above, D is a domain of ob (JVq, jects, \j/a function assigning subsets of D to the worlds in W, and I_ is as before we except for conditions assigning 1 or 2 to quantified formulas. Here introduce in principle all the problems of existence and modality; but it is worth emphasizing that those problems present no peculiar aspect for deontic logic. In fact, they are somewhat less pressing. Clearly, in modal propositional logic it is of great importance not to assume that the objects in the universe are necessarily fixed once and for all, i.e., regardless had more, of the objects it has we must allow that the universe may have or fewer, objects. On the other hand, for pure deontic logic we

466

HECTOR-NERI

CASTA?EDA

- if may assume without damage that the agents and objects are constant we are dealing with a system of duties and interdictions at a given time. even if we adopt deontic systems that Thus, there is no great distortion contain the Barcan formula and its converse as theorems. deontic contexts are wholly not referentially opaque in = sense, or they are nearly so. It is palpably clear that if a b and x Quine's ought; to do to a some action A (that involves no referential opacity), then x ought; to do A to b. And this is, of course, as it should be. Different properties of objects or persons may, certainly, yield different prima facie Furthermore, duties other considerations that conflict. But since each prima facie duty is relative to a ground, are irrelevant to it. In the case of the overriding

ought we are, so to speak, face to face with brute facts and events, after and one's the cancellation of all intentional and intensional considerations, duty is simply to alter a possible train of events, are now considered or referred to. regardless of how these

Since we are primarily interested in presenting a semantical foundation for our non-standard approach to deontic logic, we need not go into the and modality. standard problems of quantification for our non We must note, however, that our non-standard models data gathered in standard axiom system do satisfy the proto-philosophical Section 2 above. 6. Prescriptive In Section or Imperative Values.

2.2 we argued that prescriptive (and imperative) logic is two and the argument we gave does justify our axiom 01. But we still valued, owe an account of the semantical values of prescriptions. The crucial in a pos of a prescription of the membership thing is the interpretation sible deontic world Wj. (We are, presumably, quite clear on what itmeans to say that an indicative or a proposition belongs to a possible world). a prescription to a world if it is that Our view is essentially belongs is to explain endorsed at least partially in that world. And the problem is not a logical problem; is. This problem, however, what endorsement it is the analogue of the problem of determining what is true, which takes widely different forms in different cases : criteria for veridical perception, it suffices that we etc. For our present purposes scientific methodology, discuss the more Naturally, of endorsement. general characteristics are especially and deliberately some prescriptions endorsed.

ON THE SEMANTICS OF THE OUGHT-TO-DO This is so when,

467

fulfilled,

please"; Do's and Don't's. We

for instance, we issue imperatives that we intend to be at a dinner table we request "John, pass the salt, we also endorse prescriptions when we formulate rules or sets of as when also endorse the prescriptions which are enacted by of the institutions we belong. It may be also said

the rule-making bodies that we endorse the prescriptions which demand the doing of actions that are necessary for the attainment of the goals or ends we pursue. But are such that for a given prescription all of these kinds of endorsements

"X do A" we may very well neither endorse it nor endorse its denial "A" do not do A". Thus, in order to secure a two-valued logic of prescriptions we must include a closing principle that puts either A or ~A together with the endorsed prescriptions. Such principle behind the is, doubtlessly, argument developed phenomenological the principle in question is this : (J) If neither possible dorsement a prescription deontic world in Section 2.2. Our view is that

A nor its denial ~A

is endorsed

in a

only is true in Wj\ otherwise, 1. Alternative Deontic

Wj, given that determine Wj9 then :A belongs to Wj if and if A is fulfilled, i.e., the proposition to A corresponding ~A belongs to Wj.

all of the methods

of en

Systems for

the Ought-To-Do

Some philosophers have argued that deontic logic differs from proposi tional modal that hold in the latter logic in that some of the principles do not have valid deontic counterparts. For instance, many philosophers Other philosophers insist that there is a object to OtA 3 0?Avb). strong sense of 'right' in which the Tightness of an action is not identical with the non-obligatoriness of the omission of the action. Now we shall not examine these or related claims. We want to point out only that our non-standard semantical approach to deontic logic can a foundation for other systems. We can, for one thing, reject provide principle ( J) and have possible deontic worlds in which some prescriptions are missing together with their negations. We can, for another thing, limit the propositions that belong to all the possible worlds, this providing a distinction between fundamentally relevant and fundamentally irrelevant circumstances. We can, thirdly, consider deontic systems in which the worlds in W are not on equal footing, but related by weaker relations. In

468

HECTOR-NERI

CASTA?EDA

short, our non-standard approach to deontic logic is amenable to develop means of the fertile methods ment by discovered by Hintikka, Kanger, And, once again, our philosophical Kripke and Montague. understanding in full is bound to grow by the contemplation of several alternatives deployment. approach,
above.

so in view of an initial adequacy of the is particularly all the proto-philosophical data exhibited that fully satisfies This

Indiana University REFERENCES 1 Stig Kanger, Provability in Logic, Almquist & Wiksell, Stockholm, 1957; 'ANote
and Modalities' and 'On the Characterization of Modalities', both

on Quantification

2 Jaakko 27 (1961), and 'The Modes Theoria and Quantification', Hintikka, 'Modality Acta Philosophica Fennica 16 (1963). of Modality', 3 Saul in Modal The Journal 'A Completeness Theorem of Symbolic Logic', Kripke, 24 (1959), and 'Semantical of Modal I', Zeitschrift f?r mathema Analysis Logic Logic der Mathematik 9 (1963). tische Logik und Grundlagen 4 Richard and Quantifiers', Necessity, 'Logical Necessity, Ethics, Montague, Physical

in Theoria 23 (1957).

Inquiry 3 (1960).
5 J. Barkley IV, pp. Chapter 6 For a detailed deontic Morality Wayne 7 On Rosser, Logic 55-76. examination see H.-N. for Mathematicians, of the overall McGraw-Hill, New York, 1953,

statements

Casta?eda's

and the Language of Conduct State University Press, Detroit, see H.-N. and prima facie Casta?eda's conflicts of duties 'Imperatives, oughts 44 (1966). The Australasian Journal and Moral of Philosophy Oughts', Oughts 8 For see H.-N. a discussion to solve the Good Samaritan of proposals paradox, Casta?eda, 19 (1968); 'Acts, W.

connections between and imperatives Decisions and Oughts' in 'Imperatives, and G. Nakhnikian), Casta?eda (ed. by H.-N. 1967. edition, 1963; paperback Mich.,

L. ?qvist, tions', Nous 9 H.-N. Casta?eda, 10 See W. edition,

and Deontic Studies the Logic of Obligation, Calculi', Philosophical on Contrary-to-Duty 1 (1967); Nous 'Reflections Sellars, Imperatives', and Epistemic 'Good Samaritans, Obliga Contrary-to-Duty Imperatives, 1 (1967). Nous 1 (1967); and L. Powers, 'Some Deontic Logicians', 'Actions, Methods 107-117. Imperatives, of Logic, Holt, and Obligations', Rinehart Proceedings New of York, the Aris revised

totelian Society 67 (1967-1968) 45 ff.

V. Quine, 1959, pp. and Winston,