1 Introduction

Social and political theorists often talk of social structures, the positions of individuals within social structures, the effects of social structures. Common examples of social structures include, on the one hand, broad and deep social phenomena such as the wage-labor system of industrial capitalism and the heteronormative and bionormative nuclear family, and on the other hand, much more local and flexible phenomena such as the social structure of a particular institution such as a school, church, or business. Needless to say, sociologists have a multiplicity of competing theories about social structures answering different questions. However, many, including many philosophers, fail to notice social structures, and when structures are mentioned, will often find the idea mysterious. Their attention, including (and perhaps especially) their moral attention, focuses on individuals.

My goal in this essay is to develop a philosophically useful account of social structure, drawing on recent work in the social sciences. By “philosophically useful” account, I mean an account that will enable us to integrate attention to social phenomena more effectively in philosophical discussion. This will include sketching a social ontology that connects social structures with structural explanation.

Social structures are theoretical entities, postulated to do work in a social theory. There are a variety of constraints on what they can be, given the work they are introduced to do. For example:

a) They are invoked in structural explanations;

b) They enable us to identify and critique structural injustice;

c) They provide the context for human agency;

d) They are, in some sense and to some degree, constituted by relations between individuals.

My focus in this paper is on (a). The strategy is this. I’ll begin by sketching an account of structural explanation. I’ll then sketch an account of social practices and social structures. The question, then, is whether these two accounts mesh. If the account of structural explanation is plausible, does the account of social structures I provide give us an understanding of how social structural explanation works?

2 Questions and the objects of explanation

Suppose I am playing ball with my dog. I stuff a treat into a hole in the ball and throw it for him. The ball goes over the lip of a hill and rolls down into a gully. Why did the treat end up in the gully? If we imagine the trajectory of the treat alone, from a space near my hand, through an arc the air, then landing about an inch above the ground and moving at about that height down the hill until it stops, it would be a huge task to explain the particular events that determined each of its movements. A much easier explanation would be to point out that the treat was inserted into a ball that was thrown and rolled down the hill into the gully. In this latter explanation, we explain the behavior of the treat by its being part of something larger whose behavior we explain.

The explanation of the treat ending up in the gully by reference to the ball is not an inferior explanation to the more detailed one. Rather, it is in many ways a better explanation. For example, the explanation is more stable: although tracing the particular causal history of the treat’s movements down the hill gives us lots of information about that particular event, pointing to the treat’s being in the ball enables us to understand why the treat would have landed in the gully even if I had thrown the ball higher, or with a curve or spin, even if the hill had been a bit less steep, or the wind stronger. The treat would have landed there even if it had been a different ball, made of a different material. Moreover, insofar as explanations help us engage the world, the explanation relying on the ball’s movements is more useful: it provides a better model for seeing how I could intervene to prevent the treat from ending up in the gully (not throw the ball in that direction, for example, or catch up with the ball and stop it from rolling).

This simple example is intended to show that sometimes it is good and useful to explain the behavior of a thing by explaining the behavior of something of which it is a part, if it is a part whose behavior is constrained by other parts of the whole. If I had simply thrown a handful of treats in the direction of the gully, the fact that the treat in question was part of a handful of treats would do little or nothing to explain its movement because the handful is just an aggregate, not a structured whole. The very idea of a structured whole is of something whose parts are interdependent. Different structures will have different effects on their parts. If I had inserted the treat into a cube, it is very likely that the treat would not have ended up in the gully.

Although structural explanations occur in many different domains, my concern is with social structural explanations. Garfinkel (1981) offers a simple example to illustrate how structure can be not only explanatory, but serve in the best explanation:

Suppose that, in a class I am teaching, I announce that the course will be “graded on a curve,” that is, that I have decided beforehand what the overall distribution of grades is going to be. Let us say, for the sake of the example, that I decide that there will be one A, 24 Bs, and 25 Cs. The finals come in, and let us say Mary gets the A. She wrote an original and thoughtful final. (Garfinkel 1981, p. 41)

Garfinkel argues that in this case, when we ask, “Why did Mary get an A?” the answer: “She wrote an original and thoughtful final” is inadequate. Why? Because in order to earn the single A, one would have to write the best final. If the instructor had not decided to grade on a curve, many students could have earned As by virtue of writing thoughtful and original finals. Garfinkel says, “So it is more accurate to answer the question by pointing to the relative fact that Mary wrote the best paper in the class” (Garfinkel 1981, p. 41). Mary earned the A not simply by virtue of her performance, but by virtue of her performance in comparison with others and the particular grading structure that made the comparison the key factor in determining who earned the A.

What is the role of the grading structure in this explanation? On Garfinkel’s view, explanations are answers to questions, and questions establish a contrast space for possible answers.Footnote 1 In the case of Mary’s A in the class, there are several different questions, with different foci and contrast spaces. In uttering:

  1. 1)

    Why did Mary get an A?

We might be asking,

  1. 2)

    Why did Mary get an A (as opposed to a B or C)?

or we might be asking,

  1. 3)

    Why did Mary (as opposed to Bob or Susan) get an A?

As soon as we begin to list the contrasts, or ‘foils,’ it is clear that there are background assumptions that limit range of suitable answers. An adequate explanation must find a relevant difference between the focus and the foils (Lipton 1991). If, instead of (3) we were asking:

  1. 4)

    Why did Mary (as opposed to George Bush or Barack Obama) get an A?

An adequate response might be that Mary was registered in the class and took the exam, whereas neither Bush nor Obama were.

Garfinkel assumes that the correct answer to (1) must take into account the background facts about the curve and claims that “there is, therefore, an unexplained presupposition that there is exactly one A in the class” (Garfinkel 1981, p. 42).Footnote 2 On his view, a better representation of the question would be:

  1. 5)

    Given that the instructor is evaluating all and only the students in the class on an A–B–C grading system with A the highest and C the lowest and a curve that allows only one A, why did Mary get an A?

To this question, it is clear that “Because she wrote a thoughtful and original final,” is inadequate, regardless of whether the contrast is between Mary and others in the class or between Mary’s earning an A or a different grade. The correct answer is, “Because Mary wrote the best final in the class.”

On Garfinkel’s account, the information included in the ‘given’ clause can serve two different functions. First, it excludes a set of options that are not being considered relevant (in response to (2) only grades are relevant; in the case of (3) only the participants in the class who took the course for credit are relevant). Second, it can structure the remaining options. In the final exam example, the fact that the exam was graded on a curve structures the grading options so that only a certain distribution across the options is possible (1 A, 24 Bs, 25 Cs). The limitation to these distributions is a feature of the world, not just a feature of the conversational context; and the adequacy of the explanation depends on whether it can differentiate the actual distribution from the other possibilities that have that structure.

Although explanations are answers to questions, not all questions are equally good at probing the phenomenon in question to get at a good explanation. This is often because the questioner doesn’t have enough information about the events to ask a good question. Good questions depend both on the issue the questioner has in mind, and also on the phenomenon being explored (Garfinkel 1981, p. 32, pp. 56–57). Compare (1) with a question that makes Mary’s singleton A explicit:

  1. 6)

    Why did Mary (as opposed to Bob) get the A in the class?

This is, plausibly, a better question, because it makes explicit that the range of possibilities that need to be ruled out concern comparative facts (it is not enough just to provide information about how well she did, but how it turned out that she earned an A and the others didn’t).

Let’s focus now on (6). There are two ways we might represent the contrast space for this question (Garfinkel 1981, p. 44). On the first, we could consider the range of possible grades for each member of the class (the product of the individual possibility space for each member of the class = 350) and consider why it ended up that there was a single A granted to Mary: was her work so much better than everyone else’s? Did the rest of them party too hard the night before the final? But this would be a misrepresentation of the possibilities and the explanatory task, for in this class, the options were constrained so that only one person could earn an A. On the second version, the contrast space consists of the various ways in which the students might be distributed according to the pattern of 1 A, 24 Bs, and 25 Cs. This is the appropriate representation of the space of possibilities in the case under consideration. Garfinkel calls such constraints on the contrast space structural conditions. “In cases like these, the imposed structural conditions radically alter the kinds of explanations we give because they constrain and truncate the contrast spaces” (Garfinkel 1981, p. 45). The idea is that given our purposes and intentions in seeking the explanation, and the facts of the case, some ways of structuring the set of possible answers are better than others because of the possible outcomes.

What does this example show? The first lesson is that an explanation is a response to a question that either implicitly or explicitly frames what is relevant for the purposes at hand. This framing occurs through presuppositions, foci and contrasts. Secondly, if there are conditions that structure the contrast space (as in Garfinkel’s curve) so that only some possibilities are available, then an adequate explanation must be sensitive to this structure: the conditions define the set of competing cases that the explanans must rule out in order be adequate.

What does it mean, however, to be “sensitive to this structure?” Let’s return to the example of the dog’s ball rolling into the gully. We begin with a question: Why is the treat in the gully? Our answer is in terms of the ball: I threw the ball in that direction and couldn’t get to it fast enough to stop it rolling (and the treat was in the ball). Even though we appear to be asking about the treat, the best explanation will be in terms of the system or structure of which the treat is a part, given that the treat was in the ball, this is how things unfolded with respect to the ball-with-treat-in-it. In short, the treat wasn’t functioning independently. The same might be said if we ask why Mary (and not Bob) earned an A. To answer why she earned the grade she did, we need to point to the structure of which her grade was a part. Given that Mary’s performance was graded on a curve so there could only be one A, Mary earned the A because she wrote a better final than Bob. Mary’s and Bob’s grades were not independent. The explanatory emphasis in each case is not the apparent subject of the question (Mary, the treat), but the whole of which the individual is a part (the movement of the ball, the instantiation of the constrained distribution of grades with Mary on top) of which the individual is a part. The explanation of the relevant features of the whole, together with the fact that the individual is part of the whole, yields the answer to our question.

Garfinkel emphasizes that when we are considering apparently competing explanations for a phenomenon, we must be very clear what is the proper object of explanation and whether or not the explanations have the same object. Given the foregoing discussion, we should consider the following hypothesis: if a particular is part of a system or structure, then the best first step may be to indicate that it is part of the system or structure and proceed to explain what’s up with the structure. The explanation of the workings of the structure will be the best way to explain the behavior of its parts.

3 Structures, in general

It is often the case that we can explain the behavior of an object by reference to its parts: a plant grows towards the sun because the shady side grows faster than the sunny side; water boils or freezes because of the behavior of the molecules that compose it. But as we have just seen, we can also explain the behavior of something by reference to the whole of which it is a part. My dog Sparky’s tail is in the kitchen because he went there to see what was cooking; several dozen watermelon seeds landed on the deck because the watermelon they were part of fell off the picnic table.

Structures, broadly understood, are complex entities with parts whose behavior is constrained by their relation to other parts. Stuart Shapiro suggests:

I define a system to be a collection of objects with certain relations. An extended family is a system of people with blood and marital relationships, a chess configuration is a system of pieces under spatial and “possible move” relationships…A structure is the abstract form of a system, highlighting the interrelationships among the objects, and ignoring any features of them that do not affect how they relate to other objects in the system. (Shapiro 1997, p. 73)

I take Shapiro to be suggesting that my family is a system that includes particular individuals (Steve, Isaac, Zina, Sparky, me) who stand in relations such as “parent of,” “child of,” “spouse of,” “dog of,” etc. But we can abstract from this particular Yablanger system to see it as instantiating a more general structure shared by other families. We can then distinguish the individual in a system (me), from the position within the structure (parent, spouse). That is, considering the abstract relationships that form the structure, we can then distinguish occupiers of positions from the positions. Shapiro offers this characterization of the difference:

…there is an intuitive difference between an object and a place in a structure, between an officeholder and an office…There are, in effect, two different orientations involved in discussing structures and their places (although the border between them is not sharp)…We might say, for example, that the shortstop today was the second baseman yesterday, or that the current vice president is more intelligent than his predecessor…Call this the places-are-offices perspective. This office orientation presupposes a background ontology that supplies objects that fill the places of the structures. In the case of baseball defense and that of government, the background ontology is people; in the case of chess games, the background ontology is small, movable objects—pieces with certain colors and shapes. (Shapiro 1997, p. 82)

In contrast to this office orientation, there are contexts in which the places of a given structure are treated as objects in their own right…We say that the vice president is president of the Senate, that the chess bishop moves on a diagonal, or that the bishop that is on a black square cannot move to a white square. Call this the places-are-objects perspective. Here, the statements are about the respective structure as such, independent of any exemplifications it may have. (Shapiro 1997, p. 83)

So considering a family structure with places for parent and child, we can consider the places as offices for individuals—I occupy the position of parent—or we can treat the nodes or positions in the structure as objects themselves—parents are responsible for their children. Considering places as objects, we ignore the individuals that occupy the place, and focus on the relationships that hold between the places. I’ll use the term ‘node,’ e.g., the parent node in the family structure, for places-as-objects.

Notice, however, that if we distinguish the occupiers of positions from positions (nodes), then if it is given for the purposes of an explanation that an individual occupies a node in a structure (the dog treat in the ball, Mary being graded on a curve), the constraints defined by the structure’s internal relations matter in explaining the facts in question. The object of explanation becomes the node rather than the occupier of the node. The space of possibilities for the occupier of the node, qua occupier, is limited by relations internal to the structure. In Garfinkel’s terms, the structural conditions on the object, considered as occupier of node, require us to shift the proper object of explanation.

In explaining why the dog treat rolled down the hill, it doesn’t matter whether it was bacon or beef, crunchy or chewy; it does matter that the ball containing it was round. Without knowing the shape and weight of the ball, the slope of the hill, etc. we won’t be in a position to explain—in a stable and helpful way—the eventual position of the ball/treat. In explaining why Mary earned the A, it doesn’t matter where she was born, who her parents were, or what she had for breakfast: what matters is that only one A was possible, and her final was better than all the others. Even when questions direct our attention to the occupier of the position as the object of explanation (why did this treat travel down the hill in the way it did), the best answer to the question may be in terms of the structure the object occupies.

In short, sometimes our explanatory needs are very specific and we are concerned with a particular token event; a fine-grained grid will demand precise explanations, but the explanations are relatively unstable—minor changes in the sequence of events would render the explanation inapt. In other cases we are concerned with an individual as an example of a type, and the type is defined by reference to the node the individual occupies in a structure. Such explanations will abstract away from certain features of the individual and allow for “inessential perturbations” in the sequence of events (Garfinkel 1981, pp. 30–34). As a result the explanation will apply more broadly and be more stable. Moreover, once we identify the structure that constrains the behavior of the individual, we can ask new questions, demand further explanation: why is the individual within this structure? Why does this structure/set of relations exist, rather than that?

4 Dretske on structuring causes

Fred Dretske has also been interested in different levels of explanation in explaining the behavior of individuals. He draws a distinction between triggering causes and structuring causes. Dretske discusses, in particular, the causes of behavior, which he takes to be a process originating in an “inner state” C, and resulting in a bodily movement M. Although for our current purposes we don’t need to constrain the ‘C’ and ‘M’ in this way, his model is useful.

In looking for the cause of a process, we are sometimes looking for the triggering event: what causes the C which caused the M. At other times we are looking for the event or events that shaped or structured the process: what caused C to cause M rather than something else. The first type of cause, the triggering cause, causes the process to occur now. The second type of cause, the structuring cause, is responsible for its being this process, one having M as its product, that occurs now…There is a clear difference between explaining why, on the one hand, Clyde stood up then and explaining, on the other hand, why what he did then was stand up (why he stood up then). He stood up then because that was when the queen entered, or when he saw the queen enter, the room. He stood up then as a gesture of respect. The difference between citing the triggering cause of a process (the cause of the C which causes M) and what I have been calling its structuring cause (the cause of C’s causing M) reflects this difference… (Dretske 1988, pp. 42–43, my italics)

In more general terms, supposing C to be a cause and E and effect, if

C causes E.

Then C is the triggering cause of E. But we may also want to know what causes [C causes E], i.e., what causes the process of which E is a part.

  1. 7)

    Why is it that [C causes E] rather than [C causes E’]?

Dretske suggests that to answer (7) we need to find the structuring cause.

Whether we cite the triggering or structuring cause in an explanation will depend on what question we are asking and the contrasts and presuppositions of the question. Using Dretske’s example, we might ask:

  1. 8)

    Why did Clyde stand up when the Queen entered?

    With several different contrasts in mind:

  2. 9)

    Why did Clyde stand up when the Queen entered, i.e., just then (as opposed to moments before or moments after)?

  3. 10)

    Why did Clyde stand up (as opposed to stay seated, take off his hat, clap his hands) when the Queen entered?

  4. 11)

    Why did Clyde (as opposed to others in the room) stand up when the Queen entered?

To answer (9), Dretske suggests, we must give the triggering cause, i.e., the event(s) that caused Clyde to stand. But in response to (10) we should give the structuring cause. Presumably this would be to situate Clyde within a social structure: Clyde is Queen Elizabeth’s subject (he is a part in a larger whole). Locating Clyde in a structure, we shift our attention from individuals to nodes: subjects are required as a matter of etiquette (the structural conditions that limit the possibilities for nodes in the structure) to stand when the Queen enters. This is not a quirk of Clyde’s: generally subjects of the Queen will stand up when she enters the room, given the relations between Queen, her subject and the rules of etiquette.Footnote 3 Once we make explicit the background structure that constrains a subject’s (in this case, Clyde’s) behavior, we can then ask, why is standing up the polite thing to do? And why are there Queens and subjects at all?

So it appears that structures are important to explanation because they constrain behavior of individual things insofar as they occupy nodes in the structure. The structure does not simply provide background conditions for the events in question (the standing), for it is the workings of the structure that are sometimes the proper object explanation: Why stand? Why these constraints? Why these nodes? Why this structure here? Questions about nodes in structures are the proper subject matter of social science, for it is types of individuals, types of action, etc. that are of primary interest. This suggests that there need be no competition between a triggering cause and a structuring cause: both what they explain and how they explain differ.Footnote 4

5 Social structural explanation: the invisible foot

Within the philosophy of social science there is a longstanding controversy over whether all explanation must be “individualistic” (Recently, Pettit 1993; Udehn 2002; List and Spiekermann 2013; Epstein 2009, 2014). Although it is controversial what individualism requires, the general idea is that in order to explain the behavior of agents, the explanans must ultimately be in terms of their (perhaps idealized) psychological states.Footnote 5 As Frank Jackson and Philip Pettit point out, individualists take the basic problem with structural explanation to be that:

…it is far from clear how most of the factors invoked in structural explanations are supposed to affect individual human beings, in particular to affect them other than by coming into their consciousness and constituting covertly (sic?) microexplanations (Jackson and Pettit 1992, p. 110).Footnote 6

Let’s consider a potential social structural explanation in detail:

  1. 12)

    Why do women continue to be economically disadvantaged relative to men?

We might restate this with an explicit foil as follows:

  1. 13)

    Why do women continue to be economically disadvantaged relative to men (as opposed to reaching economic parity with men)?

    Biologistic explanation: women are innately disadvantaged in comparison with men in what it takes (intelligence, competitiveness, etc.) to be successful in high paying jobs.

    Individualistic explanation: women, more than men, prefer to spend time with children over being in a high-paying job (…fill in relevant beliefs and desires), so they choose to forego economic success for the benefits and pleasures of motherhood.

    Structural explanation: Women are positioned in a self-perpetuating economic structure that systematically disadvantages them.

Here is one elaboration of the structural explanation:

The Invisible Foot (Okin 1989; Cudd 2006): Imagine a couple, Larry and Lisa, who, we suppose, are equally intelligent, talented, educated, and experienced in the workplace; they have equal power in their relationship, have no prejudices about gender roles, and are equally capable of all domestic tasks and childrearing tasks. Larry and Lisa decide to have children; baby Lulu arrives. They live in a community where decent childcare is beyond their means. Moreover, let’s suppose that in this community, as elsewhere, there is a wage gap: women, on average, make only 75 % of what men make. Under these conditions, unless Larry and Lisa have special reasons to think that they are unusual in their earning capacities, it is reasonable for Larry to work full-time and for Lisa to make adjustments in her work, e.g., to work part-time, to take time off, to take a less demanding job. But in our society, “wealth determines power, domestic work is unpaid, and divorce laws do not evenly divide wealth” (Cudd 2006, p. 149). So Larry accrues greater human capital and ends up with more power in the relationship. Moreover, insofar as Larry and Lisa are typical, women on average will be poorer risks for employers who will “tend not to trust that women will stay with their careers or that if they do, they will devote the kind of time and energy to them that men will” (Cudd 2006, p. 149). As a result, “women’s jobs” that require less commitment, mobility, and experience will pay less, and women will have to prove themselves exceptional to be considered for high paying “men’s jobs.” As a result, the pattern is reinforced (Cudd 2006, pp. 148–151).

Arguably, the invisible foot is a form of structural injustice, and it is a kind of injustice that does not become apparent when considering only biologistic or individualistic explanations. In particular, the explanation focuses on Lisa and Larry as nodes in a structure, rather than simply as individuals with certain personal deficits or tendencies.

Let’s consider how one might structurally explain Lisa and Larry’s behavior:

  1. 14)

    Why did Lisa quit her job (rather than Larry)?

  2. 15)

    Why did Lisa quit her job (rather than find a different solution to the childcare problem, e.g., go part time, rely on grandparents, or excellent affordable daycare….)?

Plausibly, if we were looking for the triggering cause of Lisa’s quitting her job, we could point to a decision she made to do so, e.g., Lisa’s decision to quit her job caused her to quit her job. To gain insight, at least we have to consider further the beliefs and desires that gave rise to the decision: Lisa decided to quit her job because she believed that it would be best for her family, and wanted what’s best for her family.

Notice that this psychological explanation, (even if we offered even her more explicit reasoning), doesn’t provide a very good answer to either (14) or (15). To answer (14) it is not enough to say that Lisa quit (just) because she wanted what’s best for her family. That would be comparable to saying that Mary earned the A in Garfinkel’s class just because she wrote an excellent final (ignoring Bob’s performance and the curve). The fact is that Lisa quit her job because she chose to and Larry didn’t also choose to quit his job. It is a background structural constraint that they both can’t quit, and so Larry’s behavior, their relationship, and the limited options available are crucial to explaining her action. As before, it seems better to shift the object of explanation to the structure: Why did Lisa end up in the gully (so to speak)? Because Lisa is part of a system that includes Larry, her employer, etc., and given that Larry wasn’t going to quit, the employer wasn’t going to provide childcare, and she couldn’t just leave Lulu home alone, this was her only real option. She might have made a rational choice to quit, but it is inadequate to just point to her choice as if it occurred independently of the workings of the system.

Moreover, to answer (15) simply in terms of Lisa’s choice to stay home with Lulu is inapt. The possibility space for childcare is severely limited in Lisa and Larry’s situation; there are structural constraints that make only some options genuine possibilities. Her desires were “canalized” by the circumstances, because she occupies a node in a social structure. This is compatible with her action being rational, and autonomous. Clyde was a free agent when he chose to stand when the Queen entered. But insofar as he was the Queen’s subject and followed the rules of royal etiquette, he had no choice but to stand. Qua subject, qua node in the structure, standing was his only option; seeing him as subject explains his action. Similarly, qua mother/spouse in a highly constrained social structure, quitting is Lisa’s only real option. Lisa is part of that structure, and that’s how the structure works. As individual, Lisa could have behaved otherwise—she could have left Larry and Lulu—but as mother/spouse in the system, that was not an option.

The fact is, of course, that Lisa quit her job rather than Larry because Lisa is a woman who occupies the wife/mother node in a problematic structure of family/work relations. The structure within which she and Larry live combines facts of human (infant) dependency, a stable framework of gender relations, and a particular wage-labor system. These structural constraints limit the possibility space—this choice architecture—for both Lisa and Larry; the differences in what is available to them, given their gender, is crucial for explaining what occurs. Moreover, the explanation illuminates normative dimensions of the circumstances that would otherwise be missed. Given only the biologistic or individualistic explanations, the fact that women remain economically disadvantaged relative to men appears not to be a matter of moral or political concern: if the best explanation of women’s choices to forego economic success is that they, as individuals, desire to be caregivers of children (and the elderly), this is a choice we must respect. No intervention in the name of justice is called for, except possibly the gender disparity in wages that is built into the scenario. The structural explanation reveals, however, that there is a deeper problem than the wage inequity.Footnote 7 The “invisible foot” explanation shows that women as a group are structurally situated so that it is rational for them to choose options that keep them subordinate. (In raising the issue of subordination, I assume, with Okin (1989), that the partner in a relationship who earns less, has less work experience and credentials, and who identifies with norms of care will be disadvantaged with respect to exiting the relationship, so will be vulnerable to exploitation and abuse). Without the structural explanation, injustice is obscured.

The focus on structural constraints provides resources for capturing significant regularities: those whose choices are similarly constrained will tend to act in similar ways, even if their personal histories, psychologies, and attitudes differ, e.g., Lisa and Mona may make similar choices re childcare even if they grew up in very different circumstances and are psychologically different in many other ways. So even if in Lisa’s case the structuring cause is mediated by her beliefs about what options are available, the social structure is both a key factor in explaining her beliefs, and the pattern of women’s choices, i.e., the choices of those who are situated at the same node in the structure. This, ultimately, is what is at issue in the original question (12): why do women continue to suffer economic disadvantage relative to men? Abstracting from Lisa’s and Mona’s particular psychologies recognizing that they function at a node in a structure enables us to provide an explanation that applies to anyone occupying that position. This offers a more stable explanation, as we discussed above.

6 Social structures

So in order to play the right sort of role in structural explanation, social structures must impose constraints on our action: they provide us positions/offices within a set of relations that constitute the structure. I will sketch some broad outlines of the approach I favor (see also Haslanger 2012, Ch. 15, 17; Haslanger 2014b).

6.1 Social relations and social practicesFootnote 8

There are many constraints on action (physical, biological, psychological). How does a social structure constrain us? If physical, biological, and psychological constraints are the only forms of constraint, then it might seem that we are forced back to a kind of individualism: assuming that social structures don’t constrain us physically or biologically, they must do so by virtue of our attitudes about them. Or as Jackson and Pettit suggested: social factors affect individuals only by “coming into their consciousness and constituting covertly (sic?) microexplanations” (Jackson and Pettit 1992, p. 110). One model is that social constraints function through norms and have their effects by creating social expectations and anticipation of blame and praise (Bicchieri 2006, p. 11; cf. Martin 2009, pp. 6–7). This model of social constraint, however, is inadequate.

Social structures, as I understand them, are networks of social relations. These include relations between people: being a parent of, being an employee of, being a spouse of; they also include relations to things: cooking, owning, occupying, driving, eating, herding. Social relations, in turn, are constituted through practices. Our practices relate us to each other and to the material world; they situate us at nodes in the structure. Consider cooking:

Cooking rice is an instance of a more general practice of cooking, and regular engagement in the practice is constitutive of a social role: cook. Being a cook relates one in specific ways to other persons (not only the customer or family, but also the farmer, grocer, garbage collector, sources of recipes, including traditions, cookbooks, etc.), and also relates one in specific ways to things (foodstuffs, sources of heat, water, utensils). Cooking is only possible within a social structure that provides the ingredients, skills, tools; the norms for taste, texture and ingredients; the distribution of labor of cooks and consumers, etc.

What is a practice? Social practices are, in the central cases (though not all cases), collective solutions to coordination or access problems with respect to a resource.Footnote 9 The solution consists in organized responses to the resource.

Borrowing from contemporary anthropology (and social science more broadly), I have proposed this hypothesis (Haslanger 2012, Ch. 15, Ch. 17):

Practices consist of interdependent schemas and resources “when they mutually imply and sustain each other over time”. (Sewell 1992, p. 13; my italics)

Schemas consist in clusters of culturally shared concepts, beliefs, and other attitudes that enable us to interpret and organize information and coordinate action, thought, and affect. Schemas are public—think of them as social meanings conventionally associated with things in our social world, including languageFootnote 10—but are also internalized and guide behavior (Howard 1994; Hollander and Howard 2000). Both concepts and beliefs, in the sense intended, store information and are the basis for various behavioral and emotional dispositions. Although schemas are variable and evolve across time and context, their elements are sticky and resist updating. Resources are things of all sorts—human, nonhuman, animate, or not—that are taken to have some (including negative) value (practical, moral, aesthetic, religious, etc.). On this view, a practice exists when there are public schemas for interpreting, conceptualizing and responding to resources, and such resources are utilized and modified in order to fit the schemas. Individuals participate in practices, sometimes intentionally and knowingly, sometimes not, if their behavior accords with the schemas in engaging with the resources.

In social reality, schemas and resources are both causally (“sustain each other”) and constitutively (“mutually imply”) interdependent. Consider food, e.g., corn:

An ear of corn can be viewed as something to eat, as a commodity to be sold, as a religious symbol. In other words, we can apply different schemas to the object, and the schemas frame our consciousness and evaluation of the object. The different schemas not only offer modes of interpretation, but license different ways of interacting with the corn. Actions based on these different schemas have an effect on the ear of corn qua resource, e.g., it might be cooked for food, or the kernels removed to be shipped, or it might be dried and hung in a prominent place to be worshipped. The effects of our actions then influence the schema. If the corn sells for a good price, its value is enhanced and the farmer may seek ways to grow it more efficiently, possibly investing in new and different varieties.

Causal interdependence: schemas emerge and develop in response to resources and resources emerge and develop in response to schemas.

Constitutive interdependence: a schema for X is a way of collectively interpreting and organizing information about X’s qua resource. Schemas are constitutively defined by the resources they organize, and something’s being a resource of a particular kind depends on what schema interprets/organizes it.

6.2 Social constraints?

How do practices, then, impose social constraints? At least three kinds of factors condition engagement in a practice, e.g., my dinner making:

  • Personal attitudes, habits, dispositions; both mine and the personal attitudes of those I interact with.

  • Resources: the materials/tools available (Note that materials/tools may also include skills, time?).

  • Schemas: the collective concepts, narratives, expectations, of those in my cultural milieu.

Both schemas and resources constrain and enable my action: I’m not going to cook idli if I don’t know what idli are, if idli are just not what we eat, if I don’t have an idli steamer or the ingredients, whereas I will cook pasta because it is what we eat, I have a box on the shelf, a way to heat water, and a colander.

It might seem plausible to say that schemas are what socially constrain us. Perhaps resources impose physical constraints, personal attitudes impose psychological constraints, and shared schemas impose social constraints. But how do schemas constrain? If schemas are internalized and govern action through individual psychology (admittedly, not just beliefs and desires, for schemas affect perception, attention, cognitive associations and involve other sub-personal processes), again we seem to be left with familiar individualistic explanation.Footnote 11

However, remember that structural constraints need not be a matter of causal processes that trigger action; social constraints set limits, organize thought and communication, create a choice architecture; in short, they structure the possibility space for agency (see also Satz and Ferejohn 1994). Elucidating this possibility space and the position of an individual within the structure of that space, can explain their behavior. (Think of Mary and Bob in Garfinkel’s class, Clyde and the Queen, Lisa and Larry.) Even if individual psychology provides triggering causes of our choices and actions, what we do, and what we can do, depends on social structure within which we act. The structures socially constrain our behavior by making certain kinds of things available (or not), e.g., childcare centers, idli steamers; by providing templates of interaction that favor (or discourage) certain forms coordination with respect to a resource, e.g., share, hoard, distribute; and by canalizing our attitudes accordingly.

In short, structures beyond my attitudes and the attitudes of others create my choice architecture: I can’t cook idli without an idli steamer. This is not just a physical constraint, but also a social constraint: the artifact is not available to me in my social milieu. What food I prepare is constrained both by the social materials available and the social meanings/schemas, over and above the physical objects and individual attitudes. Such meanings are not up to the individual agent but depend on collective understandings and the resources that have been organized by those understandings. The options are constituted through our practices, i.e., the interdependence of schemas and resources.

We are constrained in many ways in acting: physically, psychologically, and socially. The constraint is social, I propose, when we occupy places in a structure, and the relations constituting the structure are relatively rigid due to the interdependence of schemas and resources. The materials and meanings that together make up our social world are not alterable simply by my thinking or intending otherwise, or by my individual action alone. What I do, and what I can do, is socially constrained. Pointing to these constraints can be, depending on the question, explanatory.

7 Conclusion

Social structures consist in a network of social relations, some of which are to other people, some of which are to non-human animals, some to things; some are conscious and intentional (marriage), some are not (consumer vulnerability, racial privilege). Although I have not explored different kinds of structures in detail, or the different ways in which structures are constituted, it would appear that offices/nodes in a structure of relations may be related causally (unemployment/crime), constitutively (what it is to be a rook is to be governed by the rook rules), or regulatively (how a subject should behave in the presence of the queen). Resources, including material objects, mediate our relations to each other within a structure. In an important sense, we organize ourselves around resources. The resources, not only schemas, structure our behavior—one cannot prepare idli without an idli steamer or ride a bicycle without a bicycle. My actions are my own, triggered by my thoughts, desires, goals. However, the resources I rely on and the meaning of my action are not mine alone but depend on the structure I’m part of. Illumination of these structures is important for explaining human action.

I have argued that an explanation of individual action in structural terms situates individuals within “offices” or nodes in a structure. We explain the behavior of the individual given their place in a structure. This offers insight into why the particular individual behaved as he/she did, but it also contributes to our understanding of the individual as the instance of a type—a type defined by the conditions for existing at that node. By carving the explanadum across a broader range of possibilities (as a type, not a token), we can achieve better, more stable, explanations. Moreover, by identifying the structure within which action occurs, we can also call for an explanation of the existence and shape of the structure. Structural explanations should not be avoided, or merely tolerated. They should be sought, for they are in many ways preferable to individualistic explanations.