The Oxford Handbook of Philosophy of Social Science 2012

The Oxford Handbook of Philosophy of Social Science
Harold Kincaid
Print publication date: Nov 2012

Print ISBN-13: 9780195392753
Published to Oxford Handbooks Online: Nov-12
Subject: Philosophy, Philosophy of Science, Social and Political Philosophy
DOI: 10.1093/oxfordhb/9780195392753.001.0001
Micro, Macro, and Mechanisms
Petri Ylikoski
DOI: 10.1093/oxfordhb/9780195392753.013.0002
Abstract and Keywords
This article, which takes a fresh look at micro–macro relations in the social
sciences from the point of view of the mechanistic account of explanation,
introduces the distinction between causal and constitutive explanation. It
then discusses the intentional fundamentalism, and challenges the idea that
intentional explanations have a privileged position in the social sciences.
A mechanism-based explanation describes the causal process selectively.
The properties of social networks serve both as the explananda and the
explanantia in sociology. Knowledge of the causal mechanisms is vital in
the justification of historical causal claims. The intentional attitudes of
individuals are also important in most mechanism-based explanations of
social phenomena. It is important to pay closer attention to how real macro
social facts figure in social scientific theories and explanations.
social sciences, mechanism-based explanation, intentional fundamentalism, causal process,

constitutive explanation, explananda, explanantia, micro–macro relations
2.1. Introduction
This chapter takes a fresh look at micro-macro relations in the social

sciences from the point of view of the mechanistic account of explanation.
Traditionally, micro-macro issues have been assimilated to the problem
of methodological individualism (Udéhn 2001, Zahle 2006). It is not my
intention to resurrect this notoriously unfruitful controversy. On the contrary,
the main thrust of this chapter is to show that the cul-de-sac of that debate
Page 1 of 36 Micro, Macro, and Mechanisms

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). (c) Oxford University Press, 2013. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford
Handbooks Online for personal use (for details see Privacy Policy).
Subscriber: University of Western Ontario; date: 14 April 2013
can be avoided if we give up some of its presuppositions. The debate about
methodological individualism is based on assumptions about explanation,
and once we change those assumptions, the whole argumentative landscape
changes.
The idea that social scientific explanations are based on causal mechanisms
rather than covering laws has become increasingly popular over the last
twenty years or so (Hedström and Ylikoski 2010). Interestingly, a similar
mechanistic turn has occurred also in the philosophies of biology and
psychology (Wright and Bechtel 2007). Until recently, the connections
between these two emerging traditions for thinking about mechanisms
have been rare. The aim of this chapter is to employ ideas developed
by philosophers of biology to address some issues that the advocates of
mechanisms in the social sciences have not yet systematically addressed.
I argue that ideas about levels of explanation and reductive research
strategies, which were originally developed in the context of cell biology and
neuroscience, can be fruitfully adapted to the social sciences. They can both
strengthen the case for mechanism-based explanations in the social sciences
and bring the philosophy of social science debates closer to social scientific
practice.
The chapter is structured as follows. In the first section, I will take a look
at recent work on mechanism-based explanation. While I suggest that the
mechanistic account of explanation presupposes some more fundamental
ideas about explanatory relevance and causation, I also argue that it
provides a fruitful tool for thinking about micro-macro relations in the social
sciences. In the second section, I will criticize a common philosophical way of
formulating the micro-macro issue and provide my own characterization that
is not dependent on the assumption that there is a unique or comprehensive
micro level. The third section introduces the distinction between causal and
constitutive explanation, and argues that this distinction helps to make sense
of the call for microfoundations in the social sciences. The final section will
take on a doctrine that I call intentional fundamentalism, and it challenges
the idea that intentional explanations have a privileged position in the social
sciences.
2.2. Mechanism-based Explanation
The idea of mechanism-based explanation has been developed

independently among social scientists (Harré 1970; Elster 1989, 2007; Little
1991; Hedström and Swedberg 1998; Hedström 2005; for a review see
Hedström and Ylikoski 2010) and philosophers of biology (Bechtel 2006,

2008; Craver 2007; Darden 2006; Wimsatt 2007). In the social sciences, the
idea of causal mechanism has been used mainly as a tool for methodological
criticism, while in the philosophy of biology the motivation has been that of
finding a descriptively adequate account of biological explanation. Despite
these separate origins and motivations, both traditions are clearly building
on similar ideas about scientific explanation. For example, both share the
same dissatisfaction with the covering law account of explanation (Hedström
2005; Craver 2007).
There is no consensus on the right definition of a causal mechanism.

Although some theorists find such a situation frustrating, I do not think this
constitutes a real problem. The entities and processes studied by different
sciences are quite heterogeneous, and it is probably impossible to propose
a mechanism definition that would be both informative and cover all the
prominent examples of mechanisms. Some disciplines, such as cell biology
(Bechtel 2006) and the neurosciences (Craver 2007), study highly integrated
systems, whereas others, such as evolutionary biology and the social
sciences, study more dispersed phenomena (Kuorikoski 2009), so it is more
plausible to think that informative characterizations of mechanisms are field
specific. The task of a philosophical account is to show how these exemplars
are related to general ideas about explanation, evidence, and causation, not
to engage in verbal sophistry. However, it is possible to give some general
characteristics of mechanisms.
First, a mechanism is always a mechanism for something; it is identified

by the kind of effect or phenomenon it produces. Second, a mechanism
is an irreducibly causal notion. It refers to the entities of a causal process
that produces the effect of interest. Third, a mechanism has a structure.
When a mechanism-based explanation opens the black box, it makes
visible how the participating entities and their properties, activities, and
relations produce the effect of interest. The focus on mechanisms breaks
up the original explanation-seeking why-question into a series of smaller
questions about the causal process: What are the participating entities,
and what are their relevant properties? How are the interactions of these
entities organized (both spatially and temporally)? What factors could
prevent or modify the outcome? Finally, there is a hierarchy of mechanisms.
While a mechanism at one level presupposes or takes for granted the
existence of certain entities with characteristic properties and activities, it
is expected that there are lower-level mechanisms that will explain them.
In other words, the explanations employed by one field always bottom out
somewhere. However, this fundamental status of certain entities, properties,

and activities for a given mechanism is only relative, as they are legitimate
targets of mechanistic explanation in another field. Of course, this chain of
explanations ends somewhere—there are no mechanism-based explanations
for fundamental (physical) processes (Hedström and Ylikoski 2010).
Although the mechanism-based account is often presented simply as an

idea about scientific explanation, the notion of mechanism is associated
with a wider set of ideas about scientific knowledge. For example, there
are ideas about the justification of causal claims, the heuristics of causal
discovery, the presentation of explanatory information, and the organization
of scientific knowledge (Ylikoski 2011). There is no doubt that these not
yet clearly articulated ideas partly explain the appeal of the approach. For
example, as I will show later in this chapter, claims about the explanatory
role of mechanisms are often confused with claims about their relevance to
the justification of causal claims (see also Kincaid, this volume).
While I think all the above ideas are important advances in understanding
explanatory reasoning in science, it is not necessary to assume that the
notion of mechanism is the ultimate solution to all problems in the theory of
explanation. On the contrary, the mechanistic theory presupposes accounts
of explanatory relevance, causation, and the nature of generalizations that
provide the basis for mechanisms. The notion of mechanism should not be
treated like a black box. I have argued elsewhere (Hedström and Ylikoski
2010; Ylikoski 2011) that if the mechanistic ideas are combined with the
theory of explanation developed by James Woodward (2002, 2003), we can
get quite far in solving these problems. While for the present purposes we
do not have to consider in detail the relation between mechanisms and
generalizations, some comments on explanatory relevance are in order as
later arguments depend on it.
A mechanism-based explanation describes the causal process selectively.

It does not aim at an exhaustive account of all details but seeks to capture
the crucial elements of the process by abstracting away the irrelevant
details. The relevance of entities, their properties, and their interactions is
determined by their ability to make a relevant difference to the outcome
of interest. If the presence of an entity or of changes in its properties or
activities truly does not make any difference to the effect to be explained,
it can be ignored. This counterfactual criterion of relevance implies that
mechanism-based explanations involve counterfactual reasoning about
possible changes and their consequences (Ylikoski 2011). A natural way to
understand these causal counterfactuals is to understand them as claims

about the consequences of ideal causal interventions (Woodward 2003,
2008). The causal counterfactual tells us what would have happened to the
effect if the cause had been subject to a surgical intervention that would not
have affected anything else in the causal configuration. An advantage of
the interventionist account of causation is that it allows talking about causal
dependencies in every context where the notion of intervention makes
sense. Unlike some other theories of causation, such as various process
theories, it is level-blind and applicable to special sciences such as cell
biology or sociology.
2.2.1. Mechanisms and Reductive Explanation
One of the distinctive features of the mechanistic approach to explanation is

that it reorients the issues related to reductionism and reductive explanation.
In one sense the mechanistic way of thinking is thoroughly reductionist: It
attempts to explain activities of mechanisms in terms of their component
parts and their activities, and then subjects the component mechanisms
to the same treatment. In this sense, the reductive research strategy has
probably been the single most effective research strategy in the history
of modern science. However, there is another sense in which mechanism-
based explanations are clearly nonreductionist: Although they do refer to the
micro level, they do not replace or eliminate the higher-level facts nor the
explanations citing them. Rather than serving to reduce one level to another,
mechanisms bridge levels (Darden 2006; Craver 2007; Wright and Bechtel
2007; Richardson 2007; McCauley 2007; Wimsatt 2007).
The mechanistic account of reductive explanation differs significantly from

the traditional philosophical accounts of intertheoretical reduction that
conceive reduction as a derivation of one theory from another (Richardson
2007; McCauley 2007). The mechanical account of reductive explanation
does not start with a strongly idealized picture of a discipline-wide theory
that contains all knowledge about its level. Nor does it conceive reduction
as a deductive relation between such theories (or their corrected versions).
Rather, reductive mechanistic explanations are constructed piecemeal with
a focus on particular explanatory targets. While there is an assumption that
everything is mechanistically explainable and a presumption that ultimately
all mechanistic accounts are mutually compatible, there is no overarching
effort to combine them into one grand theory that would cover all the
phenomena that the scientific field studies. Also, contrary to the traditional
accounts that conceive reduction as elimination or replacement, the
mechanisms are inherently multilevel. The components and their operations

occur and are investigated at one level, whereas the mechanism itself and its
activities occur and are investigated at a higher level. In this sense accounts
of mechanisms often have the character of interfield theory (Darden 2006).
This makes it difficult to characterize the reductive understanding provided
by mechanical explanations as deductive relations between independent
theories.
The mechanistic stance also gives reasons for rethinking the notion of
levels. According to the traditional layer-cake conception, there is a neat
hierarchical layering of entities into levels across phenomena, and the
scientific disciplines (e.g., physics, chemistry, biology, psychology, sociology)
are distinguished from each other by the level of the phenomena that they
are studying (see Oppenheim and Putnam 1958). From the mechanistic
point of view, this way of thinking unnecessarily drives together levels of
nature and science, and misleadingly suggests that the levels are both
comprehensive and the same independently of the investigative context
(Craver 2007). The actual scientific disciplines do not match neatly with the
metaphysical picture of levels of organization or reality. And while there are
many problems in a serious characterization of the metaphysical picture of
levels, there do not seem to be any particularly good reasons to accept such
a metaphysical constraint for an account of scientific explanation.
The notion of the levels of mechanism plays an important role in the

mechanistic account but is free from many of the traditional assumptions
about levels. The levels of mechanisms are perspectival in the sense that
the levels are dependent on the explanatory target. Macro-level facts are
explained by appealing to micro-level processes, entities, and relations,
but these items belong to the micro level just because they are required
for the full explanation of the macro fact, not because they belong to some
predetermined micro level. Whatever is needed for explaining the macro fact
is regarded as belonging to the same level. However, there is no guarantee
that these components would always be at the same level in all possible
explanatory contexts. Nor it is obvious that the micro-level entities and
processes that account for these components would be in any simple sense
from the same level. For every hierarchy of mechanisms, there is a clear
hierarchy of levels of mechanisms, but these levels are local. There is no
reason to assume that separate hierarchies of mechanism levels would
together produce the neatly delineated and comprehensive levels of nature
assumed in the traditional layer-cake model (Craver 2007).

These views have a number of interesting consequences for traditional
ways of thinking about reductive explanation and the explanatory role of
microfoundations in the social sciences. For example, once we give up
the outdated deductive model of theory reduction, many of the traditional
fixations of the methodological individualism debate simply become
meaningless. For example, there is no need to provide individualistically
acceptable redefinitions of macro-social notions because the explanation
of macro facts is no longer conceived as a logical derivation. Similarly,
the search for any bridge laws between theories becomes pointless.
This has the consequence that the key anti-reductionist argument about
multiple realization loses much of its significance. From the point of view
of mechanistic explanation, multiple realization is simply an interesting
empirical observation that does not pose any serious threat to the possibility
of explaining macro properties in terms of micro properties and relations. Just
as the sciences have learned to live with the fact of alternative causes, they
can learn to live with the phenomenon of multiple realization.
The advocates of the mechanism-based approach in the social sciences have

noticed some of these consequences. For example, they have largely given
up the old ideas about reductive explanation and have instead emphasized
the importance of microfoundations (Elster 1989; Little 1991). However, I do
not think that all the implications of the mechanistic perspective have been
taken into account in the philosophy of social sciences. This is visible, for
example, in the fact that quite often the mechanistic approach is associated
with methodological individualism (Elster 1989). Similarly, much of the
debate about micro-macro relations is still focused on arguments that are
based on a premechanistic understanding of reductive explanation (Sawyer
2005; Zahle 2006).
The aim of this chapter is to sketch what a consistently mechanistic way to

think about micro-macro relations would look like and to show that some
of the key presuppositions of the traditional debate about methodological
individualism should be given up. One of these is the assumption of a
comprehensive, unique, and privileged individual level. The notion of
comprehensiveness refers to the idea that there is a consistent and well-
defined individual level that is sufficient to cover all social phenomena and
that would serve as a reduction basis for all nonindividual social notions.
Uniqueness refers to the assumption that in all social explanations, the micro
level would always be the same level, for example, the level of intentional
rational action. Finally, the notion of privileged refers to the presumption that
explanations in terms of this special level have some special explanatory

qualities that set them apart from explanations from other levels. In the
following, I will challenge all three assumptions and argue that once they are
given up, we can approach the micro-macro issues in the social sciences in a
more clear-headed manner.
2.3. Rethinking the Macro
A popular argumentative strategy among anti-individualists has been

to borrow ideas from the philosophy of mind. They are inspired by the
arguments for nonreductive materialism, so they build their argument
based on an analogy with the mind-brain relation. Given that these
arguments are not very mind specific—it is a general practice just to talk
about M- and P-predicates—their appeal is understandable. The ideas of
supervenience and multiple realization seem to provide a neat way to
argue against reductionism, at least if one accepts the traditional idea of
derivational reduction. While there are reasons to suspect that the notion of
supervenience is less illuminating than is often assumed (Horgan 1993; Kim
1993) and that the traditional view of reduction does not completely collapse
under multiple realization (Kim 1998), we can set these issues aside as their
relevance presupposes a premechanistic account of reductive explanation.
Here I want to focus on the mind-brain analogy as I think it is misleading.
The mind-brain analogy is inappropriate because it mischaracterizes the

nature of the social scientific micro-macro problem. The central problem
in the philosophy of mind is to figure out how the explanations provided
by psychological theories that employ mental concepts are related to
the accounts of the brain’s working provided by the neurosciences. The
challenge is to relate two levels of description that are fundamentally talking
about the same thing. The (nondualist) antireductionist position does not
typically challenge the causal sufficiency of the neural-level facts. The setup
is quite different in the social scientific micro-macro debates.
The problem in the social sciences is not that of bridging a comprehensive

and exhaustive individual-level understanding of social processes (the
analogue to the idealized knowledge of the brain) to a more social or
holistic description (that would be analogue to the idealized psychological
theories employing the mental vocabulary). It is typical for anti-individualists
to challenge the causal sufficiency of individual facts. They often claim
that the facts about individuals allowed by the individualist are either not
sufficient to account for all social facts or the individualists are cheating by
accepting facts that are not properly individualistic. This is because the issue
is not really that of relations between two comprehensive (and potentially

competing) levels of description, but that of seeing how local facts about
individuals and their social interactions are related to large-scale facts about
groups, organizations, and societies. So, the relation is really more like the
one between the whole brain and its parts than the mind and the brain.
While this contrast is useful for highlighting the inappropriateness of the
mind-brain analogy, I do not want to develop it further as there are many
problems with the organ-society analogy. It is better to skip all the brainy
analogies and to take a fresh look at the micro-macro problem as the social
scientists face it.
A useful starting point is the observation that macro social facts are typically
supra-individual: They are attributed to groups, communities, populations,
and organizations, but not to individuals. There might be some attributes
that apply both to individuals and collectives, but typically macro social
properties, relations, and events are such that they are not about individuals.
Another salient feature of many social micro-macro relations is the part-

whole relationship. One way or another, the macro social entities are made
of their constituting parts. Usually this relation of constitution is more
than mere mereological aggregation or simple material constitution. First,
many social wholes are composed of a heterogeneous set of entities; there
are intentional agents, their ideas, and material artifacts. Second, in all
interesting examples of social wholes, the relations between the components
play an important role. (Similarly, often the relations between social wholes
and between the social whole and its environment are also important.)
However, the important thing is that the part-whole relationship makes
it possible to see the micro-macro relation as a question of scale: The
difference between micro and macro is the difference between small- and
large-scale social phenomena.
I do not propose that we can simply define the micro-macro contrast as an

issue of scale. All differences in scale do not constitute a meaningful micro-
macro relation, and the heterogeneous nature of macro social facts makes
it difficult to characterize the additional requirements for their defining
features. However, I do want to suggest that it provides a fruitful way to
think about micro-macro relations and an antidote for the tendency to see
parallels in the philosophy of mind.
Thinking of the micro-macro issue as an issue of scale makes it possible

to conceive of it as being without a unique micro level. Whereas the
contrast between “individual” and “social” levels is categorical, the contrast
between small and large is relative and allows a continuum of various

sizes. Depending on the application, the micro entities could be individuals,
families, firms, or groups. This flexibility is in accordance with the way social
scientists think. They do not assume that micro is always about one specific
set of entities.
Another consequence is that whether an attribute is a macro or micro

property depends on what it is contrasted with. A friendship relationship is
a macro property from the psychological point of view, but a micro property
when considered from the point of view of the social networks within a
community. Rather than being set a priori, the contrast between micro and
macro depends on one’s explanatory interests. For example, international
politics and organizational sociology construct the micro-macro contrast
quite differently. In the former, states and other organizations are often
treated as individuals, whereas in the latter, the organizations and their
properties are the macro reality to be explained. Similarly, an economist
studying market processes can treat firms and households as the micro level,
while for disciplines such as industrial organization and family sociology, they
are the macro items that require explanation.
From the point of view of a mechanistic philosophy of science, this

flexibility is not surprising. The same dependence of levels on epistemic
concerns is also observable in the biological sciences. The cell biologists
or neuroscientists do not think in terms of comprehensive or unique
micro levels either. The levels of mechanisms found there depend on the
explanatory concerns, not on a priori ontological considerations. This is not
worrisome for the mechanistic point of view, as the key assumption is that
whatever is found at the micro level can always be turned to a macro-level
explanandum for another set of enquiries.
The social macro properties do not constitute a unified kind, so it makes

sense to characterize them with a sample of examples rather than with a
general definition. The following classification of typical sociological macro
social properties is not intended to be exhaustive of sociology or the social
sciences in general. There are many parts of macro social reality that fall
between my four categories. However, I hope the four examples can be used
to illustrate the applicability of the scale perspective.
2.3.1. Statistical Properties of a Population
A major concern for sociology is the various statistical attributes of

populations. Among these are distributions and frequencies. Sociologists are

interested in both distributions of attributes to various kinds of individuals
and distributions of individuals with certain attributes to social positions
and spatial locations. For example, when they are studying the ethnic
segregation of cities, comparing societies in terms of inequality, or describing
the social stratification of a society, they are attempting to account for
distributions. Another relevant property of distributions are frequencies.
Sociologists are interested in typical, rare, dominant, or marginal behaviors,
beliefs, or attitudes within a specified population. Similarly, they are
interested in ratios of attributes such as unemployment or incarceration
within the population. So, when sociologists are studying changes in
racial prejudices over time, comparing the level of conformism between
communities or tracking the changes in the level of union memberships, they
are interested in explaining frequencies.
All these statistical macro social properties are inferred (or estimated) from
data about the members of a population. There is no other way to access
them. However, it does not make any sense to attribute these properties
to individual units. Another important thing about these macro social facts
is that the units of these statistics do not have to be individuals; they can
as well be families or firms. It is noticeable that statistical macro properties
are in no way dependent on the members’ beliefs and attitudes about them.
The members of the population can have false, or even crazy, beliefs about
distributions and frequencies that characterize their own society.
While the statistical properties of populations usually only serve as

explananda in the social sciences, they do have some legitimate and
nonreducible explanatory uses. For example, in the cases of frequency-
dependent causation (e.g., cases in which the causal effect of an individual
having a certain property depends on the frequency of that property in the
population), the statistical facts are the crucial difference makers. Similarly,
in many social scientific explanations, the correlations between various
variables (for example, wealth, education, taste, and place of residence) play
an important role in accounting for individual differences in behavior and
attitudes. Both of these cases are quite easily conceived as cases of larger-
scale facts influencing smaller-scale phenomena, while other ways to think
about levels are not as natural.
2.3.2. Topologies of Social Networks within a Population
Sociologists are also interested in relations and interactions between

individuals. When considered together, these relations constitute networks

of social relations within the population. A social network can be regarded
as a map of all of the relevant ties between the members of a specified
population. When sociologists are studying the spread of information
within an organization, comparing groups with respect to their level of
network clustering or analyzing the brokering opportunities of an individual
occupying a structural hole (i.e., a position between two networks that are
not otherwise connected), they are examining social networks.
The importance of social networks is increasingly being recognized in

the social sciences, and social network analysis is becoming increasingly
popular in various social sciences. Social network analysis is based on the
observation that networks have many interesting (formal) properties, such as
centralization, cohesion, density, and structural cohesion (Scott 2000). While
the social network is inferred from knowledge about individual relationships,
the properties of the network are prototypical macro properties. It does not
make any sense to apply these attributes to individual nodes of the network.
Similarly to statistical properties, the units of network analysis are flexible.
There is no requirement that the nodes of the network (the members of the
population) are persons. They can also be groups, families, organizations, or
even states.
The properties of social networks serve both as the explananda and the
explanantia in sociology. As an example of the latter, consider the notion
of a structural hole (Burt 1992), which is used to explain the differences in
agents’ ability to access information and in their opportunities to influence
social processes. In these explanations the structure of the network plays
an irreducible role, and it is quite natural to think of the social network as
a large-scale social phenomenon influencing local interactions between
individuals. In contrast, it is very difficult to think about them in terms
of social and individual levels. As social networks are attributes of the
population, it would be quite a stretch to call social networks individual
properties. But if they are macro-level properties, what would be the
individual-level properties that could be regarded as their bases? Collections
of relevant individual relations, one might suggest, but that would be just a
vague way to talk about networks. Things are simpler if one does not have to
bother with such questions. A network is simply a more extensive entity that
is constituted by more local relations and it can have properties that are not
properties of its components.

2.3.3. Communal Properties
By communal properties I refer to social scientific notions that apply to

specific communities, but not to isolated individuals. Among these notions
are such things as culture, customs, social norms, and so on. For example,
cultural differences are primarily between groups, not between individuals.
Similarly, social norms and customs are properties of communities—
attributing them to solitary individuals does not make sense. Many of these
notions do not have precise definitions, and their explanatory uses are often
confusing (Turner 1994; Ylikoski 2003), but they do have an important role in
the social sciences.
While communal properties are attributed to groups, they are quite

straightforwardly based on facts about individuals. Underlying these notions
is the idea that the members of a group share certain beliefs, expectations,
preferences, and habits. However, it is crucial that the sharing of these
individual attributes is not purely accidental: The members have these
individual properties because the other members of the group have them.
The sharing of these properties is due to continuing interaction. For example,
the existence of a social custom presupposes that the novices learn specific
expectations and habits when they become members and that the members
of the group stick to these expectations and habits because others also do
so. Underlying the (relative) unity of a culture are facts about the shared
origins of the ideas, values, and practices of the members and their constant
interaction with each other. Similarly, the cohesion of a culture is based on
the frequency of interactions with the group and the rarity of interactions
with outsiders, not on any kind of higher-level influence on individuals.
Descriptions of customs, social norms, and cultures are always based

on idealization and abstraction. Members of a community never have
exactly the same ideas, preferences, or routines. That would be a miracle,
given what is known about human learning and communication (Sperber
2006). There is always some variation among the members, no matter how
comprehensive the socialization processes are. However, these idealized
descriptions are still useful. They draw attention to features of the group that
are typical and salient when it is contrasted with some other group.
Although communal properties, as I have described them, are tied to a

social community defined by frequent interactions, the boundaries of these
communities are fluid. This makes it possible to describe culture on various
scales—for example, on the levels of a village, a local area, and a nation.

However, descriptions on larger scales are bound to be more abstract and
less rich in detail as individual variation takes its toll. The same flexibility that
characterizes statistical and network properties applies also to communal
properties, which can also be attributed to nonpersonal units. For example,
it is possible to describe social norms that govern interactions between
organizations.
When we consider communal properties as idealizing abstractions from

shared individual properties, there is no need to refer to them as any kind
of autonomous level of reality. They just describe more extensive facts
than descriptions of the individual attitudes, habits, and preferences that
constitute them. The scale perspective also appears natural when the
explanatory use of communal properties is considered. For example, when
we are explaining the behavior of an individual by appealing to social norms,
we are referring to larger-scale facts about the group members that are
causally relevant to the micro-level behavior. There is no need to postulate
a separate realm of norms to understand what is happening. It is just that
the expectations and responses of the other group members influence the
individual’s judgments about appropriate behavior.
2.3.4. Organizations and Their Properties
Organizations such as states, firms, parties, churches, and sport clubs

are important parts of the social reality. While the community that is the
basis for communal properties is not often clearly demarcated, a clear
demarcation is often the case with organizations. They usually have specified
criteria for membership, at least for the operational members. They also
have rules that define the rights and duties of members and the roles of
various functionaries. These (written or nonwritten) rules make it possible
for organizations to have stability and continuity, so that it makes sense to
talk about their continuing existence when their functionaries are replaced
and the members change. Furthermore, many organizations exist (and are
defined) in the context of other organizations, so one has to pay special
attention to context when attempting to make sense of organizations.
Organizations as entities can have many properties that are not properties
of their members. They can even have goals that are not the personal
goals of their members, and some organizations are treated as legal
persons. This has convinced many that organizations are real entities that
should be treated as a separate ontological category. I do not have strong
opinions about issues of ontological bookkeeping, as it is remembered

that organizations are human artifacts that are always made of persons,
their ideas about the rules, and often, of material artifacts. Whatever the
organization does, is done by its members in its name. It is of crucial social
importance whether an action, for example, a questionable comment, was
made as a representative of an organization or as a private person. But
these are facts about the status attributed to the behavior, not about the two
completely different entities producing the behavior.
When a person causally interacts with an organization, she interacts with

other persons (although this interaction is increasingly mediated via material
artifacts such as ATM machines). There is no downward causal influence
from a higher level. Everything happens at the same level; it is just that
the intentional attitudes and relations of a larger group of people are
important to the details of the local situation. Similarly, the influence of the
organization on its members happens through other members, no matter
how high up some of the members are in the organizational hierarchy.
While the rules (and their interpretation by others) are external to any
individual person, there is no need to posit them as a separate ontological
category. These observations suggest that even in the case of organizations,
the layer-cake model of the social world is not very illuminating. What is
interesting about organizations is the habits and mental representations of
their members, the resources they control as members of the organization,
and their (materially mediated) interactions, not some higher ontological
level.
Again it is good to return to real social scientific questions. They concern

issues such as: How do large-scale collective enterprises—for example,
organizations—manage (or fail) to achieve certain things? What kinds of
unintended consequences do these collective activities have? How does a
membership in such collective enterprises influence the individual members?
The explanatory answers to these questions often refer to organizations and
their properties, but there is no problem in conceiving them as large-scale
things influencing smaller-scale things or other large-scale things.
These examples of macro social facts suggest a kind of flat view of society
in which the difference between micro and macro is one of scale, not of
different levels. The large-scale facts about distributions, frequencies,
interactions, and relations have an irreducible explanatory contribution
to make, but there is nothing comparable to the mind-brain relation. As a
consequence, the metaphor of levels that underlies the layer-cake model
does not really help to make sense of the issues that social scientists

addressing social macro facts are facing. Giving it up will have a number of
beneficial consequences.
First, there are some philosophical advantages. As I will argue in the next
section, once we give up the image of levels, we get rid of the problem
of causal exclusion that arises from the image of causally competing
levels. There is no problem of downward causation as there are only causal
influences from large-scale things to small-scale things and descriptions of
large-scale things at various levels of abstraction. The problem is replaced
with the more down to earth problem of explanatory selection: Under which
description can we formulate the most robust claims about counterfactual
dependence? Secondly, we no longer have to face the problem of finding an
acceptable definition of the comprehensive individual level so that we can
argue for or against methodological individualism. We can start analyzing
real social scientific explanations instead and focus our attention on the
possible contributions that large-scale things make to those on a smaller
scale and what kinds of causal mechanisms mediate these influences.
This change in framing also has some advantages when considering relations
between disciplines. The division of labor between psychology and the
social sciences is justified by differences in scale and the importance of
large-scale relations and interactions, not in terms of independent and
autonomous levels of reality. This guarantees that the social sciences will
never be reduced to psychological sciences. However, thinking in terms of
scale also cuts down the false aspirations of disciplinary autonomy. When
the social scientists are denied their own autonomous level of reality, the
ideal of completely psychology-free social science becomes less appealing.
It should be an empirical matter whether the details of human cognition
matter for social explanation. It might be that is some cases it makes good
mechanistic sense to incorporate some processes on the sub-personal level
in the explanatory theory. I will return to this possibility in the final section.
2.4. Causation, Constitution, and Microfoundations
One prominent idea in the recent philosophy of biology debate about

mechanisms has not been employed in the philosophy of social sciences
debate.1 This is the distinction between causation and constitution. Although
the difference between constitutive and causal explanation has been noted
earlier (Salmon 1984; see also Cummins 1983), it has only recently become a
topic of systematic study (Craver 2007).

Both causation and constitution are relations of dependence (or
determination), and they are easily confused. However, there are some
crucial ontological differences. Causation is a relation between events;
it is about changes in properties. Causation takes time, so we talk about
causal processes. Finally, causation is characterized by the asymmetry of
manipulation: The effect can be manipulated by manipulating the cause, but
not the other way around (Woodward 2003).
In contrast, constitution relates properties. The properties (and relations) of

parts constitute the properties of the system (sometimes also the relations
to the environment are important). The whole is made of its parts and
their relations. Unlike causation, constitution does not take time, and we
do not talk about the process of constitution. Furthermore, the relata of
constitution are not “independent existences” (as Hume called them).
For this reason we cannot characterize the relation of constitution with
the help of the asymmetry of manipulation. For example, the molecular
structure of glass constitutes its fragility: To be fragile is to have a particular
molecular structure; the fragility is not a consequence of the molecular
structure. However, there is another sort of asymmetry: the asymmetry of
existence. The parts preexist the system in the sense that the parts can exist
independently of the system, but the system cannot exist independently of
its parts (although the system can exist independently of particular parts).
An interesting sort of regress characterizes both causation and constitution.

In the case of causation, we talk about chains of causation. This is based
on the idea that for every event that is a cause, there is another event that
is its cause. A similar idea applies to constitution; we assume that all parts
can be further decomposed into their parts and their organization. We could
call these chains of constitution. Now a tricky question is whether there
exists a first cause that is not itself caused, and a similar problem can be
stated concerning the ultimate building blocks of reality, but in this context
we can leave them aside. There is no danger that such ultimate things will
show up in the social sciences. However, these regress properties create
chains of explanations, which are relevant from the point of view of the social
sciences. The crucial thing in this context is to understand that although
there is always an explanation for every social scientific explanatory factor,
this does not imply that their explanatory status depends on us knowing
the explanation for them. Both in the case of causation and constitution, an
explanation presupposes that the explanans facts are the case, not that we
have to have an explanation for those facts. I will return to this issue in the
next section.

Explanation is about tracking relations of dependence. Although
metaphysically the relations of constitution and causation are quite
different, in terms of explanation the basic principles are quite similar. Both
explanations attempt to track networks of counterfactual dependence. A
causal explanation tells us how the antecedent events and their organization
(timing and location) bring about the event to be explained. In contrast, a
constitutive explanation describes how the properties of the components and
their organization give rise to the system’s properties.
In both cases we are looking for the difference-makers: The criterion of

explanatory selection is counterfactual. As the precise explanandum is best
characterized in contrastive terms (why x is the case rather than x*), we
are interested in the differences that would have made the difference we
are interested in (Woodward 2003; Ylikoski 2007; Northcott this volume).
In the case of causation these differences are in antecedent events; in
the case of constitution these differences are in the properties of parts (or
in their organization). Also in both cases it makes sense to ask a further
question: Why does the counterfactual dependence hold? The answers to
these questions will in both cases draw from the same body of mechanical
knowledge, so it is understandable that in the philosophy of biology debates
both explanations are called mechanical explanations. So, despite the
important metaphysical differences, the same basic ideas about explanation
can be applied to both cases.
Not only are the principles of explanatory relevance similar, so are the
explanatory questions. This leads easily to confusion. Consider the question:
“Why is this glass fragile?” The question is ambivalent: It could either mean
“How did the glass become fragile?” or it could mean “What makes the glass
fragile?” The first question is causal; the latter question constitutive. The
answer to the causal question will tell us about the causal history of the
glass—it will specify the crucial features of the process that led to the object
being fragile rather than robust. The answer to the constitutive question will
not focus on earlier events. It will detail the relevant aspects of the object’s
molecular structure that makes it fragile. So while the explanation-seeking
questions may look the same, the request for explanatory information is
quite different. Without a clear understanding of the differences between
causation and constitution, some confusion is bound to occur. This is also
the case in philosophy of social sciences. For example, it is quite a different
thing to explain how a regime became stable than to explain what makes
it stable. While some of the facts cited by both explanations might be the
same, they are addressing different explananda: One is focused on how

the causal capacity was acquired and the other on the basis of that causal
capacity. A social scientist is usually interested in both questions, but she
should not confuse them with each other.
For all social macro properties, one can ask both constitutive and causal why-
and how-questions. (Although for some statistical properties the constitutive
questions are relatively trivial.) The first sort of questions asks how the
macro properties are constituted by the micro-level entities, activities, and
relations. The aim is to track how the details of macro-level facts depend
on the micro details. The question is often how the macro facts would have
been different if some of the micro facts had been different in some specific
way. These questions can also be characterized in terms of interventions:
How would the macro facts change if some of the micro facts were changed?
Notice that here intervention is a causal notion (all change happens in time),
but the dependence of interest is constitutive.
A clear example of constitutive explanation is an explanation for the

difference in the problem-solving capacities of two groups. The crucial
difference might be in the properties of the members, such as their
intelligence or social skills. Alternatively, the pivotal factors might be the
informal social norms that characterize the interactions within the group or
its formal organization. Of course, the explanation may also be found in some
combination of these factors. Just like in this example, the usual explananda
of constitutive explanations are causal capacities and dispositions of the
whole. The constitutive explanation tells us what gives the whole (population,
group, organization, or society) those properties, and the answer is found in
the causal capacities of the parts and their organization.
The explanantia in constitutive explanations are always at the micro level.

As the explanation attempts to capture what the whole is made of, an appeal
to the properties of the whole does not really make sense. In this sense, the
methodological individualists, and other reductionists, have been on the
right track. On the other hand, the explanation of macro properties does not
in any way diminish their reality: The wholes are as real as their parts. This
implies that those methodological individualists who have suggested that
a micro explanation somehow eliminates the macro properties are either
metaphysically confused or just choosing their words badly. The talk about
macro reducing to micro makes as little sense as the talk about reducing
effects to their causes.
The causal questions about the macro social properties are concerned
with their origin, persistence, and change. These explanations are tracking

counterfactual dependencies between events. How would have the outcome
been different if some of the causes had been different in some specified
manner? What kind of difference would an intervention on some antecedent
facts make? The explanantia in these causal explanations are always
antecedent events.
This is the context in which confusion between constitution and causation

can create trouble. If we are considering simple causal statements about
causal dependence, individualists tend to make the claim that the causes
have to be at the micro level. However, nothing in the notion of causation
implies that the real causal work is always to be found at the micro level. Of
course, the notion of constitution implies that every time we have a cause
at a macro level, we also have micro level facts that constitute it. If we stick
to the counterfactual criterion of explanatory selection, as I think we should,
there is no a priori reason to privilege micro-level causes (Woodward 2003,
2008). It is sufficient that there is an appropriate counterfactual dependence
between the macro variable and the explanandum. Of course, in many cases
the justification of a claim about this causal dependence might require some
knowledge of the underlying mechanisms. However, this observation about
the justification of a causal claim should not be confused with the claim itself.
Similarly, although adding mechanistic details to the explanation will involve
references to micro-level processes, this does not imply that the macro facts
will lose their explanatory relevance. They will still be possible difference-
makers and legitimate explanatory factors. In other words, although the
information about the relevant mechanistic details improves the explanation
significantly, it does not remove the causal relevance of the initial invariance
involving macro-level facts.
In the counterfactual account of causal relevance, the location of explanatory

relevance at the micro or macro level is a contingent matter that depends
on the explananda that one is addressing. There is no reason to assume
that the most invariant counterfactual dependence (with respect to the
contrastively specified explanandum) will always be found at the micro level.
Similarly, one has to give up the often presented suggestion that levels of
explanation should match so that macro would always explain macro and
micro would always explain micro. The issues of explanatory relevance
(how the explanatory factors are selected, at which level of abstraction
they are described, etc.) are always determined by the facts of the case
and the details of the intended explanandum, not by generic philosophical
arguments.

2.4.1. The Proper Role of Microfoundations
Is the above argument about the legitimacy of macro-level causal facts

compatible with the mechanistic call for microfoundations? I want to
argue that it is fully compatible with the core ideas of mechanism-based
thinking. Contrary to the common assumption, the point of mechanistic
microfoundations is not that we have more real causes at the micro level,
but to have a better grasp of the explanatory dependence underlying the
causal relation involving macro variables. Consequently, the advocates of
mechanism-based explanations should not call into question the reality of
macro-level causal relations. Instead, they should emphasize the importance
of microfoundations for understanding these dependencies. There are a
number of reasons why microfoundations are important.
First, all causal relations involving macro properties are mechanism-

mediated causal relations. Understanding how the dependence involving
macro variables is constituted helps to understand why that particular
dependence holds (Ylikoski 2011). It also integrates the piece of causal
information contained in the macro-level generalization to other pieces of
explanatory knowledge (Ylikoski and Kuorikoski 2010). This is certainly a form
of explanatory understanding that we should be interested in if we take the
notion of explanatory social science seriously.
However, the utility of this information is not limited to the expanded

theoretical understanding. It also often tells about the conditions under which
the causal dependence in question will hold. There are three dimensions
to this knowledge. First, there is knowledge about the range of values
of the explanandum variable that are possible without the dependence
breaking apart. Second, there is knowledge about the sensitivity of the
dependence to changes in background conditions. Finally, there is possible
knowledge about alternative interventions that could bring about similar
effects. Without knowledge of these issues, the explanatory use of the
macro-level explanatory generalization can be very risky business. It is very
difficult to extrapolate to other cases without understanding the background
mechanisms (Ylikoski 2011; see also Cartwright, this volume, Kincaid, this
volume).
Apart from an expanded understanding and the security of an explanatory

claim, the insight into the underlying mechanisms might also help to
improve the explanatory generalization. With the help of a mechanistic
understanding, one might be able to make the explanandum more precise or

to reformulate the explanatory generalization in such a manner that it allows
a broader range of values of the explanandum variables or background
conditions (Ylikoski 2011).
Figure 2.1 Macro-Micro Linkages
These considerations justify the presumption that microfoundations are

important for proper explanatory understanding. However, they do not
demolish the explanatory relevance of macro facts. On the contrary, they put
them in the right context as the mechanisms bridge the large-scale micro
facts to causal interactions between persons and to their decision-making
processes. I think this is the point James Coleman (1990) attempted to make
with his often misunderstood graph.
Following Hedström and Swedberg (1998, 23), I refer to the arrows in figure
2.1 as situational mechanisms (arrow 1), action-formation mechanisms
(arrow 2), and transformational mechanisms (arrow 3). The situational
mechanisms describe how social structures constrain individuals’ actions
and cultural environments shape their desires and beliefs, the action-
formation mechanisms describe how individuals choose their preferred
courses of action among the feasible alternatives, and the transformational
mechanisms describe how individual actions produce various intended and
unintended social outcomes.
Coleman was critical of nonmechanistic explanations that remain at the

level of macro regularities. However, there is no reason to assume that he
was denying the causal relevance of macro social facts. Rather, his point
was to make it clear that proper sociological understanding requires that we
understand both the mechanisms by which large-scale social facts influence
the local decision-making processes of individual agents (the situational
mechanisms) and the mechanisms by which individual actions create and
influence macro social facts (the transformational mechanisms). He was
calling for mechanisms that bridge the levels, not just descriptions that
somehow reduce the macro facts to individual level facts. Only when we

understand the relevant mechanisms, do we have a satisfactory theoretical
grasp of the social phenomena in question.
Coleman’s criticism of Weber’s (partial) explanation of the emergence of

modern capitalism in Western Europe illuminates these points. Weber started
with an idea that was commonplace in late nineteenth-century Germany:
There is a close connection between Protestantism, entrepreneurism, and the
rise of capitalism. To substantiate this vague explanatory suggestion, Weber
asked what changes the emergence of Protestantism brought about in the
beliefs, desires, and communal practices of individual agents. This question
has both causal and constitutive dimensions that are not clear in Coleman’s
analysis. However, Coleman’s focus is on Weber’s second causal question:
How did these changed life practices of individuals influence economic
activities and institutions and how did these changes in turn facilitate the
formation of modern capitalism? Coleman’s central point was that Weber
was not clear enough about this last passage of the causal chain. He was not
able to give a sufficiently clear account of the transformative mechanisms
that connected the Protestant ethic to the rise of modern capitalism. In other
words, Weber was not able to show how the changes at the micro level (the
life practices of Protestants) bought about a major macro-level outcome (the
early forms of modern capitalism). As the crucial mechanism is lacking, so is
the legitimacy of Weber’s causal claim about history.
Here it is important to see the difference between the justificatory and

explanatory roles of mechanisms. Coleman’s analysis shows why it is
legitimate to challenge Weber’s causal claim. Knowledge of the causal
mechanisms have an important role in the justification of historical causal
claims, so pointing to the missing details of the causal chain constitutes a
challenge to the legitimacy of the causal claim. However, this criticism of
a singular causal claim does not imply that Coleman generally considers
macro-level facts to be nonexplanatory or causally impotent. He is simply
challenging the justification of this particular historical hypothesis.
2.5. Intentional Fundamentalism
Arguments for methodological individualism often appeal to the special

explanatory status of intentional explanations. I call this position intentional
fundamentalism. According to intentional fundamentalism, the proper
level of explanation in the social sciences is the level of the intentional
action of individual agents. The intentional fundamentalist assumes that
explanations given at the level of individual action are especially satisfactory,
fundamental, or even ultimate. In contrast to explanations that refer to

supra-individual social structures, properties, or mechanisms, there is
no need to provide microfoundations for intentional explanations. They
provide rock-bottom explanations. In other words, according to intentional
fundamentalism, the intentional explanations of individual actions are
privileged explanations.
Although intentional fundamentalism can take various forms, it is often

related to rational choice theory. French social theorist Raymond Boudon
(1998, 177) expresses the idea clearly: “When a sociological phenomenon is
made the outcome of individual reasons, one does not need to ask further
questions.” The idea is that in the case of supra-individual explanations
there is always a black box that has to be opened before the explanation
is acceptable, but in the case of intentional explanation there is no such a
problem: “The explanation is final” (Boudon 1998, 172). Diego Gambetta
appeals to same sort of finality (1998, 104): “Not only will a rational choice
explanation be parsimonious and generalizable; it will also be the end of the
story.”2
My claim in this section is that intentional fundamentalism is not compatible

with the causal mechanistic account of explanation. As intentional
fundamentalism is often advocated by rational choice theorists and as
many believe that rational choice explanations are the best examples of
mechanical explanations in the social sciences, this incompatibility claim is of
some interest. If my argument is valid, it suggests that the relation of rational
choice theory and a mechanism-based philosophy of science requires some
rethinking. It also implies that one common argument for methodological
individualism is much less credible than is commonly assumed.
2.5.1. The Regress Argument
To make sense of intentional fundamentalism, we should start with

the explanatory regress argument for methodological individualism.
Methodological individualists often make the case that nonindividualist
explanations are either explanatorily deficient or not explanatory at all.
At most, they allow that explanations referring to macro social facts are
placeholders for proper (individualistic) explanatory factors. In this view,
the explanatory contribution of supra-individual explanations is at best
derived: They are explanatory because they are (in principle) backed up
by a truly explanatory story. This is the regress of explanations argument:
Unless grounded at the lower level, explanations at the macro level are not
acceptable. The underlying general principle is the following:

[P] A genuine explanation requires that the explanans is itself
explained or is self-explanatory.
In short, the explanatory buck has to stop somewhere.
The principle [P] is general, and it raises the possibility of an explanatory

regress that is only halted at a fundamental (physical) level. This would be
highly unintuitive, so for the intentional fundamentalist the buck stops at the
level of (self-interested) rational intentional action. This level is treated as
inherently understandable, as shown in the above quotations from Boudon.
The inherent intelligibility of intentional action explains why the search for
microfoundations should stop at the level of the individual. The special status
of intentional explanation also makes the explanatory regress argument
safe for the methodological individualist: He can use the argument’s full
force against anti-individualists who cannot make a similar claim about a
privileged status, and it does not challenge the legitimacy of his favored
explanatory factors.
The fundamentalist argument for individualism fails for a number of reasons.

The first reason is that the principle [P] is not valid. The explanatory relation
between the explanans and the explanandum is independent from the
question of whether the explanans is itself explained. An explanation of X in
terms of Y presupposes that Y is the case, but it does not presuppose that Y
is itself explained. Of course, it would be great also to have an explanation
for Y, but this is a separate issue from the legitimacy of the explanatory
relationship between Y and X. The distinctness of these issues implies that
the regress does not begin.
Why whould anyone believe in [P]? One plausible suggestion is the following:
The belief in [P] arises from a straightforward confusion between justification-
seeking and explanation-seeking why-questions. It makes sense to ask how
well justified are those things that one appeals to in justification of one’s
beliefs. It also makes sense to ask whether one is justified in believing the
things that one appeals to in one’s explanation. However, justifying one’s
belief in Y is not the same as explaining why Y is the case.
2.5.2. Intentional Explanations without a Special Status
Another reason for the failure of the regress argument is that intentional
explanations lack the special properties assumed by the argument. If
one accepts the mechanistic account of explanation, as many advocates
of rational choice sociology do, such a special status does not make

any sense. The assumption that human deliberation is a black box that
should not be opened is more in line with nineteenth-century hermeneutic
romanticism than with causally oriented social science. Of course, the chain
of mechanistic explanations will end somewhere (if there is such a thing as
a fundamental level), but that stopping point is not the level of individual
rational action.
A mechanistic explanation appeals to micro-level processes, but nothing in

the notion of mechanistic explanation implies that these micro things would
always be facts about the intentional actions of individuals. Mechanisms that
cite supra-individual entities or properties are certainly possible (Mayntz
2004). For example, various filtering mechanisms that are analogical to
natural selection are difficult to understand other than as population-wide
processes, and when the units that are selected are organizations (for
example, firms), it is natural to conceive the mechanism as supra-individual.
Similarly, the crucial parts of the explanatory mechanism could well be
located below the level of intentional psychology. For example, various
facts about human information processing—for example, implicit biases
(see Kelly and Mallon, this volume)—could well be relevant for explanatory
understanding of intentional action. There is no valid reason to give up
mechanistic thinking in the case of intentional action.
Another reason to challenge intentional fundamentalism is the implicit

realism of mechanistic thinking. For mechanists, explanation is factive.
It is not enough that the explanation saves the phenomenon: It should
also represent the essential features of the actual causal structure that
produces the observed phenomena. So, if the explanation refers to the
goals, preferences, or beliefs of agents, the agents should indeed have those
mental states. Mere as-if storytelling does not suffice for a mechanistic
explanation as it does not capture the relevant parts of the causal process.
This realist attitude goes against the instrumentalist attitude common
among many rational choice theorists. The fact that one can rationalize
any behavior does not imply that those rationalizations are also the correct
causal explanations for those behaviors. Similarly, the human fluency in
coming up with intentional accounts for our behavior is not a reason for
regarding them as superior explanations.
It is important to understand the limited nature of my argument. I am not

denying that intentional explanations are, and will be, an indispensable
part of the social scientific explanatory repertoire. For me, intentional
explanations are legitimate causal explanations. Furthermore, the intentional

attitudes of individuals play an important role in most mechanism-based
explanations of social phenomena. The only thing I am challenging is the
supposed special explanatory status of intentional or rational accounts of
human action. In the mechanistic account of explanation, the importance of
certain sorts of explanatory factors is not a basis for their privileged status.
Neither should my rejection of intentional fundamentalism be regarded as

a wholesale attack on the use of rational choice theory. For many social
scientific purposes, a rather simple version of intentional psychology is
both preferable and sufficient. For example, when one is attempting to
make sense of social complexity, it is understandable that social scientists
attempt to keep the psychological assumptions of their models very
simple. Such idealizations are fully legitimate if they do not lead to a
gross misrepresentation of the causal mechanism under consideration.
However, the practical necessity of these idealizations does not constitute a
justification for accepting intentional fundamentalism.
Furthermore, my argument should not be regarded as an argument against

the claim that there should exist a division of labor between the social
sciences and the sciences of cognition. However, it follows from the flexibility
of mechanistic levels that the boundaries of this division of labor are
adjustable and not fixed. It is inherent in the idea of mechanistic explanation
that all the gaps between levels of analysis are ultimately to be bridged by
mechanistic interfield theories. So the challenge for the social sciences is not
to define their objects of study in such a way that they are in no way touched
by psychological sciences, but to look at ways in which social and cognitive
mechanisms can be meaningfully combined. This is not as easy as it sounds,
as recent attempts to combine neuroscience and economics show (Kuorikoski
and Ylikoski 2010).
2.6. Conclusions
In this chapter, I have attempted to show what consequences the

mechanism-based account of explanation would have on issues traditionally
discussed under the title of methodological individualism. Borrowing
some ideas developed by philosophers who have studied the mechanistic
explanation in the biological sciences, I have argued that we should give up
the notion of a unique, privileged, and comprehensive individual level that
has been a presupposition of the individualism debates. In addition, I have
argued that rather than employing metaphors borrowed from the philosophy
of mind for micro-macro relations, we should pay closer attention to how
real macro social facts figure in social scientific theories and explanations.

There the micro-macro issue is more an issue of bridging large-scale social
facts to small-scale social interactions rather than that of finding a way to
see relations between autonomous levels of reality.
References
Bechtel, William. 2006. Discovering Cell Mechanisms. The Creation of Modern

Cell Biology. New York: Cambridge University Press.
Find This Resource
• Find it in your Library
• Worldcat
• Google Preview
Bechtel, William. 2008. Mental Mechanism. Philosophical Perspectives on

Cognitive Neuroscience. London: Routledge.
Find This Resource
• Worldcat
• Google Preview
Boudon, Raymond. 1998. “Social Mechanisms Without Black Boxes.” In Social

Mechanisms: An Analytical Approach to Social Theory, Peter Hedström and
Richard Swedberg, eds., 172–203. Cambridge: Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Burt, Ronald. 1992. Structural Holes: The Social Structure of Competition.

Cambridge, MA: Harvard University Press.
Find This Resource
• Worldcat
• Google Preview
Coleman, James. 1990. Foundations of Social Theory. Cambridge, MA: The

Belknap Press.
Find This Resource
• Worldcat
• Google Preview

Craver, Carl. 2007. Explaining the Brain: Mechanisms and the Mosaic Unity of
Neuroscience. Oxford: Clarendon Press.
Find This Resource
• Worldcat
• Google Preview
Cummins, Robert. 1983. The Nature of Psychological Explanation.

Cambridge, MA: Bradford/The MIT Press.
Find This Resource
• Worldcat
• Google Preview
Darden, Lindley. 2006. Reasoning in Biological Discoveries. Essays on

Mechanisms, Interfifield Relations, and Anomaly Resolution. Cambridge:
Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Elster, Jon. 1989. Nuts and Bolts for the Social Sciences. Cambridge:
Find This Resource
• Worldcat
• Google Preview
Elster, Jon. 2007. Explaining Social Behavior. More Nuts and Bolts for the
Social Sciences. Cambridge: Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Gambetta, Diego. 1998. “Concatenations of Mechanisms.” In Social

Mechanisms: An Analytical Approach to Social Theory, Peter Hedström and
Richard Swedberg, eds., 102–24. Cambridge: Cambridge University Press.
Find This Resource
• Worldcat

• Google Preview
Harré, Rom. 1970. The Principles of Scientific Thinking. London: Macmillan.

Find This Resource
• Worldcat
• Google Preview
Hedström, Peter. 2005. Dissecting the Social. On the Principles of Analytical

Sociology. Cambridge: Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Hedström, Peter, and Peter Bearman, eds. 2009. The Oxford Handbook of
Analytical Sociology. Oxford: Oxford University Press.
Find This Resource
• Worldcat
• Google Preview
Hedström, Peter, and Richard Swedberg. 1998. “Social Mechanisms: An

Introductory Essay.” In Social Mechanisms: An Analytical Approach to Social
Theory, Peter Hedström and Richard Swedberg, eds., 1–31. Cambridge:
Find This Resource
• Worldcat
• Google Preview
Hedström, Peter, and Petri Ylikoski. 2010. “Causal Mechanisms in the Social
Sciences.” Annual Review of Sociology 36: 49–67.
Find This Resource
• Worldcat
• Google Preview
Hempel, Carl. 1965. Aspects of Scientific Explanation. New York: The Free
Press.
Find This Resource

• Worldcat
• Google Preview
Horgan, Terence. 1993. “From Supervenience to Superdupervenience:

Meeting the Demands of a Material World.” Mind 102: 555–86.
Find This Resource
• Worldcat
• Google Preview
Kim, Jaegwon. 1993. Supervenience and Mind. Cambridge: Cambridge

University Press.
Find This Resource
• Worldcat
• Google Preview
Kim, Jaegwon. 1998. Mind in a Physical World. Cambridge, MA: The MIT Press.
Find This Resource
• Worldcat
• Google Preview
Kincaid, Harold. 1996. Philosophical Foundations of the Social Sciences.

Analyzing Controversies in Social Research. Cambridge: Cambridge
University Press.
Find This Resource
• Worldcat
• Google Preview
Kuorikoski, Jaakko. 2009. “Two Concepts of Mechanism: Componential Causal

System and Abstract Form of Interaction.” International Studies in the
Philosophy of Science 23 (2): 143–60.
Find This Resource
• Worldcat
• Google Preview
Kuorikoski, Jaakko, and Petri Ylikoski. 2010. “Explanatory Relevance Across

Disciplinary Boundaries—The Case of Neuroeconomics.” Journal of Economic
Methodology 17 (2): 219–28.

Find This Resource
• Worldcat
• Google Preview
Little, Daniel. 1991. Varieties of Social Explanation: An Introduction to the

Philosophy of Social Science. Boulder, CO: Westview Press.
Find This Resource
• Worldcat
• Google Preview
Mayntz, Renate. 2004. “Mechanisms in the Analysis of Social Macro-

Phenomena.” Philosophy of the Social Sciences 34 (2): 237–59.
Find This Resource
• Worldcat
• Google Preview
McCauley, Robert. 2007. “Reduction: Models of Cross-Scientific Relations

and Their Implications for the Psychology-Neuroscience Interface.” In The
Handbook of Philosophy of Science. Philosophy of Psychology and Cognitive
Science, Paul Thagard, ed., 105–58. Amsterdam: North Holland.
Find This Resource
• Worldcat
• Google Preview
Oppenheim, Paul, and Hilary Putnam. 1958. “Unity of Science as a Working

Hypothesis.” In Concepts, Theories, and the Mind—Body Problem, Minnesota
Studies in the Philosophy of Science II, Herbert Feigl, Michael Scriven, and
Grover Maxwell, eds., 3–36. Minneapolis: University of Minnesota Press.
Find This Resource
• Worldcat
• Google Preview
Richardson, Robert. 2007. “Reduction without the Structures.” In The

Matter of the Mind. Philosophical Essays on Psychology, Neuroscience, and
Reduction, Maurice Schouten and Huib Looren de Jong, eds., 123–45. Oxford:
Blackwell.
Find This Resource

• Worldcat
• Google Preview
Salmon, Wesley. 1984. Scientific Explanation and the Causal Structure of the
World. Princeton, NJ: Princeton University Press.
Find This Resource
• Worldcat
• Google Preview
Sawyer, R. Keith. 2005. Social Emergence. Societies as Complex Systems.

Cambridge: Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Scott, John. 2000. Social Network Analysis. A Handbook. 2nd ed. London:
Sage.
Find This Resource
• Worldcat
• Google Preview
Sperber, Dan. 2006. Explaining Culture: A Naturalistic Approach. Oxford:

Blackwell.
Find This Resource
• Worldcat
• Google Preview
Thagard Paul. 1999. How Scientists Explain Disease. Princeton, NJ: Princeton
University Press.
Find This Resource
• Worldcat
• Google Preview
Turner, Stephen. 1994. The Social Theory of Practices. Tradition, Tacit

Knowledge and Presupposition. Cambridge: Polity Press.
Find This Resource

• Worldcat
• Google Preview
Udéhn, Lars. 2001. Methodological Individualism: Background, History and

Meaning. London: Routledge.
Find This Resource
• Worldcat
• Google Preview
Wendt, Alexander. 1998. “On Constitution and Causation in International

Relations.” Review of International Studies 24 (5): 101–17.
Find This Resource
• Worldcat
• Google Preview
Wimsatt, William. 2007. Re-Engineering Philosophy for Limited Beings.

Piecewise Approximations to Reality. Cambridge, MA: Harvard University
Press.
Find This Resource
• Worldcat
• Google Preview
Woodward, James. 2002. “What Is a Mechanism? A Counterfactual Account.”

Philosophy of Science 69 (3): S366–S377.
Find This Resource
• Worldcat
• Google Preview
Woodward, James. 2003. Making Things Happen. A Theory of Causal

Explanation. Oxford: Oxford University Press.
Find This Resource
• Worldcat
• Google Preview

Woodward, James. 2008. “Mental Causation and Neural Mechanisms. In Being
Reduced.” Essays on Reduction, Explanation, and Causation, eds. J. Hohwy
and J. Kallestrup, 218–62. Oxford: Oxford University Press.
Find This Resource
• Worldcat
• Google Preview
Wright, Cory, and William Bechtel. 2007. “Mechanisms and Psychological

Explanation.” In The Handbook of Philosophy of Science. Philosophy of
Psychology and Cognitive Science, ed. Paul Thagard, 31–79. Amsterdam:
North Holland.
Find This Resource
• Worldcat
• Google Preview
Ylikoski, Petri. 2003. “Explaining Practices.” Protosociology 18: 316–30.

Find This Resource
• Worldcat
• Google Preview
Ylikoski, Petri. 2007. “The Idea of Contrastive Explanandum.” In Rethinking

Explanation, eds. Johannes Persson and Petri Ylikoski, 27–42. Dordrecht:
Springer.
Find This Resource
• Worldcat
• Google Preview
Ylikoski, Petri. 2011. “Social Mechanisms and Explanatory Relevance.” In

From Social Mechanisms to Analytical Sociology, ed. Pierre Demeulenaere,
154–72. Cambridge: Cambridge University Press.
Find This Resource
• Worldcat
• Google Preview
Ylikoski, Petri, and Jaakko Kuorikoski. 2010. “Dissecting Explanatory Power.”

Philosophical Studies 148 (2): 201–19.
Find This Resource

• Worldcat
• Google Preview
Zahle, Julie. 2006. “Holism and Supervenience.” In The Handbook of

Philosophy of Science. Philosophy of Anthropology and Sociology, eds.
Stephen P. Turner and Mark Risjord, 311–41. Amsterdam: North Holland.
Find This Resource
• Worldcat
• Google Preview
Notes:
(1) . There are some exceptions. For example, Wendt (1998) distinguishes
between causation and constitution. However, his discussion of constitution
is very confused. His notion of constitution covers not only the constitution
of causal capacities, but also causal preconditions, definitions, and other
conceptual relations. The standard philosophy of science notion that I am
using is limited only to the constitution of causal capacities.
(2) . The key issue here is not whether these authors would ultimately
subscribe to intentional fundamentalism. I am only claiming that in these
passages they argue as if intentional fundamentalism is correct.


The Oxford Handbook of Philosophy of Social Science 2012

Uploaded by

Copyright:

Available Formats

The Oxford Handbook of Philosophy of Social Science 2012

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Oxford Handbook of Philosophy of Social Science 2012

Uploaded by

Copyright:

Available Formats

The Oxford Handbook of Philosophy of Social Science

Print publication date: Nov 2012

Micro, Macro, and Mechanisms

Abstract and Keywords

social sciences, mechanism-based explanation, intentional fundamentalism, causal process,

This chapter takes a fresh look at micro-macro relations in the social

Page 1 of 36 Micro, Macro, and Mechanisms

2.2. Mechanism-based Explanation

The idea of mechanism-based explanation has been developed

Page 2 of 36 Micro, Macro, and Mechanisms

There is no consensus on the right definition of a causal mechanism.

First, a mechanism is always a mechanism for something; it is identified

Page 3 of 36 Micro, Macro, and Mechanisms

Although the mechanism-based account is often presented simply as an

A mechanism-based explanation describes the causal process selectively.

Page 4 of 36 Micro, Macro, and Mechanisms

2.2.1. Mechanisms and Reductive Explanation

One of the distinctive features of the mechanistic approach to explanation is

The mechanistic account of reductive explanation differs significantly from

Page 5 of 36 Micro, Macro, and Mechanisms

The notion of the levels of mechanism plays an important role in the

Page 6 of 36 Micro, Macro, and Mechanisms

The advocates of the mechanism-based approach in the social sciences have

The aim of this chapter is to sketch what a consistently mechanistic way to

Page 7 of 36 Micro, Macro, and Mechanisms

2.3. Rethinking the Macro

A popular argumentative strategy among anti-individualists has been

The mind-brain analogy is inappropriate because it mischaracterizes the

The problem in the social sciences is not that of bridging a comprehensive

Page 8 of 36 Micro, Macro, and Mechanisms

Another salient feature of many social micro-macro relations is the part-

I do not propose that we can simply define the micro-macro contrast as an

Thinking of the micro-macro issue as an issue of scale makes it possible

Page 9 of 36 Micro, Macro, and Mechanisms

Another consequence is that whether an attribute is a macro or micro

From the point of view of a mechanistic philosophy of science, this

The social macro properties do not constitute a unified kind, so it makes

2.3.1. Statistical Properties of a Population

A major concern for sociology is the various statistical attributes of

Page 10 of 36 Micro, Macro, and Mechanisms

While the statistical properties of populations usually only serve as

2.3.2. Topologies of Social Networks within a Population

Sociologists are also interested in relations and interactions between

Page 11 of 36 Micro, Macro, and Mechanisms

The importance of social networks is increasingly being recognized in

Page 12 of 36 Micro, Macro, and Mechanisms

By communal properties I refer to social scientific notions that apply to

While communal properties are attributed to groups, they are quite

Descriptions of customs, social norms, and cultures are always based

Although communal properties, as I have described them, are tied to a

Page 13 of 36 Micro, Macro, and Mechanisms

When we consider communal properties as idealizing abstractions from

2.3.4. Organizations and Their Properties

Organizations such as states, firms, parties, churches, and sport clubs

Page 14 of 36 Micro, Macro, and Mechanisms

When a person causally interacts with an organization, she interacts with

Again it is good to return to real social scientific questions. They concern

Page 15 of 36 Micro, Macro, and Mechanisms

2.4. Causation, Constitution, and Microfoundations