Political concepts are central to the enterprise of political science. The concepts we use shape the world we see. Without solid conceptual foundations, the edifice of political science is insecure. If we fail to develop clear and precise concepts, our theoretical insights and empirical discoveries will fail to be clear and precise, too. This article reviews major pitfalls for conceptual analysis as well as the fundamental challenges to concept formation and conceptual innovation in the study of politics.
- Conceptual Commitments
- Language Acts
- The Triangle
- Conceptual Disorders
- Conceptual Opacity
- Conceptual Confusion
- Conceptual Ambiguity
- Conceptual Vagueness
- Definitional Defects
- Conceptual Instability
- Conceptual Abuse
- Conceptual Lumping
- Concept Formation
- Semantic Fields
- Empirical Boundaries
- Analytic Frames
- Conceptual Structures
- Conceptual Cores
- Family Resemblances
- Diminished Subtypes
- Contested Concepts
- Conceptual Innovation
- New Definitions
- New Concepts
- New Terms
- Semantic Constraints
In contemporary political science, concept formation is often regarded as a distraction, a mere prelude to serious research that is given scarce attention. Scholars sometimes ignore conceptual disputes, resolve them by fiat, or delegate their resolution to political philosophers. At the same time, a strong tradition of self-conscious and systematic concept analysis, resting on the pioneering work of Giovanni Sartori, David Collier, and others, does exist in the discipline. This article offers an analytical synthesis that weaves together insights of conceptual debate in both philosophy and political science.
Since its origins in ancient Greece, Western philosophy has been debating the nature and meaning of concepts. For centuries, thinkers tried to resolve one fundamental problem: the relation between the world and the mind, the objective and the subjective, things and ideas. They conceived the mind as a mirror and concepts as mental images of the outside world, as cognitive representations of objective realities that uphold the fragile correspondence between the two worlds. In the mid-20th century, the so-called linguistic turn in modern philosophy, brought about by authors like Ludwig Wittgenstein and John Austin, redefined the basic coordinates of concept analysis. It shifted the axis of conceptual debate from cognition to language and from language as a system of representative symbols (“Platonism”) to language as a medium of social action (“pragmatism”).
According to the classic conception of language, concepts are our basic units of thought. According to a pragmatic understanding of language, concepts are our basic units of (linguistic) action. In this perspective, concepts are not interior images that correspond to external realities but practical tools that allow us to do many things, many more than just putting vivid labels on inanimate objects. They allow us to threaten and promise, to bless and condemn, to give orders and to request favors, to express tenderness and anger, to know and believe, to contract marriage and christen ships, and so on. Designating objects “out there” in the external world (reference) is just one linguistic function among innumerable others.
As practitioners of social science, too, we do more than offer aseptic statements about the world. In our texts and speeches, we do more than describe and explain, more than refer to facts and associations between facts. We laud and criticize colleagues, highlight and downplay themes, support and refute arguments, persuade and dissuade readers, and so forth. However, while reference is not everything, it does play a leading role in the social sciences. All types of “speech acts” (John Searle) characteristically contain referential elements. They refer to something, be it in the physical world of objects, the social world of norms and interaction, or the subjective world of emotion and cognition. Arguably, articulating empirical and theoretical propositions constitutes the nucleus of our linguistic activities. It is what we are supposed to do with social science concepts: developing descriptive and explanatory inferences, making and breaking claims about the social word. In these linguistic performances—our primary speech acts in the social sciences—reference is key. We need our concepts to perform referential roles. We need them to grasp concrete realities in abstract terms. Classical philosophy was centrally concerned with one specific purpose of language: its referential role. We should not be surprised to see that conceptual discussions in contemporary social sciences, too, privilege traditional reference over other linguistic roles.
If concepts are means of action, their meaning does not derive from their correspondence to objective realities but from their practical roles in linguistic communication. In Wittgenstein’s famous dictum, “The meaning of a word is its use in the language” (1952/1968, sec. 43). Language is a medium of social communication. Its rules of usage and meaning are public, not private. Our shared knowledge about the meaning of a word derives from our shared linguistic practices. As competent language users, we know what others know about the meaning (the conditions of legitimate use) of a concept. As responsible language users, we accept the meaning of a concept (its conditions of legitimate use) when we use it and accept that others can hold us to account for using it. As in other realms of social action, responsibility means that we accept the consequences of our deeds. Take the standard example of a promise. If I promise you x, I understand the meaning of promise making and accept its conditions of validity. Among other things, I understand and accept that x is a future action that benefits you, that I am able to perform it, that my promise obliges me to perform it, and that I actually intend to perform it (Searle, 1969, chap. 3). If I promise, yet violate any of these conditions of validity that constitute the meaning of promises and in consequence fail to carry out x, you can hold me accountable.
When we employ concepts more specifically as means of propositions (the primary form of speech acts in the social sciences), we use them as carriers of general claims about the empirical phenomena they are referring to. When applying them to concrete cases, we subscribe to these claims. We commit ourselves to their truth (applicability). If I call a man a friend, a thief, or a left-wing dictator, I articulate (and thus embrace) certain (contextually understood) claims about my relationship to him, his relationship to alien property, or the form and substance of his exercise of state power. In case of doubt, confusion, or contestation, I must be ready to justify my conceptual choices and accept the consequences. The bundles of claims we commit ourselves to when employing a concept comprises its meaning. Often these claims and commitments remain implicit. Formal definitions serve to make them explicit (Robert Brandom, 2000).
In political science, we still have to assimilate the insights of pragmatic philosophy. Our discussions of concepts, as far as they take place, still tend to be anchored in the classical distinction between mental creations and real objects. In addition, we tend to reify both sides of the mind–world distinction. We tend to treat both concepts and their referents as if they were things. The result might be described as a kind of double false consciousness. We tend to misrepresent social reality as well as our representations of reality.
(a) The Reification of Reference. Concepts are abstractions, not proper names. They do not serve to designate particular objects but classes of objects. On the referential side, our paradigms of objects are still concrete, material things with observable properties. Very few objects of political research correspond to this model. The realities we study are symbolic. Our concepts are not generalizations from observed properties but abstractions of symbolic realities.
(b) The Reification of Concepts. On the conceptual side, we tend to treat our abstractions, too, as if they were tangible objects, fixed in time and space. Employing the language of factual propositions, we tend to ask what a concept is (and is not) and what its essential attributes are (and are not), as if comprehending the concept required discerning its visible properties.
According to the widely used tripartite conception of concepts developed by Charles Kay Ogden and Ivor Armstrong Richards (the Ogden-Richards triangle), concepts are symbolic entities that consist of three elements: their meaning (connotation or intension), their referents (denotation or extension), and a term or word (their name). This conception of concepts, introduced by Sartori into political science, is still indebted to the notion of concepts as symbolic intermediaries between mind (as location of meaning) and the objective world (as location of reference). It is therefore vulnerable to the reification of both reference and meaning. Still, it serves well to understand the contingent nature of conceptual commitments (the element of choice in the relations between terms, meaning, and reference) and in general offers a set of useful distinctions to analyze the formation and de-formation of concepts (which will guide parts of our subsequent discussions).
Conceptual discussions in the social sciences often carry therapeutic ambitions. In the pursuit of clarity and precision, they strive to cure scientific language of the multiple disorders that are thought to afflict ordinary language. The tradition of conceptual analysis in political science that was initiated by Sartori and his colleagues in the 1970s and is continued today most prominently by David Collier and John Gerring (2009) partakes of this therapeutic project. Ordinary language is not generally defective, though. It is as clear and precise as speakers need it to be. Still, scientific language differs from ordinary language in some fundamental regards. Among other things, it is written in form and literal in style; it involves a strong commitment to truthful, transparent, and evidence-based argumentation; it aims at generating general knowledge; and it demands the development of a common specialized vocabulary within the academic community. Most of these distinctions are normative, not empirical. They do not give social scientists a mandate to remedy the deficiencies of ordinary language, but they do involve the professional obligation to craft a shared specialized vocabulary that steers clear of major conceptual disorders and malpractices.
Everyday linguistic communication unfolds on the basis of implicit meaning. Neither do speakers offer formal definitions of the words they use, nor do their interlocutors ask for them—except when their shared understandings turn problematic and when communicative irritations arise, which are instances of incongruence between the concrete application of concepts and their social meaning that we take for granted. You promised to be punctual and are an hour late. Is this your notion of punctuality? You say we are friends, but you left me alone in the face of danger. Is this your idea of friendship?
In the social sciences, we run higher systemic risks of breaking through the thin ice of implicit understandings. Our key concepts are often complex and contested, and we cannot take for granted that others comprehend them in the same manner as we do. Linguistic transparency is, therefore, our first obligation in the social-scientific use of concepts. Karl Marx remarked once that he needed three volumes to explain the concept of capital. We need not go that far in explicating the core concepts we use in our research. Concise formal definitions will often suffice. Yet if we fail to make explicit our central conceptual commitments, our theories and findings cannot contribute to the construction of common knowledge, only to the accumulation of fragmentary statements whose interrelations are uncertain. Conceptual opacity engenders opaque research.
As the descendents of Noah in ancient Babylonia set out to build a premodern skyscraper (“a tower, whose top may reach unto heaven”), God, alarmed by their capacity of monolingual coordination, intervened to “confound their language, that they may not understand one another’s speech” (Genesis 11:7). More than anything else, the concept analytical tradition of Sartori has been concerned about conceptual confusion. According to its disciplinary diagnosis, the builders of comparative political science are afflicted by a similar confusion of tongues as the architects of the Tower of Babel. Lacking discipline and coherence, schools and scholars are speaking past each other with different, mutually incomprehensible vocabularies.
If concepts form triangles of terms, meanings, and referents, conceptual confusion may arise from three sources: confusing relations between terms and meanings (ambiguity), confusing relations between meanings and referents (vagueness), and confusing stipulations of meaning (definitional defects).
To ensure unequivocal associations of words and conceptual commitments “the golden rule is to have one word for each meaning” (Sartori, 1984/2009, p. 113). Confusion may arise if we have several words for one concept (synonyms) or one word with various meanings (homonyms).
Genuine synonyms pose no problem for communication. They enrich our vocabulary and help us avoid tedious repetition. The troubles arise from fuzzy synonyms—neighboring or overlapping terms that are situated in a disorderly semantic field and whose exact relations remain unspecified. Political science is replete with such terms. While language users employ them (loosely) as synonymous, it is unclear whether they actually do carry equivalent conceptual commitments. For instance, institutions are often defined as rules, and vice versa. Yet we do not know to what extent we can treat the two as interchangeable concepts. In the best of cases, fuzzy synonyms share a recognizable semantic core but differ in their precise connotations (additional shades of meaning). For example, genocide and ethnic cleansing may refer to the same murderous acts, yet the former maintains analytic distance, while the latter adopts the hygienic discourse of the assassin.
Although words frequently carry multiple meanings, in ordinary language, homonymy does little to disturb our ability to communicate. In everyday talk, context determines meaning. If it fails to do so, we can always ask for clarification. In the social sciences, by contrast, homonymous terms are more corrosive to communication. If scholars attach incongruent meanings to key terms that define their fields of inquiry, they will fail to create common knowledge. Instead of studying one class of phenomena, they will be studying different subjects under the same name. The unity of their field of research will be apparent only, a nominal delusion veiling the substantive fragmentation of their research. For instance, if some hold the goal of “democratic consolidation” to be the prevention of authoritarian regression while others take it to lie in the achievement of democratic deepening, comparative inquiries into the conditions of democratic consolidation will address qualitatively different substantive problems.
Concepts are abstractions. They allow us to speak about empirical phenomena in general rather than marvel at them one by one, in puzzlement over their uniqueness (which we could not grasp anyway without a prior notion of generality). In the social sciences, we want concepts to be precise, to circumscribe clearly the realm of phenomena to which they apply. Vague concepts fail to do so. They leave the relation between conceptual claims and empirical objects indeterminate. They do not allow us to decide which phenomena lie inside and which outside their realm of application. Political actors often apply such elastic concepts of unclear denotation as weapons of political struggle. For instance, candidates would condemn “vote buying” by their adversaries without making clear what kind of acts they are referring to. Are they condemning any campaign promise that offers voters material benefits in the future? Squeezed between conflicting expectations, political contenders often benefit from evasiveness. Scholars, by contrast, cannot leave their readers guessing what they are talking about. Leaving its empirical referents in the dark, vague concepts lead our research into obscurity.
Our semantic definitions are sources of confusion if their relations to words and things are confusing. They are confusing, too, if their internal structure is defective. Definitional defects may be manifold. Our definitions may be contradictory or tautological. They may be incomprehensible or prone to provoke misunderstandings. They may ignore standard rules of classification by offering classificatory schemes that fail to be exclusive, exhaustive, and one-dimensional. Or they may confound levels of abstraction within taxonomical orders of classes and subclasses. As a matter of fact, there is nothing easier than to be confusing. A moment of distraction, a slip of the tongue, a typographical mistake, and the meaning of what we meant to say dissolves in mist.
In the social sciences, linguistic stability permits continuity in research. The diffusion of new theoretical approaches tends to involve major shifts in our vocabulary. That is what new theories often are: new vocabularies and redescriptions of the world. The waves of conceptual instability they induce tend to reshape collective research agendas. They tend to disrupt established lines of inquiry even when substantive concerns remain essentially the same. Each time we replace our theoretical vocabularies, we tend to reinvent the wheel of empirical research. For instance, the new literature on the political economy of political regime change ignores earlier debates on capitalism and democracy, the new literature on state failure ignores the earlier literature on state building, and so on. Unless we develop and conserve the ability to translate between theoretical languages, destabilizing our core categories can be deeply damaging to the much cherished accumulation of knowledge.
The tools of language are open to almost limitless forms of abuse. Two symmetrical strategies of reality distortion are of particular interest to political scientists: conceptual stretching (Sartori) and conceptual masking. Conceptual stretching involves the application of (often value-laden) concepts to cases that lack essential characteristics of these concepts (as when a corrupt politician describes himself as an honest man). Conceptual masking involves the description of (morally relevant) cases through neutral concepts that disguise the essential characteristics of these cases (as when a bank robber describes himself as a common customer). The former puts forward semantic overstatements in which concept application betrays fundamental conceptual commitments. The latter puts forward semantic understatements in which concept application denies fundamental factual realities.
These forms of conceptual abuse are surely more frequent and more severe in politics than in political science. Modern authoritarian regimes, masters of linguistic abuse, routinely practice both. Stretching the notion of popular rule beyond recognition, they tend to portray themselves as higher forms of democracy, as when Augusto Pinochet described his form of military dictatorship in Chile as “protected democracy.” The totalitarian regimes of the 20th century were notorious in inventing quasi-neutral technical terms and bureaucratic acronyms to camouflage the unspeakable atrocities they committed against humanity. For instance, the Nazis described their factories of assassination as “concentration camps,” the Soviets under Stalin their officially inexistent colonies of slave labor and death as “Gulag,” “the zone,” or simply and enigmatically “the other side.”
Both malpractices destroy conceptual validity. They sever the link between connotation and denotation, between conceptual commitments and factual applications. In instances of conceptual stretching, speakers claim too much, and realities negate the essence of the concepts they use (often, their moral essence). In instances of conceptual masking, speakers claim too little, and concepts deny the essence of the realities they face (often, their moral essence).
Names allow us to designate individuals, concepts classes of individuals. More general concepts capture larger classes, more concrete concepts smaller ones. According to the well-known “ladder of abstraction” introduced by Sartori, the number of defining attributes of a concept (connotation or intension) and the number of its referents (denotation or extension) are inversely related. At a high level of abstraction, concepts carry few defining attributes and cover many cases. At low levels of generality, they contain numerous defining features and apply to few cases.
In scholarly research, just as in ordinary language, we have to choose the level of conceptual abstraction that seems appropriate for our purpose. If we talk too abstractly in everyday life, our interlocutors may get irritated: Come down, be concrete, we don’t want to hear generalities! If we are overly specific and draw excessively fine distinctions, they may well get impatient, too: Focus on the relevant, stop splitting hairs! In William Ockham’s famous formulation, “You need no razor to cut butter.”
In crafting social-scientific concepts, we have to seek a pragmatic balance between our ambitions of theoretical generalization and our needs for analytical differentiation. As in everyday interactions, we may err on either side when choosing our levels of conceptual abstraction. If we aim too high and employ excessively general concepts that obliterate “differences that make a difference” (Gregory Bateson), our critics will accuse us of conceptual lumping (Sartori). If we aim too low and choose excessively specific concepts that trace irrelevant distinctions, our critics will reproach us with conceptual splitting.
Overall, to the extent that the social science community is amenable to “linguistic therapy” (Umberto Eco) and avoids developing conceptual pathologies, it strengthens its collective capacity to communicate effectively. Social-scientific language demands more than conceptual health and discipline, though. It also requires theoretical and conceptual creativity, grounded in linguistic competence and empirical knowledge.
Concept formation is the systematic development and explication of the core claims to which we commit ourselves when applying a concept. It requires us to understand ordinary and specialized uses of the concept, map its location within its semantic field, situate it within empirical realities and analytical frames, understand its structural properties, choose our semantic commitments, and choose the term that best resonates with its meaning.
If the meaning of a word lies in its use, we need to comprehend the usage of a word if we wish to comprehend the semantic commitments it involves. The first question to ask concerns usage in ordinary language. In English, this question has a straightforward answer: the Oxford English Dictionary (OED). Initiated well before the advent of 20th-century pragmatic philosophy of language, the OED is a monument to a pragmatic understanding of language as a medium of social practice. Alien to prescriptive or regulatory pretensions, it meticulously registers “the meaning of everything” (Simon Winchester) by documenting concrete instances of word usage across centuries of linguistic development.
The second question concerns the scientific usage of a word. According to Sartori’s seminal “Guidelines for Concept Analysis,” the semantic “reconstruction” of a concept starts with a review of the relevant scholarly literature. Unlike the OED, Sartori does not recommend tracing concepts in usage in order to uncover the implicit claims that can be inferred from their practical applications. Rather, he directs scholars to compile lists of explicit definitions (until they get bored by repetition and redundancy), enumerate the attributes included, and bring them “into some meaningful kind of organization” (Sartori, 1984/2009, p. 120). Meaningful organization, indeed, is the central task of concept formation and the most difficult one, as it escapes rules and recipes. To begin with, it requires us to place our concepts in their linguistic contexts.
If we wish to understand a concept properly, we must not analyze it in isolation. We need to map the semantic field it inhabits and locate it in the web of relationships that connects it with its conceptual neighbors. Unless we know what the concept shares with proximate concepts and what separates it from them, we cannot grasp its specificity. Neighboring concepts often share large intersecting circles of meaning, yet carry semantic nuances that are relevant for our research purposes. For instance, if we set out to study civil wars, we step into a rich semantic field of overlapping, yet not identical, concepts, such as guerrilla war, revolution, rebellion, ethnic violence, regional violence, organized violence, political violence, state failure, anarchy, political disorder, and political fragmentation. Semantic cartography, surveying and mapping the “systems of terms” (Sartori, 1984/2009, p. 142) that constitute semantic fields, provides a relational and comparative understanding of such clusters of concepts. It helps us grasp better not only the central connotations of interrelated concepts but also their finer shades of meaning that may be decisive in choosing one concept over another.
In shedding comparative light on the precise meaning (intension) of related concepts, the analysis of semantic fields also helps clarify their precise referents (extension). To the extent that we grasp the differences and similarities in the substantive claims neighboring concepts contain, we grasp the differences and similarities in the empirical phenomena they refer to. For instance, in the semantic field of civil war, some concepts, such as political violence, are situated at high levels of abstraction and include violent actions outside contexts of societal warfare (e.g., terrorism), while others, such as regional violence, refer to more narrow categories of violent conflict. Some concepts, such as ethnic violence, involve claims about the motives of violence, while others do not. Some, such as guerrilla war, emphasize the presence of organized actors, while others, such as political disorder, emphasize the dissolution of central authority. Some concepts, such as state failure, attribute agency to the state, while others, such as civil war, distribute it among societal actors.
All these semantic differences involve empirical differences. They point to different empirical phenomena. Understanding these differences allows us to understand the empirical scope of concepts—their bounded territories. It gives us an idea of what they include and exclude. Depending on what we want to see and what we decide to ignore, it allows us to choose and use the concepts most akin to our analytical interests. Concepts, writes Gary Goertz (2006), are “theories about the fundamental constitutive elements of a phenomenon” (p. 5). Within the semantic field of civil war, we will select our concept of choice depending on the empirical dimensions our theories designate as relevant—the organized nature of civil wars, their political motives, their outcomes, their causal association with state power, or their membership in the general category of political violence. Sometimes, our inherited vocabulary does not trace the precise distinctions we are interested in. Still, our semantic maps allow us to visualize the configurations of empirical boundaries as previous concept users have found them relevant. Though at times incomplete and inconclusive, they open up a first dialogue between concepts and cases, mediated by theory.
When scholars reconstruct the scientific usage of their core concepts in relation to proximate concepts (semantic fields), they may come up with long lists of contending or coexisting, divergent or convergent terms, definitions, and applications. In and of themselves, such listings are of little analytic use. The key challenge is to order them in a manner that resonates with theoretical traditions and empirical concerns in relevant fields of research, and this is easier said than done. The standard recommendation sounds simple: First identify underlying analytical dimensions and then show how different concepts and uses of concepts differ along these dimensions. But which are these underlying dimensions? They do not just lie around, self-evident and open to simple inspection. How then can we find them, and how can we construct them?
The answer is perhaps disappointing: Concept formation is not a bureaucratic enterprise but a constructive one (like social science in general). It is not about the mechanical application of rules, be it of logic or language, but about the creative process of abstraction. In our efforts to make sense of divergent definitions and uses of concepts, we have to renounce the comfort of rules. This is the weak point, the structural lacuna, regarding guidelines for concept formation: There is none for this crucial task.
As a first step, it is always helpful to ask about units of analysis. If authors talk about democracy, is it regimes or states they are talking about? If they talk about the rule of law, is it individual decisions or judicial systems they are talking about? Clarity about the kinds of “objects” different usages of a concept strive to grasp often helps us locate them at different levels of abstraction. However, it does not tell us anything about the substantive claims different usages carry. We are not biologists looking at elephants or cows, tangible, objective phenomena whose observable characteristics make them easy prey to the classic logic of classification per genus et differentiam. Identifying the general commonalties that determine their genus and then the specific characteristics that determine their species still involves a good deal of complexity and controversy within the biological community and yet in principle requires little “sociological imagination” (C. Wright Mills). But this is not so in the social sciences. If we have determined, for example, that students of “ideology” alternatively refer to structures of thought, language, or behavior, we still do not know what kind of structures they are talking about. We still have to discern (somehow) the analytical dimensions that distinguish different claims about the defining structures of ideologies, such as internal coherence, external differentiation, sophistication, factual accuracy, abstraction, hierarchy, stability, dogmatism, sincerity, and consciousness (Gerring, 1997).
Scholars do not, however, enter the process empty-handed. They are not naive observers of unstructured data. They know the literature, they know the facts and the theories, and they know the language and the paradigmatic cases that define their field of study. The analytical dimensions they introduce (as well as the analytical dimensions their predecessors introduced in the first place) to bring “meaningful order” into conglomerates of contending definitions are anchored in their theoretical and empirical knowledge. It is such theoretical as well as empirical anchorage that makes the meaningful, fruitful, useful organization of semantic fields possible.
Once we have succeeded in bringing analytical order into multiple uses of a concept, we can reconstruct its structural properties. Conceptual structures are configurations of conceptual commitments. Once we have understood the claims authors commit themselves to when using a certain concept, two questions ensue: How strong are their conceptual commitments? And how do those commitments they consider binding (the “essential features” of a concept) relate to each other?
The strength of our commitments to the conceptual claims we articulate varies by degrees. Some claims we hold to be indispensable across contexts. They constitute the core meaning of a concept. Other claims we deem to be secondary and contingent. They form the peripheral and contextual connotations of a concept. According to the classical approach to concept analysis, from Aristotle to Sartori, if we wish to comprehend a concept, we need to identify the former, its semantic core. In the face of multiple uses of a concept, the semantic core is located at the intersecting area of those claims competent concept users declare to be binding (“necessary and sufficient”). Staking out a common ground of binding conceptual commitments (“the core concept”) often allows us to distinguish “narrow,” “thin,” or “minimal” definitions (that limit themselves to the core) from “broad,” “thick,” or even “maximal” ones (that go beyond, up to envisioning ideal-typical instances of the concept).
According to the classical conception of concepts, if different uses of a term do not share common semantic ground, they do not count as instances of the same concept. They appear as instances of multiple concepts. Modern philosophers of language (reinforced by more recent psychological research on typicality effects) have shed doubt on the notion that our ordinary usage of concepts is governed by the strict membership rules of necessary and sufficient conditions. They have contested the notion that concepts carry semantic cores. Rather than committing themselves to a fixed set of indispensable claims, they have suggested that concept users often commit themselves to a more open set of mutually substitutable claims. Wittgenstein (1952/1968, sec. 66–71) introduced the idea of family resemblances to describe concepts that are defined by an ensemble of alternative attributes rather than sharing a core of necessary attributes. In set-theoretic terms, the relevant features of family resemblance or radial concepts do not form intersections but unions. In terms of classical logic, they are not linked by the operator AND (necessity) but by the operator OR (substitutability) (Goertz, 2006, chap. 2).
It seems indeed to be the case that ordinary language users routinely apply empirical concepts, such as fruit and furniture, to concrete objects on the basis of their closeness to typical examples (“prototypes”). These intuitive applications are not based on an invariable set of claims all competent language users subscribe to. However, the use of family resemblance seems less frequent (and less compelling) in the social sciences. When students of politics use the notion of family resemblance, they commonly apply it not to the highest level of abstraction (the definition of general properties of a concept) but to lower levels of abstraction (the definition of constitutive dimensions or the observation of concrete instances of a concept). Concepts described as family resemblances often do seem to share an abstract semantic core (at a high level of generality), even if either their constitutive dimensions or their observational indicators are mutually substitutive (at lower levels of generality).
As a matter of fact, the observation of family resemblances seems to be generally dependent on the prior comprehension (at least vaguely and implicitly) of an abstract semantic core. It is only because we possess a general notion of their common nature that we can discern the elastic observational resemblance of certain classes of cases. Otherwise, we would perceive no more than superficial similarities among disjointed phenomena. Arguably, this is even true for the two paradigmatic concepts of “families” and “games” from which Wittgenstein derived his notion of family resemblances.
Consider families. The sociological literature offers numerous overlapping definitions of families and subtypes of families. For instance, the modern nuclear family typically includes spouses and their dependent children. The notion of family resemblance, however, does not explicate the meaning of the concept of family. It does not respond to the semantic question of conceptual essence but to the phenomenological question of empirical appearance. It does not pretend to identify families and distinguish them from other social groups on the basis of their general characteristics but to identify the members of particular families on the basis of their physical appearance. Family members may look alike, but they are not members of a family (neither in general nor in particular) because they look alike.
In cases of family resemblance, empirical referents may lack relevant characteristics of a concept and still represent genuine instances of it. In cases of “diminished subtypes” (David Collier & Steven Levitsky, 1997), empirical referents do not fully possess the relevant characteristics of a concept and therefore do not represent genuine instances of it. Classical typologies do not differentiate among members of a category. Objects are either in or out. If they are in, they are equal members of full standing. Diminished subtypes take into account that all members are equal—but some are less equal than others.
Diminished subtypes arise from continuous multidimensional concepts. When all constitutive dimensions of a concept are held to be essential, the full absence of any dimension involves the absence of the phenomenon in question. However, if constitutive dimensions are not dichotomous, but continuous, cases may lie somewhere in between full presence and full absence on any specific dimension. Situated in the “gray zone” (Goertz) of one constitutive dimension, these cases are still recognizable members of a general class of phenomena. Yet they are less than full members. Due to their structural deficiencies, they appear as distant, damaged, distorted representatives. By adding qualifiers (adjectives) to the original concept, we can avoid conceptual stretching (fraudulent claims to full membership in a category) and point to their specific deficiencies. The beauty of diminished subtypes lies in their diagnostic precision.
For example, if democracy requires competitive elections under universal suffrage, the absence of either electoral competitiveness or electoral inclusiveness renders a political regime nondemocratic. Yet governments may impose partial restrictions on either electoral competition or electoral participation that are not severe and systematic enough to involve the absence of either dimension. Such bounded, ambiguous constraints may turn electoral regimes into diminished subtypes of democracy. In the face of bounded restrictions on competition, we might speak of controlled democracies. In the face of bounded suffrage restrictions, we might speak of exclusionary democracies.
Diminished subtypes of multidimensional concepts may arise from the limited presence of one (or some) of their essential dimensions. They may also arise from the full absence of desirable, yet nonessential, attributes. Diminished subtypes of democracy often seem to express structural deficiencies whose benchmarks are not the core of democracy but the ideal of democracy. They refer to democratic regimes that are in full accordance with democratic minimum standards, yet fail to fulfill more stringent expectations to democratic governance. For example, in clientelist democracies, citizens lack the programmatic orientation, and in apathetic democracies, they do not show the participatory enthusiasm we expect from high-quality democracies. In a symmetrical manner, as empirical cases may be underperforming in relation to a particular standard, they may also be overperforming, thus giving rise to “enhanced subtypes.”
The notion of family resemblances introduces some degree of flexibility and fuzziness into the classical idea of essential attributes. The notion of “essentially contested concepts,” formulated by British philosopher Walter Bryce Gallie only a couple of years after the publication of Wittgenstein’s Philosophical Investigations, involves a potentially more radical objection to the classical idea of conceptual cores. According to Gallie’s (1956) seminal paper, all concepts are open to contestation, yet some are “essentially” contested insofar as “their proper use . . . inevitably involves endless disputes about their proper uses” (p. 169). Which is the source of such intrinsic and irresolvable conceptual disputes? There are two plausible answers. One points to the complexities of concept application, the other to inner tensions that lie at the very core of some concepts.
Gallie’s (1956) account focuses on concept application. For concepts to become essentially contested, he says, they must be multidimensional—“internally complex”—and normative—“appraisive” (pp. 171–172). Competent speakers recognize and value the various dimensions of the concept. “Any explanation of its worth must therefore include reference to the respective contributions” of each dimension (p. 172). Yet, despite their abstract consensus on the fundamental component parts of the concept, speakers are likely to weight them differently and apply them differently. They are likely to dispute their relative importance as well as their practical implications under changing circumstances. For instance, actors may agree that modern democracy rests on the principles of majority rule and constitutional government but still disagree about their rank order (which is more important) and their operationalization (how they are to be put into practice).
According to a more radical reading, essential contestation affects the very core of concepts, not just their application. It arises when concepts involve irresolvable inner contradictions and when they are founded on genuine dilemmas, paradoxes, or impossible ideals—“conflicting imperatives” (Reinhard Bendix). Consider the ideas of democracy and the rule of law. The etymological root of democracy (“rule of the people”) assigns citizens the double, paradoxical role of rulers and ruled. Their powers cannot be maximized simultaneously. In addition, modern democracy arguably embodies fundamental principles (e.g., popular participation and constitutional restraints) that are mutually supportive as well as (at times) mutually subversive. Balancing the trade-offs they impose demands more than establishing simple orders of priorities. The concept of “the rule of law” suggests a mode of impersonal domination where formal rules govern, rather than human beings—the government of laws, not people. Yet the abolishment of human judgment in the exercise of power constitutes an impossible ideal. These inner tensions are constitutive for concepts such as democracy and the rule of law. Since they are irresolvable, they feed an inexhaustible stream of legitimate disputes that are irresolvable as well. In this perspective, appraisive concepts are likely to be essentially contested if they are essentially self-contradictory.
Language is a social medium of communication and is public and restrained by shared rules of grammar and semantics. At the same time, it is the quintessential medium of personal expression, free and open to variation and innovation as no other societal institution. Ask James Joyce and Ernst Jandl, for example. In the social sciences, our license to innovative uses of language is more limited than, say, in poetry or insult. Still, within constraints, we do enjoy important margins of conceptual innovation.
In the face of contending definitions, scholars cannot rely on the common, accepted usage of a concept. They have to meddle with the rules of the language game by formulating their own conceptual commitments. The particular definitions they develop may relate in various ways to preexisting definitions. They may privilege one usage against others (selective definitions), they may tie new bundles of conceptual claims by accepting some but shedding others (eclectic definitions), they may strive to express underlying commonalities among multiple conceptions (synthetic definitions), or they may embrace new claims that alter the central connotation of the concept (deviant or original definitions). The last move overlaps with the creation of new concepts.
When do we need new concepts? The answer is very simple: when we wish to draw distinctions that we had not drawn before or when we wish to grasp commonalities that we had not grasped before. Sometimes we wish to seize new empirical phenomena, and sometimes we would like to adopt new perspectives on old phenomena. In the study of politics, we are continually confronted with novel realities; continually, we are trying to see together what political actors tend to look at in isolation; and continually, we are developing theories that redefine the relevant boundaries of the political world. Hence, our incessant demand for conceptual creativity.
Like the political realities they try to capture and the political theories they strive to express, new concepts seldom emerge de novo as radical breaks with the past. More often than not, scholars introduce new concepts by modifying old ones. The repertoire of incremental conceptual innovation is broad. Authors may craft new concepts by (a) redefining the substantive meaning of a given concept, (b) importing concepts from other languages or scientific disciplines, (c) remodeling the ladder of abstraction by adding new distinctions or removing established ones, (d) introducing diminished or enhanced subtypes, or (e) changing the property space of a concept (by altering the distance between conceptual poles, shifting their location, or introducing intermediate categories).
When we have reviewed the various uses of a concept and its semantic neighbors and when we have understood the configuration of conceptual claims others accept as binding and chosen those we do, we sometimes face the task of “selecting the term that designates the concept” (Sartori, 1984/2009, p. 123)—sometimes but not always. Sometimes we have terminological choices; sometimes we do not.
The names of grand concepts in political science, such as justice, power, and rationality, the terms we use to designate them (and have used to designate them for years, decades, or even centuries), are given and fixed at present. Scholars often disagree about the precise meaning they associate with certain concepts designated by certain terms. They embrace contending “conceptions” of the concepts under dispute. The tight coupling between terms and concepts, however, prevents their semantic disagreements from spilling over into terminological disagreements. Switching the term under discussion would involve switching the concept under discussion. Authors who discuss, for instance, the concepts of justice, power, and rationality either discuss these concepts under these names, or they discuss something else. If they move even to neighboring terms, such as fairness, authority, and intelligence, they move into different (albeit contiguous) conceptual terrains. When terms and concepts are welded together through strong bonds of semantic history and when the former represent the latter without equivalent substitutes, terminological choices precede semantic debates. By choosing a term, we bring the concept it stands for into focus, which then allows us to partake in ongoing disputes about disputed aspects of its meaning.
When concepts are less deeply anchored in history and theory, the sequence can be inverted. We can first determine the substantive claims we are interested in and then settle on appropriate names, either by selecting among available terms or by crafting new ones. If we articulate our conceptual claims in a precise manner, we put ourselves in a position of selecting the precise terms whose connotations correspond most closely to the substance of our concept. Semantic fields that are densely populated with near-synonyms of similar standing (with none of them dominating all the others) offer the most latitude for fine-tuning our terminological choices. For instance, if we study the consolidation of political regimes, we may choose alternative terms that lie in its semantic vicinity yet emphasize diverging substantive concerns. If we wish to stress the duration of regimes over time, we may talk about continuity, endurance, or persistence. If we wish to stress their ability to weather systemic crises, we may talk about resilience, viability, or sustainability. If we wish to stress the process character of consolidation, we may talk about stabilization, institutionalization, or entrenchment.
While refined concepts ask for refined vocabularies, new concepts demand new terms. Whenever the diffusion of new theories (e.g., game theory) or the emergence of new realities (e.g., electoral autocracies) induces waves of conceptual creativity, they are accompanied by waves of terminological innovation. The rise of game theory has brought a whole new vocabulary into political science that includes notions such as backward induction, bounded rationality, perfect equilibrium, incomplete information, cooperative games, focal points, mixed strategies, and so on. The rise of electoral authoritarian regimes since the end of the Cold War has led comparative scholars to propose a broad assortment of labels designed to capture these novel political systems, such as hybrid regimes, semiauthoritarian regimes, inconsistent regimes, multiparty autocracies, competitive authoritarianism, and institutionalized dictatorship. Overall, driven by changing theories as well as changing realities, terminological innovation is a pervasive phenomenon in political science.
Linguistic innovation, in political science as elsewhere, is always anchored in linguistic tradition. While deviating from tradition, it cannot cut itself loose from it. Conceptual innovations in political science, whether they involve new definitions, new concepts, or new terms, are therefore inevitably constrained innovations (whether tightly or loosely). They are constrained by the semantic past of concepts (etymology and conceptual history) as well as by their semantic present (ordinary and specialized usage). They are also constrained by the systemic logic of semantic fields. Modifying individual nodes in the web of interdependent terms that constitutes a semantic field reverberates throughout the entire web. Any pretension to remake our conceptual tools has to recognize this twin linguistic reality that concepts are rooted in their semantic past and present and embedded in their semantic environment. Semantic constraints are not straitjackets. They leave room for critique and creativity, for selective changes and selective continuities—some more, some less. Yet if conceptual innovations depart too sharply from established usage and create arbitrary associations between words and meanings, they will fail to serve as effective tools for communicating novel insights. They will be outright incomprehensible, liable to provoke systematic misunderstandings, or devoid of resonance and thus condemned to oblivion.
“Clear thinking requires clear language,” Sartori (1984/2009, p. 102) wrote a quarter of a century ago. To begin with, it requires clear thinking about language. If we learn to incorporate conceptual self-awareness into our canon of methodological sophistication, we will do better theory and better research. We may not reach the impossible ideal of a fully transparent, clear, and precise technical language, and as a scientific community, we may be too diverse and competitive to build another tower of Babel. Yet we will put the edifice of political science on more solid foundations.
- Austin, J. L. (1962). How to do things with words. Oxford, UK: Oxford University Press.
- Bevir, M., & Kedar, A. (2008). Concept formation in political science: An anti-naturalist critique of qualitative methodology. Perspectives on Politics, 6(3), 503–517.
- Brandom, R. B. (2000). Articulating reasons: An introduction to inferentialism. Cambridge, MA: Harvard University Press.
- Collier, D., & Gerring, J. (Eds.). (2009). Concepts and method in social science: The tradition of Giovanni Sartori. New York: Routledge.
- Collier, D., & Levitsky, S. (1997). Democracy with adjectives: Conceptual innovation in comparative research. World Politics, 49(3), 430–451.
- Gallie, W. B. (1956). Essentially contested concepts. Proceedings of the Aristotelian Society, 56, 167–198.
- Gerring, J. (1997). Ideology: A definitional analysis. Political Research Quarterly, 50(4), 957–994.
- Goertz, G. (2006). Social science concepts: A user’s guide. Princeton, NJ: Princeton University Press.
- Murphy, G. L. (2004). The big book of concepts. Cambridge: MIT Press.
- Sartori, G. (2009). Guidelines for concept analysis. In D. Collier & J. Gerring (Eds.), Concepts and method in social science (pp. 97–150). New York: Routledge. (Original work published 1984)
- Searle, J. R. (1969). Speech acts: An essay in the philosophy of language. Cambridge, UK: Cambridge University Press.
- Wittgenstein, L. (1968). Philosophical investigations (3rd ed.; G. E. M. Anscombe, Trans.). New York: Macmillan. (Original work published 1952)