The Image: Hard vs. Soft

Discussion of a division between “hard” and “soft” approaches to politics is pervasive within the discipline. It is also the primary focus of the literature on the discipline. Almond (1988) explicitly presents a discipline which can be mapped along two dimensions: a methodological dimension, a hard-soft dimension, and an ideological dimension, a right-left dimension. At the extremes exist four “tables” that are “brightly lit and visible”: the soft left (e.g., critical theorists), hard left (e.g., socialist and dependency theorists), soft right (e.g., Straussian theorists), and hard right (e.g., rational choice theorists and/or quantitative researchers). Most political scientists are not part of these four tables, but they remain the main means of dividing the discipline. Of these two dimensions, the methodological dimension is the one on which much of our divisions emerge:

The image of political science that emerges from Almond's work is one with those using hard science on one end and those that prefer (or at least appreciate) softer approaches to politics at the other.

Almond's (1988) thesis was a catalyst for a renewed discussion on the state of political science. Part of this discussion included a panel at the APSA Annual Meeting and subsequent series of essays in PS (Monroe et al. 1990). The general response was to accept with some caveats Almond's diagnosis and move on to his prescription for a cure—to identify, if possible, a core that will reunite the disparate sects within political science. Almond, as part of the roundtable discussion of his original work, states that this core should be political theory, which “ought to codify the history of the discipline, the development of central ideas, concepts, and theories” (1988, 35). Theory does not function in this way because it “has, with few exceptions, been traumatized by the diffusion of scientific aspirations and methods into political science, and seduced by simplistic temptations of 1960s and 1970s thinking” (1988, 35). Overall, the members of this roundtable found Almond's diagnosis of the discipline accurate but disagreed over whether what, if anything, should be done to (re)unite the discipline.

The hard-soft model is also implicit in earlier histories of the field. Lepawsky (1964) examined the pre-professional era of political science (prior to the creation of APSA) and presents a distinction between scientific and traditional, context-specific approaches to the study of politics. Somit and Tanenhaus ([1967] 1982) recount the move from a pre-scientific study of politics toward a new, more scientific paradigm. Ricci (1984) likewise focuses on the rise of a science of politics that pulled political science from its philosophical roots. These and other histories present an image of the discipline that is divided between the hard, scientific and the softer, more ecumenical approaches.

More recent work continues to present the discipline as divided along a hard-soft dimension. Pye finds that “neither content nor method provides an ‘essential core’ for the discipline as a whole” (1990, 4). The reason for this is the division between what he labels “scientific” and “humanistic” approaches to the study of politics:

Lowi (1992) similarly describes the discipline as including, if not defined by, “the three hegemonic subdisciplines of our time—public opinion, public policy, and public choice” (1). Together, these subdisciplines are very similar to the groups Almond identifies as the hard right. Each of these political scientists locates the hard-soft dimension as the primary division within political science.

Empirical verification of this image of the discipline is rare, but in an innovative study of political scientists' views of the discipline's academic journals, Garand and Giles (2003) find two divisions within political science. One is the hard-soft division. Analyzing the mean evaluations of 16 well-recognized journals, Garand and Giles find statistically significant differences between qualitative, quantitative, and mixed-approach scholars for 11 of these journals. They also find, however, differences between scholars in different subfields. There are significant differences between scholars of American politics, comparative politics, international relations, and political theory for seven of the journals. These findings support the claim that there is a hard-soft division in political science, but they also suggest that a subfield division may also exist.

Overall, the hard-soft image is widely accepted but rarely empirically verified in the literature on political science. Such verification is critical. As the dominant image presented in the literature, the hard-soft model categorizes political science by method rather than by subject. If this is an adequate description of the discipline, then political scientists who make choices about how to organize editorial boards, departments, and curriculum need to be sensitive to the representation of both hard and soft approaches to political science if they are to represent the discipline well. But what if this hard-soft model does not reflect the actual organization of the discipline?²

If the hard-soft model is incorrect, then political scientists need to reconsider how they tell the history of the discipline, how they organize their departments, how they select members for editorial boards, and how they choose to construct curriculum.

Organization of Political Science

Before modeling political science, one should define what political science is. The hard-soft model has defined political science ontologically and historically. Both have an appropriate place in the literature, but we must differentiate between an ontology of political science and a sociology of political science, what Holden (2000) refers to as the “intellectual realm” and the “organizational realm.” The former is what sociologists of science might refer to as a system of ideas: our theories, models, concepts, and scholarship. The latter is our social system; it is how we organize our departments, associations, curriculum, and publications (King 1980).³

A sociological definition of political science is consistent with the definition proposed by Joseph Cropsey: “political science is what political scientists do” (Monroe et al. 1990, 42). This understanding of the discipline moves the discussion away from the ontology of political science to its sociology. We may continue to debate the ontology of political science, but we should be able to agree that there is such a thing as political science, and one way to understand it is to study how it is organized.

This sociological view of political science provides the starting place for an empirical model of the discipline. Such a model should 1) include within the discipline anyone who identifies as a political scientist, and 2) count political science research as anything that these people do as scholarly activities.

Modeling political science should include, first, a description of how political scientists organize their scholarly activities. The dominant way that this is done is by forming associations that are open to anyone who wishes to identify themselves as a political scientist. The largest and most inclusive of these associations is the APSA, which currently has over 15,000 members drawn from over 80 countries. Within APSA, organized sections bring together scholars who have similar professional goals (c.f. Losco 1998). These organized sections may be formed around any topic or grouping so long as they meet a threshold number of members. Members are not required to join organized sections, but a majority joins at least one. One way, then, of modeling political science is to examine the pattern of membership in these organized sections.

A second approach is to model political science scholarly activity. This could be done by examining published work by political scientists, but this proves very difficult to do because it requires that the researcher choose a priori what political science should be. Questions of which books and journals to include within political science are impossible to answer unless the researcher imposes his or her definition of the discipline on the literature. For example, the Library of Congress lists 483 political science periodicals. To evaluate which of these journals is political science would require a paring down of this list, as was done by Baum et al. (1976) and Young (1995). Such a list could be reduced to only those that political scientists are familiar with, use, or admire. Yet, such a list would include publications that are in a distinct discipline such as the American Sociological Review and publications that are not peer-reviewed such as Foreign Affairs (Garand 1990). Moreover, it would exclude those scholars who choose books as the primary means of publishing their scholarly work, which would bias the selection against theory and qualitative research.

An alternative approach is to move away from publications and examine current scholarly activity as presented and discussed at the APSA Annual Meeting. This has three advantages over other approaches. First, it would include all material selected by the membership of APSA (through its division chairs). As already discussed, APSA is the best representation of what political scientists are; thus, the scholarly work presented at the APSA Annual Meeting is by political scientists for the benefit of other political scientists. Second, it allows minority approaches to political science to be represented even if they are excluded from mainstream (and possibly hegemonic) publications. Finally, the allocation of panels is distributed among divisions that, like organized sections, represent groups created by political scientists themselves, not groups classified by a researcher. By examining the pattern of research presented at APSA Annual Meetings, one can provide an empirical description of what political scientists do that is difficult, if not impossible, to provide using other methods.

Modeling Political Science

The hard-soft model of political science implies a geographic representation of the discipline. It uses terms like dimensions, continuums, and location to describe political science. Keeping with this tradition, I estimate a statistical model of political science that allows us to see the differences between political scientists and their scholarly work. This is done using multidimensional scaling (MDS), which provides a graphical representation of the dissimilarities between objects (Rabinowitz 1975). That is, it maps objects onto a two-dimensional space so that objects that have a great deal in common are positioned close to one another and those that have little in common are located at a greater distance from each other. MDS graphs provides a very useful way of understanding otherwise complex relationships.

This method provides a means of empirically testing the hard-soft model. If political science is separated along methodological lines, then we should see (literally) groups with similar approaches to politics located close together; groups that have different approaches to political science should be farther apart. MDS is also useful because it is a barefoot empirical approach to modeling political science. The only data used are those on the membership in APSA's organized sections (as of February 2004) and the number of joint panels at the 2003 Annual Meeting in Philadelphia. Any patterns that are created are based on the similarity of membership between sections or the similarity in subjects discussed at the Annual Meeting. No other data or measures are used.⁴

Two mappings of political science result—one of the organized sections based on membership and one of the APSA Annual Meeting divisions based on joint panels.

The Organizational Realm

MDS provides a graphic representation of the differences and similarities between groups of political science and their scholarship. The mapping appears to be a reasonable image of political science organization.⁵

One indication of this is that both scalings are similar despite being based on different data. The organized section scaling and panel divisions scaling appear to be mapping political science along the same dimensions. One way to know this is to examine the location of the organized sections and their corresponding divisions on both maps. The correlation between the location of the organized sections on the two mappings is 0.70 (p < 0.01). On the second dimension the correlation is lower, in part because the second dimension is not as meaningful for the divisions scaling. The correlation between the locations on the second dimension of both scalings is 0.47 (p < 0.01). Thus, whatever the dimensions represent, both scalings produce a model of political science that is similar for both the organization of political scientists and their scholarly work.

What is this first dimension? An ocular examination of the pattern on the first dimension rejects the hard-soft model that dominates the literature. As shown in Table 1, political scientists and their research are primarily organized around common subjects of inquiry, not approach to politics or methodology. At one end of the organized section mapping are local American politics: state politics and policy, urban politics, legislative studies, public administration, and presidency research. At the other end are international security and arms control, international history and politics, conflict processes, and human rights. The same pattern is seen for the divisions mapping, where panels on constitutional law and on elections and voting behavior are at one end of the first dimension and panels on international security and arms control and on conflict processes are on the other. In the middle are subjects such as comparative politics and interdisciplinary studies that bridge local domestic politics and international state-level politics.

Further evidence of this interpretation is presented in Table 2, which presents the correlation between the location of each organized section and some of its characteristics that were not used in estimating these locations. I include two sets of characteristics. The first measures some of the organizational features of the sections: the number of members and the cost of joining each section. Neither is related to location of the organized section on this first dimension. The second set of characteristics is based on the self-identified subdiscipline of the membership. This information is gathered from members when they join APSA (members can choose two subdisciplines and can change their selection at any time). The location of an organized section has the highest correlation with the proportion of the section's membership that studies American politics and the proportion that studies international relations. The more Americanists that a section includes, the more to the left/lower the section is located. The more that a section is made up of IR scholars, the more to the right/higher the section is located. Also related is the percentage that study comparative politics (more toward IR end of dimension) and those that study either public administration or public policy (more toward American end of dimension). Unrelated to this first dimension are the proportions that study political philosophy or the proportion that study methodology. These correlations support the interpretation that this dimension is not hard-soft, but based on the subject of inquiry.

These results raise doubts about the validity of the hard-soft model. Political science is a diverse discipline that covers topics ranging from the rise of empires to the timing of answers in public opinion surveys. Political scientists may study several subjects, but the link between them is not a common methodology but a connection between the political phenomena studied. Methodology is important, but it is a secondary dimension. This second dimension improves our understanding of the relationship between organized sections and research panels. Unlike the first dimension on which groups had common subjects, the second dimension separates hard, scientific approaches to politics from soft, interpretative approaches. Organized sections that are “soft” include foundations of political thought, politics and literature, ecological and transformational politics, and law and courts. At the hard end of the dimension are political methodology, political communication, political psychology, conflict processes, and elections, public opinion, and voting behavior. This hard-soft dimension is not simply a qualitative-quantitative dimension, as the qualitative methods organized section is in the middle of the dimension. Moreover, there are other groups such as comparative politics and presidency research that often use qualitative (but still empirical and scientific) methods.

This interpretation is also supported by the correlations between the location on this second dimension and the membership characteristics of the organized section. The highest correlation with this second dimension is between the proportion that identify themselves as methodologists and the proportion that identify themselves as political philosophers. The more methodologists in an organized section, the more “hard” their location; the more philosophers in a section, the more “soft” their location. There is a division between hard science and soft studies of politics, but it is a secondary division in political science.

The model presented in this article offers a new image of political science. Political science is diverse and the first dimension represents the differences in the type of political phenomena studied. Political scientists and their research are connected by a common subject of inquiry, not approach to understanding or discussing politics. The hard-soft dimension is less important but does constitute the second dimension, with humanistic approaches to politics at one end and scientific approaches at the other.

Mapping Political Science

The coordinates listed in Table 1 provide the location of each organized section and division, but since these groups are not of equal size a simple mapping of their location is insufficient if one wants to model political science. One needs a way of viewing the concentration of political scientists and their scholarship. Fabrikant and Buttenfield (2001) show that this can be done by using cartography to show the density of objects mapped onto a space similarly mapped using MDS. I follow a similar technique by using the location of the sections and divisions and their size to show the concentration of political scientists and their scholarship. The result is a topographic mapping of these concentrations, where darker shading indicates a greater concentration and lighter shading indicates less concentration. Figure 1 is the topographic map of the concentration of political scientists using the organized section data. Figure 2 is the map of the concentration of political science scholarship using data on the APSA Annual Meeting division panels. Together these panels provide a view of the concentration of political science.

Both figures provide similar maps of the concentration of political scientists and their scholarship, respectively. In the organized section map the horizon is dominated by three areas of concentration. The first and largest is the concentration of comparative politics scholars. The second is a collection of interdisciplinary scholars, what some might call the “… and Politics” groups: religion and politics, women and politics, history and politics, and others. Finally, there is a third peak in the area covered by American politics scholars. This peak is in the largest region in terms of area. Unlike comparative politics scholars who are organized together, there is not an “American politics” organized section that concentrates American politics scholars. Instead, the American politics area covers a wider area that is linked to public policy and public administration at one end and wraps “up” and “around” to political methodology. On the other side of comparative politics is international relations, which is more disconnected from the other subfields. Finally, political theorists inhabit an island separated from the larger continent of political science. It is closer to the interdisciplinary subfields but remains distinct and disconnected.

The map of the scholarship of political scientists is very similar, though there are some interesting differences because there are more divisions representing more subfields, particularly other studies connected to traditional theory such as normative theory and constitutional law. In this map there is a greater concentration of political philosophy connected to the rest of scholarship, though still far from being the core that some advocate. If there is a core, it is the combination of comparative and interdisciplinary subjects that make up the center of the map. This map also differs because there is a small but united concentration of political economy research that lies between comparative politics and international relations. Interestingly, the international studies subfield is separated from the rest of political science scholarship.

In both maps, the concentration of political science is in the scientific or hard section. The sections and divisions that are dominated by humanistic or soft scholarship are distant and less concentrated than the mass of political science. Such a distinction has been noted by many, including Monroe et al. (1990) who discusses this using the metaphor of geography and topography:

Theory may be an island, but political science is a discipline whose major division appears to be organized around common questions, not common methods or approaches to the study of politics.

How is political science organized? The model of political science estimated in this analysis and mapped in Figures 1 and 2 provide two possible answers. First, there is a hard-soft dimension to political science, but it is secondary to the primary dimension—the type of politics that is studied. Political scientists have organized themselves so that they are more related to other political scientists who study similar events and concepts, not to those who study politics in similar ways. Second, most political scientists and political science scholarship approach politics as scientists, not humanities scholars. Scholars who study traditional theory represent a smaller and largely disconnected area of political science.