You are on page 1of 10

The Formal Analysis of Normative Concepts

Author(s): Alan Ross Anderson and Omar Khayyam Moore


Source: American Sociological Review, Vol. 22, No. 1 (Feb., 1957), pp. 9-17
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2088759 .
Accessed: 22/01/2015 10:06

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.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.

http://www.jstor.org

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
THE FORMAL ANALYSIS OF NORMATIVE CONCEPTS 9
ence between visual images and external facts. makers of contemporary science are em-
The mental image or idea was thought to broiled in metaphysical controversiesno less
be like a mirror image or portrait or possi- real because they are usually ignored.
bly a map which reproduces exactly or in In a sense, then, this present discussion of
some proportion the features of the object the role of "systems" and "models" in sci-
represented. The idea was true in proportion
as it reproduced the objected represented.... ence, a discussion admittedly much too am-
The naive pictorial theory of correspondence bitious, has reached a point not far from
was accordingly rejected by the more critical where it was left nearly a half-century ago
mechanists, and numerous attempts have by Aliotta. Scientific theory, which holds
been made to deal with the problem in terms that reality is a function of the observer and
of a symbolic theory of correspondence. The the observing process, maintains a philo-
hypothesis of symbolic correspondence, how- sophically idealistic position. This is, of
ever, creates a transition which has led many course, by no means a damning fact, except
mechanists to suggest that it is not corre- when unawareness of it starts from and
spondence that is important in the truth of a leads to unexamined generalizations. Aliotta
sentence or formula, but the predictive power
pointed out, "The scientific concept is some-
of these to produce expected results. The
truth of a formula is its workability.29 thing more than a mere summary of percep-
tions; it is not an abridged experience, but
The difficulties raised by the correspondence an idealized experience, and its fruitfulness
test of truth of a model are compounded lies in its ideal character."30 Thoughtfully
when one turns, with Pepper, to the other pursued, this theme makes it possible on
tests of truth and particularly when one the one hand to avoid the equation which
realizes that each dominant metaphor or asserts that the model is reality and enables
model seems to insist on its own special test
us, on the other, increasingly to incorporate
of its own truth. The point is that the model- reality into the model, into some model.
29 Stephen Pepper, World Hypotheses, A Study
30 Antonio Aliotta, The Idealistic Reaction
in Evidence, Berkeley and Los Angeles: University
of California Press, 1942, Ch. IX. against Science, London: Macmillan, 1914, p. 402.

THE FORMAL ANALYSIS OF NORMATIVE CONCEPTS *


ALAN Ross ANDERSONAND OMAR KHAYYAMMOORE
Yale University

logic has been


LTHOUGH mathematical indefinitely many systems can be con-
applied in several empirical disci- structed; but whether such tools are to be
plines,' logicians have only recently of use to sociologists depends in part on
begun to develop formal systems that are the amount of interchange across discipli-
in
likely to be of special interest to sociologists. nary boundaries. This article is written
From the point of view of modern logic, the hope of enlisting the support and in-
terest of investigators treating problems to
*This paper reports one aspect of a more gen- which such notions as rule, norm, obligation,
eral research program supported by the Office of
Naval Research, Group Psychology Branch, Con- etc., are central.
tract No. SAR/Nonr-609(16). Permission is granted Concepts of this kind are of obvious im-
for reproduction, translation, publication and dis- portance for sociological theory; almost all
posal in whole or in part by or for the U. S. Gov- sociologists make use, in one way or an-
ernment. [The Review retains copyright authority
for all other uses.] other, of such notions. But in spite of the
1 See for example the references given in Omar widespread use of normative concepts, very
Khayyam Moore and Scarvia B. Anderson, "Mod- little is known of their logic. For the most
ern Logic and Tasks for Experiments on Problem part logicians have restricted their attention
Solving Behavior," The Journal of Psychology, 38
(July, 1954), pp. 151-160. to descriptive discourse (variously called

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
10 AMERICAN SOCIOLOGICAL REVIEW
"informative," "factual," "cognitive," etc.), Wright,6 Feys 7 and Prior.8 Space does not
as opposed to prescriptive discourse (or permit a general review of these systems,
"emotive," "normative," "expressive," etc.). though each of them merits consideration.
Powerful technical tools have been developed We have selected for initial discussion a
for the analysis of the former, but there system which is particularly relevant to
have been only a few attempts to provide sociology, namely von Wright's Deontic
analogous machinery for the latter. What- Logic.9 But before discussing this formalism,
ever the reasons for the neglect of the it may be helpful to outline informally a
prescriptive may be, this situation seems problem typical of those on which the logi-
unfortunate. We all do, as a matter of fact, cal analysis might be expected to throw light.
draw conclusions from a variety of state- Some sociological theorists have main-
ments that may, broadly speaking, be called tained that under certain circumstances
"normative." We reason from moral pre- there is a "strain of consistency," and
cepts, from imperatives, from commands, under other circumstances a tendency to-
from statements of legal responsibility and ward inconsistency, among the norms of a
obligation, and from sets of rules for games given social order. An empirical test of such
like chess, bridge, and football. a thesis presupposes a reasonable explica-
These facts suggest that there are grounds tion of the concept of "consistency" as used
for optimism concerning the possibility of in such analyses. In many instances it is
developing a formal logic for norms.2 There difficult to see precisely what is meant by
have been many informal discussions of nor- "consistency" as applied to sets of norms.10
mative discourse, centering usually around Despite the vagueness and ambiguity of the
imperatives; but there have been very few term "consistency," however, we would ex-
attempts to give a rigorous and formal char- pect to find some uniformity in its applica-
acterization of the logical structure of the tion by sociologists. Consider, as illustra-
family of related notions consisting of legal tions, the following three cases.
or moral obligations, prescriptions, sets of I. If a set of norms both obligates and
directions, recipes, fiats, technological com- forbidsan agent to executea given act under
mands, and the like. Among these are Mally,3 the same circumstances,the set of norms
Menger,4 Hofstadter and McKinsey,5 von would no doubt be regardedas inconsistent.
1I. If a set of norms (1) makes it obliga-
tory to do act B if act A is done, (2) permits
2 It should be emphasized that the chief value
act A, but (3) forbidsact B, then again the
of formalizing a logic of norms, or indeed any logic set wouldno doubtbe held to be inconsistent.
at all, does not lie simply in the use of shorthand
notational devices, though an economical symbolism
It might be maintained,that is, that doing
can be helpful. The idea is not to "translate" the permitted never commits an agent to
from one language into another, but rather to doing the forbidden,within a consistent set
express propositions in a language the logical struc- of norms.
ture of which is explicit and definite. An adequate
6 G. H. von Wright, "Deontic Logic," Mind,
formalization makes it possible to check in a de-
tailed and rigorous way just what follows from 60 (January, 1951), pp. 1-15.
what, and such rigorous checking is an obvious 7 Robert Feys, "Expression modale du 'devoir-

prerequisite for the use of the axiomatic approach. etre,'" The Journal of Symbolic Logic, 20 (March,
For an especially clear discussion of the sense 1955), pp. 91-92.
in which systems of formal logic (including 8 A. N. Prior, Time and Modality, Oxford: The

mathematics) are "languages," see Alonzo Church, Clarendon Press, 1956.


Introduction to Mathematical Logic, Volume I, 9 "Deontic" logic is the logic of the modes of
Princeton: Princeton University Press, 1955, pp. obligation. Expositions of von Wright's system
47-68, especially footnotes 111 and 116. may be found in von Wright, loc. cit.; or in G. H.
3 E. Mally, GrundgesetZe des Sollens, Graz: von Wright, An Essay in Modal Logic, Amsterdam:
Leuschner and Lubensky, 1926. North-Holland Publishing Company, 1951; or in
4 Karl Menger, "A Logic of the Doubtful. On A. N. Prior, Formal Logic, Oxford: The Clarendon
Optative and Imperative Logic," Reports of a Press, 1955.
Mathematical Colloquium, 2nd series, No. 1, Notre 10 It should be emphasized that "consistency,"
Dame: University Press, 1939, pp. 53-64. in such contexts, frequently has only a vague
5 A. Hofstadter and J. C. C. McKinsey, "On analogical resemblance to "consistency" in the vari-
the Logic of Imperatives," Philosophy of Science, ous logical senses. For the latter see Alonzo Church,
6 (October, 1939), pp. 446-457. op. cit., pp. 108-109.

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
THE FORMAL ANALYSIS OF NORMATIVE CONCEPTS 11
III. Considera set of norms which (1) be performed in any case," none of the
makes act B or act C obligatoryif act A is terms "obligatory," "permitted," "forbid-
done, (2) permits act A, but (3) forbids den," are applicable to them?
both act B and act C. To say that such a Evidently what is required is a system
set of normsis inconsistentamountsto say- of concepts that will handle not only the
ing that an act which commits an agent to easier problems, but will also enable us to
forbiddenalternativeswould itself be forbid-
den by a consistentset of norms. deal effectively with complex and/or un-
familiar cases. The system of deontic logic
For such simple cases, it seems that com- proposed by von Wright provides, among
mon sense suffices to decide whether norma- other things, a tool for treating just such
tive systems are consistent (though of course questions as have been outlined above.
we are all familiar, from recent develop-
ments in science, with the fact that common FORMAL DEONTIC LOGIC
sense may well be inadequate even for Von Wright's system assumes a universe
problems which appear to be easily re- of discourse consisting of "acts," or per-
solved). But in any event it is clear that haps better, "act-types." The word "em-
common sense does not take us very far. bezzlement," for example, refers to a kind
The following case, although undoubtedly of incident, or a class of incidents, rather
much simpler than many situations which than to any particular occasion on which
actually arise, will provide an illustration. something is embezzled. As variables for
IV. Considera set of normsfrom which it "act-types" (for brevity, now, simply
follows that (1) it is obligatorythat if act "acts") he uses "A," "B" "C" . . . In
D be done then act B not be done, (2) it addition to the atomic acts over which the
is forbiddenthat if act A is done then act variables range, there are various combina-
C is done, (3) it is obligatorythat if act A tions of acts:
is done then either act B or else both act Correspondingto any act A there is a.
C and act D be done, and (4) it is per- negation-actof A, symbolized-A. ~A is that
mitted that act D be done. Should (1)-(4) act which is performedby an agent if and
be taken as evidence that the set of norms only if he does not performA; to use von
is inconsistentor not? Even if it is clear to Wright'sexample,"The negation of the act
the reader how this question should be an- of repayinga loan is not repayingit."
The conjunction-act,symbolized(A & B),
swered, it is obvious that realistic cases of
is that act which is performedby an agent
much greater complexitymight arise. More- if and only if he performsboth act A and
over, it would be valuable to have a clear act B.
statement of principles in accordancewith The disjunction-act (A v B) is that act
which any case, simple or complex,could be which is performedby an agent if and only
decided. if he performseitheract A or act B (or both).
The implication-act(A -- B) is that act
But complexity is not the only problem. which is performedby an agent if and only
Some apparently simple questions may be if it is not the case that A is performedand
puzzling and difficult to resolve, in part, B is not.
because they are rarely raised. For example, The equivalence-act(A <---- B) is that act
suppose that a set of norms obligated an which is performedby an agent if and only
agent to perform some act which is impossi- if he performs both A and B, or neither
A nor B.
ble for causal or perhaps even logical rea-
sons. Should such a set of norms be re- The only primitive deontic concept is
garded as consistent? If not, then what "permission"; the others (i.e., obligatory,
should be said about logically impossible 1" An example of a
acts? Are they to be forbidden? Or should cannot be performed, iscontradictory act, one that
smoking-and-not-smoking
we treat them as permitted? Or should they (at the same time by the same person). This sort
perhapsbe regardedas indifferent?Or should of contradictory act is not to be confused with,
we say that, inasmuch as contradictory acts e.g., smoking in defiance of a "No Smoking" sign,
or with sometimes smoking and sometimes not
are vacuous in the sense that they cannot smoking.

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
12 AMERICAN SOCIOLOGICALREVIEW
indifferent, forbidden) are defined in terms Wright remarks, indifference is a narrower
of this primitive concept. "Permission" is concept than permission. Everything indif-
symbolized by "P," and the result of apply- ferent is permitted, but not conversely; ob-
ing "P" to the name of an act is a formula ligatory acts, e.g., are permitted but not
of the system. That is, "PA" is to be in- indifferent.
terpreted as expressing the proposition that The definitions above make use of con-
acts of kind A are permitted. (The notion nectives drawn from the propositional calcu-
of permission is, of course, always to be lus as well as from the calculus of acts. And
understoodas relative to some set of norms.) the application of the propositional calculus
Similarly, "PEA" is to be understood as to the deontic propositions yields many
saying that it is permitted not to do A; etc. formulae which are theorems in virtue of
"P" together with the act-types thus pro- their propositional structure alone. For ex-
vide a collection of deontic propositions. As ample, (PA-- PA), i.e., if A is permitted
was remarked previously, there is available then A is permitted. However, the interest-
adequate technical machinery for manipu- ing theorems from the sociological point of
lating propositions, and von Wright brings view are those that depend for their status
to bear on the deontic propositions the as theorems on the deontic concepts, and on
classical two-valued propositional calculus. the two principles governing them, which
This calculus contains analogues of the op- we quote from von Wright:
erations just described for the calculus of I. The Principle of Deontic Distribution.
acts.'2 The other deontic concepts may then If an act is the disjunction of two other
be defined in terms of the concept of per- acts, then the propositionthat the disjunc-
mission as follows: tion is permittedis the disjunctionof the
propositionthat the first act is permittedand
Def. 1. OA= df APIA
FA = df -PA the propositionthat the second act is per-
Def. 2.
mitted.14
Def. 3. IA = df (PA) & (P~A).
II. The Principleof Permission.Any given
Definition 1 says that an act A is obliga- act is either itself permittedor its negation
tory ("O") if and only if it is not permitted is permitted.'5
not to do A; definition 2 says that an act A The formal analogue of Principle I is
is forbidden ("F") if and only if A is not
permitted; and definition 3 says that an P(AvB) d---> [(PA) v (PB)],
act A is indifferent ("I") 13 if and only if
that is, if it is permited to do A or B (i.e.,
it is permitted to do A and also permitted if it is permitted to execute the disjunction-
to do MA (i.e., not to do A). As von
act: A or B), then either it is permitted to
12That is, corresponding to any proposition p do A or it is permitted to do B; and con-
there is a denial U-p which is true if and only if p versely. The formal analogue of Principle
is false, and for any two propositions p and q II is
there is a conjunction (p & q) (which is true if and
only if both p and q are true), a disjunction (p v (PA) v (P-A),
q) (which is true if and only if at least one of p that is, either it is permitted to do A, or it
and q is true), an implication (p -> q) (which is
true if and only if it is not the case that p is true is permitted to do A.
and q is false), and an equivalence (p<--->q) (which With the aid of the three definitions and
is true if and only if p and q are both true or both the two principles, von Wright develops a
false). Expositions of this calculus may be found decision procedure16 as a method of de-
in any standard textbook in symbolic logic. Of
course the symbols "..I", "&",, etc., are here used
in a different sense, but no ambiguity need arise 14 Von Wright, "Deontic Logic," op. cit., p. 7.
since, for example, "I" applied to an act-name 15 Ibid., p. 9.
always yields an act-name, whereas negation 16There are in general two ways of specifying
(symbolized the same way) applied to a proposi- the set of theorems of formal systems, either by
tion always yields a proposition. means of an axiomatization, or a decision proce-
13Von Wright discusses this concept, but does dure. An axiomatization consists of a list of axioms,
not specify notation for it; we introduce a special together with one or more principles of inference
symbol for "indifference." For expository purposes, that lead to theorems. A decision procedure is an
trivial changes and occasional departures from effective method for determining whether or not
extreme rigor will be found here and elsewhere. an arbitrarily given expression of the system is

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
THE FORMAL ANALYSIS OF NORMATIVE CONCEPTS 13
termining which formulae are theorems. The with mentioning two more general consid-
system therefore shares with axiomatic erations relevant to von Wright's logic.
formulations the property of providing a In the first place, leaving aside for the
rigorous (and in this case effective) method moment the fact that certain expressions
of testing expressions for "logical truth." are provable as theorems, the formalismpro-
The illustrative examples mentioned earlier vides a clear explication of a number of
can easily be tested by von Wright's method, normative concepts. In this respect it may
and it develops that in each of the cases, be compared with the Hohfeldian analysis
(I), (II) and (III), the formal results ac- of such notions as duty, demand-right,
cord with the presumed common sense solu- power, immunity, etc.19 That analysis at
tion. In case (IV), von Wright's decision this level is not without value may be seen
procedure shows that the set of norms is from Hoebel's application of the Hohfeldian
inconsistent.17 concepts to primitive law.20
As regards the puzzling questions con- In the second place, the deontic concepts
cerning the deontic status of impossible acts, of von Wright's system have close formal
it appears that several alternatives are avail- connections with other logical concepts, a
able. Von Wright discusses this question, fact first pointed out in recent literature by
and proposes a principle of deontic con- von Wright himself. In fact it was the ob-
tingency: "A tautologous act is not neces- servation of formal similarities between the
sarily obligatory, and a contradictory act notions "obligatory," "permitted," "indif-
is not necessarily forbidden."18 This prin- ferent," "forbidden," and the extensively
ciple is consistent with the other two studied concepts "necessary," "possible,"
principles mentioned above, but it is not "contingent," "impossible," which led von
the only way to resolve the problem. In a Wright to formalize a deontic logic. The
system to be mentioned below, for example, formal similarities extend to other concepts
application of the deontic categories is as well, as summarizedin the table of modal
limited to contingent states of affairs (i.e., categories. (See Table 1.)
those which are neither necessary nor im-
possible). TABLE 1. MODAL CATEGORIES
*
It is not feasible to discuss the alterna-
Alethic Existential Epistemic Deontic
tives available in detail without giving a
complete and rigorous formulation of the a. necessary universal verified obligatory
systems. Since it lies beyond the scope of b. possible existing unfalisfied permitted
this paper to present these systems qua c. contingent partial undecided indifferent
formal systems, we will content ourselves d. impossible empty falsified forbidden

* This classification appears in G. H. von


a theorem. Although for some systems there are Wright, An Essay in Modal Logic, Amsterdam:
both axiomatizations and decision procedures, the North-Holland Publishing Company, 1951, p. 2.
two techniques are in general independent; some We have again made slight changes for expository
systems have axiomatizations and no decision pro- reasons. In particular,von Wright did not supply
cedures, others have decision procedures and no terms to designate two of the modes, on the
axiomatizations. The decision procedure for von groundsthat no suitableEnglish words were avail-
Wright's system consists roughly of a truth-table able. We have inserted "partial"and "unfalsified,"
analysis of expressions in disjunctive normal form. by fiat. So far as we know, this use of "partial"
There is as yet no axiomatization of the system in is new.
print, though it would be a simple exercise to
produce one. The alethic modes are predicated of prop-
17 A complete truth-table analysis of example
ositions, in such contexts as "It is necessary
(IV), carried out in the manner proposed by von
Wright, would involve initially a two-valued truth-
that p," and "It is imposible that p,"-p
table with 216 rows (some of which would be being any proposition. These modes have
deleted). The amount of labor involved in testing
such expressions can be vastly reduced by adapt- 19 W. N. Hohfeld, Fundamental Legal Concep-
ing methods of Alan Ross Anderson, "Improved tions as Applied in Judicial Reasoning and Other
Decision Procedures for Lewis's Calculus S4 and Essays, edited by W. W. Cook, New Haven: Yale
von Wright's Calculus M," The Journal of Sym- University Press, 1923.
bolic Logic, 19 (September, 1954), pp. 201-214. 20 E. Adamson Hoebel, The Law of Primitive
18Von Wright, "Deontic Logic," op. cit., p. 11. Man, Cambridge: Harvard University Press, 1954.

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
14 AMERICAN SOCIOLOGICALREVIEW
been treated in the systems of Strict Impli- of relations among the norms. However,
cation due originally to Lewis,2' and further these relations by no means exhaust the
studied by McKinsey,22 von Wright,23 and interesting and important aspects of norma-
others. The existential modes are regarded, tive systems. Two crucial aspects of such
in many important contexts, as properties systems are the following. (1) There is a
of classes. For example, the statement that need for a method of relating norms to the
there are no rich professors can be con- system of social sanctions or penalties that
strued as saying that the class of rich pro- support them. (2) It is important to relate
fessors is empty. The existential modes are both norms and sanctions to possibilities for
handled in the theory of quantification, for action.
which one locus classicus is Principia Mathe- 1. It is, of course, customary to couch
matica.24 The epistemic and deontic modes discussion of norms in terms of the conse-
were first investigated formally by von quences of conforming to, and deviating
Wright. It is clear that the epistemic modes, from, their demands. The lack of such con-
like the alethic, apply to propositions. Von sideration is not to be regarded as a defect
Wright, as previously noted, takes the de- in von Wright's approach; it is in fact com-
ontic modes as applying to acts, though mon in logical studies to isolate concepts
other alternatives are available, one of which and consider only some of their formal
will be discussed in the next section. properties. Von Wright's system simply ex-
Among the formal similarities noted by hibits the methodological principle of divid-
von Wright are the following (where the ing difficulties in order to conquer them
letters refer to entries in the table): what- individually, a practice characteristic of
ever is a is b, whatever is b is not d, and twentieth century studies in mathematics
whatever is c is neither a nor d. To put the and logic. But this lacuna does point to the
last relation more concretely: a contingent need for an expansion or elaboration of
proposition is neither necessary nor impos- von Wright's system so as to take adequate
sible; a partial class is neither universal nor account of sanctions.
empty; an undecided proposition is neither Although it is clear that the notions of
verified nor falsified; and an indifferent act sanction must somehow be got into relation
is neither obligatory nor forbidden. Of with the deontic logic, methods of effecting
course these examples by no means exhaust this rapprochement have not yet been ex-
the relations among modes in a given cate- tensively explored. However, some work has
gory. Moreover, there are systems of logic, been done toward analyzing the relation be-
some of which von Wright discusses, which tween commands (rather than norms) and
combine concepts from two or more cate- sanctions.
gories. In our opinion, a more adequate con- In an article primarily concerned with
ceptual scheme for the analysis of normative commands and imperatives, Bohnert sug-
concepts would involve a combination of gested a formal analysis of the role of
the modal categories listed above. "penalties" in imperative discourse.25 In
brief, he recommends that commands be
FURTHER DEVELOPMENTS treated as ellipses for disjunctive proposi-
tions, one part of which is ordinarily un-
Given a social group with an empirically
verbalized, but nevertheless understood in
specified normative structure, von Wright's the context of
utterance of the command.
logic would clearly facilitate the analysis The locution
"Do A," for example, is to be
21 C. I. Lewis, A Survey of Symbolic Logic, understood as elliptical for the declarative
Berkeley: University of California Press, 1918. proposition "Either you will do A, or else
22 J. C. C. McKinsey, "A Solution of the Deci-
S." where "S" is understood as referring to
sion Problem for the Lewis Systems S2 and S4, a penalty or sanction that the recipient of
with an Application to Topology," The Journal of
Symbolic Logic, 6 (December, 1941), pp. 117-134. the command will presumably attempt to
23 Von Wright, An Essay in Modal Logic, op.
cit. 25 H. G. Bohnert, "The Semiotic Status of
A. N. Whitehead and Bertrand Russell, Prin-
24 Commands," Philosophy of Science, 12 (July, 1945),
cipia Mathematica, Cambridge: Cambridge Univer- pp. 302-315. We have made inessential changes in
sity Press, 1910, 1912, 1913. his notation to facilitate subsequent discussion.

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
THE FORMAL ANALYSIS OF NORMATIVE CONCEPTS 15
avoid. Alternatively "Do A" might, depend- of playing, these possibilities being circum-
ing on the context, be understood as an scribed only by the (non-normative) rules
ellipsis for "Either you will not do A, or of the game, and not by their chances of
else R," where now "R" refers to a reward success or failure. From among these possi-
which the recipient of the command will bilities, certain strategies are selected as,
presumablyattempt to attain. roughly speaking, most likely to win.
Bohnert's suggestions have not been The case seems to us to be similar in
axiomatized, but they show at least one analyzing normative systems. If we think
way in which commands can be related to of possibilities for action as being bounded
penalties so as to yield propositions. This by non-normative"rules"consisting of "laws
feature of Bohnert's analysis is a distinct of nature," then the social norms stipulate
advantage, since it makes possible the ap- that certain of these possibilities will, if
plication of the usual logic of propositions actualized, lead to rewards or penalties.
to commands.26Other ways of relating com- There are extensively studied formal sys-
mands to propositions have been proposed,27 tems suitable for the analysis of possibili-
but Bohnert seems to have been the first to ties (and the other alethic modes). Pioneer
formulate such a relation in a way that work on this subject was done by Lewis in
brings in penalties and rewards explicitly. 1918,30 and since then the systems S1 to
It would not be difficultto treat Bohnert's SS 31 S6 32 57 and S8,33 M,34 and T,35 to
suggestion axiomatically,28and the resulting mention only a few, have been constructed
"logic of sanctions," as we might call it, and discussed extensively. These modal sys-
would constitute a rigorous formal theory tems provide alternative explications of the
on a par with von Wright's deontic logic. The concepts of possibility, necessity, contin-
obvious next step would be to construct a gency, etc., and this kind of analysis is
formal theory that encompasses both these required for the clarification of the vague
systems so as to bring out the relations be- notion of "possibilities for human action."
tween obligations, the imperative sentences Just as deontic logic makes a contribution
which express these obligations, and the to the analysis of normativesystems, and the
sanctions associated with them. "logic of sanctions" could also contribute
2. It is clear that the possibilities of hu- to this end, so the logic of "possibility,"
man action can be studied without regard taken together with the foregoing, may shed
either to particular realizations of these pos- light on the problem. What is wanted is a
sibilities, or to the consequences of acting formal frame of reference rich enough to
in any particular way. As an illustration of incorporate at least these three facets of
this point we may take the theory of games.29 normative systems.36From the formal point
The analysis of any game begins with a
complete characterizationof all possible ways 30 Lewis, op cit.
31 C. I. Lewis and C. H. Langford, Symbolic

26 Traditionally, commands have not been re- Logic, New York: The Century Company, 1932.
garded as propositions, primarily on the grounds 32 M. J. Alban, "Independence of the Primitive
that we are not inclined to ask of a command such Symbols of Lewis's Calculi of Propositions," The
as "Mail this letter" whether it is true or false. Journal of Symbolic Logic, 8 (March, 1943), pp.
Certain writers have evinced an extreme reluctance, 25-26.
for philosophical and grammatical reasons, to ad- 33 Sdren Hallden, "Results Concerning the De-

mit any very close relation between propositions cision Problem of Lewis's Calculi S3 and S6," The
and commands. In our opinion this is an open Journal of Symbolic Logic, 14 (January, 1950),
question, to be decided by constructing logical pp. 230-236.
systems whose utility can be tested in scientific 34 Von Wright, An Essay in Modal Logic, op.
practice. cit.
27 See for example R. M. Hare, The Language 35 Robert Feys, "Les Logiques Nouvelles des
of Morals, Oxford: The Clarendon Press, 1952. Modalit6s," Revue Neo-scholastique de Philosophie,
28 In formalizing Bohnert's proposal, care would 40 (November, 1937), pp. 517-553, and 41 (May,
be required to avoid, if possible, such counterintui- 1938), pp. 217-252.
tive results as those mentioned by Menger, loc. cit., 360f course there are other important aspects
p. 60. of normative systems; there are, for example subtle
29 John von Neumann and Oskar Morgenstern, problems concerning the relations between possi-
Theory of Games and Economic Behavior, Prince- bilities for action, and beliefs about these possibili-
ton: Princeton University Press, 1947. ties. There are a number of ways in which the

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
16 AMERICAN SOCIOLOGICAL REVIEW
of view, (1) and (2) above motivated the present O'M* in detail, but it may be of
construction of systems more comprehen- interest to mention the way in which the
sive than von Wright's upon which we will deontic modes are defined. We use the letter
commentbriefly. "M" to represent "possibility"; that is,
The system O'M*37 has the following fea- "Mp" is read "it is possible that p."
tures. In addition to the usual primitive O'p= df [(Mp) & (M-p)]
Def. 4.40
notions of alethic modal logic,38 0'SM takes & -M (uP & -S) .
only the notion "sanction" symbolized "S,"
as primitive. The basic deontic modes are In this definition of obligation "O'," the
defined with the help of this notion in such first clause, [(Mp) & (Ma-p)], says that
a way that von Wright's system (with ap- p is contingent, i.e., possibly true and possi-
propriate qualifications) emerges as a sub- bly false. That is, it is possible to act in
system of O'M*.The only axiom mentioning such a way as to make p true, and also in
the sanction states that the sanction is con- such a way as to make it false. The last
tingent; i.e., it is possible to behave in such clause M ( Up & -S) says that it is im-
a way that the penalty or sanction will occur, possible that p and the sanction should both
and also in such a way that it will not occur. be false; thus if p is false, then the sanction
The deontic modes are construedas applying S is true. This amounts to saying that p
only to contingent propositions; that is, (1) is obligatory if and only if it is contingent,
propositions p, q, r, . . . rather than acts, and failure to do what is required to make
are the elements of the universe of dis- p true would lead to the sanction.
course,39 and (2) the question of what to Similarly,we define:
do with contradictoryand tautologouspropo- F'p = df [(Mp) & (M-p)]
Def. 5.
sitions is answered by limiting the range of & AM (p &AS).
deontic modes to propositions that are
contingent. That is, p is forbidden if (as before) p is
It is beyond the scope of this paper to contingent, and moreover, it is impossible
that p should be true and S false. In other
systems we are discussing might be extended so
as to cover such questions. Epistemic modal oper- words, it is impossible to act in such a way
ators may prove of value in this connection, and as to make p true without incurring the
one obvious and important way in which these sanction.
systems should be extended is by adding quantifi-
cation theory (a generalization of the existential Def. 6. I'p = df M (p &-S) &M
modes). (up & -S).
37This system is closely related to a group of
new systems described and elaborated in Alan Ross That is, p is indifferent if it is possible
Anderson, The Formal Analysis of Normative Sys- that p should be true and S false, and it
tems, Technical Report No. 2, Contract No. is also possible that p should be false and
SAR/Nonr-609( 16), Office of Naval Research,
Group Psychology Branch, 1956. S false. I.e., it is possible to act so as to
38The underlying alethic system of O'M* is in make p true, without incurring the sanction,
fact von Wright's system M (shown to be equiva- and also possible to act so as to make p
lent to Fey's system T by Boleslaw Sobocinski, false, without incurring the sanction. (The
"Note on a Modal System of Feys-von Wright,"
The Journal of Computing Systems, 1 (January,
contingency of an indifferent p is a logical
1953), pp. 171-178). Choice of alternative alethic consequence of the definition of indifference,
logics yields systems somewhat different from and hence need not be explicitly stated.)
O'M*. The exposition here of O'M* does not reflect With these definitions the relations ex-
various "reductions" effected for the family of pressed in Def. 1-3 all become provable as
systems OM of Anderson ibid.
39The letter "p," for example, is a variable
theorems of O'M-; moreover, the two prin-
ranging over propositions. In this context, the ciples enunciated by von Wright are also
propositions may be about "acts," in von Wright's
sense. 40 Definitions 4-6 are arranged to facilitate com-

This point should not be construed as meaning parison with definitions 1-3. We here use "O"' in-
that "acts" are unimportant, or have somehow stead of "O", etc., to distinguish these notions
been done away with; taking propositions as ele- from those of von Wright. An analogue of von
ments of the universe of discourse, is simply an Wright's primitive notion of permission is defined
alternative interpretation, worth investigating. There in O'M* as follows:
may well be others, e.g., propositional functions. P'p = df Map & M (p & -S)

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions
THE STATISTICAL MEASUREMENT OF AGREEMENT 17
theorems of O'M*.41This means that von area of application of such formal systems.
Wright's system is a subsystem of O'M*; It is hoped that the foregoing discussion
and that some account is also taken of will have suggestive value for workers in
Bohnert's suggestions. It would be point- other sociological fields, and will stimulate
less to try to characterize O'M* further in interest in current research in mathematical
such a brief compass; formal systems always logic, especially modal logic. This research
require intensive study if their explanatory is not remote from the daily affairs of soci-
power is to be fully appreciated. ologists. For instance, whenever instructions
If the "explanatory power" of such con- are given for filling out a questionnaire,
structs seems highly abstract or remote, the commands expressing obligations are in-
situation should be compared with con- volved. More generally, any adequate socio-
temporary physical theories, where mathe- logical theory must encompass, in our opin-
matical theories much more recondite than ion, the concepts norm, obligation, etc. It
those presented here have a direct and im- is therefore a matter of importance to de-
mediate bearing on empirical research. And velop sound techniques for analyzing norms
just as many mathematical developments and systems of norms.
were initiated by empirical problems, so the
development of O'M* was motivated by the SUMMARY
hope of throwing light on some empirical The development of an adequate theoreti-
problems in small-group research. More cal structure for sociology will in all likeli-
specifically, in experiments currently being hood require interdisciplinary co-operation
conducted in the Yale Interaction Labora- between sociologists and those working in
tory, opportunities are available for study- the formal sciences if it is to proceed in a
ing, under well-controlled conditions, small maximally fruitful way. The purpose of this
groups in the process of developing norma- article is to bring to the attention of soci-
tive structures. A clear and accurate account ologists recent work in mathematical logic
of these interactional processess calls for a which has direct relevance for their research.
precise and rigorous conceptual framework Von Wright's deontic logic offers promising
within which to characterize the behavior leads for the analysis of normative struc-
of the group. tures; the family of systems of which O'M*
Small-groups research is of course only is a member provides a more comprehensive
one, and not necessarily the most important, framework, taking account of the role of
41 As applied to contingent propositions, that is; penalties and possibilities for action vis-a-vis
this qualification applies elsewhere as well. norms.

THE STATISTICAL MEASUREMENT OF AGREEMENT


W. S. ROBINSON
University of California, Los Angeles

A NUMBER of statisticalproblemsof im- portance of agreement as a statistical con-


portance in social research logically cept in its own right.
require the measurementof the agree- The idea of agreementappears in a variety
ment (rather than the correlation) between of situations, among which are problems in-
two or more variables. These problems are volving: the interchangeability of measures,
often handled individually as they arise and as in the substitution of an index for a
on an ad hoc basis. Appropriate techniques, more fundamentalmeasure; measurementof
however, are available for handling them, the reliability of an instrument such as a
though they have hitherto seen use mainly test or scale; measurement of the "objec-
in non-sociological fields. The purpose. of tivity" or lack of bias of a concept as applied
this paper is to discuss these techniques in to observational material by different ob-
a general setting, and to indicate the im- servers; comparison of observed with theo-

This content downloaded from 128.235.251.160 on Thu, 22 Jan 2015 10:06:10 AM


All use subject to JSTOR Terms and Conditions

You might also like