Things of

BY ANDREW ABBOTT

I N this paper, I shall argue that it is wrong to look for

boundaries between preexisting social entities. Rather we
should start with boundaries and investigate how people create
entities by linking those boundaries into units. We should not
look for boundaries of things but for things of boundaries.

Interests in Boundaries

Problems of things and boundaries have arisen in both areas

of my research: studies of professions and temporality.
I conceived of the professions as living in an ecology
(Abbott, 1988). There were professions, and turfs, and a social
and cultural mappingthe mapping of jurisdictionbetween
those professions and turfs. A change in this mapping was the
proper focus of studies of professions and happened most
often at the edges of professional jurisdictions. These edges
could be studied in the three arenas of workplace, public, and
state.
All of this presupposed much about boundaries of profes-
sions, of turfs, indeed of jurisdictions themselves. About
boundaries, I presumed that they could be specified, that they
did in fact separate professions, and that they were the zones
of action because they were the zones of confiict. And, indeed.

*This paper was first given as ?. lecture at the Asilomar conference on organizations
at the kind invitation of Richard Scott. I thank him for the opportunity to develop
these ideas. I thank Robert Gibbons for pointing me towards coalition theory and
Michael Wade for discussions of runaway selection.

SOCIAL RESEARCH, Vol. 62, No. 4 (Winter 1995)

858 SOCIAL RESEARCH

i presumed a spatial structure to these boundaries, as did the

many people who attacked my theory for covering mainly the
Exceptions in the lives of professions and not accounting for
the stable life "at the core of a profession." Here was a notion
of professions as convex bodies, with secure heartlands deep
behind the boundary territories.
Beyond these implicit presuppositions about boundaries, I
jaresumed something much more profound. In arguing mainly
about interprofessional conflict, I took for granted the
existence of the professions doing the confiicting. This had
been necessary, of course. One has to presuppose something,
and if I made interprofessional confiict the focus of attention,
the bodies in confiict were the obvious things to presuppose.
But when I argued about the emergence of professions in
border territories, or about the gradual dissolution of
professions without jurisdictions, or about the transformation
of professions via amalgamation and division, I was taking for
granted the notion of acting bodies called professions, capable
of being split or joined, capable of coming into or losing some
kind of permanent existence.
A somewhat similarly linked set of issues about boundaries
and entities arose in my work on temporality. One of my
Central concerns was how processes of different temporal sizes
go together. The substantive issue again involved professions
(Abbott, 1982). Why were there very few psychiatrists working
in mental hospitals in the 1930s when psychiatry had begun as
the profession of asylum doctors in the late nineteenth
century? I considered a number of responses to this question.
One had to do with an annual mobility model of the behavior
of doctors entering and moving within the nervous and mental
disease area. Another had to do with the much slower growth
of exciting local communities of neurologists and psychiatrists
in major cities, communities that changed the very conditions
of mobility. Yet another had to do with gradual changes in
psychiatric knowledge, moving over a fifty-year period towards
psychologism and Freudianism. And a final analysis invoked
 THINGS OE BOUNDARIES 859

changes in social control that took more than a century to fall

into place.
It was easy to set up these four explanations for the
psychiatrists' move, but hard to imagine how they went
together. If one assumed that knowledge change was the
central causal force, then it would determine the mobility
choices of psychiatrists in, say, the qiunquennium 1880 to
1885, even though it itself would not be measurably complete
until 1920 or 1930. Obviously, that could not be. But it was
equally ridiculous to think that such grand, contextual changes
never mattered at all. Clearly, all these forces were working
through the same present and yet contriving to somehow work
independently.
All this drove me toward an intensely processual view of
social structure. But if one took such a processual viewas did,
for example, Ceorge Herbert Mead and Herbert Blumerthe
problem of entities became acute. Was a social entity a merely
accidental stability in a process, a kind of standing wave? Were
boundaries in fact literally ever-changing and, hence, not in
any real sense boundaries at all?
Now it is easier to explain stasis as an emergent phenome-
non in a fundamentally changing universe than vice versa.
Social theories that presume given, fixed entitiesrational
choice being the obvious current examplealways fall apart
over the problem of explaining change in those entities, a
problem rational choice handles by ultimately falling back on
biological individuals, whom it presumes to have a static, given
character. But it is very nearly as difficult to account, in a
processual ontology, for the plain fact that much of the social
world stays the same much of the time. Here, too, is the
problem of entities and boundaries.
In both these research areas, then, there arose questions of
the relation between entities and boundaries and questions
about the conditions under which social entities can be said to
come into or leave existence. I propose an answer to the latter
question by proposing an answer to the former.
860 SOCIAL RESEARCH

Boundaries into Entities

I begin with a basic assertion about the relation of

boundaries and entities: social entities come into existence
when social actors tie social boundaries together in certain
ways. Boundaries come first, then entities.
Let me restate this assertion in concrete examples. In the
view proposed here, a geographical state is a set of frontiers
which are later linked into what topologists call a closed Jordan
curve (a continuous single boundary that defines an inside that
is nowhere continuous with an outside). An organization is a
set of transactions that are later linked into a functional unit
that could be said to be the site of these transactions. A legal
corporation is a set of market (and other) relations that are
later linked in a certain, specified fashion. A profession is a set
of turf battles that are later yoked into a single defensible
position in the system of professions.
The major alternative to the position I am advocating takes
the relation between boundaries and entities as a synchronic,
even logical relationship. In this view, boundaries are a logical
correlate of thingness and vice versa. Therefore, indeed,
saying that a set of closed boundaries exists is logically
equivalent to saying that a social thing exists. This common
view obviously cannot provide a temporal account of the
origins of social entities.
It is not surprising that we never start with boundaries. The
prototypical entity in modern social thought is the biological
human being. We think about social entities as overgrown
versions of such biological individuals, and, thus, have become
accustomed to think that social entities have essences like
biological individuals, that they have some internal plan or
thingness or Aristotelian substance.' Moreover, we assign to
human individuals a self-other boundary guaranteed by
centuries of Cartesian philosophy and cannot imagine such
boundaries without human entities, an inability all too easily
generalized to the level of social entities.
 THINGS OF BOUNDARIES 861

I am here suggesting that we reverse the whole fiow of

metaphor. Rather than taking the individual human being as
metaphor for the social actor, let us take the social actor as
metaphor for the individual human being. Not only is there
much biological evidence for thisthe world is full of
organisms like slime molds and jellyfish that appear to be
individuals but are actually societiesbut also, under such an
assumption, the fruitful belief that there might be social
boundaries without social entities becomes possible.
Let me now turn to the logical issues of imagining
boundaries without there being any entities for those
boundaries to be the boundaries of. It will be helpful here to be
formal. In algebraic topology, spaces are understood in terms
of neighborhoods. The neighborhoods of a point x are
arbitrarily defined "parts of the universe that are near x." In
real (Cartesian) space, neighborhoods are typically circles or
spheres or hyperspheres, loci such that all points in the locus
are within a certain Euclidean radius of x. But, in general,
given a set of neighborhoods and a set M, a boundary point of
M is a point x such that every neighborhood of x contains at
least one point in M and one point in the complement of M.
(The complement of M is the rest of the universe once M is
deleted.)
Note that this formal definition of boundaries requires the
prior existence of entitiesthe set M and its complementin
terms of which the boundaries can be logically defined. Yet the
direction of the definitions could be reversed. One could in
principle define the neighborhood system and the potential
boundary set and then construct the set of which the (potential)
boundary set is the actual boundary. That is, given an
assignment of points to neighborhoods and a complete list of
potential boundary points, we would say that an entity M exists
if we can find some assignment of all the points in all the
neighborhoods to either M or the complement of M such that:
(1) each point has a unique location (in M or the complement
of M), and (2) given these locations, the points in the potential
862 SOGIAL RESEARGH

boundary set all do in fact meet the definition of boundary

points vis a vis M. This would define a set purely in terms of its
property of being a thing that possesses a boundary and is a
perfectly legitimate definition. But note that with the
definition so constructed there is no guarantee of uniqueness.
There might be several ways of assigning the points to a set
(not M, but N, O, P, . . . ) and its complement (N', O', P', . . . )
that fulfilled the definitional conditions. Indeed, these
different "assemblies of boundaries" would be many if we
allowed the definition of the neighborhood system to fiuctuate.
Thus, in formal topology, boundaries and entities are more
or less logically equivalent. Either one could be primal. But in
the logical sequence from neighborhood system to definition
of boundary points to definition of set we see a logic of
increasing specification that could easily be regarded as
temporal, an account of the emergence of entities.
To postulate the temporal priority of boundaries, however, I
must come up with a definition of boundary that makes sense
even when there is nothing to bound. To do this, I shall
replace the concept of set membership with the more general
notion of "difference of character." Thus, I shall define a point
X as a boundary point in space S if every neighborhood of x
contains at least two points that differ in some respect: not one
in M and one not in M, but simply two points that differ in
some respect. (Note that the boundary point is defined "in a
space S" rather than "of a set M.") In the simple case, this
difference will be a single known propertycolor, gender,
creed, education. In the more complicated (and more likely)
case, it will be a combination of properties or dimensions of
difference.
I have not mentioned "boundary of" anything. These points
are simply what we might call "sites of difference." Also note
the assumptions, here, of "differences" and of some kind of
atomic unit (the point) to which those differences can
appertain.
These two assumptions are critical and problematic. The
 THINGS OE BOUNDARIES 863

"differences" are things that emerge from local cultural

negotiations. That is, local interaction gradually tosses up
stable properties defining two "sides." These are not necessar-
ily category labels, although the traditional argument (which
starts with entities) would presuppose that fact. Eor me, the
central requirement is rather that these differences be local
and interactional. But to whom or to what unit do these
properties appertain? It would be easy to slide into the
traditional argument: these units are pre-existing entities
individual people, for examplewho bring different, endur-
ing qualities to interaction. In this case, my whole position
turns into an elaborate micro to macro translation argument.
I wish to claim something more radical. I want to emphasize,
with Blumer and other strong interactionists, that the units
might be anythingpeople, roles, physical locations, shards of
prior social entities, and so on. Of course, any social interaction
begins with what is in part a soup of preexisting actors and
actions. But interaction is not merely the actors' way of
reproducing themselves. This is the seductive assumption that
fools both functionalism and rational choice theory into
accepting a social ontology that by making stasis primary loses
its ability to explain change. If we would explain change at all,
we must begin with it and hope to explain stasiseven the
stable entity that is the human personalityas a byproduct.
Previously-constituted actors enter interaction but have no
ability to traverse the interaction inviolable. They ford it with
difficulty and in it many disappear. What comes out are new
actors, new entities, new relations among old parts.
What are these parts? Is this a hidden assumption that there
really are enduring, atomic units? It is not if we hold the world
to be "a world of events" (Mead, 1932, p.l). The parts are then
events, instantaneous and unique. That some events have
stable lineages, thereby becoming what we call "actors," is
something to be explained, not something to be assumed.^
Having thus extended the logical foundations of my
position, however, I readily admit that the example I shall use
864 SOCIAL RESEARCH

in this paperthe constitution of social work as a social

entitydoes happen to be an example that seems to fit into a
micro to macro translation framework. But that accident
should not mislead the reader. I could as easily have inverted
the example, for the production of what are usually called
micro entities (the personality, for example) takes place by
precisely the same process, but with what we usually call macro
entities playing the part of sites of difference. (I shall discuss
this example briefly later. An analogous view of personality
was held by Simmel.) Micro and macro are not equivalent to
real and emergent. Interaction and events are real; both micro
and macro entities are emergent. The world is a world of
events.

An Example

Let me move from algebraic topology and high theory to

social reality in the late nineteenth century. In 1870, social
work did not exist; the phrase did not exist, the set of activities
did not exist, the "thing" did not exist. There were some
Activities being done that would eventually be done by social
workers, but these were not being done by any group of people
in particular, nor were they aggregated even at the simplest
level into the tasks they would be when social workers did them.
For example, there were hospitals. Doctors, nurses, and others
working in those hospitals sometimes contacted authorities at
other institutions about patients they were discharging. But
this work was not organized into a systematic set of tasks done
by some one role in particular, much less was that role
articulated with "similar roles" in other social institutions. Or
again, "friendly visiting" by the wealthy to the homes of the
poor existed as a behavior; however, but it was not articulated
with anything like a systematic view of charity but was rather
seen as an outgrowth of earlier gentry-type obligations.^
Moreover, there were some tasks that would eventually be
 THINGS OF BOUNDARIES 865

done by social workers that were not even imagined. Nobody

was investigating family structures with an eye to their impact
on health. Nobody was thinking about vocational education
outside the trades themselves. Nobody was applying psychiat-
ric concepts like stress to everyday life.
By 1920, all this had changed. The word social work was old
and established. There were professional societies of social
workers. There were schools for social work. There were
journals for social work. There was an employment exchange
and a clearly defined labor market. There were institutions
hiring large numbers of such "social workers" and employing
them in positions of that name. Most importantly, there was a
fairly clear turf, a set of things to do.
If we then ask ourselves, "When did social work emerge?" we
find that that question immediately disappears behind the
more ominous question of "what does it mean to say that social
work emerged?" Now, we can address this latter question by
turning it into the question of "what was the order in which
certain institutions that we take as characteristic of social work
emerge?" Then at least we know what came before what. This
sequence-of-institutions view at least moves us from a static
view of a profession's origins to a narrative one.'' But the
narrative is deceitful. For we build our narrativesat least our
historical narrativesfrom back to front. We start with what
we know emerged and then seek its origins. But history is lived
front to back. Things emerge not from fixed plans, but from
local accidents and structures.
In particular, the social work turf itself, the shape of the
things to do and of things done, was by no means the relatively
given factor that it appears in, say, the case of medicine. Nor
should its origins be intuited by reasoning back from what
appeared in the 1920s. Rather, we must ask ourselves why
things like probation and kindergartens, which were originally
part of this area of activity, disappeared from it in the final
"thing" that emerged as social work.
Even worse, by the time we see major institutional events in
866 SOGIAL RESEARGH

social work, we are well past the real moment of structuration,

the moment when the very shape of tasks began to become
fixed. Eor example, the first local social work school arrived in
New York in 1898. The first professional associations arrived
in the teens, typically specialty associations in areas like medical
and psychiatric social work. The social work labor exchange
also appeared in the teens, as one of the major activities of
what would slowly become the national professional association
in reality as well as name.
But the real action was long before. The area of charities
and welfare began to take its first loose shape within the
National Conference of Charities and Corrections, which
began in 1874. At that original meeting were doctors, lawyers,
clergymen, wealthy men and ladies, university faculty, and a
whole variety of others. Under its consideration were things
like vocational education and probation, as well as lunacy,
tuberculosis, venereal disease, alcoholism, unemployment,
child welfare, and who knows what else. The early institutions
for social welfareinstitutional churches, settlement houses,
dozens of individual charities and charity societiesarticulated
between these and many other problems and services.
Settlement houses provided kindergartens, cooking classes,
adult education, and vocational guidance as well as several
varieties of what we now think of as social services.
To the people of that time, all these different services, which
seem to us like things that belong in schools and jails and
hospitals and other such places, made sense together in one
site. The emergence of social work as an entity was (that is, can
be defined as) the separation of those things into tasks that fell
under social work and tasks that fell elsewhere. And it is my
contention here that the separations themselves emerged as
independent, unconnected boundaries long before it made any
sense to speak of social work as a social entity.
Let me now return to the formal exposition. I had defined
boundaries not as points all of whose neighborhoods contain
both some inside and some outside, but rather simply as points
 THINGS OE BOUNDARIES 867

all of whose neighborhoods contain sites of difference. I did

not insist that all those differences be the same.
Such sites of difference are very common. Indeed, in many
social states of affairs, they are quite random. Early social work
is quite typical in this regard. If we look at John Mohr's data
(1992, 1995) on New York charities organizations, we find that
many adjacent charities differed in many different ways. His
data happen mainly to involve client differences, but other
data show many other kinds of differences.
These locally random sites of difference become proto-
boundaries only when they line up into some kind of extended
opposition along some single axis of difference. Thus, we
might have two kinds of people who do some task and find
that difference lining up into a systematic difference across
many work sites or across several types of institutions. The
kind of boundary that emerged between systems analysts and
programmers in the 1970s and 1980s is an example of this
difference. It appeared and reappeared independently in
organization after organization. In this case, we have an
extended set of boundary points that began to take on a special
reality by virtue of the precedence of one or a limited number
of types of difference. The names were conveniences to
describe the boundary, which emerged well before there was
any really systematic social reality to the entity "systems
analysis" or "programming." It is rather that a dimension of
differencein this case over how one approaches a computing
problememerges across a number of local settings to
produce a proto-boundary.
Note, however, that I have assumed, in the word "local,"
some kind of adjacency structure to the social space involved.
That is, we imagine some kind of metric of propinquity that
places institutions or areas. Examples of these metrics might be
professional mobility between areas, or career structure
linkages between areas, or relations of division of labor, or
client exchange, or whatever. It seems to me that we should for
the present lei?ve the nature of propinquity open. I simply
868 SOCIAL RESEARCH

underline that I have assumed some kind of propinquity

measure, which may be independent of the dimensions of
difference.
I am now poised to present a conception of the origin of
entities. In my social work example, we have a zone of social
space that we may loosely call welfare space or social order
space in which various proto-boundaries are set up. These
might involve gender, training, or prior profession. It is the
yoking of these proto-boundaries that makes the entity "social
work." Note that it did not really matter what these boundaries
were, at first. They began as simple, inchoate differences.
They were not boundaries of anything, but rather simple
locations of difference. They were not associated from one
workplace to another; they were not consistent from one client
type to another; they were not necessarily stable over time.
For example, kindergartens began in the 1880s to be
conducted by people with a much wider variety of back-
grounds than had appeared in the earlier Frobelian move-
ment: some came from education, some from volunteering,
some from churches. And these people also differed in some
cases by gender, by class, by level of education. Rapid
expansion from the earlier, smaller movement drove this
differentiation. But the important distinction that emerged
was in levels of special training; the older, specially-trained
Frobelians were overwhelmed by less trained workers coming
into kindergartens via the settlement houses. But in the
probation field the dimension of difference was different. In
probation, the main difference was among clients, who were
themselves differentiated by the state laws governing proba-
tion, and this difference of clients in turn drove a difference in
origin and orientation of people working with those clients:
adult probation largely demanded legal professionals, while
child probation was dominated by the newer aims of the child
welfare movement.
Both kindergartens and probation thus became sites of
difference, and, hence, boundaries in my sense, but the
 THINGS OF BOUNDARIES 869

differences were not similar across the two areas or even across
given instantiations of a type of institution in a single area. In
other parts of social welfare, a gender proto-boundary
emerged. The best example of such a boundary was in
psychiatric social work, an area in which men (psychiatrists)
and women (psychiatric social workers) did largely the same
kinds of things under different professional banners. In other
areas, what mattered was a similar opposition between people
who had connections to churches and those who did not.
Friendly visiting itselfthe very root of social workwas such
an area.
Social work as an entity came into existence when various
social agentsthe leaders of the settlement and charities
organization movements, the heads of state boards, the
superintendents of institutionsbegan to hook up these sites
of difference into larger proto-boundaries and then into larger
units. (Other agentsparticularly leaders of other environing
professions and proto-professionsdid so as well, the most
important of these being the newly powerful occupation of
school superintendents.) That is, social work emerged when
actors began to hook up the women from psychiatric work with
the scientifically trained workers from the kindergartens with
the non-church group in friendly visiting and the child
workers in probation. All those people were placed "within"
social work, and the others ruled outside it. An image was then
developed to rationalize this emerging reality as a single thing.
Unfortunately, in the process of making such a hook-up,
certain areas (like probation) may have ultimately proven too
distant, in some sense, to have one of their parties included in
the emerging thing called social work.
This is not to argue that in some cases, along other
dimensions, a single boundary may not have been crucial. One
"edge" of social work illustrates this well. An important
boundary in home economics, industrial education, and
kindergartens was that between services that were school-based
and those that were settlement-house based. All of these areas
870 SOCIAL RESEARCH

could have ended up "in" social work. But by linking the

"school" sides of these boundaries together, school administra-
tors achieved a much more secure location of certain welfare
Subjects into the school curriculum than did the emerging
social work leaders into their own institutions.
Note that these various proto-boundaries may well reach
clear out of what a functionalist might regard as social order
space. Gender differences are an obvious example. A gender
opposition in kindergartens, for example, was tied directly to
similar oppositions in school systems via those particular sites
where kindergartens had been started in schools. The social
world is a crazy quilt in several dimensions, with many local
regularities and edges, some of which peter out in quite small
areas, and some of which run clear out of sight.
The making of an entity is simply the connecting up of these
local oppositions and differences into a single whole that has a
quality which I shall call "thingness" (which I will examine
ishortly). In a great many cases, this connecting up is a matter
of conscious agency. The process involved in creating the
medical profession in mid-nineteenth-century England is a
good example. There were four or five potential candidates
for inclusion in the whole thing to be called medicine: the
apothecaries, who had begun by selling medicine but were now
Organized into a tight, self-examining group; the chemists,
with their skill in drug manufacture; the surgeons, with their
physical expertise; and the physicians, with their university
degrees. The preceding unities of these groups had broken up
because a new dimension of differenceexpertise certified by
direct examinationwas created by the apothecaries. On this
new dimension, the kinds of things available to differentiate
physicians from surgeons and apothecariesthings like uni-
versity degrees and certain kinds of training periodsfailed to
provide continuing, effective differentiation. That is, the new
dimension of difference collapsed preexisting differences. In
this new way of looking at the medical world, the only
effective, sharply differentiable entity one could create was one
 THINGS OF BOUNDARIES 871

that would include physicians, surgeons, and apothecaries,

which is precisely the group that became the modern British
medical profession in 1856.
Note that what the apothecaries did was to change the way
one looked at the space of differences. It was as if they forced
a three-dimensional world into two dimensions, and in doing
so moved themselves from the periphery into the center of the
medical group. As a result of their act, other professions and
actors beyond the professions moved to create a new unity.
The emergence of social entities is often such an act, a drawing
together of things. It could arise in any of the sources of
actioncharisma, tradition, optimization, accident, monoma-
nia, value rationality. We should remain explicitly agnostic
about which is actually involved.
I need now to consider the term "linking up" or "yoking"
that I have used throughout to refer to the connection of
boundaries. What does it mean to say that an organization is a
set of transactions which are later linked into a functional unit
that could be said to be the site of these transactions? Or that a
legal corporation is a set of market relations that are later
linked in a certain, specified fashion? Or that a profession is a
set of turf battles that are later yoked into a single defensible
position in the system of professions?
Yoking means connection of two or more proto-boundaries
such that one side of each becomes defined as "inside" the
same entity. There seem to be two ways in which this can be
done. In the example of British medicine just mentioned, the
introduction of examinations had the effect of destroying a
previous dimension of differencebroadly speaking, that of
class, both of practitioner and clientthereby bringing things
close together that had previously been far apart. In formal
terms, such a yoking is a projection from a social space of
higher dimensionality to one of lower dimensionality. (For
example, Monterey, California, is right next to Monterey,
Tennessee, if one ignores the difference in longitude, because
the two are at the same latitude36 15 N.)
872 SOCIAL RESEARCH

This kind of yoking is, I am persuaded, the central form

used when a social space is already filled with entities, when a
division of a social space into entities is already established and
institutionalized in some way. Under such circumstances, the
only ways to radically change arrangements in a social space
are to delegitimize old differences or to emphasize new ones.
The former strategy yokes entities together, the latter divides
them.
But when a social space is empty or, rather, unstructured, as
was the case with the area of social order and welfare in the
late nineteenth century, yoking means literal connection of
boundaries. That is, to create social work, a group of actors
hooked up the female side of friendly visiting with the
nonchurch-affiliated side of the provision of social services
with the nonmedical side of patient work in hospitals, and so
on. The founding definitions of social work included one
group from each of these disputes, placing that group "inside"
the entity-to-be.
This second kind of entity emergence can be envisioned in
two ways, one of which looks backward to the sources of my
argument, the other of which depends on terms yet
uninvestigated. In terms of my earlier definitions, the
emergence of an entity is the assemblage of various sites of
differenceboundaries in the loose sense defined earlier
into a set of boundaries in the topologically strict sense,
boundaries that define an inside and an outside. But the work
of creating an entity must also be seen as the work of
rationalizing these various connections so that the resulting
entity has the ability to endure, as a persistent thing, in the
various ecologies in which it is located.

Thingness

This brings us to the issue of thingness, entity-like quality,

endurance, or whatever we wish to call it. In the processual
 THINGS OF BOUNDARIES 873

ontology that I am here setting forth, the central quality of an

entity is endurance. If "the world is a world of events," in
Mead's ringing phrase, then what distinguishes entities is their
property of repetition, of being events that keep happening in
the same way. Repetition could arise either internally, through
some structure of causes that internally regulates "enduring
events," in which case I shall speak of internal reproduction.
Or it could arise from an external structure, an ecology that
leaves no real room for change in the individual, in which case
I shall speak of ecological reproduction.
But entity-ness seems to me to go beyond mere recurrence.
If one recalls my original question about the mutual effects of
historical processes of different sizes, it is clear that what makes
an historical event important is its independent standing as a
site of causation, as a thing with consequences. It is this
independent causal authority of large-scale and small-scale
events that makes theorizing the social process so difficult. And
what is true for large and small events is true also for large and
small entities, which are, in my argument, a subclass of events.
Thus, a second crucial property of entities is their ability to
originate social causation, to do social action. But action here
must be defined broadly, not merely as Weberian subjective
action, but rather as any ability to create an effect on the rest of
the social process that goes beyond effects that are merely
transmitted through the causing entity from elsewhere.
An entity is, therefore, something more than a standing
wave. It acquires, somehow, a coherence or internal autonomy.
If that coherence is lacking, if we have pure ecological
reproduction to which is added no internal solidity, it may be
less useful to think of a given recurrent event as an entity. At a
minimum, we must distinguish between such "internal" and
"ecological" forces promoting entities.
For a set of examples, let me return for a moment to the
theory of occupations. It is pretty clear that our ideal type of
an occupation includes three things: a particular group of
people, a particular type of work, and an organized body or
874 SOCIAL RESEARCH

structure, other than the workplace itself, capable of some kind

of reproduction. The high professions and the guilds are, of
course, the archetypical examples of such occupations. But we
can easily imagine social entities that lack one of the three
attributes, but that are, in fact, real social entities in the sense
that they can be continuously reproduced and can have
independent causal consequences.
First, suppose that we have a particular group of people who
are doing a particular kind of work, but who are not
organized. Such quasi-occupations typically appear in the
formative years of strong form occupations. But they may also
be permanently created, if there are organizational forces
laws against combinations of workers come to mindthat
prevent worker organization. But we can also imagine a
particular group of people with an organized body and no
work. Usually this situation arises from technological change
or from displacement by other groups or from loss of demand.
The railroad engineers illustrate the first of these, psychic
mediums the second, and the clergy, at various times, the
third. We can call such groups workless occupations. And,
finally, we can imagine an organizational structure tied to a
particular body of work, but no consistent group of people
doing that work. These can be called turnover occupations, for
their chief characteristic is the intense turnover of workers
within them. The nineteenth-century railroads offer examples
of this but more characteristic are modern life-cycle occupa-
tions, occupations served by individuals only at certain points
of the life cyclefiight attendant in the early years of fiying,
for example.
I, thus, have four types of occupations so far: strong form
occupations, quasi-occupations, workless occupations, and
turnover occupations. All of these are entities, in the sense that
they can persist, and that they can have causal consequences
for adjacent social groups. It is clear that we are willing to
think of them as social things.
But suppose I ask whether there are occupational entities
 THINGS OF BOUNDARIES 875

with only one of the foundational properties of an occupation.

What does it mean to think about a particular group of people
in the work world, without either an organization or an area of
work? How would such a group be identified? Such a situation
could arise as an occupation was disappearing. An example
might be canal boatmen, who up until a few years ago existed
in dwindling numbers in rural New Jersey and Pennsylvania.
There was an annual reunion but no work and no organiza-
tion. It was an occupation of memory and, in fact, had no
causal consequences for anybody. Similarly, it makes little
sense to think of a mere organizational form surviving without
work to do or a consistent set of members. Surely this is not a
social entity in any real sense.
But the third moment of occupationan area of work to
doexists in some people's minds as a social thing. It is true
that one might think of disaggregated tasks, at a very low level,
as existing in discrete chunks, independent of particular
people or organizations. (To imagine them aggregated is the
enticing trap of simple functionalism.) Thus, in the social
welfare area, things like teaching a mechanical arts class or
contacting another welfare institution or directing a program
for small children could all be seen as simple tasks. But how
these tasks are connected up is precisely what constitutes the
making of a social entity. It is all very well for the functionalists
to say "doctors have control of everything that helps make the
body well," thereby identifying the area of wellness. But this
area did not exist ex ante; one has only to think of the exclusion
from it, by the medical profession's definition of its turf, of the
single greatest determinant of wellnessdietwhich the
profession has been quite content to leave to uninstructed
family members for years. So this single dimensionof
taskis also no warrant for defining an entity.
We can then draw an effective line between entities and
non-entities in occupational life. I would like to raise two issues
about this distinction, however.
First, the argument presented here is basically an argument
876 SOCIAL RESEARCH

about tiling, about dividing up a space in some way. Taking it

as a general model for the creation of social entities seems to
assume that the typical process of entity emergence in social
life is division of labor or turf. To what extent do we in fact
wish to assume this process? Or are there rather several types
of origins for social entities, one of which is division of labor? It
is true that processes like amalgamation and division can be
Understood within a boundaries-into-entities format. Division,
for example, is simply making an internal difference into part
of an external boundary, putting that internal difference on
the same footing as the inside/outside distinction.
But it would be harder to so construe cloning, the typical
mechanism favored by the new institutionalists (like John
Meyer) to account for the origin of entities. Suppose we create
a new franchise or a new university. Are these not simply
clones of existing structures, and, thus, does their origin not lie
rather in imitation than in association of differences? In a
broader sense, cloning can be construed as a version of role
theory, and thereby directly connected to the notion that social
reality is produced by plans and scripts put into place by actors,
a la Parsons. To me, however, it seems more useful to think
about such scripting as a phase in the construction of entities.
That is, in the view presented here, scripting is one of several
ways of conducting the action that pulls together a set of
boundaries into a social entity.
Scripting cannot be seen as independent of boundaries
because no social entity ever takes shape in a vacuum.
Ecological constraints are always in position, and, thus,
scripting alone can never produce an entity. Take the extreme
case of the production of a human personality within a family.
A first child enters a complex environment of two adults, with
various boundaries placed between them in various dimensions
dnd directions. And those of us who have children know well
that the child's personality emerges as a pulling together of
various existing oppositions. The Oedipus complex is precisely
a contest over such a difference. There are many others.
 THINGS OF BOUNDARIES 877

arising in different differences that divide the parents in

different ways. From a hooking-up of these differences comes
a new social entitythe child's personalitythat in turn
restructures the existing boundaries of the relationship
between the adults, just as successive children will redraw the
cozy threesome of parents and only child. A child's personality
emerges as an assemblage of various sides of various sites of
difference between parents and child, parent and parent, and
child and other children.^
Thus, it seems best to retain the notion that the prestructure
of an entity lies in the creation of zones of difference within
the social process or social space. These zones of difference
gradually shape into proto-boundaries, which are then yoked
by some kind of activity into an entity. If this proto-entity is to
persist, it must have both internal reproduction and some kind
of causal authority. However, we must recall that this process is
not in any way necessary. There are many ways a given set of
boundaries could have been structured into an entity, and
what matters is simply that the resultant entity have internal
reproduction and causal authority, not that it be optimal in any
way. Rather, it must satisfy. Boundaries are always being set up
within groups, but only occasionally do these fall into
defensible or coherent possible entities. Sometimes, action or
accident takes advantage. But there is no guarantee of being
"best."
I would like also to make another claim about the process of
boundaries-into-entities. It seems to me quite significant that
entity-status among occupations involves more than one
dimension of difference or structure. It may well be that social
entities cannot exist without the tension provided by the
differing pulls of different structural dimensions. That is, what
gives entities their structural resilience is their defensibility,
their endurance in several different dimensions of difference.
In a simple form, this might be seen as a sort of overlapping
cleavages argument. Strength lies in overlapping cohesiveness
of several kinds. Yet, at the same time, an entity's causal
878 SOCIAL RESEARCH

influence or extent may reflect vast extensions along certain

particular dimensions. That is, while it might seem that
compactness (or heartlands, or whatever metaphor we choose)
would best allow an entity to reproduce and defend itself
against redefinition out of existence by other entities, what
produces causal authority may well be connection across long
reaches of the social world.
The professions offer numerous instances of this phenome-
non. A good example would be actuaries, who have a tightly
organized, rigidly controlled, and quite small occupation.
Entry is tightly structured, careers are tightly structured, task is
tightly structured. In all three basic dimensions of people,
work, and organization, actuaries are sharply bounded from
the rest of the professional labor force. Yet it is in a sense
because of this that their influence is so small. By contrast,
consider the accountants whose "profession" is porous to the
point of absurdity, whose careers lead in quite diverse
directions, and whose task areas include heavily contested
zones like tax law, unstructured areas like management
consulting, and dying heartlands like public auditing. The
accountants are, in fact, far more causally effective than the
actuaries because they can bring their force to bear in so many
different arenas, where they play so many different roles. It is
precisely the structure of long tenuous boundaries anchored to
a few more or less secure heartlands that enables accounting to
be powerful. As in the Padgett/Leifer argument, openness
provides strength (Padgett and Ansell, 1993; Leifer, 1991).
Just as for Leifer, skill means never having to make rational
choices, for entities, strong causal effect means never being
pinned in any one heartland.
This strength refers to endurance, to the temporal dimen-
sion of entities, as well. Rigidity provides short-term safety but
long-term vulnerability, as is well shown by the rigid but
self-defeating temporal structure of the English lawyers when
compared with the open and tenuous layout of the Americans
(Abbott, 1988, Ch. 9).
 THINGS OF BOUNDARIES 879

Conclusion

In closing, I shall mention the relation between this view of

social reality and certain others. I have already mentioned the
new institutionalism, a view emphasizing isomorphism of
entities and production of isomorphic entities both by
ecological processes (so-called "coercive isomorphism") and by
cloning (a version of "mimetic isomorphism").'' I have
elsewhere argued that much of the new institutionalism has
failed to make the leap to a process theory of reality (Abbott,
1992), a failure embodied in part in the emphasis on cloned
institutions. A simple reproduction model is far from the
precarious, processual social world envisioned here.
Another theory of entities identifies them as congeries
around an ideal type. Lakoff (1987) and others, for example,
believe social entities are constructed by defining an ideal type
and then identifying various entities that resemble it as
members of its class. "The professions" (as a class) is an
excellent example of such a phenomenon, with medicine and
law as the ideal types. But I believe this process of naming and
metonymy comes much later, long after the structuration
processes discussed here. Archetypal class images are central at
that later moment but do not play an important role in
creating entities in the first place, except insofar as scripts play
the role I earlier assigned to them. As in board games like go
and chess, by the time the structures are clear enough to be
labeled and discussed, they have long since been established
(Leifer, 1991.)
Another related theory is that branch of game theory
dealing with coalitions.^ To be sure, there are obvious ways in
which game theoretic concepts differ from the general view of
social reality presented here. The world of game theory is not
one of general fiux and indeterminacy, of unbounded games
in indefinite temporalties between players who themselves can
be reconstructed and whose interests and resources differ in
undefined and crosscutting ways. Nonetheless, coalition theory
880 SOCIAL RESEARCH

is among the most indeterminate in game theory: its

foundations complex, its implications unclear, its solution
concepts conflicting. There is some sense in which thinking of
entities as "coalitions of units" accords well with the view
presented here, particularly in a micro to macro context.
A final relevant theory is the theory of what we might call
attractor structures. In social life, we often have situations in
which runaway processes destroy the middle ground of
indeterminacy around a group. A group's interest in compli-
ance with extreme rules by its members reduces free riding
and, hence, increases rewards for remaining members, who
further strengthen the extreme rules, and so forth. Iannac-
cone (1994) notes such a mechanism among churches. Similar
runaway evolutionary processes leading to stable "entities" (in
this case whole social structures) are discussed by Michael
Wade and others (Breden and Wade, 1991; Wade, 1995).
These models are all organized by the same intuitive idea:
under certain conditions becoming a well-defined group
strongly increases the rewards to constituents of that group.
This intuition presumes a micro/macro character I have tried
to avoid, but the overall structure of the theoretical argument
is still closely related to the arguments made here.
I have set forth the view that social entities are often
secondary to social boundaries. It is unclear whether this rule
holds only in rare casesunformed turfs like social welfare in
the late nineteenth centuryor is the general rule accounting
for entities in social life. Addressing that question, however,
means embedding this claim within a larger, general theory of
social structure and temporality. Such a task takes me beyond
the bounds of this piece.

Notes

' In the extreme version (for example, in event history models), we

are willing to think about existence as an attribute and to allow
 T H I N G S OF BOUNDARIES 881

ourselves to imagine things that "are" but that lack the attribute of
existence, as if the world were made up of myriads of potential
entities, some of which have the predicate of actual existence.
^ In this position, I am following Mead and Whitehead. See Abbott,
1994.
^ For sources on social work, see the references in Abbott,
1994.
"Cf. Abbott, 1991.
^ This argument shows why replacing traditional role theory with a
role theory based on structural equivalence is useful. Recipes and
scripts can be proposed, but ecological constraints select among
them. The result is a hybrid "role theory" that consists of a dialogue
between scripts and surrounds.
*^ Both terms are from Dimaggio and Powell, 1983.
^ On coalitions, see Myerson, 1991, Ch. 9.

References

Abbott, A., "The Emergence of American Psychiatry," unpublished

dissertation, University of Chicago, 1982.
Abbott, A., The System of Professions (Chicago: University of Chicago
Press, 1988).
Abbott, A., "The Order of Professionalization," Work and Occupations
18 (1991): 355-84.
Abbott, A., "An Old Institutionalist Reads the New Institutionalism,"
Contemporary Sociology 21 (1992): 754-56.
Abbott, A., "Temporality and Process in Social Life," paper
presented at the Fifth Norwegian National Sociology Confer-
ence, Tjome, September 1994.
Abbott, A., "Boundaries of Social Work or Social Work of
Boundaries," Social Service Review 69 (1994): 545-569.
Breden, F. and Wade, M. J., " 'Runaway Social Evolution,' "Journal of
Theoretical Biology 155 (1991): 323-37.
Dimaggio, P. J. and Powell, W. W., "The Iron Cage Revisited,"
American Sociological Review 48 (1983): 147-60.
Iannaccone, L. R., "Why Strict Churches are Strong," American
Sociological Review 99 (1994): 1180-1211.
Lakoff, G., Women, Fire, and Dangerous Things (Chicago: University of
Chicago Press, 1987).
Leifer, E. M., Actors as Observers (New York: Garland, 1991).
882 SOCIAL RESEARCH

Mead, G. H., The Philosophy of the Present, A. E. Murphy, ed. (Chicago:

University of Chicago Press, 1932).
Mohr, J., "Community, Bureaucracy, and Social Relief," unpublished
dissertation, Yale University, 1992.
Mohr, J., "Reinventing Asylums," paper presented at the American
Sociological Association, August 1995.
Myerson, R. B., Game Theory (Cambridge, MA: Harvard University
Press, 1991).
t*adgett, J. F. and Ansell, C. K., "Robust Action and the Rise of the
Medici," American Journal of Sociology 98 (1993): 1259-1319.
Wade, M. J., "The Ecology of Sexual Selection," Evolutionary Ecology
9 (1995): 118-24.