83
CU1LTURE AS AN INFORMATION SYSTEM
E. A. Hamme l
Unirversity of California, Berkeley
The forms and subtleties of human communication have long engaged the
interest of social observers, analysts and actors alike. Attention has been
directed to the functioning of language in the communication of fact and of
emotion, of the importance of gesture and attitude, and of the ways in which
speech is amplified and conditioned by the social context in which it occurs.
It is clear, too, that the control and manipulation of communication is im-
portant in the establishment of specific role relationships between individ-
uals,
An extension and further consideration of the last point, which has
been eloquently examined in the wTorks of Simmel (19L9) and Mo,f.an (1959),
among others, forms the subj,ec of +this essay.  If we nay- dsefine oociety for
our purposes here as a sysrtem of interrelated statuses and culture as a set
of patterned behaviors peculiar to them, we may regard all of culture as a
kind-of communication with a Single general function. The functlon of cul-
ture, in this view, is that of an information system which provides a vocabu-
lar-y of messages identifying statuses in the social system, for the actQrs in
that system. To say this is to do only a little mrace than restate the truism
that cultural behavior is patterned and that certain ways of acting and speak-
ing are proper to certain statuses in a given society.  It is the view of cul-
ture as a consciously employed              -s^rma>eo system, hcwev;r, wit`h sone of the
properties of infornation Systems -,w1hich' may give 9 :. e te uruism fter utility.
The designation of' statues bv acto-r inter se is a yithin g but simple and
automatic. and it is tiheir use of iulture as if it were an information system
(regardless of whrat eltse it may be) that inte-ests tus here.
One of the most rnltable features of information systems is that they
may differ in their degree of o-ganizzation, znat is. the degree to which
specific items of information are Intercorrelated.  ihe degree of organiza-
tion of such systems is measured, in information theory, bv the numiiber of
discrete items (expressed as binary digits) which must be "sent" in order to
transmit the messages in the repertoire of the system. The degree of organi-
zation is expressed as the entropy of the System, such that entropy varies
inversely with the degree of organization.   Consider, for example, a simple
information system designed to send only the messages "dog,"' "bear," tlmonkcey,
and "lgorilla," and assume that no other messages can exist. Assume further
that sender and receiver share the same information system. The system can
contain all four messages as discrete items of information and would thus
require "space" for all four items. If the structure of messages was so
ordered, however, that two items only were sent, and these in a particular
order, one could transmit "carnivore" or not transmit it as the first segment
and transmit " small" or not transmit it as the second segmentrl. Although each
message would have to contain space for two ordered items, the system as a
whole would have to contain only these two, rather than four, but could still
transmit all four messages. The organization of the second systen is higher,
and its entropy therefore lower, than that of the first.  The degree of
84
entropy also reflecLs the uncertainty of the receiver at any point in time as
to the specific nature of the message just received or that to be received
next; that is, it is also a measure of ambiguity.
Considerations of this kind are also applicable to cultural communica-
tion. Two actors must perceive their own and their alter's statuses with some
agreement if they are to interact efficiently. If they do not, the communica-
tions between them can be appropriately considered as "tnoise."  If the entropy
of their mutual information system is low, their respective statuses are
quickly identified, and instrumental social action can proceed.   In some rath-
er simply organized societies, in which the number of statuses is limited,
recognition of respective status may be immediate; it may, for example, de--
pend only on identification of sex and age.   (In fact, the actors may already
know each other, but we are not concerned with that instance here.) In more
complex societies, where the number of statuses is greater, more communica-
tion is necessary between strangers in order to establish the status relation-
ship between them. Beyond this, and in both cases, further communication
will often take place in order to establish role and subrole rel-ationships,
perhaps down to the last detail of personal style in interaction.   Social sys-
tems cle.    <an vary in this way, in the number of statuses and roles they
contain, or in our analogy, in the number of messages that may be transmitted.
Cultural systems, however, can also vary in the number and organization of
items of information that need to be imparted in order to signal a given
status relationship.  Some social systems, or segments of systems, such as
military organizations, are rather highly structutted, and the items of infor-
mation, such as insignia of rank, specif'y a complex message with a minimum of
vocabulary.  Others, like groups at a cocktail party, require many items of
information in order to achieve Specificatilon of relationships.
From one point of view, low ent--opy and efficiency of communication
are virtues. But a social system is more complex than a computer, and from a
second point oW view the flexibility of most social systems demands redun-
dancy and ambiguity in the items of theqir cultural systems, precisely in or-
der that status relationships may be manipulated by actors. With a reason-
able degree of entropy in the cultural system, that is, with the existence of
ambiguity, it is possible to delay precise definition of status relationships
between two actors. With redundancy it is still possible to identify these
relationships with precision and force once they have been established. A
cultural StC!rt must thus be extremely flexible; it must be capable of defin-
ing some relationships with rapidity and precision but of avoiding the defini-
tion of others altogether, and of playing all points of the scale between.
But this second point of view, of constant and free manipulation of
signals and of the delights of ambiguity, can be assumed in practice only by
the most cynical and self-possessed of actors. If only by introspection, one
would suggest that cultural systems were also characterized by a constant psy-
chological strain for consistency of messages, for definition of relation-
ships, and thus for some stability of position and security for actors. On
other grounds as well a strain for definition is assured, for in the absence
of such definition a society would not operate at all; in fact it would not
exist. Further, no society is without differential evaluation of statuses or
concepts of super- and subordination of social positions. At least among
superordinates one might expect to find insistence on status definition, vis-
a-vis subordinate alters, else their relative position in the social struc-
ture would be lost. All in all, the forces for consistency, efficiency, and
organization of the symbolic system probably outweigh the requirements for
redundancy and ambiguity, so that an unequal balance is achieved, with ambi-
guity remaining as a spice in the stew.
Having begun this kind of analogizing, I will press it further, first
inserting an apologia for a dangerous and often fruitless mode of thought.
Analogical reasoning is sufficiently common so that it might.be regarded as
the basis of all generalization and learning. It ought not to be rejected
out of hand as disreputable, as some would do, but should be examined in
particular cases to see if its results are useful. Our problem here is to
see whether a series of analogies with the concept of entropy are at all use-
ful in understanding certain common cultural phenomena. Entropy, as a con-
cept of communications theory, is a perfectly good, rigorous, and mathemat-
ically precise bool for the analysis of communications systems. As a concept
of statistical mecrianics (where it differs in specific rature, but not in
general intent) it is equally precise.  It cannot be that precise in its ap-
plication to the analysis of cultural systems, for reasons to be set forth
later. Nevertheless, I would not reject the general implications of these
concepts (in communication and in mechanics) as an entering wedge into con-
cerns proper to the social scientist. We4might well discard the wedge after
its use, but let us retain it for a time.
From what has been said above, positing the particular -relationship
of culture to society as that of an information system to a series of possi-
ble messages to be transmitted, several things follow.   The nore complex a
society is in its struciture of statuses. the more complex must be it.s cultur-
al code for the transmission of status identification between actors.   If
manipulation of statuses is important (and I believe at always is), the cul-
tural code must always be a littile more complex than the seAies of messages
it is to transmit; it      haust hve a degree of entropy sufficient to allow both
ambiguity and redundancy.   _t can thus provide emphatic messages about status,
misleading ones, and some messages about statuses that do not even exist.-
The very richness of its vocabulary may in fact allow the development of new
statuses, if it does not stimulate them. The elaboration of human culture
and of society, often seen as being of a piece, exhibit in this view a partic-
ular kind of interrelationship which may aid in understanding their co-
ordinated, self-reinforcing growth. (See also Hammel 1962, for suggestions
on-a similar relationship between behavior and values.)
The process of expansion and growth just noted is, by and large, a
secular one. There is another, often equally unidirectional, which neverthe-
less exhibits some of the characteristics of cyclical change in that it often
appears, in the structural sense, to go nowhere at all. This is the "trickle
effect," or that very large component of intra ocietal diffusion which occurs
in a downward direction in social hierarchies.    (It is in fact possible to
argue that all intrasocietal diffusion involves this "trickle" or, in psycho-
logical terms, prestige emulation, but there is no need to force the issue
here.) Again, the concept of entropy may be analogically useful. First, let
us consider why the trickle effect, in a structural sense, occurs at all. To
do this, by our analogical route, we must introduce the concept of entropy in
statistical mechanics. There, entropy also refers to the degree of disorgan-
86
ization of a substance (say, a gas), but the arguments relating to it specify
that the entropy of a body cannot decrease without some expenditure of energr.
In fact, the entropy of most real systems tends to increase over time, since,
If they change at all without input of energy, they can only become more dis-
organized.
Are there any useful features in this concept for social scientists?
If there are, they seem to stem from the fact that the analogy between cul-
turl systems and information systems is partially false in a particular way.
The concept of entropy In communications refers to systems which are station-
ary and ergodic--unchanging In the sense that the frequencies of different
signals emitted do not change, on the average, with time. Quite clearly, cul-
ture Is a non-stationary source, even by definition. It does change. But,
as an information system it also has a particular degree of organization.
Thus, using both analogies, If it changes at all over time, without arny input
of energy, it must become a more disorganized information system.    Cultures
need not change in this non-directed way, but they often do, if only because
of         committed in the acting out of roles or in educational transmssion
of the rules of behavior, The conscious employment of ambiguity by actors
must also play its part. In the absence of constraints and corrective action,
then, culture must have an inherent tendency to fall apart. The entropy of
statistical mechanics will carry us only this far in analogy; it will not
"explain" why the trickle effect in any initial state system starts down the
hiezrarcb  rather than up.  In order to explain that feature (in an Initial
state), we would have to assme that any point on a hiearcby was in some way
more organized than any subordinate pointo    We can, howevers see how such a
process of downward "trickling" might be maintained once it has begun.
If  the cultural system of a hierarchically organized set of statuses
becomes less coherent and less representative of the set of statuses, it can
only mean that the statuses are then less distinguishable from one another In
the messages that are ecchanged between actors. If there indeed exists the
strain for consistency posited earlier, actors will make an effort to re-
establish the correspondence between culture and social system by reorganiz-
ing the content of their messages. They will, in other words, expend energy
in o rder to reduce the Informational entropy of the cultural   system.  Their
tactics can take a variety of forms.
First, all actors may agree that the incumbents of statuses should
practice only that behavior appropriate to their status     Second,s  as va ria-
tion of the first technique, superordinates my take steps to prevent the ac-
quisition of their status symbols by subordinates, even though they them-
selves do not reject their own acquisition of behavior appropriate to statuses
above their own. Effort expended in these ways tends to preserve the integ-
rity of the cultural system. The popularity of these means to stability Is
evident In the multiplicity of devices emloyed     Sumptuary laws are ubiqui
tous, and indigation at their transgression is often violent and supported
by supernatural sanction. To the actors threatened by this kind of cultural
disorgnization, violation of sumptuary laws Is truly an attack on God.
(Communism Is atheistic for more than just the obvious reason.) There are,
however, more subtle means of preventing intrasocietal diffusion. The
smbols of status may simply be difficult to acquire in the sense that acqul-
87
sition presupposes a certain worklng minimum; money, in a capitalistic econ-
omy, is a perfect example. Further, some kinds of qualifications for status
can never be acquired but can only be bestowed. The fact that membership in
a kin group and other closed rules of recruitment are so common, particularly
at the tops of social hierarchies, is testimony to their effectiveness.
Despite these efforts against creeping entropy, other means are al-
ways necessary. They relate not to the prevention of disorganization but to
its correction. Such correction can only take the form of acquisition of
some new form of behavior as appropriate to a status, to replace that behav-
ior which has been lost through intrasocietal diffusion. It is here that the
downward direction of such diffusion may be explainable. At any point inter-
nal to a hierarchy, where such new behavior must be acquired as a corrective,
there is ordinarily only one possible source--the statuses above. Given hu-
man evaluation of statuses and of their customary behavior, it is too much to
expect that actors will replace t>2e lost distinctive features of their sta-
tuses with cultural materials from belo-w them, particularly since the infe-
rior patterns may be recognlzed as former customary behavior of their own.
(An exception to this point will be noted below.) By and large, actors will
acquire cultural materials both which are appropriate to statuses above them
and with which they are familiar. If the behavior they borrow has been pub-
lic, they may borrow 'it from far up the scale (if t-hey can afford to acquire
it).  If it has been private, but they nevertheless know of it, they must
borrow it from a status more nearly adjacent to their own. Songs and dances,
which are cheap and public, move rapidly through hierarchies. Airplanes and
Cadillacs, which are public and expensi-ve, move more slowly. Sexual habits,
private and inexpensive, move even more slcwly. (My imagination and sense of
decorum will not cooperate to provide examples of behavior which are both
private and expensive.)
These phenomena have been discussed f-rom the standpoint of preserva-
tion and siLtrengthening of existiq ng ctural iiormation sysTems. Nothing, in
fact, has been said about status hunger, about emulation, ornly about a desire
to preserve the structural status quo. This is a conservati-ve attitude, to
be sure, but the most radical departures from expectations can be traced to
conservative forces if one wishes. Consider, for example, a marked exception
to the general rule of downward diffusion in social hierarchies-- the folk-
loristic revival.  Boston Brahmins dance to Cotton-Eye Joe, and Peruvian debu-
tantes sample sebiche and step to the marinera. There are many explanations
for such adoptions of lower class behavior by the elite.  (Not a few Peruvi-
ans eat sebiche because they like the taste.) But from a systemic viewpoint,
and with attention to the markedly nationalistic context in which such folk-
loristic revivals occur, it is clear that the conservative view may again
have its utility.  Elites, in the modern world, have much in common--so much,
in fact, that they are often hard put to distinguish themselves one from
another.  Often, they will compete in their rates of borrowing from some
mutually prestigious source, such as Paris. But even Paris has its limits as
a supplier. Ultimately, intercommunicating national elites can only distin-
guish themselves by borrowing from those segments of their own populations
which are normally the most isolated and therefore culturally most different--
their peasantries.
88
A borrowing of this kind from below in fact contradicts our earlier
remarks only in part, since it does not occur between adjacent points in the
hierarchy. It is as if the route of diffusion were a loop, from the elite by
stages to the plebes, and then directly again to the elite--from the minuet
to the reel. In its final folkloristic jump back to the elite, the "peasant"
custom is more often than not reintroduced by experts in nationalism and does
not constitute a direct borrowing at all.
Invention is of course another possible answer to the problem of the
differentiation of elites, but it is rare.  Significantly, it seems to be con-
centrated in the hands of artists and members of the intelligentsia in modern
society, actors occupying an anomalous social position.  Below the "top" of
the hierarchy in many dimensions of value, they nevertheless hold superordi-
nate positions in particular aspects of their culture. Like others, they
lose the ability to communicate the nature of their status as soon as others
borrow from them.  Scorning to borrow from below, yet rejecting much of the
behavior which signalizes generally superior status, they must create en-
tirely new patterns to revalidate their positions.  They, too, will borrow,
but when they borrow for themselves one often finds the curious phenomenon of
scorn for their own peasantry and avid borrowing from someone else's.
If the concept of entropy is at all useful in considering these phe-
nomena, its employment should lead to the formulation of testable hypotheses
that could not conveniently stem from another set of concepts, or which at
least seem obvious from the nature of entropy and the arguments following
from it.
(1) If the general effects of entropy are as given, so that intra-
societal diffusion tends to occur down hierarchies (even if they are multiple
and complex), outright innovation should be commonest at the tops of such
hierarchies, as should borrowings from other cultural systems. Internal bor-
rowing should be most frequent in the body of a hierarchy, and innovation and
external borrowing rarer. In general, the evidence of observation seems to
be in accord.  Most new Items in a complex culture start at or near the top
of at least one dimension of value and work down.
(2) If diffusion indeed acts as a stimulus to innovation at the top,
societies in which intrasocietal diffusion is common should exhibit a greater
rate of innovation than others. Since such diffusion should occur more fre-
quently when hierarchies are multiple, complex societies should exhibit a
greater rate of innovation than simple ones. A similar conclusion is evident
when one considers that complex societies must, of necessity, have a more
complex cultural information system, with greater opportunities for error and
ambiguity (both conscious and unconscious) in the transmission of status-
signalling messages. That higher degree of entropy, in itself, should stimu-
late efforts at reorganization, part of which would result in innovation.
The greater innovative activity of complex societies also seems confirmed by
observation.
(3) Hierarchies have one characteristic which is frequently ignored--
that is, they have a bottom. Consider the fate of this bottom in intrasocie-
tal diffusion. It obtains new cultural elements via the trickle effect but
passes them on to no one, since there are no subordinates. At any other
89
point in the hierarchy, two things can happen to an item of the cultural
information system:  it can be diffused downward, or it can pass out of exist-
ence. Both processes would stimulate replacement. At the bottom, however,
only one of these--outright disappearance--can occur. The stimulus for re-
placement must on that account also be less frequent. If that is so, less
borrowing from above should occur, so that the next point above in the hier-
archy would longer retain its integrity as a unified set of cultural signals.
These effects should be felt up the hierarchy, perhaps in diminishing degree.
If these speculations are correct, some negative feedback is present in intra-
societal diffusion, a feedback which acts as a brake on an otherwise self-
reinforcing process. General corroboration of these speculations may perhaps
be found in the greater traditiornalism and reluctance to change found in
lower segments of social hierarchies.
In defense of the proposition that these speculations have not been
idle, I can only suggest that they may arrange a variety of cultural and
social phenomena in a relatively unified way, so that their interrelation-
ships may be compared along a single dimension. That more than the straw man
of entropy is involved in culture change and in manipulation of information
cannot for a moment be denied, but to think of it as being at least partially
involved is an aid to the sociological imagination. Apart from its general
analogical utility, however, the concept of entropy seems to pose many opera-
tional difficulties. In a most rigorous view, the analogy is useless unless
it can be applied precisely, and that does not seem possible at the moment.
First, social systems are not stationary sources of information, so that the
concept of information entropy applies only to ideal systems of cultural in-
formation. Second, the complexity of messages, unit by unit and through
various levels of contrast and context, is so great as to defy measurement.
Even these difficulties could be overcome if one were to regard the concept
of entropy as proper only to a model of ideal sociocultural systems, with all
actual implementations only as approximations.  But there is a more serious
difficulty--that of the nature of the message.
In a telegraph system, the output message is a telegram. It can be
objectively compared with the message specified at the source to see if it is
the same or different. Indeed, in all of the communication systems in which
the concept of entropy is legitimately applied, the messages are in some
sense objects in the real world. But in the speculations offered, the mes-
sages are constructs, either of the analyst or of the actors in a social
system. We may quite properly calculate the number of bits necessary to
send the message "dog,t because we can see it printed out, d-o-g. We are
not concerned whether our code, be it 1011 or anything else, specifies a
"real" dog but only the word, 'dog." All of the calculations of information
theory are concerned with messages in natural (and sometimes artif icial)
languages which have some objective existence in themselves. If we are to
extend these concepts to the cultural language which describes status, we
must be sure that it is a language with objective validity. The problem
occurs with natural languages as well, but there it is possible to determine
whether or not the words of the language designate some identifiable refer-
ential object or not and whether or not actors are in accord about such
designations. (At least one can ask whether the words, spoken, correspond
to certain written forms that can be identified.) With statuses as our
referential objects, the anchor is much less secure. If we depend on native
90
conceptions of social status we are culture-bound and faced with the diffi-
culty of investigating a communication system only by means of itself. If we
attempt to use a universal scientific language of description, we find that
no such language exists, except for restricted spheres of culture and in the
most general terms.
It should not be thought, however, that the enterprise is hopeless.
The complexity of social and cultural phenomena will be described in time, as
analysts develop a culture of investigation which enables them to ignore
those aspects of events which are, for any particular theory, irrelevant. It
is not too much to suggest that the concept of culture as an information sys-
tem may aid in that process. By focusing attention on the descriptive prob-
lem of just what bits of information are transmitted in behavior and on the
inferential one of identifying the status messages involved in these signals
it may lead to further refinement of our consideration of the activities of
Maxwell's real demon.
NOTES
In anthropology, some of the more recent investigations and reflec-
tions in these matters are found in Hall (1963), Gumperz (1961), Hymes (1962).
*Versions of this paper were presented at the 61st Annual Meeting of
the American Anthropological Association, Chicago, 1962, and to the Stanford
Anthropology Club, January. 1963. I am indebted to the Center for Advanced
Study in the Behavioral Sciences for providing the time and freedom to dis-
cuss and develop some of these ideas, and particularly to D. H. Hymes,
T. D. McCown, and J. H. Rowe for their criticisms. Rowe has also written on
some of the phenomena considered here (1962); the basis for this article was
in fact laid in seminar discussions in 1957 (see Rowe 1962).
3For general presentations of the nature of information theory see
Cherry (1961) and Pierce (1961).
4Pierce (1961) is particularly (and rightly) critical of confusions
between the entropy of information theory and that of statistical mechanics,
as well as of their incautious analogical usages in other fields.   There re-
mains, however, a vague similarity between these entropies and some of the
phenomena discussed here. This article, in fact, attempts to take advantage
of just those tantalizing ambiguities and similarities, with apologies to
Pierce's justifiable demands for true rigor in application.
HFor discussion of the nature and functions of the trickle effect see
Fallers (1954), Rowe (1962), Tarde (1903). For some empirical examples in
Peru see Hammel (1961).
91
REFERENCES
CHEERRY, C.
1961
On human communication. New York, Science Editions. (Qrtginalt publ shed
by the MoIoT Press and John Wiley and Sons, 1957.)
FALLERS, L.
1954   A note on the "trickle effect."  Public Opinion Quarterly 18:,314.321,
GOFFMAN  E.
1959   The presentation of self In everyday life.  New York.
GUMPERZ, J. J.
1961   Speech variation and the study of Indian civilizationo  American Anthro-
pologist 63:967.988.
HALL, E. T.
1963   A system for the notation of proxemic behavioro  American Anthropologist
65:01003.10260
HANMEL, E. A.,
MS-a   Intrasocietal diffusion and the sociology of entropy.  Paper presented
at the 61st Annual Meeting of the American Anthropological Association,
November 1962.
M1iwb  Culture as an information system.  Paper presented to the Anthropology
Club, Stanford University, January 1963.
1961   Wealth, authority and prestige In thie Ica valley, Peru.  Albuquerque,
University of New Mexico Publications in Anthropology, No. 10.
1962   Social rank and evolutionary position in a coastal Peruvian village.
Southwestern Journal of Anthropology 18sl99-215.
HYMES, Do Ho
1962   The ethnography of speaking.  In Anthropology and human behavior,
T. Gladwin and W. Co Sturtevant  eds. Washington, D.Co. The Anthro-
pological Society of Washington.
PIERCE, JO Ro
1961   Symbols, signals and noise:  the nature and purposes of communication.
New York, Harper and Bros.
ROWE 9 J, H.
1962   A. social theory of cultural changeo  Berkeley, The Kroeber Anthropolog-
ical Society Papers, No. 26, pp. 75-80.
SIMMEL, G.
1949   The sociology of sociability, E. C. Hughes, trans. American Journal
of Sociology 55:254-261.
TARDE, G. de
1903   The laws of imitation, Elsie C. Parsons, trans. (from the 2d French
ed. of 1895). New York, Henry Holt. (First French ed. published In
1890..)
VEELEN, T.
1953   The theory of the leisure class. New Yorks, entor Bookso   (Originally
published In 1899, the Macmillan Co.)