. 3
( 7)


Williams, 2005.
66 A Realist Philosophy of Social Science

forget anything. The chief matter of understanding is exotopy of the one who does
the understanding “ in time, space, and culture “ in relation to that which he
wants to understand creatively . . .
In the realm of culture, exotopy is the most powerful lever of understanding.
It is only to the eyes of an other culture that the alien culture reveals itself more
completely and more deeply (but never exhaustively, because there will come
other cultures, that will see and understand even more).25
In what follows, we offer that there will be no paradox in holding that
the social scientist could have a better understanding of the society than
a member “ otherwise, indeed, it would be hard to see the point of doing
social science. There are two reasons for this. First, the ethnographer may
be obliged to make an effort at articulating that which is not articulated by
members. Competent members must know enough to carry on activities
“ they have practical knowledge even if they would often be unable to offer
this discursively. We can think of qualitative research as aiming at getting
clear about what actors do know, both discursively and non-discursively.
Thus, an ethnographer may, after inquiry, know more than members do
exactly because she has uncovered the implicit rules, recipes, and norms
that are implicated in everyday activity. This is also one of the virtues of
a comparative approach.
Getting a handle on members™ beliefs and understandings of their world
is but a necessary ¬rst step for both understanding and explanation in
social science. Because she can take a second step and provide an under-
standing of the social mechanisms that explain typical activities in that
society, her understanding is still greater.26
This last observation, to be developed in the next chapter, suggests an
opportunity not available to natural science. Improving our understand-
ing of the natural world can certainly help us to better adjust, even to
better in¬‚uence the outcomes of its processes. But we cannot change the
processes of the natural world; we cannot make gravity cease to exert its
effects, even if we can build powerful engines that propel rockets into
space. But improving our understanding of the social world does give us

25 Quoted from Sahlins (2004: 5). Sahlins is quoting Bakhtin from Tzvetan Todorov,
Mikahil Bakhtin: The Dialogical Principle (Minneapolis: University of Minnesota Press,
1984). See also, Patricia Hill Collins, who writes: “Each group speaks from its own
standpoint and shares in its own partial, situated knowledge. But because each group
perceives its own truth as partial, its knowledge is un¬nished. Each group becomes better
able to consider other groups™ standpoints without relinquishing the uniqueness of its
own standpoint or suppressing other groups™ partial standpoints” (2000: 330).
26 Bourdieu and Giddens share in arguing that the work of symbolic anthropology
(e.g., Geertz), symbolic interactionists, and those in¬‚uenced by Schutz, for example,
Gar¬nkel, provide incomplete accounts, stopping at what is termed here, “the ¬rst step.”
For Bourdieu and Giddens, there is a need, as here, to try to explain why the social world
is the way it is. Foucauldian genealogy may well ¬t in here as well.
Explanation and understanding in the social sciences 67

the chance to change social reality in ways more congenial to our values
and interests.

Social science as emancipating
While activity requires that the members have practical knowledge, it
does not require that they have a grasp of the mechanisms that produce
and sustain the beliefs of members. Thus, what role does mass media
play in forming belief? How is this related to mechanisms of the polit-
ical economy? To voting behavior? Nor does action require that all the
beliefs that sustain these mechanisms be true. Indeed, the reproduction
of a practice may require that members have false beliefs about the prac-
tice that their activity sustains. If it can be shown that some of the beliefs
essential to the reproduction of practices are false, distorted or otherwise
inadequate, that conditions are not what they seem and that consequences
were unexpected, agents will have grounds for changing their practices.
Consider the belief that males are superior. Is this belief essential for the
practices that de¬ne the traditional patriarchal family? But if (as people
increasingly appreciate) this belief is false, people have good reason for
changing their behavior and thereby altering the inherited roles and rela-
tions. It is not that the social scientist has inserted his (or her) values into
the inquiry. It is rather that there are inherent practical implications that
follow from seeing how beliefs enter into the constitution of a practice
and then asking whether these beliefs are true or false. Establishing the
truth or falsity of our beliefs about the world is, it must be agreed, the task
of any science. For the social sciences, this has an added emancipatory
potential (Giddens, 1984; Bhaskar, 1979).

Problems and objections
What are the problems with the foregoing theory of society? It will be
pro¬table to focus on typical criticisms of structuration theory. Remark-
ably, perhaps, Giddens has been read both as a structuralist determinist
who effectively denies agency, and as a voluntarist who “emphasizes the
control we exercise over our worlds.”
Thus, Richard Ashley (1989), in¬‚uenced by post-structuralist criti-
cism, sees Giddens as lapsing into structuralist determinism. He argues
that Giddens makes two moves. First, his narrative dichotomizes “the
utter arbitrariness of history” and “the structures of social totality whose
form theory represents and whose continuity theory narrates.” Second,
“knowing agents” are located “at the frontier of this already established
opposition; as beings who, behind their backs, are constituted in re¬‚ection
68 A Realist Philosophy of Social Science

of the structure of the social totality and who, looking forward, ¬nd nar-
rative signi¬cance only insofar as they administer historical contingency
and bend it to the reproduction of the structure that constitutes them”
(Ashley, 1989: 277). Thus, the dependence of structure on practice is
a mere supplement “ “a way of rendering the structuralist escape in the
face of contingent events that threaten to undo a structure™s supposed
hegemony in the determination of what history means” (Ashley, 1989:
277). According to Ashley, then, Giddens effectively subordinates “the
dependence of structure on practice” to the “dependence of practice on
structure.” Putting aside the darkness of the prose, Ashley sees Giddens
as obliterating agency and seems to have managed this by giving struc-
ture a reading characteristic of much deterministic social science: as both
oversocializing the individual and by giving structure the sole causal role
in action.
But far more of his critics have read Giddens as a voluntarist (and
subjectivist).27 On this view, because structure is but virtual, there are no
objective constraints on action. This line of argument, like the previous
one, trades on giving structure a critical causal role and we need to say
more about this. But two points regarding constraints on action may
quickly be re-emphasized.
First, there are the objective material constraints which severely limit
us because we are organic beings in a physical environment. These are
both powerful and usually underestimated (or ignored). They provide the
background causes of everything we do “ and suffer.
Second, there are the objective constraints imposed by history. Agents
reproduce and transform structure in acting: they do not create it; they
do not, as state of nature theory typically assumes, agree to form a society
ex nihilo, nor do they ex nihilo create the structures in terms of which they
will act. Since they can create only with the materials at hand, the legacy
inherited is profoundly constraining. But, for structuration theory, in con-
trast to a determinist understanding of structure, actors reproduce and /
or transform structure by acting. Hence, there is always both change,
often unintended, and stability. Nor for structuration theory, echoing C.
Wright Mills, is there a general theory of social change. Constraints on
action, like the capacities for transforming structure, are historically vari-
able, both in relation to the material conditions and inherited institutional
circumstances but also in relation “to the forms of knowledgeability that
27 A very good recent example of the latter is the work of Michael Burawoy (1998: 15),
who, while valuing re¬‚exivity and ethnographic depth, holds that for Giddens, “in the
end, intuitive notions of structure evaporate and we are left with a voluntarist vision that
emphasizes the control we exercise over our worlds.”
Explanation and understanding in the social sciences 69

agents possess about these circumstances” (Giddens, 1984: 179). The
charge of structural determinism and of voluntarism denies this.
Third, as regards members of societies, there are the constraints (and
enablements) imposed by the lottery of life. We do not choose our parents;
nor, accordingly, do we choose our time and place in history and society.
Both these generate objective constraints, best understood as “placing
limits upon the range of options open to an actor, or plurality of actors,
in a given circumstance or type of circumstance” (Giddens, 1984: 177).
Moreover, as Durkheim noted, these appear as “facticities,” or as Marx
said, as “natural.” But this is not inconsistent with the idea, argued here,
that “ ˜society™ is manifestly not external to individual actors in exactly
the same sense as the surrounding [natural] environment is external to
them” (Giddens, 1984: 172).
A further range of criticisms regards the relation and ontological sta-
tus of rules and resources, the two elements that, according to Giddens,
constitute structure. It is unfortunate that the idea of rules, which has
been so important in the philosophical literature, has misled some other-
wise careful readers of Giddens. Sewell (1992), for example, offers that
the “cultural schemas” would be a far more preferable term, since it
provides opportunities not available to the usual meanings of rules. Cul-
tural schemas are the meat of cultural anthropology (and, one should
add, work by ethnomethodologists and others, for example, Goffman)
and opens a richness which includes an array of “various conventions,
recipes, scenarios, principles of actions, and habits of speech and ges-
ture,” in addition, then, to “the sorts of things spelled out in statutes,
proverbs, liturgies, constitutions and contracts” (Sewell, 1992: 8). We
can welcome this suggestion (whether or not we take it as a clari¬cation
or an extension of Giddens™s idea). But critically, for Sewell, these latter
“more formally stated prescriptions” should be considered “resources”
not rules. And for him, resources are not virtual, but actual.
Giddens™s view that resources are virtual has troubled a host of writ-
ers who, like Sewell, agree that rules (schemas) are properly understood
as but virtual. For these writers, resources, unlike rules, have an objec-
tivity. Of course, statutes, constitutions and so forth, actually exist. But
this misses the point. The marks on the paper, like the paper, exist inde-
pendently of us, but they are statutes only because of our beliefs, beliefs
that we act on. These are, in Searle™s terminology, institutional facts,
concept- and activity-dependent. Moreover, none of these more formally
stated prescriptions is self-interpreting “ as would be admitted. That is,
writing down the rules of grammar does not make them less virtual in
Giddens™s sense: they are actualized concretely only in action.
70 A Realist Philosophy of Social Science

Sewell suggests another argument for making resources actual rather
than virtual. “To say that schemas [but not resources] are virtual is to
say that they cannot be reduced to their existence in any particular prac-
tice or any particular location in space and time: they can be actualized
in a potentially broad and unpredetermined range of situations” (1992:
8). But it is hard to see why this is not true of resources which also
empower agents “in a potentially broad and unpredetermined range of
Sewell rightly notes that for Giddens, “resources can be anything that
can serve as a source of power in social interactions” (1992: 9). In
an effort to extend and clarify this, Sewell offers that such sources of
power may be human or non-human. Non-human resources are indepen-
dently existing objects, for example factories owned by capitalists; human
resources are strength, knowledge, etc. Remarkably, he omits one™s place
in social relations, for Giddens surely the most critical resource of power
(Giddens, 1984: 25“26, 83“86, 89). But as above, a building is a factory
owned by capitalists only because of an array of institutional facts, only
because members have an array of concepts and beliefs in terms of which
they act. Indeed, it is just these which make a person who owns a fac-
tory a capitalist and which, accordingly, gives him power lacked by the
A more plausible version of this criticism has been made by some
interpreters of Bhaskar™s very similar “Transformational Model of Social
Activity” (TMSA). While Giddens has noted that his view assumes a real-
ist ontology, he has been less speci¬c than Bhaskar on what this entails
for social science.
In agreement with Giddens, Bhaskar holds that structures are contin-
ually reproduced (transformed) and “exist only in virtue of and exer-
cised only in human agency (in short they require active ˜functionaries™)”
(1979: 51). “Such a point, linking action to structure, must both endure
and be immediately occupied by individuals . . . [T]he mediating sys-
tem we need is that of positions (places, functions, rules, tasks, duties,
rights, etc.) occupied (¬lled, assumed, enacted) by individuals, and of the
practices (activities, etc.) in which, in virtue of their occupancy of these
positions (and vice versa), they engage” (Bhaskar, 1979: 51).
Plainly, Bhaskar™s emphasis is on social relations as sources of power,
but unlike Giddens, for him, what Giddens calls rules and resources
are collapsed “ perhaps wisely. These differences have been thought to
be inconsequential.28 But some of his interpreters, at least, have insisted
28 I am one of those who thought so. More lately, I have been convinced that there is an
unresolved tension in Bhaskar™s work on this issue. For an excellent account, see Varela
and Harr´ , 1996.
Explanation and understanding in the social sciences 71

that, unlike Giddens, Bhaskar is ¬rmly materialist and maintains a strong
objectivist view of social relations.
Thus, Porpora insists on a “difference between a materialist and ide-
alist approach to reality” turning on a difference between “a concept
of social structure as an objective reality and a concept of structure as
an intersubjective reality” (1989: 202). Thus, “what Giddens means by
structure are cultural rather than material conditions.” Porpora illus-
trates his point with the argument about poverty. One side attributes
poverty to “cultural factors, to the resocialization of each new generation
of poor people into rules and norms and ways of thinking that perpetu-
ate poverty” (Porpora, 1989: 202). This seems very much like Giddens.
On the other side are those who ¬nd the causes of poverty to be “objec-
tive circumstances of the social position the poor ¬nd themselves in,” a
feature of which is the absence of cultural capital, which, says Porpora,
Giddens can also acknowledge. But “another feature of the objective cir-
cumstances . . . relates to the distribution of jobs or social positions in
society . . . [W]hat we are talking about here are relational properties
of a social system . . . [and] the causal effects of those relationships on
the life chances of the poor.” He concludes: “Ultimately, we are talking
about those relationships as precisely the sort of external constraints on
action, the existence of which, as we have seen, Giddens wishes to deny”
(Porpora, 1989: 207).
This very typical response probably resonates with many structural-
ist writers, but it is quite plain that, typically, it bifurcates structure and
culture, subjective and objective. Institutional facts are as objective as
brute facts. The present level of unemployment is an objective fact even
while it depends upon features of the world that require special human
institutions for their reality, and even if, as noted, these depend upon col-
lective intentionality “ otherwise intersubjectivity.29 As regards external
constraints on action, what is intended is safe enough. There are no jobs,
so one cannot get one. But that fact is not external to the activities of
persons; it is not a brute fact but an institutional fact.
It was part of Giddens™s project to transcend the now familiar culture /
structure dichotomy. Rules are cultural, but as he insisted, they “cannot
be conceptualized apart from resources.” Resources generate capacities,
powers, including relational power, or power over others and these are
sustained by rules which provide both meaning and sanctions to activ-
ity. Plainly, the capitalist / wage labor relationship distributes resources
unequally, but that the capitalist owns the factory and that the worker

29 Contrary to Searle who speaks of “ontologically subjective features” of things it would
be better to speak of “ontologically intersubjective features” of things.
72 A Realist Philosophy of Social Science

works for wage, requires an extensive range of institutional facts which
members sustain in their activities.
Of course, unemployment puts persons lacking cultural capital at a
deep disadvantage. But the cause of employment is not social forces,
market forces or social structure. It is the unintended product of the
reproduction of capitalism by agents working with materials at hand “
capitalists, workers, consumers, government of¬cials and so on.
Of course, if social structure is rei¬ed “ made independently real “ it
could be causal, but not only is this ontologically dubious, it is not a move
available to the position defended by Porpora. A similar argument has
been made by Margaret Archer (1995) who argues, following Bhaskar,
that structure and culture are both emergent properties and as such are
“bearers of causal powers.” Still, quoting Bhaskar, Archer insists that
“the realist is committed to maintaining that ˜the causal power of social
forms is mediated through social agency™ ” (1995: 195). The question to
ask is this: why postulate the existence of structure or culture as causally
relevant if, to be causally effective, these must be mediated by social
This suggests another move. Paul Lewis (2000) has rightly noted that
the analysis of causality developed by Harr´ and Madden (1975) and
assumed in this volume is not an appropriate framework if one insists that
social structures have causal ef¬cacy. As Lewis says, “social structures are
not ef¬cient causes and hence are not powerful particulars (as the latter
are understood by Harr´ and Madden)” (2000: 257). He agrees, to be
sure, that persons as agents are “powerful particulars”: as agents, they
can make things happen. Accordingly, as he notes, to sustain a notion of
causality for social structure, one needs another framework. Following
suggestions by Bhaskar, Lewis offers that social structure can be thought
of as a material cause of social action. Thus, following Aristotle, the slab
of marble that is fashioned by the sculptor is a material cause of the
¬nished work. On the present account of the foregoing, agents do work
with materials at hand, but this is to be understood in terms of both
their human capital and the actually existing social situation in which
they ¬nd themselves. Both enable and constrain their actions. It is true
also that the marble enables and constrains the action of the sculptor:
were she working with some other materials “ for example, wood “ she
would be enabled and constrained differently. On the analysis offered in
the present account, these differences are the consequences of the causal
properties of marble and of wood. But as is plain enough, the slab of
marble does exist independently of persons and their actions and thus
can have causal powers. As Lewis would not deny, social structure is
Explanation and understanding in the social sciences 73

concept- and activity-dependent. But if so, the analogy fails. Indeed, it is
dif¬cult to see how a cause can be a material cause if it is not a powerful
The same problem would seem to arise with those recent arguments
that advocate realist views of mechanisms, but hold that social structure
is an emergent property with causal powers (Sawyer, 2003; Wight, 2003).
Bunge rightly insisted that “mechanisms are processes in concrete (mate-
rial) systems, whether physical, social, technical or some other kind . . . By
contrast, the conceptual and semiotic systems have compositions, envi-
ronments, and structures but no mechanisms” (2003: 191).
Of course, organizations have properties that cannot be ascribed to
their members. Bureaucracies, for example, are dif¬cult to dislodge, oper-
ate impersonally and are often painfully slow in getting to a conclusion.
But ¬rst, these properties are not causal and second, they are explained
by looking at how bureaucracies are organized and what enables and con-
strains individuals acting in them. Indeed, as we argue in the following
chapter, we understand bureaucracies by providing the social mechanism
that explains these properties.
On the other hand, one need not accept Harr´ ™s linguistic characteriza-
tion of social phenomena as “generated in and through conversation and
conversation-like activities” (Harr´ and van Langenhove, 1999: 10). Of
course, these are pertinent, but unless conversations and conversation-
like activities are but extended metaphors for a whole range of transac-
tions and interactions, of which some are not self-conscious, too much
is omitted (see May, 2002). Perhaps here Bunge™s admonition (above) is
relevant. Nor need we accept Harr´ ™s view that concepts like “class” can-
not refer to objective social structures and are but taxonomic categories
used to classify and label people and practices (Harr´ and Varela, 1996).
Like brute facts, institutional facts can be theorized. As noted, “class” is
a theoretical term which, by abstraction, might well explain a structured
pattern of practices. More generally, skeptical of a slip toward Durkheim,
Harr´ has been reluctant to extend his realism to institutional facts. His
account, accordingly, is vulnerable to a materialist critique of the sort
mounted by Porpora and Archer.
An account of institutional facts as the background of all human action,
along with a ¬rm acknowledgement of the material and historical con-
straints of virtually existing social structure, is quite suf¬cient to provide
all the bene¬ts of structuralist insights without any of the manifest dis-
advantages of attributing causality to a concept- and activity-dependent
social structure. Indeed, much would seem to be lost. In addition to the
potential incoherence of a theory of causality which allows that something
74 A Realist Philosophy of Social Science

that is not a powerful particular can have causal ef¬cacy, we risk the rei¬-
cation of social structure and the tempting slide into determinism and
the loss of agency.30
On the present view, agents, nature and natural happenings are causes “
with profound consequences on what happens in society. “Social forces”
is a misleading metaphor “ and a profoundly destructive one at that. The
idea that it intends to capture can easily be unpacked in terms of social
processes generated by the activities of persons working with materials at
hand. Indeed, it will be a critical part of our argument that the generative
mechanisms of macro-outcomes must be theorized in terms of the actions
and interactions of persons. The analogy of atoms to the properties of
molecules is quite exact, except that the atoms of social outcomes are
irreducibly social persons. This is the topic of the chapter that follows.

30 See Sewell™s treatment of this. He argues, plausibly, that “in spite of his devastating
attacks on Cartesian and L´ vi-Straussian ˜objectivism™ . . . Bourdieu™s own theory
has fallen victim to an impossibly objectivized and overtotalized conception of soci-
ety” (1992: 15). The problem, of course, is exactly analogous in Parsonian structuralist
4 Agents and generative social mechanisms

In those sciences we termed “abstract,” theory provides representations
of the generative mechanisms, including hypotheses regarding ontology,
for example, that there are atoms, and hypotheses regarding causal pro-
cesses, for example, that atoms form molecules in accordance with prin-
ciples of binding. We noted also that a regression to more fundamental
elements and processes also became possible. So quantum theory offers
generative mechanisms of processes in molecular chemistry. Typically,
for any process, there will be at least one mechanism operating, although
for such complex processes as organic growth there will be many mech-
anisms at work. Theories that represent generative mechanisms give us
understanding. We make exactly this move as regards understanding in
the social sciences, except that, of course, the mechanisms are social. As
with complex natural processes, typically, there will be many mechanisms
at work. As in the physical sciences, the theorizing of mechanisms in the
effort to understand is not the only task of social science. As we argued,
understanding presupposes good description, both quantitative and qual-
itative. Finally, we will need to consider the problem of explaining events
and episodes. This is developed in chapter 5.
The foregoing has also argued that persons are the dominant1 causal
agents in society “ even while, of course, they work with materials at
hand. It follows, accordingly, that in the social sciences, the generative
mechanisms of social outcomes are the actions of persons and no fur-
ther reduction is either plausible or demanded. That is, for purposes of
inquiry in the social sciences (excepting here experimental psychology),
the fundamental unit of analysis is the person (understood as above) “
the half-truth of methodological individualism.2

1 Dominant causal agents “ not exclusive causal agents, since there are also critical non-
social natural causes at work in society.
2 Experimental psychology is variously understood, but as conceived here its problem is to
identify the mechanisms which produce powers of “mind,” including cognition, memory,

76 A Realist Philosophy of Social Science

Social mechanisms
The idea that the explanatory goals of social science require theories of
social mechanisms is hardly new, even if the idea is very often an unar-
ticulated background assumption of studies. This is fairly evident among
writers we think of as doing agent-centered work, for example, symbolic
interactionists, Erving Goffman, rational choice theorists and the recent
work of James Coleman. But although less noticed, it is also at least in
the background of many more historically oriented writers, for example,
Barrington Moore, Marshall Sahlins, E. P. Thompson, Richard Sennett
and Jonathan Cobb, Charles Tilly, William Sewell, Arthur Stinchcombe,
Stephan Vlastos, Michael Burawoy, Mark Granoveter, Pierre Bourdieu,
Anthony Giddens, Raymond Boudon, John Elster and many others.3
More recently, it has played a central role in the provocative debate gen-
erated by Margaret Somers over the role of general theory in historical
sociology4 and the idea has recently been rearticulated by a number of
very recent writers.5 This development is most encouraging even though,
to be sure, there remains considerable disagreement even among advo-
cates of the idea of social mechanisms as to what exactly this involves. We
begin with an illustration by means of a concrete study, Willis™s Learning
to Labor (1981). Willis was not self-consciously employing Giddens™s
metatheory,6 nor did he offer a formal model of the key social mech-
anism at work in explaining why working-class kids get working-class
jobs. But his account is a superb illustration of how structuration theory
leads easily to the idea of explanatory social mechanisms.

learning, perception, emotion, etc. See Manicas and Secord, 1984. As noted, it is not
engaged in “explaining behavior” and is not, strictly speaking, a social science. Learning
is a psychological mechanism: what is learned is social.
3 Thus, Bourdieu writes that the task of sociology is to “uncover the most profoundly
buried structures of the various social worlds which constitute the social universe, as well
as the ˜mechanisms™ which tend to ensure their reproduction or their transformation”
(Bourdieu and Wacquant, 1992: 7).
4 See appendix C, “Rational Choice Theory and Historical Sociology.”
5 See especially, the collection of essays in Hedstrom and Swedberg (eds.), 1998; McAdam,
Tarrow and Tilly, 2001; more recently, the special issue of Philosophy of the Social Sci-
ences, 34, 2, 2004 and the panels of the American Political Science Association meeting
of August 2003 with papers by Bennett, Mahoney and Gerring. These are available
on the Internet: http://www.asu.edu/clas/polisci/cqrm/Bennett APSA 2003.pdf. Replace
“Bennett” with “Mahoney” and “Gerring” to access the other two papers.
6 It has been argued that Willis did not need structuration theory to produce his study, that
the in¬‚uence of Marx and Marxists is evident in his work. This hardly needs to be denied.
But structuration theory was not born ex nihilo. It was, it seems, exactly the re-thinking of
Marxism which led Willis, Giddens and Bhaskar to such similar conclusions. As Giddens
has said, his work is an extended gloss on the famous text in Marx™s 18th Brumaire, that
men make history but not with materials of their own choosing.
Agents and generative social mechanisms 77

Learning to labor: an example
Willis de¬nes his task as follows: “The dif¬cult thing to explain about
how middle class kids get middle class jobs is why others let them. The
dif¬cult thing to explain about how working class kids get working class
jobs is why they let themselves” (1981: 1). Indeed, if Willis is correct,
it is “too facile” to say that they have no choice, and misleading to say
that they are “socialized” for those jobs. On the contrary, those who end
up taking the worse jobs are active participants in constituting a culture
that effectively prepares them for those jobs. It is this mechanism that
interests Willis.
In Part I of Learning to Labor Willis offers a very rich ethnography
of a working-class school. The description focuses on “the lads” who
articulate a counter-school culture which in its most basic dimension is
“entrenched general and personalized opposition to ˜authority™ ” (1981:
11). The “conformists” or “ear™oles” have “a visibly different orientation.
It is not so much that they support teachers, rather they support the idea of
teachers” (1981: 13). The teachers, ¬nally, recognize that their authority
“must be won and maintained on moral not coercive grounds.”
The lads have ample resources with which to resist: “a continuous
scraping of chairs, a bad tempered ˜tut-tutting™ at the simplest request,”
“comics, newspapers and nudes under half-lifted desks melt into elusive
textbooks,” and more. To be sure, the lads know “the rules.” It is thus
that they can so successfully avoid outright confrontation and manipulate
them to serve their own purposes. But even more important, within the
“space won from the school and its rules” the lads have created a “multi-
faceted” implement of their culture. Called “having a laff,” it is used to
de¬ne the group, “to defeat boredom and fear, to overcome hardship and
problems “ as a way out of almost anything” (1981: 29). Striking in this
regard is their discovery of Gar¬nkeling: “Let™s laugh at everything he
says,” “Let™s pretend we can™t understand and say, ˜How do you mean?™
all the time.”
The lads also de¬ne themselves against girls and ethnic minorities.
Women, for the lads, are both “sexual objects and domestic comforters.”
Girlfriends are called “the missus.” But while “mum” is the model for the
girlfriend, she “is de¬nitely accorded an inferior role: ˜She™s a bit thick,
like, never knows what I™m on about™ ” (1981: 45).
Finally, since this is a working-class school, Willis provides an account
of the “shop¬‚oor culture” which is the domestic context for the students
and thus a potential resource for them. He ¬nds two critical features:
¬rst, “a massive attempt to gain informal control over the work pro-
cess” and second, a disdain for theory: “The shop¬‚oor abounds with
78 A Realist Philosophy of Social Science

apocryphal stories about the idiocy of purely theoretical knowledge. Prac-
tical ability always comes ¬rst and is a condition for other kinds of knowl-
edge” (1981: 56). It is plain that these work against the school and provide
a critical contrast to middle-class environments and schools. As Willis
writes: “When the middle class child is thrown back on to his indige-
nous culture, instead of ¬nding strengthening and con¬rming opposi-
tional themes there, he ¬nds the same ones” (1981: 76).
The critical difference, then, between “the lads” and the “ear™oles” is
that the latter have accepted the critical principle of the teacher / pupil
relation, the idea of teaching as a “fair exchange: knowledge for respect,
guidance for control.” Of course, this is supported and sanctioned in
many ways, beautifully developed by Willis.
In Part II, then, Willis attempts to explain the key aspects of what he has
described. Critical theoretical concepts are introduced. “Penetration” is
de¬ned as “impulses with a cultural form towards the penetration of the
conditions of existence of its members and their position with the social
whole but in a way which is not centred, essentialist or individualist”
(Willis, 1981: 119). This ¬rmly realist formulation demands a great deal.
In the ¬rst place, and importantly, members are not trying to achieve
penetration, however much “practical consciousness” may reveal that
they have some understanding of the conditions of existence. Second “
and this is a problem shared by social scientists who speci¬cally seek such
understanding “ penetration is, at best, partial, since there are always
limitations: “blocks, diversions and ideological effects which confuse and
impede the full development and expression of these impulses” (1981:
Indeed, the fact that members may lack any sort of adequate discur-
sive knowledge of what is going on and may still have practical knowledge
which, if properly understood, shows that they have achieved at least par-
tial penetration is the basis for the failure of survey research. As Willis
(1981: 122) writes, “direct and explicit consciousness” “may well re¬‚ect
only the ¬nal stages of cultural processes and the mysti¬ed and con-
tradictory forms which basic insights take as they are lived out.” It is
only by fully immersing oneself in extended interaction that one may dis-
cover what is really known by members. Moreover, methods which rely
on verbal or written responses cannot distinguish “attempts to please
the other, super¬cial mimicry, earnest attempts to follow abstract norms
of, say, politeness, sophistication or what is taken as intelligence” from
comments and responses offered in ongoing activity which “have a true
cultural resonance” (1981: 122).
Willis then seeks to explain and assess the beliefs and actions of
the agents in his study. On his view the lads™ rejection of school and
Agents and generative social mechanisms 79

opposition to teachers is a consequence of their penetration of the “teach-
ing paradigm.” They know better than “the new vocational guidance what
is the real state of the job market.” Thus, they have a “deep seated scep-
ticism about the value of the quali¬cations in relation to what might be
sacri¬ced to get them” (1981: 126). More, the lads make “a real penetra-
tion of what might be called the difference between individual and group
logics and the nature of their ideological confusion in modern education”
(1981: 128). In the school and in the culture, “it is never admitted that
not all can succeed.” Finally, and even more profoundly, since the grasp
of this reality leads them to assume that they will be doing the least skilled
forms of labor, they make the further penetration into the fundamental
features of capitalist production, that “the measure of abstract labour
is . . . time” (1981: 135).

The lads™ indifference to the particular form of work they enter, their assumption
of the meaninglessness of work not what kind of “right attitude” they take to it,
and their general sense of the similarity of all work as it faces them, is the form
of a cultural penetration of their real conditions of existence as members of class.
(1981: 136)

These are, however, but partial penetrations and by no means are
they suf¬cient to make these youths into politically active radicals. Willis
argues that these penetrations may be seen as a rejection of convention-
ally constituted individualism. But individualism is not defeated in itself,
but “for its part in the school masque where mental work is associated
with unjusti¬ed authority, with quali¬cations whose promise is illusory”
(1981: 146). The upshot is the reverse polarization of the manual / men-
tal labor distinction and the consequent rejection of all that school might
offer. But, argues Willis, this re-evaluation of manual labor depends upon
sexism: “Manual labor is associated with the social superiority of mas-
culinity, and mental labor with the social inferiority of femininity” (1981:
148). Indeed, “we may say that where the principle of general abstract
labor has emptied work of signi¬cance from the inside, a transformed
patriarchy has ¬lled it with signi¬cance from the outside . . . The brutality
of the working situation is partially re-interpreted into a heroic exercise of
manly confrontation with the task” (1981: 150). To be sure, this youthful
re-evaluation need not be permanent. It suf¬ces that it lasts long enough
to effectively trap them forever.
Willis notes that while this goes some way toward explaining why
all do not aspire to the “rewards and satisfactions of mental labour,”
it is easy enough to see how insights into their future in the world of
work would lead people to refuse to work at all. Willis had in mind
here West Indians who have inherited a culture of wagelessness and
80 A Realist Philosophy of Social Science

poverty, but black Americans with similar legacies might also seize this
Before concluding this section, several further features may be noted.
Willis™s school was working class in a fairly straightforward sense: the
students were all from families of manual laborers. This fundamental
structuring mechanism is critical, but largely in the background of his
account. Similarly, Willis makes no effort to explain how it is that some
of the youths in his study become lads and some become conformists. Dif-
ferentiation, the process by which this occurs, is left largely unanalyzed.
Since individuals are biographically unique and many contingencies may
well enter into an individual™s decision to conform or not, this is perhaps
as it must be. As in any science, there are limits to explanation. Willis is
clear that the answer is not in parenting: “Parents have their own complex
and creative relations to class themes and in no sense press their children
into a simple standard working class mold” (1981: 73). Rather, working-
class values and problems “ including the need for cash “ are materials
for differentiation. Thus, the idea of socialization as that is used in main-
stream sociology played no role in Willis™s account. As he was at pains
to show, “social agents are not passive bearers of ideology, but active
appropriators who reproduce existing structures only through struggle,
contestation and partial penetration of those structures” (1981: 175).
Similarly, dominant ideology theses misidentify what is at issue. Not
only is the functional neatness of such theory totally rejected, but much
of the critical cultural material is not mediated downward from dominant
groups, Rather, it comes from “internal cultural relationships,” for exam-
ple, the working-class af¬rmation of manual labour. On Willis™s view, ide-
ology does “naturalize” what is conventional and potentially fragile, and,
crucially, it does “dislocate.” Thus, in liberal culture, there is a pervasive
emphasis on the differentiation of occupational possibilities for youth and
on the range of opportunities these provide for individual satisfactions.
But since these ideas do not convince those who doom themselves for
manual labor, “the effect of its thrust is reversed and acts centripetally,
not to make jobs various, but to decentre the cause of their sameness”
(1981: 163).

It is . . . no one™s fault that work is boring and tiring and mostly meaning-
less . . . Instead of a centred world of oppression from a speci¬c and determinate
social organization of thought, production and interests[,] we have the naturalistic
world of a thousand timeless causes. (1981: 163)

The political consequences of this are obvious. But we should empha-
size also that both the lads and the conformists tended to end up doing
similar work, even if the conformists can believe that since they are
Agents and generative social mechanisms 81

especially equipped with quali¬cations, they should be in “better” jobs
than, and to be a “different kind of person from, ˜the lads™ ” (1981: 152).
Indeed, as Willis says, “once such a division is founded in the working
class . . . it massively legitimates the position of the middle class: not
capitalism but their own mental capacities keep them where they are”
(1981: 152). If you™re so smart, why ain™t you rich?
At a very high level of abstraction and simpli¬cation, the social mecha-
nism which explains why working-class kids get working-class jobs looks
something like this:
1. By virtue of their location in prevailing social relations, working-class
kids believe that, despite what they are told, there is little real oppor-
tunity for them to get out of the working class.
2. Accordingly, drawing on materials available to them, they resist.
3. Accordingly, they do not succeed in school, often dropping out.
4. Accordingly, they are unquali¬ed for anything but working-class jobs.
5. A ¬nal unintended consequence also follows: criticism of outcomes is
diffused and the existing social distribution of jobs is legitimated.

The structure of social mechanisms
We can use this example to ¬‚esh out more formally the structure of
social mechanisms. I here follow a groundbreaking essay by Gudmund
Hernes (1998). Hernes identi¬es two sets of abstract elements in a social
mechanism: a set of assumptions regarding the speci¬cation of the actors
and a set of assumptions regarding “structure.” The ¬rst set is generated
by providing answers to the following questions: (a) What do they want?
(b) What do they know? (c) What can they do? and (d) What are their
In our example, working-class kids and the two subgroups, the lads and
the ear™oles are the key actors (but of course, there are many others, the
teachers and parents among them). The ¬rst question, (a), is answered
by identifying the preferences, purposes and goals of the actors. Answers
to the second question, (b) What do they know? was, in the Willis study,
the most dif¬cult to answer, but this probably is typical. As we have
argued, while actors have practical knowledge, not only are the condi-
tions and consequences of action not generally available to them, but
there are problems even regarding their preferences and goals. Worse,
beliefs may be unacknowledged, unstable and context-bound. (We com-
mented on this as regards the dubious bene¬ts of conventional survey
research.) The answer to question (c), What can they do?, is a question
of what powers they have. In our formulation (which is hardly sacrosanct),
it will be answered in terms of rules and resources, the key elements of
82 A Realist Philosophy of Social Science

social structure as conceived here. Again, we need to keep in mind that
resources involve rules and thus rules and resources are not concretely
disconnected. Resources can range, of course, from the personal capac-
ities of the actors to the capacities they have by virtue of their positions
in social relations. Identifying the actors as working-class kids is full of
promissory notes regarding the latter. Finally, as regards (d), What are
their attributes?, Hernes has in mind their sex or race, and in some con-
texts, their health, or other special attributes pertinent to the mechanism
being theorized.
The second set of assumptions he terms “structure assumptions.” But
not surprisingly, given the idea of the duality of structure, we can expect
considerable overlap between these assumptions and the assumptions
attributed to actors. Indeed, it is not entirely clear that even for analyt-
ical reasons, these are usefully distinguished. Thus, Hernes asks: What
are the states actors can be in? He notes that this includes “positions
or roles taken.” These were, of course, critical to the understanding of
“what can they do” (above). Structure assumptions include “the number
of other actors, the number of relations they can enter . . . the alternatives
they confront, options they face, or constraints they encounter” (Hernes,
1998: 94). Some of these, he continues, as just noted, include “norms,
rules and laws.” He adds, importantly, that a key question “will usually
be whether such states remain constant or whether they are subject to
change.” These are all important speci¬c considerations for understand-
ing but it may well be that the distinction is misleading. There are two
sorts of problems here.
First, Hernes remarks that the ¬rst set of assumptions regarding actors
corresponds to “methodological individualism,” but plainly, his individ-
uals are persons and are robustly social beings. There is no attempt
to reduce social predicates to predicates of individual psychology. As
Bhaskar has rightly said, “the real problem appears to be not so much how
one could give an individualist explanation of behavior, but that of how
one could even give a non-social (that is, strictly individualist) explanation
of individual, at least characteristically human behavior . . . A tribesman
implies a tribe, the cashing of a cheque, a banking system” (1979: 35).
Moreover, while actors are assumed to have purposes and to be rational,
there is no commitment here to some version of rational choice theory. As
Boudon argues (in an essay in the Hedstrom and Swedberg volume), since
it cannot accommodate a host of beliefs “which are a normal and essential
ingredient of many social actions” (1998a: 183), rational choice theory is
fatally ¬‚awed. Keeping in mind considerations already adduced (above),
his alternative “cognitivist model” “ perhaps misleadingly named “
is perhaps suf¬cient. It “supposes that actions, decisions and beliefs are
Agents and generative social mechanisms 83

meaningful to the actor in the sense that they are perceived by him as
grounded on means” (1998a: 191). This is, of course, the key insight of
the work of Weber, extended by Schutz, properly understood.
There is, accordingly, nothing amiss in focusing on agents and their
capacities, provided we are talking about situated social beings. But there
is the danger of making social structure causal. Hernes argues that some
mechanisms “are based on what could be dubbed ˜collapsed actors™ ”
(1998: 94). For this model, “no speci¬c assumptions are made about
what the actors want, know or have.” But it is highly doubtful that there
are any outcomes in which this is possible, even if we include non-social
causes as structures, for example, contagious diseases. As Hernes says,
disease is a critical causal component of what happens to people. But even
so, any sort of explanation will require a mechanism which includes not
merely the biological causes, but the condition and actions of people who
are responding to the threat of disease. That is, there are no outcomes
where, as Hernes puts it, “the actors are just objects.” Thus, one can apply
“a standard diffusion model” in which “the infection rate is proportional
to the number of haves and have-nots,” but even on this model, the actors
are not fully collapsed since explaining outcomes requires assumptions
about the actions of persons, that at the very minimum, they are not
acting to be infection-free. In arguing that we can have a mechanism in
which “the structure will overwhelm the actors whatever assumptions
are made about them” (1998: 94), Hernes effectively adopts a structural
determinism in which agency disappears altogether. But this is never the
case. For agents, there are always choices “ however restricted they may
be. As regards the infection rate, for example, decisions by agents will
play a critical causal role in the outcomes. Thus, will they be fastidious
regarding cleanliness, avoid congested areas, and so on.
It is worth pausing here to emphasize that structuralist explanations
tend to be satisfying exactly because they acknowledge that the situation
of actors is critical to understanding what they do. But the greatest advan-
tage of thinking in terms of social mechanisms, as developed here, is to
acknowledge that agents, nevertheless, remain the key players. Not only
does this reinforce the idea that agents sustain, reproduce and transform
structure, but as well, it calls attention to them in the effort to explain
whether and how “structure” is changing or not changing. As will be
argued in the next chapter, we need Louis XVI to explain the French
Revolution, but we need also to understand the mechanisms involving
the actions of peasants and nobles, of sans culottes and bourgeois.
Perhaps a more interesting case is an effort to explain increased divorces
and extramarital affairs. We can theorize a mechanism in such a way that
opportunities for extramarital affairs vary with labor force participation.
84 A Realist Philosophy of Social Science

Thus, as the proportion of sexes becomes more equal, the number of
potential couples increases. Here two kinds of actor are assumed. As
above, we need to make some strong assumptions about their goals and
beliefs if we are to reach any outcome. We need to assume that the per-
sons are sexually motivated, that they have beliefs about how desires can
be satis¬ed, and so on. That these may be taken for granted does not alter
the logic of the mechanism. Typically, accounts omit much that is essen-
tial to explanation, but reasonably taken for granted. As Hernes rightly
notes, this mechanism (like the former one) can be elaborated to make
it more realistic. Thus, one might distinguish different subgroups with
different beliefs and purposes, between, as he says, the Philanderers and
the Purehearted.

Abstraction, representation and realism
This suggests a deeper issue that needs clari¬cation. Consider Hernes™s
de¬nition of a social mechanism:

A mechanism is an intellectual construct that is part of a phantom world which
may mimic real life with abstract actors that impersonate humans and cast them
in conceptual conditions that emulate actual circumstances. A mechanism like
a model is a stripped-down picture of reality; it is an abstract representation that
gives us the logic of the process that could have produced the initial observa-
tion . . . Mechanisms are the virtual reality of social scientists. But it is the stuff
of which the world of the social scientist is made: This arti¬cial, manmade world
of mechanisms is real“real virtuality. (1998: 78)

This formulation needs careful gloss. Epistemologically speaking, all
theory is a representation of a reality, an intellectual construct, and it
is always abstract: it can never catch the full-bodied reality. Indeed, we
would not want it to. We strip down reality to get at the bare bones exactly
because faced with the complexity of concrete reality, understanding
requires that we identify abstractly the pertinent causal mechanisms. The
chemist is interested in concrete salt as NaCl, and the mechanism regards
the movements of electrons of the theoretical entity NaCl. Assuming then
that ordinary salt is mainly NaCl we can explain its dissolving in water.
But if we accept the theory, we accept that the generative mechanism
is real. That is, not only could it have produced the outcome, but hav-
ing ruled out alternative explanations, we believe that it did produce the
Social mechanisms, like the social structures which are the product
and medium of action, are real, but they are not independently real;
their existence is dependent on the beliefs and actions of persons; hence
Agents and generative social mechanisms 85

they are but “virtually real.” (Hernes is not interested, we may judge, in
relating his account to Giddens™s. Nonetheless, it is plain that it coheres
neatly “ as I think it must.)
Moreover, the construction of “abstract actors” “ in Alfred Schutz™s ¨
formulation, “homonculi” or “typical actors” “ are abstractions from real
persons, representations of them qua some signi¬cant attributes which
they have. In our example, Willis constructed theoretical actors based on
his ethnographic materials. The lads are all different in all sorts of ways,
but as regards the mechanism being theorized, they share in a set of
typical attributes.7 Thus, given an abstracted representation of their con-
dition, the concrete behavior of real ¬‚esh and blood individuals becomes
intelligible. As with causal mechanisms in nature, we get understand-
ing not predictive ability. That is, we understand the process by which
working-class kids tend to get working-class jobs, but we cannot pre-
dict that Sam, for example, will identify with the lads or that he will
end up in a working-class job. For reasons particular to Sam™s biogra-
phy, he might be a great success in business. In other words, the mech-
anism explains the generalization: working-class kids get working-class
But we must resist an instrumentalist interpretation of social mecha-
nisms, typical of mainstream economics. As noted in chapter 2, on this
view, the assumptions of the mechanism need not be realistic at all. That
is, not only need there be no real persons with all the attributes of the
construction, but the assumptions can be contrary to facts known about
them. Thus, neo-classical price theory assumes that ¬rms and consumers
have complete knowledge, are consistent maximizers and so on. On this
view, since prediction and explanation are thought to be symmetrical,
good predictions are thought to be good explanations, it is hard to see
how manifestly false assumptions about persons and their conditions can
yield explanations of real concrete outcomes.8 Of course, it may not be
easy to know if the assumptions of the mechanism are true. Willis™s study
is convincing because he gives us good reason to believe that the attributes
of the actors identi¬ed by him are true of the lads.
Using the Willis example, there are a number of other important
observations which we can make regarding social mechanisms: as in the

7 These are not, accordingly, ideal-types as these are usually understood. The attributes in
the construction are true of “the lads,” etc.
8 Lawson (1997) rightly insists that conditions for the so-called “method of successive
approximation” cannot be satis¬ed in the case of neo-classical theory. Also see chapter
6, below. Similarly, Weber seems to have assumed that neo-classical theory provides an
ideal-type, but the same problem arises: it is not just that there are no markets satisfying
these conditions, but that there never could be.
86 A Realist Philosophy of Social Science

natural world, in the social world, mechanisms seldom, if ever, operate in
isolation.9 There are a host of other mechanisms at work which are either
connected or partially constitutive of those which are the focus of Willis™s
account. This is, to be sure, a problem for theory. In the foregoing exam-
ple, we can identify at least the following as plausible candidates:
(a) Mechanisms which give us an understanding of the working of cap-
italism and thus explain why it is quite impossible that everyone
succeed. These mechanisms will be highly abstract. Willis simply
assumes that not everyone can succeed, but Willis employs Marx™s
analysis of abstract labor in his account of the behavior of the lads.
(b) Mechanisms which (as part of the foregoing) provide an understand-
ing of the working of labor markets, for example, how networks func-
tion, credential barriers, the role of the reserve army and so on. (Com-
pare here, of course, important work by Collins (1979), Tilly and
Tilly (1997), Granoveter and Tilly (1988) and many others.)
(c) Mechanisms which give us an understanding of schools, including
mechanisms which generate neighborhoods, and accordingly, mech-
anisms which explain the class distribution in the school and which,
in turn, explain the consequent peer structure, the attitudes of teach-
ers in schools with predominantly working-class students, and the
objective outcomes regarding levels of additional education, and sub-
sequent job distributions.10
(d) Mechanisms which give us an understanding of materials available to
students in the construction of their beliefs, including mechanisms
of identity formation which include, in turn, mechanisms of peer
formation and mechanisms at work in households, especially, in this
instance, how gender attitudes and attitudes of parents toward white-
collar work produce belief.
Willis offers powerful hints about many of these in the course of his
narrative, including the mechanisms of identity formation, but some are
simply taken for granted, such as the generative mechanisms of capital-
ism and the mechanisms which explain the existence and conditions of
working-class schools. Plainly, this is to be expected.
But it is easy to see also that the mechanism which is the focus of Willis™s
study can be applied widely, albeit with differences in the speci¬cs. A
similar mechanism is at work in several studies of drug use among ghetto

9 See McAdam et al., 2001: 27, citing Gambetta, 1998. McAdam and Gambetta offer
accounts of mechanisms which differ from mine.
10 There is, of course, a host of good work on this even while there is a strong tendency
to search for single causes or to suppose that quantitative methods can provide relative
importance of “factors.” For a recent excellent review, see Rothstein, 2004.
Agents and generative social mechanisms 87

youth. Following earlier work by Terry Williams (1989), Phillipe Bourgois
(1997), having immersed himself in an inner-city neighborhood, dis-
cerned a mechanism at work in which the parties were “frantically pur-
suing the American Dream.” As in the Willis study, the local dynamic is
structured by the dynamics of the international political economy, taken
for granted by Bourgois. It provides the objective conditions which are the
starting point for the development of the model. The critical point is to
understand how inner-city youth understand the situation they ¬nd them-
selves in, and why they do as they do. This requires an ethnography from
which we can construct the typical actors and their characteristic goals
and capacities. Typically, “the underground economy and the culture of
terror are seen as the most realistic routes to upward mobility” (1997: 70).
But while the abstract model is not ¬lled in in speci¬c detail, and remains
incomplete, it does give us considerable understanding. In particular, it
does not “account for the explosive appeal of a drug like crack . . . This
involves the con¬‚ation of ethnic discrimination with a rigidly segmented
labor market, and all the hidden injuries to human dignity that this
entails . . . It involves in other words, the experience of many forms of
oppression at once, or what I call ˜conjugated oppression™ ” (1997: 72).
That is, in terms of the foregoing analysis, several mechanisms “
including pharmacological “ are at work.
A similar approach offers understanding of the often claimed link
between drug abuse and violent crime. Extensive ¬eldwork by Goldstein,
et al. (1997) led them to develop three different explanatory models. The
psychopharmacological model is the most straightforward: it offers that
drug use causes temperamental changes in individuals which lead to vio-
lence. The economic compulsion model offers that craving drugs, persons
feel compelled to engage in economic crimes to ¬nance their drug use.
Here there is a clear goal by users along with a judgment on effective
available means. The systemic model “suggests that violence stems from
the exigencies of working or doing business in an illicit market “ a context
in which the monetary stakes can be enormous but where the economic
actors have no recourse to the legal system to resolve disputes” (1997:
116). This model joins neatly with the Bourgois model. The authors then
offer some statistical data to test the models. Examining a sample of 414
homicides, they show that only 7.5 percent were caused by the effects of
drugs, 2 percent were motivated by economic gain and 39.1 percent were
clearly the outcome of the systemic factors, violence between dealers or
dealers and users. One might notice here that 47.5 percent of the sam-
ple (which reports only homicides) were not drug-related. There are, of
course, also the mechanisms which produce and reproduce an ideology
88 A Realist Philosophy of Social Science

regarding drug use, an ideology which mysti¬es reality and which is pro-
moted by perhaps even well-intentioned media.11

Promissory notes
Ogbu™s now classic analysis of “caste minorities” (1978) offers another
useful example which need not be developed here. On this model racism
¬gures hugely. Of course, to speak of racism is to offer but a promissory
note, to be ¬lled in with an account of the social mechanisms which pro-
duce and reproduce racist outcomes. Promissory notes very often serve
as quasi or suggestive explanations in the social sciences. Sometimes,
the promissory note is left entirely empty and no mechanism is iden-
ti¬ed. The reader is left to imagine one. But as regards understanding
the processes of nature, in social science, promissory notes have some
explanatory value.
More confusing is their appearance in contexts where their explana-
tory value is inconsistent with the explicit explanatory effort of the writer.
Thus, in his much discussed book, Bowling Alone (2000), Robert D. Put-
nam seeks to explain “the collapse and revival of American community,”
the subtitle of his book. He offers that in each of several domains, “we shall
encounter currents and crosscurrents and eddies, but in each we shall
also discover common, powerful tidal movements that have swept across
American society in the twentieth century” (2000: 27). This is surely
causal language and we might easily suppose that there is some pervasive
mechanism at work here, e.g. commodi¬cation, globalization, urbaniza-
tion. Presumably these explain Americans™ “engagement in their life of
their communities,” a process that was reversed “a few decades ago “
silently, without warning”. Thus, one might argue that as large corpora-
tions and well-funded interest groups came to dominate civic life, indi-
viduals were simply not permitted access to participatory institutions.
Indeed, an account of such mechanisms is not found anywhere in the
book. Rather, for Putnam, social capital becomes a sort of intervening
variable: when it is strong so too is community, and conversely. He offers
a wealth of generalizations which test degrees of social capital. These,
he notes, rest “on more than one body of evidence” (2000: 26). Chap-
ter 3 then offers “a wide range of possible explanations” for changes in
social capital “ “from overwork to suburban sprawl, from the welfare state
to the women™s revolution, from the growth of mobility to the growth of
divorce” (2000: 27). Again, one might say that these are promissory notes,
yet to be ¬lled in. But again, we are disappointed. For him these factors

11 For some discussion of this, see Morgan and Zimmer, 1997 and Glassner, 2000.
Agents and generative social mechanisms 89

represent correlations not mechanisms. And the assumption is that where
there is a strong correlation, we have an explanation; where it is weak, the
factor is not important in the explanation. He concludes: “some of these
factors turn out to have played no signi¬cant role at all in the erosion of
social capital, but we shall be able to identify three or four critical sources
of our problem” (2000: 27). For example, “pressures of time and money”
are not signi¬cant causes since his quantitative analysis shows that cor-
relations of these variables are not signi¬cant. His “best guess” is that
“no more than 10 percent of the total decline is attributable to that set
of factors” (2000: 283). Rather, “generational change” is “the powerful
factor,” “accounting for perhaps half of the overall decline” (2000: 283).
But the fact that generational change is correlated with measures of social
capital tells us nothing about causality. Thus, what are the changes in the
beliefs and conditions of persons of different generations, why did they
change, and how do these result in changes in “social capital ”? No effort
is made in this direction.
The point here is not to raise questions about the generalizations or the
evidence cited for them, but rather to illustrate a critical point of differ-
ence in strategies of explanation. Generalizations, including signi¬cant
correlations, provide neither explanation nor understanding; they need
explaining. Oddly, there are places in the book where, in passing, Putnam
not only uses the idea of a social mechanism but, inconsistent with his
explicit methodology, he provides a sketch of one. Thus, he argues that
social capital may be better than medication in ¬ghting illness and trau-
mas (2000: 289). He writes: “to clarify how these mechanisms operate
in practice, consider the following stylized example, which while tech-
nically fabricated, depicts reality for many parents. Bob and Rosemary
Smith, parents of six-year-old Jonathan live in an urban community . . . ”
(2000: 289). He provides a sketch of their beliefs “ for example, they
support public education and like the diversity of the public school “
and of the existing conditions, for example, the school is a shambles. He
then pursues the logic of the mechanism by which social capital is the
critical factor in whether they will succeed in founding a PTA, and how,
if they succeed, this will give new resources, and so forth.12
Sometimes the promissory note has more promise since some of the
mechanism is suggested. Examples are very easy to ¬nd, often in con-
texts which make reference to processes, or dynamics, or sometimes
to the logic of a process or system. Thus, writers speak of “urbaniza-
tion,” “centralization,” “mobilization,” “state building,” “monopolistic

12 For a series of essays which do make the effort to give us this understanding, see McLean
et al. (eds.), 2002.
90 A Realist Philosophy of Social Science

competition,” and so on. Typically, these will involve a complex of mech-
anisms. Thus in glossing an argument from Tocqueville, Boudon notes
that “the macroscopic statement, ˜centralization is a cause of agricultural
development™ appears as entirely acceptable, because it is supported by
[an] individualist analysis” “ a social mechanism. As he says, “though
centralization is a complex process, it is identi¬ed with precise ˜param-
eters™ that affect the situation of decision making of the actors, here the
landlords” (1998b: 823). Boudon may be generous in holding that the
“parameters” are “precise,” but perhaps they are as precise as they need
to be.
Sometimes, promissory notes are offered as explanations with negligi-
ble effort to ¬ll in the details, to identify the typical actors, their motives,
their resources and their relations. Thus, Tilly (1992) offers that “the
processes that accumulate and concentrate capital also produce cities.”
He writes:

To the extent that the survival of households depends on the presence of capital
through employment, investment, redistribution or other strong link, the distri-
bution of population follows that of capital . . . Trade warehousing, banking,
and production that depends closely on any of them all bene¬t from proximity
to each other. Within limits set by the productivity of agriculture, that proxim-
ity promotes the formation of dense, differentiated populations having extensive
outside connections “ cities. (1992: 153)

The logic here is fairly straightforward. Persons need to be employed to
maintain life. Businessmen seeking to minimize costs will seek environ-
ments that promote this. Persons seeking employment, accordingly, will
gravitate toward those environments. Filling in even these rough pieces
of the story was perhaps needless “ depending upon our interests. It is
important to emphasize this. Understanding and explanation are prag-
matic notions so that what is demanded will be a function of what is
needed and wanted. It may well often be the case that a promissory note
will be suf¬cient.
Still, there are real dangers in failing to ¬ll in the promissory notes,
but especially, the danger of suggesting a spurious explanation. Thus,
racism includes a variety of social mechanisms. The most straightforward
mechanism is simply to show that those in a position to exclude have
decided to exclude persons on the basis of race and that nothing prevents
them from doing so. Likely involved in this mechanism is the belief on the
part of the excluders that those to be excluded are inferior in this way or
that. But evidence may show that the actors do not hold such beliefs and
that no decision has been made to exclude. It would be easy to conclude,
accordingly, that racism is not involved in the outcome. But there are
Agents and generative social mechanisms 91

more subtle forms of racism often lumped together under the heading of
“institutional racism.”
Unfortunately, “institutional racism” is a misleading term for the phe-
nomenon since it suggests that the outcome can be understood by exam-
ining the social mechanisms at work in some institution, for example, a
corporation or university. But this inappropriately restricts the problem.
It is not hard to see that, as regards many outcomes, the problem regards
the causal relations of several institutions. Perhaps a better term would be
“systematic” or “historical racism.” Thus, nobody needs to decide that a
neighborhood school will be disproportionately African-American. The
mechanisms here are interrelated but familiar enough: disproportion-
ately poor African-Americans seek low-income housing which exists in
neighborhoods which are already disproportionately low-income African-
American. Of course, unexplained in this sketch is the fact that African-
Americans are disproportionately poor. And again, explaining this will
require some other interconnected mechanisms, including the Ogbu
hypothesis, the work of Claude Steele and many others.
Thus, what Rosabeth Moss Kanter (1977) (following Wilbert Moore)
calls “homosexual reproduction” is a mechanism relevant to explain-
ing disproportionate numbers of white males in upper management
positions. Assume ¬rst (and safely) that there is a mechanism which
explains why managers earn more than janitors. (There are, it is impor-
tant to notice, alternative mechanisms which can account for this. One, of
course, is the familiar neo-classical model.) Then, in the corporate struc-
tures in which managers function, because of irresolvable uncertainties
affecting their positions and roles, social similarity becomes important.
Looking for the “right sort of person,” a white, male, Ivy league-educated
manager, for example, will prefer a white, male, Ivy leaguer to work for
and with him. As Mills had earlier observed: “To be compatible with
the top men is to act like them, to look like them, to think like them;
to be of and for them” (1959: 141). This effort to minimize uncertainty
by reproducing oneself in the workplace need not be self-conscious. The
actors need not be self-conscious racists or sexists, even if behavior leads
to racist and sexist outcomes.
Similarly, there are mechanisms which explain the higher incomes
of college graduates in comparison to those lacking college education.
(Again, there are the mechanisms detailed by neo-classical economics
and there are competing accounts “ to be considered in chapter 6.) For
present purposes, we need only note the fact that persons with bache-
lor™s degrees earn more, on average, than those who lack such degrees.
Similarly, there are complicated mechanisms which explain who gets
degrees, but we can assume that ability to perform on standardized tests is
92 A Realist Philosophy of Social Science

relevant to admission to institutions of higher education. Claude Steele
(2004) has suggested a social / psychological mechanism (related to the
array of self-con¬rming mechanisms) which helps to explain poor per-
formance in standardized tests by African-Americans. Because of widely
prevalent attitudes that they are cognitively inferior, these groups learn
to lack self-con¬dence in their preparation and capacities and this causes
them to perform poorly. Steele tested the mechanism through a series
of experiments in which randomly selected African-American students
from Stanford were told that the test measured personal attributes.
They were reminded of their race by asking them to check this off
on the questionnaire. Another group of black students, also randomly
selected, were told that the test was simply psychological research and
no mention was made of race. The ¬rst group did appreciably poorer
than the second group. Other studies of women and Asians have con-
¬rmed Steele™s theory of “stereotype vulnerability.” The mechanism
does not, however, explain why loss of con¬dence has the effects that it
evidently has.

Mechanisms as providing the micro-foundations
of the macro
Putting aside psycho-sociological mechanisms of the sort just noticed, it
is important to notice that as analyzed here, social mechanisms do not
divide into the macro and the micro. As analyzed here, they link the micro
and the macro “ or if you will, they provide the micro-foundations for
the macro. Thus, aggregated capitalist unemployment is understood in
terms of the decisions of agents in corporations, ¬rms and labor markets.
That is, on the present view, the macro / micro view is untenable and
all mechanisms assume typical agents engaged interactively in producing
outcomes. Social mechanisms can, however, be theorized as applying
locally or globally and thus in terms of varying degrees of abstraction.
For example, the mechanisms of capitalism are highly global and thus
highly abstract. Marx generated his model by abstraction from mid-
nineteenth-century British capitalism, but he is clear in seeing that soci-
eties with different histories could be capitalist, or, following Adam Smith
(and well before Wallerstein), one could also apply the model to a global
capitalist system. In Marx™s model, there are only two sets of actors, cap-
italists and wage workers, de¬ned relationally. There are no families, no
schools, no banks, no gender or racial differences; while everyone is a
potential consumer, there is no one engaged in marketing or advertising;
¬nally, government functions only to establish the legal and infrastruc-
tural conditions of a monetary economy. Flesh-and-blood agents are
Agents and generative social mechanisms 93

entirely absent, but certain beliefs and motivations, derived by abstrac-
tion, are attributed to them as typical actors. For example, wage workers,
structurally compelled to sell their labor power, know that they will not eat
unless they can secure employment. Capitalists know that in production,
surplus is derived from value-adding labor power. But for Marx, neither
group fully penetrates the conditions of capitalist reproduction, in par-
ticular that commodities are fetishized “ that a relation between persons
manifests itself as a relation between things. Marx offers also a sketch of
theory of how this occurs. Although these mechanisms (and others built
on them) apply in any capitalist society, there will be huge differences in
actual capitalist societies, precisely because as actors are more concretely
theorized, other mechanisms and other beliefs and motivations will be
compounded in generating outcomes. This is tantamount to applying
the model more locally, from, say, a region, to a nation, to a production
facility in Dearborn, Michigan.
An excellent example is the work of Dipesh Chakrabarty (1989). Fol-
lowing Marx, for Chakrabarty, there is a capitalist social relation only
if the labor power of workers is experienced as a commodity in Marx™s
carefully discriminated sense. This is not, to be sure, a subjective phe-
nomenon. It is as objective and real as anything can be. Indeed, given
the conditions outlined in Marx™s model, it is quite inevitable. That is,
the entire analysis of fetishism is meant to show how it is possible “ and
necessary “ that workers actively and uncoercively reproduce a system in
which they are exploited. The conditions for this are not merely juridical,
the wage-form, but, as Chakrabarty says, laborers must live in accor-
dance with norms de¬ning “formal freedom,” and “equality before the
law,” rights, as E. P. Thompson had argued, which were the rights of “a
free-born Englishman” “ “as Paine had left him or as the Methodists had
moulded him” (Thompson, 1978: 221). It was thus that Marx believed
that England, a society “where the notion of human equality has already
acquired the ¬xity of popular prejudice,” was the best place to decipher
the logic of capitalism.
In Chakrabarty™s analysis, the jute workers of Bengal were certainly
wage laborers, but they were not proletariat in Marx™s sense. If this is
correct, the implications need to be pressed. Brie¬‚y, Marx builds assump-
tions regarding culture “ the politics of “equal rights” “ into his model in
Capital. These assumptions simply do not obtain in India. Accordingly
(and quite apart from the critical facts of colonialism), we can hardly
expect that capitalism in India would take the same form as, presum-
ably, it did when Marx was writing about England. Moreover, because
of this cultural difference, it is not surprising that class consciousness did
not emerge among Indian jute workers. More generally, while there is
94 A Realist Philosophy of Social Science

the high abstraction, capitalism, concretely, there are only capitalisms,
differing as least as much as there are differences in historical experience.
Tilly™s “Coercion, Capital and European States” (1992) provides
another convenient example. Tilly™s aim is to offer a way to under-
stand state-building in Europe. He begins with a highly abstract the-
ory which represents “the logics of Capital and Coercion.” These are
mechanisms which are meant to apply trans-historically. Thus, we can
divide populations into those who have capital (understood by Tilly in
non-Marxist terms) as “any tangible resources, and enforceable claims
on such resources” (1992: 153),13 and those lacking such. “Capitalists,
then, are people who specialize in the accumulation, purchase and sale of
capital.” “They occupy the realm of exploitation” vis-` -vis those lacking
such resources. Surpluses produced are captured by capitalists. Simi-
larly, by virtue of their position in social relations, persons or groups can
dominate others. Tilly makes no effort to spell this out in any further
detail, but it is easy to see that as regards both mechanisms, some very
elementary assumptions are being made: for example, that those lacking
the means of life must produce for those who have the resources. Nor
is Tilly interested in theorizing the wide variety of mechanisms which,
in speci¬c times and places, allow for the accumulation and capture of
surplus. Feudal mechanisms in Western Europe were different from the
mechanisms of capitalism (as Tilly would agree). His interest here is at a
higher level of abstraction.
In the concrete world, the mechanisms work conjointly, but effec-
tively using the tools of comparison. Tilly offers that one can theorize
three processes which produce states: “a coercion-intensive process, a
capital intensive process, and a capitalist coercion path.” Thus, “in the
coercion-intensive mode, rulers squeezed the means of war from their
own populations and others they conquered, building massive structures
of extraction in the process” (1992: 164). In their phases as tribute-
taking empires, this path was taken by Brandenburg and Russia. “In the
capital-intensive mode, rulers relied on compacts with capitalists . . . to
rent or purchase military force, and thereby warred without building vast
permanent state structures. Typically, city-states and urban federations
took this path. Finally, rulers can do some of each, typically producing
full-¬‚edged national states earlier than the coercive-intensive and capital-
intensive modes did” (1992: 164“165).

13 Marx is de¬nite in restricting capital to a social relation between owners of the means
of production and wage workers who are compelled to sell their labor. To be sure, in
all historical societies surpluses are captured by those who own or control the means of
production. But how this is accomplished requires some speci¬c, different mechanisms.
A “general theory” of exploitation would not be very informative. See appendix C.
Agents and generative social mechanisms 95

Another mechanism is introduced (without explicit notice): “Driven by
pressures of international competition (especially by war and preparation
for war), all three paths eventually converged on concentrations of capital
and of coercion all out of proportion to those that prevailed in AD 990”
(1992: 165).
The account is highly abstract but nevertheless illuminating, going well
beyond the claim, for example, that the modern state is the product of
war “ or of capitalist development. But as Tilly recognizes, if this provides
abstract understanding of state-building, the devil is in the details. In his
appeal to Lorenzo, Machiavelli saw that his city-state could not stand up
to the aggrandizing motives of the new “empires,” but we need Italian
and English history to see why England achieved fully-¬‚edged nation-
state status by the seventeenth century and Florence did not. We need
more than mechanisms here, a problem to be considered in chapter 5.
Work by Stinchcombe (1998) provides another useful illustration both
of the idea that a fundamental mechanism may be employed in different
institutional contexts, and of the putative macro“micro gap. He theo-
rizes a mechanism titled “monopolistic competition” which by virtue of
strong analogy, explains “the continuity of status of corporations in mar-
kets, or universities in prestige system, [and] of world power systems”
(1998: 207). The basic mechanism needs to be stated at a relatively high
level of abstraction to cover these divergent contexts, and Stinchcombe
does not make much effort to articulate in detail the model in terms of the
speci¬c types of actors and the structures they are working with. Thus,
he argues that “in ¬elds of markets, prestige systems, and world systems,
some organizations perform better than others, and they always do this
not by becoming rentiers choosing their investments. Instead, they orga-
nize networks of collective action, create networks of suppliers, build or
buy capital resources, and give people incentives to do all these successful
performances” (1998: 270). While it is sometimes convenient to speak
of organizations as agents, it is clear that it is the top management of
these organizations who, by virtue of prevailing conditions suggestively
sketched, are enabled and motivated to “appropriate the bene¬ts of their
competence as long as the opportunity continues to pay off, or if com-
petitors develop competitive competences so that monopoly is no longer
defensible” (1998: 271). Stinchcombe™s account, of course, is more illu-
minating than the bare bones summarized here, but when all is said and
done, a good deal more might be said about the mechanisms he has
By contrast, the mechanisms analyzed in Goffman™s Asylums (1961)
apply pointedly only to “total institutions.” Goffman™s model is not
made explicit, but it is far richer than Marx™s model exactly because
96 A Realist Philosophy of Social Science

it is quite concrete. We can identify the key elements. Goffman identi¬es
two entirely antagonistic types of actors standing in a well-de¬ned social
relation: the managers and the managed (professionals versus clients, staff
versus inmates). By virtue of their place in these social relations, there are
resources available which enable the construction of their identities and
their social relations, including having appropriate credentials and dress,
along with a number of speci¬c capacities characteristic of the institution.
Abstractly, the managed must be constructed as something less than a
full person, while the manager is constructed as competent to “treat” the
managed. Thus, “social distance is typically great and often formally pre-
scribed” (Goffman, 1961: 7). Each of the two parties has goals (which
“provide a key to meaning”) and each has a system of beliefs (for the
managers, an interpretative scheme which includes a theory of human
nature). For each group, there are structured capacities for achieving their
goals. Typically, the managed undergo morti¬cation, the construction of
a different self; role dispossession (1961: 14“15); “personal defacement”
(1961: 20“21); “contaminative exposure” (1961: 23); and a diminishing
of the capacity of the managed to control action. A mechanism employed
in the stripping of power is “looping,” where a disruptive response from
an agent becomes the target of the next attack (1961: 35“36). But as
with Willis, the managed also have resources. Resistance by them takes
on a number of forms, including contesting the meaning of rules, “frater-
nization,” and “playing it cool” (1961: 61“65). Institutional ceremonies,
including, for example, a newsletter produced by inmates, an annual
party and an open house, are regular events in the life of the institution.
These are intended to produce a joint commitment to the of¬cial goals,
even if everyone “on the inside” knows better.
Goffman very convincingly shows how the beliefs of actors, true and
false, promote behaviors which have as their outcome the reproduction of
an institution in which there is a manifest disjunction between the of¬cial
goals of the institution and the actual outcomes, and how, as in the Willis
account, actors unintentionally act in self-defeating ways that sustain the
conditions of their own oppression.
Goffman™s model is of a total institution, but it offers insights into “near
total institutions,” for example, a boarding school, and even to institu-
tions not nearly as “total,” for example, a factory or university. Extended
in this way, of course, the model needs to be amended to address the
differences in the real concrete between these sorts of institutions. Bosses
and classroom teachers do not stand in the same relation to the man-
aged as staff to inmates. But there will be social mechanisms at work in
these places that can be identi¬ed and that gives us an understanding of
Agents and generative social mechanisms 97

Worth mention here are the family of mechanisms generally termed
“self-ful¬lling prophecies” or as Schelling (1998) suggests, “self-realizing
expectations.” As he says, a coffee shortage, an insolvent bank, and going
early to get a seat, would seem to be explained by the same mechanism.
The critical step is the fact that acting on the pertinent expectation is
suf¬cient to produce the outcome: if people expect a coffee shortage,
many will engage in hoarding, and there will indeed be a coffee shortage.
Similarly, if people believe that a bank is insolvent, they will withdraw
their savings, ultimately rendering the bank insolvent.

Generalization, abduction and assessing theories
of social mechanisms
It was noted that generalizations do not explain, that they need explain-
ing. It is easy enough to see also that generalizations14 will be the point
of departure for a theory of a mechanism. That is, where there is some
pattern or regularity, there are but two possibilities: either the regularity is
the product of some mechanism or combination of mechanisms at work,
or it is not. So compare, “A relatively small proportion of children from
poor neighborhoods in the UK continue into higher education,” and “For
the past three months, as the prices of real estate in Honolulu went up,
so too did movie attendance.” Both propositions might be true, but there
is little reason to believe that there is a mechanism at work which could
explain the latter. The correlation is wholly accidental, perhaps a statis-
tical anomaly. Where the regularity seems not to be accidental, Lawson
(1997) suggests the idea of a “demi-regularity,” or “demi-reg” for short.
He de¬nes it as “a partial event regularity which prima facie indicates the
occasional, but less than universal actualization of a mechanism or ten-
dency, over a de¬nite region of time-space” (1997: 204). “Demi-regs”
prima facie suggest a mechanism exactly in the sense that based on what
we know, the connection is not likely to be accidental. Of special inter-
est, then, are what he calls “contrastive demi-regs.” He gives a number
of examples, some commonsensical, some not so obvious: “Women look
after children more than men do.” “Average unemployment rates in the
western industrial societies are higher in the 1990s than the 1960s,” “In
the 1990s UK ¬rms are externalizing or ˜putting out™ more parts of the
production process than twenty years ago,” “Government persons tell
more lies in war-time.” As Weber rightly noted, there are countless num-
bers of these functioning both in ordinary life and in more sophisticated
14 Generalizations here include both universal and statistical assertions, “All Fs are Gs,”
“Most Fs are Gs” are patterns and regularities which can always be expressed in these
98 A Realist Philosophy of Social Science

science. We can offer a number of important observations regarding such
First, theory construction does not begin from nothing. Not only is it
problem-driven, but the theorist has a stock of knowledge which will be
the materials of the effort. Struck then by what is an interesting (and well-
established) “demi-reg,” inquiry into the possible mechanism or mecha-
nisms begins.
Second, the demi-regs may well be the product of descriptive work,
either quantitative or qualitative. Stephen Kemp and John Holmwood
(2003) have argued that identifying unknown patterns is a particularly
important task of statistical techniques. Thus, drawing on work by Stew-
art and his colleagues (Stewart et al., 1980) they ask whether there is a
mechanism to explain the unclear relationship between class background
and type of school with the number of years a student pursued education.
Kemp and Holmwood argue that regression techniques were effectively
employed to show that the strongest pattern regarded class and high-
status schools, a pattern not discernible without the use of these meth-
ods. But, of course, on the present view, the inquiry could not stop here.
It is unfortunate that so much solid descriptive work is so often mistak-
enly taken to be explanatory when it is not (see appendix A). Given the
identi¬ed pattern, the problem now becomes what explains it. Indeed,
some aspects of the pertinent mechanism would seem to be involved in
some of our previous examples. The interested reader might well test her
theoretical ingenuity.
Third, contrastive demi-regs force inquiry into looking for differences
which point to the probable causally relevant features. “[We] notice the
effects of sets of structures through detecting relatively systematic differ-
ences in the outcomes of prima facie comparable types of activities (or
perhaps similar outcomes of prima facie different activities in different
space-time locations, or differences in types of position-related activities
on comparable space-time locations, and so forth” (Kemp and Holm-
wood, 2003: 208“209). So, as is obvious enough, differences in domes-
tic responsibilities between men and women suggest powerfully that we
need to understand the mechanism which explains the existing division
of labor. There are, no doubt, mechanisms of gender discrimination at
work, but as with racism, these need to be spelled out and con¬rmed.
Similarly, increased unemployment rates suggest differences in produc-
tivity or rates of pro¬t which in turn suggest changes in mechanisms
explaining productivity or the rate of pro¬t.
The idea of contrastive demi-regs is at the bottom of considera-
tions regarding comparison, long recognized to be a tool of macro- and
Agents and generative social mechanisms 99

historical sociology. That is, where outcomes are different, we seek dif-
ferences in the causes. Comparative inquiry provides opportunities to
identify the pertinent mechanisms at work in one case but not the other,
or to identify pertinent differences in a similar mechanism which explains
the differences in outcome (see chapter 5 and appendix C).
Finally, as in natural science, the mode of inquiry here is neither deduc-
tion nor induction, but what C. S. Peirce called “abduction.” Given a
demi-reg, can we identify the causal mechanism which explains it? And
if there are several plausible mechanisms, can we arrive at some valid, if
still fallible, conclusion? A host of dif¬culties attend this, including, as
already noted, the fact that experiment is generally not possible in the
social sciences.15 The absence of the possibility of controlled experiment
is an important difference between the natural and social sciences, but it
need not lead to the conclusion that a human science is quite impossible
(Collier, 1994).
Two lines of argument may be noted. First, there is the sort of evi-
dence produced by Goldstein, et al. (above) to test their three different
explanatory models. This is more or less direct. Second, among compet-
ing explanatory mechanisms, there are different consequences and these
are testable. Hernes offers a wonderful example: the effort to explain why,
as reported by Norwegian media, women are stung by wasps more often
than men. It shows clearly how, on realist grounds, a theory of generative
social mechanisms might be tested.
15 So-called “natural experiments” are not experiments in any useful sense. But there are
what are sometimes termed “quasi-experiments.” An excellent example is the longitu-
dinal study, “Lifetime Effects: The High/Scope Perry Preschool Study Through Age
40.” As summarized by David L. Kirp, “From a group of 123 South Side neighborhood
children, 58 were randomly assigned to the Perry program, while the rest, identical in
virtually all respects, didn™t attend preschool. Most children attended Perry for two years,
three hours a day, ¬ve days a week. The curriculum emphasized problem-solving rather
than unstructured play or ˜repeat after me™ drills. The children were viewed as active
learners, not sponges; a major part of their daily routine involved planning, carrying out
and reviewing what they were learning. Teachers were well trained and decently paid,
and there was a teacher for every ¬ve youngsters. They made weekly home visits to par-
ents, helping them teach their own children.” “Random assignment is the research gold
standard because the ˜treatment™ “ in this case, preschool “ best explains any subsequent
differences between the two groups.” Data was collected every year from age 3 through
11, then at ages 14, 15, 19, 27 and 40. The results are quite remarkable in terms of every
relevant outcome: literacy, completion of high school, crime, and marriage and divorce
rates. Indeed, at age 40, “nearly twice as many have earned college degrees (one has a
Ph.D.). More of them have jobs: 76 per cent versus 62 per cent. They are more likely to
own their home, own a car and have a savings account. They are less likely to have been
on welfare. They earn considerably more “ $20,800 versus $15,300.” See David Kirp,
“Life Way After Head Start,” New York Times Magazine, November 21, 2004. But, of
course, we remain unclear as to what in the experience of these students explains these
differences in outcomes. It will certainly be a complicated story.
100 A Realist Philosophy of Social Science

He offers four possible explanations:
(1) The Rambo theory: “Women are a more tender species than men . . . For
a real man it would be disgracefully effeminate to call a doctor for a
dinky distress” (Hernes, 1998: 76). The mechanism has it that men
fail to report bites.
(2) The outdoors theory: “Women spend more time in the open air than
men, walking their babies and playing with their children.” This
mechanism involves gender differences in roles.
(3) The hysteria theory: Women panic when they see a wasp, agitating
them to sting. Men do not panic. The mechanism here makes women
the cause of increased biting.


. 3
( 7)