A theoretical model is a tool that can promote theory construction. In this chapter, we discuss what a model is and how it can be used to improve theory, including how to represent theories in model form and how to derive theoretical statements from models. Along the way, we look at different types of models, how to evaluate models, and future directions in model building, all in an effort to show how models can be a major implement in the theory builder's toolkit. We will use a variety of models of the communication process to help illustrate our points, which apply equally well to model building throughout the social sciences.
Referring to the use of the word model in the social science literature, Kaplan (1964b) noted “the confusing and often confused usage of the term” (p. 267). Sometimes the term is used to refer to a strictly “physical model,” such as a wind tunnel. Sometimes it is used to refer to “any theory … presented with some degree of mathematical exactness and logical rigor” (p. 267). In such a case, the terms model and theory are being used simply as synonyms for each other. Sometimes the word is [Page 108]used to represent “a model of a theory which presents the latter purely as a structure of uninterpreted symbols,” what Kaplan called a “formal model” (p. 267). Perhaps the most famous formal model of all time is the exquisitely simple yet profoundly revealing E = MC2, representing the equivalence of energy and matter that underlies Einstein's theory of relativity. Sometimes the term is used for “presenting a conceptual analogue to some subject-matter,” what Kaplan called a “semantical model” (p. 267). This is perhaps the most common usage of the term in the social sciences.
Neuliep (1996) suggested that we distinguish between “scale models” and “conceptual models” (p. 30). A scale model replicates an object from which it differs only in size. A conceptual model involves a change in medium and often is quite abstract as it attempts to represent physical, psychological, and logical processes. Neuliep noted that “because communication is a psychological process, communication models are conceptual models” (p. 30).
There are two useful modifications we might make to Neuliep's definitions. First, let's make a distinction, not between scale models and conceptual models, but between physical models and conceptual models. Most of us were made familiar with physical models in our childhood, when we played with miniature airplanes and miniature kitchen ovens. A scale model is a type of physical model that differs from that which it models only in size. Though many physical models do differ in size from that which they model, they also differ in other ways. For example, a model of a ship may be smaller than that which it models, but it also may lack working sails, engines, rudders, or other essential parts of a full-scale ship. It is a physical model but not a scale model.
Second, though most communication models are conceptual rather than physical in nature, some communication models are physical. For example, an enlarged model of the human mouth and throat might be used to demonstrate principles of speech. Likewise, a miniature model of a classroom might be used to demonstrate how seating arrangements affect communication patterns. If this model differs from an actual classroom only in size, it is a scale model as well as a physical one.
Therefore, we may have a communication model that is strictly a physical model, and if it differs from that which it models only in size, then it also is a scale model. Most social science models, however, are not physical but conceptual. In fact, most definitions of the term model found in the social science literature indicate their conceptual nature:
[Page 109]Deutsch (1952): A model is “a structure of symbols and operating rules which is supposed to match a set of relevant points in an existing structure or process” (p. 357).
Bill and Hardgrave (1973): “A model is a theoretical and simplified representation of the real world. It is an isomorphic construction of reality or anticipated reality” (p. 28).
McQuail and Windahl (1993): “For our purpose, we consider a model as a consciously simplified description in graphic form of a piece of reality. A model seeks to show the main elements of any structure or process and the relationships between these elements” (p. 2).
Wallace (1994): “Modeling is the process of developing and providing an abstraction of reality, i.e., a model” (p. 1).
Baran and Davis (1995): “Any representation of a system, whether in words or diagrams, is a model” (p. 251).
Neuliep (1996): “A model is a graphic representation of an object or process” (p. 29).
Though at first glance there might appear to be significant differences across these six definitions, they are due mainly to the decision to limit the term for some reason. For example, Neuliep (1996) and McQuail and Windahl (1993) imposed the restriction that the model be in graphic form, whereas Baran and Davis (1995) chose to emphasize that a model can be verbal, as well. Lasswell (1948) presented what many communication students regard as one of the earliest and most influential communication models, writing that “a convenient way to describe an act of communication is to answer the following questions:
In Which Channel
With What Effect?” (p. 117).
Baran and Davis (1995) considered this a “model” of the communication process (p. 253), whereas McQuail and Windahl (1993) considered [Page 110]it a “formula” that could be “transformed” into a model by drawing boxes around “Who,” and “Says What,” and so forth, labeling each box as “Communicator,” “Message,” and so forth, and then drawing arrows from one box to the next (p. 13; see Figure 7.1).
However, far more impressive than the differences among these definitions are their similarities. There is general agreement that a model applies to an object or a process. That which is modeled also is sometimes referred to as a system or structure. For example, we could have models of two different classroom structures, one suitable for a small graduate seminar and another for a large undergraduate lecture course. We also could have models of the communication processes that predominate in each of these different structures or systems. Thus, feedback might be a more prominent feature of a model of a small seminar course than of a large lecture course.
There also is general agreement that a model is a simplified representation (abstractions being by definition simplifications). Although only Bill and Hardgrave's (1973) definition explicitly refers to the representation as being isomorphic, meaning there is a one-to-one correspondence between what is being modeled and the model itself, that is implied in most of the other definitions as well. Thus, an isomorphic map of a college campus, showing its buildings, walkways, roads, parking areas, and other elements, is a model that could be followed to get easily from one place to another.
We might say that a model simply represents a portion of reality, either an object or a process, in such a way as to highlight what are considered to be key elements or parts of the object or process and the connections among them. A model is not a mirror image of reality but merely makes salient certain aspects of reality. A model helps us focus on some parts and connections among those parts while ignoring other parts and connections. It is this simplifying and focusing that makes models particularly valuable as theory-building tools.
As noted earlier, the terms model and theory are sometimes used as if they were interchangeable. For example, in discussing how theories in the social sciences tend to cover only limited contexts rather than broad, sweeping portions of reality (more characteristic of theories in the natural sciences), Hanneman (1988) states,
[Page 111]Most social scientists' models deal with particular phenomena, or narrow classes of phenomena. That is, they are theories of the “middle range.” Fewer of our models are useful to understanding the similarities and differences across wide ranges of patterns of social behavior. (p. 16)
Here, Hanneman was using model and theory synonymously, something not all that uncommon in the literature.
Noting that model is sometimes used as a synonym for theory, Kaplan (1964b) rightly asked, “If ‘model’ is coextensive with ‘theory,’ why not just say ‘theory’?” (p. 264). The answer is that although they are sometimes confused, there is a good reason to keep the two terms conceptually distinct. A theory is a set of systematically related generalizations [Page 112]suggesting new observations for empirical testing. As such, the purpose of a theory is to explain or predict. A model does not explain or predict anything. We might say that the purpose of a model is to describe and imagine.
Though a model is not a theory, a model can be used to represent a theory. As Neuliep (1996) noted, “Theorists use models because they can describe and simulate physical, logical, or conceptual processes that may not otherwise be observable or presentable” (p. 30). He gave the example of theories of listening. Because listening is a psychological phenomenon impossible to touch, a model can provide a valuable method of indirect observation. Neuliep stated that models enable theorists to illustrate, delineate, and depict the structural features (i.e., what the object or process looks like—its form) and functional features (i.e., what the object or process actually does—its purpose) of their theories in varying degrees of abstractness and detail. Some models may be very detailed and literal and others rather general and abstract. No matter how detailed or literal a model is, however, it is nothing more than a description of an object or process. If we want to understand how the object or process works, we need something more—a theory.
Even though a model cannot explain or predict, it can help us advance theory. According to Bill and Hardgrave (1973),
A model, by itself, is not an explanatory device, but it does play an important and directly suggestive role in the formulation of theory. By its very nature it suggests relationships.… The jump from a model to a theory is often made so quickly that the model is in fact believed to be a theory. A model is disguised as a theory more often than any other concept. (p. 28)
Actually, a model does more than merely suggest relationships; it implies relationships. By making relationships explicit, a model can serve as a useful springboard for theoretical developments. As Deutsch (1952) noted, a model implies judgment of relevance, and that, in turn, implies a theory about the thing modeled. In building a model, a model builder chooses certain elements to include while ignoring others and makes certain connections among elements while ignoring others. These judgments are rarely made in a theoretical vacuum. Often, in fact, theories and models exist in a sort of symbiotic relationship, with theories nourishing models, which may then cultivate theories.
[Page 113]Theories can speak to models, and models can speak to theories. If the dialogue gets frenetic enough, it is sometimes difficult to keep track of the different parties. To put it another way, theories and models can make beautiful music together, and when they dance it is hard to tell which is leading and which is following. If distracted by their skillful teamwork, most people don't really care. As Severin and Tankard (1997) commented, a model “is often confused with theory because the relationship between a model and a theory is so close” (p. 45).
In discussing how theories and models differ, Harvey and Reed (1996) stated:
Models, as opposed to theories, are well-formed metaphors and analogies. They do not claim to express the truth of the world, but merely to provide heuristic insights. While theories claim to actually explain reality, models are only partial, fictitious constructions. They seek a language of “as if,” not “what is.” But if models can make few explanatory claims, they are rich in the conceptual materials upon which they can draw and are freer to organize those materials in a manifold of different directions. (p. 309)
Noting that theories are constricted by “a formal operationalist logic or the presuppositions of a well-articulated paradigm,” Harvey and Reed pointed out that models are unconstrained by such fetters: “Modeling freely participates in acts of imagination to produce a wide range of alternative insights to old problems” (p. 309). Compared to theories, which tend to be formidable and sometimes even forbidding, models can be more casual than formal and can entice us to toy with them.
Despite their potential playfulness, however, models are serious business. Models help build theory, but they do so mainly by maiming and murdering. Kaplan (1964b) noted that theories and hypotheses often are ill-defined, vague, and uncertain, “at home only in the twilight regions of the mind, where they are safe from sudden exposure” (p. 268). Models, on the other hand, are “conscious, explicit and definite; there is nothing ghostly in their appearance or manner; they look healthy even up to the very moment of their death” (pp. 268–269). In this regard, the model saves us from a certain self-deception: “Forced into the open, our ideas may flutter helplessly; but at least we can see what bloodless creatures they are. As inquiry proceeds, theories must be brought out into the open sooner or later; the model simply makes it sooner” (p. 269).
In discussing what models are and how they differ from theories, we have already touched upon some of the important uses of a model. Deutsch (1952) neatly organized the uses of a model into four different but related functions.
First, a model can help us organize data. It can show similarities and connections among its parts not previously recognized. Because a model is intended to show the major elements of a structure or process and the relationships among them, it can keep us focused on the issues at hand and, at the same time, relate the particulars of our work to that of others working in the same areas but with different models. For example, Lasswell (1948) used his model of an act of communication mentioned earlier to organize the “scientific study of the process of communication” (p. 117). Thus, he stated, the “Who” referred to “control analysis,” the “Says What” referred to “content analysis,” the “In Which Channel” referred to “media analysis,” the “To Whom” referred to “audience analysis,” and the “With What Effect” referred to “effect analysis” (p. 117). Lasswell's model of communication thereby served the useful purpose of organizing the study of communication.
Second, a model can help us make predictions. By suggesting relationships we may not have thought about before, a model can lead us to testable hunches. As McQuail and Windahl (1993) noted, a model can “be a basis for assigning probabilities to various alternative outcomes, and hence for formulating hypotheses” (p. 2). They offered a “transmission model of news learning” based on a model of the psychological effects of television viewing on individuals that was originally proposed by Comstock, Chaffee, Katzman, McCombs, and Roberts (1978) (Figure 7.2). This news learning model evoked a number of useful predictions, such as “The probability of a particular item being ‘processed’ by a receiver as potential information depends on two main factors: its being affectively arousing and attention-gaining; its being selected as relevant or interesting” (McQuail & Windahl, 1993, pp. 86–87). According to this model, much news will be received (“actual exposure”), but it will be “scanned without registering any cognitive or emotional effect” (p. 87). Say McQuail and Windahl, “Audience reach or attention may be recorded, without any process of interpretation taking place. Items which are not processed cannot be comprehended or have learning effects” (p. 87). Just looking at a model and thinking about its parts and connections can generate good hypotheses.
[Page 115]Third, a model can be a helpful heuristic tool, a pedagogical device that encourages students to find out things for themselves. It can be an effective communication device between teacher and student, making complicated and ambiguous information simpler and clearer. Like a [Page 116]good blackboard outline of a class lecture, a model can help students follow material more easily. Noting how science is both a cooperative and a cumulative enterprise, Kaplan (1964b) stated, “The model allows the scientist to make clear to others just what he has in mind” (p. 269). As Holmes and Hundley (1997) noted, “Basic course textbooks usually include communication model illustrations to clarify and reinforce the verbal descriptions” (p. 2).
A course on mass communication theory could be organized in a number of ways. For example, it might be organized along the lines of Lasswell's (1948) model of communication, dealing with the five types of analysis he identified. It also could be organized from the perspective of the mass communicator and could focus on the obstacles he or she faces in trying to reach an audience with a message. These obstacles include properties of the mass communicator, such as lack of journalistic experience, poor language skills, and lack of transmission power, as well as properties of the audience, such as lack of access to the message, selective attention, and selective retention (McCombs & Becker, 1979). A “barriers model of communication” can focus the attention of journalism students on the challenges a journalist faces in trying to reach an audience with an intended message and what the journalist can do to overcome these limitations (Figure 7.3).
Fourth, a model can help us make measurements. As Severin and Tankard (1997) put it, “If the processes that link the model to the thing modeled are clearly understood, the data obtained with the help of a model constitute a measure, whether it be a simple ranking or a full ratio scale” (p. 46). McCombs and Shaw's (1972) agenda-setting theory was modeled by McQuail and Windahl (1993) in a way that illustrated the theory's basic proposition that “matters given most attention in the media will be perceived as the most important” (p. 105) (Figure 7.4). Like the bars in a histogram chart, the model uses varying lengths of bars to indicate how much attention the media give to different issues and uses varying sizes of the letter X to indicate the importance of different issues to the public. The model brings to mind the idea of measuring differences in both media attention and public perception in terms of relative ranks.
This view of the agenda-setting process, however, has been considerably modified as advancements in agenda-setting theory have been made. For example, it has been found that media attention to obtrusive issues (ones with which the public has direct personal experience) is not as influential on public perception as is media attention to unobtrusive issues (Zucker, 1978). A model incorporating this theoretical development [Page 117]might use distinguishing symbols such as bars made of broken rather than solid lines or bars of different colors to visualize this difference between obtrusive and unobtrusive issues. Because size has been used in the earlier model to represent quantitative differences, the added [Page 118]symbolic representation brings to mind that the difference between obtrusive and unobtrusive issues is regarded not as quantitative but as qualitative. In other words, it is a nominal difference and not one capable of distinction on an ordinal, interval, or ratio scale. A well-constructed model may help us see what kinds of measures are needed and how to design appropriate tests.
Thus, a model can not only suggest that two or more things are related but also indicate how they are related and how we might examine their relationships. The failure of the earliest models of agenda setting to account well for certain empirical findings led to the explorations of the nature of issues that produced the obtrusive-unobtrusive issues distinction. When examining data in terms of McQuail and Windahl's (1993) model, one could see clearly that the theory's expectation that longer bars would be associated with larger Xs did not always obtain. From examination of the nature of the failures, the [Page 119]importance of issue unobtrusiveness emerged. That refinement of the model was then tested and found to account better for the findings than did the original model. As Kaplan (1964b) noted, one of the greatest advantages of models is that
they allow a systematic exploitation of failure.… Now models are often used, not in the expectation of immediate success, but in the hope of successive identification of particular causes of failure, so that an acceptable theory can gradually be developed. (p. 274)
Though models can help theory builders organize, communicate, brainstorm, and suggest measurements, they do have notable limitations. Sometimes models are criticized because they are so simplistic that they appear to devalue that which they attempt to model. Because humans are almost always elements in social science models, some humanists see all models as demeaning. Further, because models by definition cannot represent all that it means to be human, some regard them not merely as inadequate and unhelpful but as distorting and dangerous because they trivialize human experience (Baran & Davis, 1995, p. 256). Model builders generally respond to such criticism by saying that they can refine models on the basis of research findings.
A second criticism of models is that they tend to concentrate attention on what is most observable and to view these observables as performing most effectively when they serve the overall system. Critics complain that this leads to maintaining a status quo bias (Baran & Davis, 1995, p. 256). The model builder's response to such criticism generally is that not only must the value of each part be assessed in terms of its contribution to the whole, but the entire model must always be grounded in historical reality: that is, placed in the context of its time and place. Take, for example, the case of the standard computer keyboard, which is known as the “Qwerty keyboard” because these are the letters found at the top left of the keyboard configuration. Because the Qwerty keyboard has been used for many years, it might be assumed that it is highly functional. “If it weren't efficient,” the reasoning goes, “we wouldn't use it.” However, the Qwerty keyboard was actually developed at a time when manual typewriters would jam if keys were pressed too quickly. The Qwerty keyboard was therefore designed to [Page 120]slow down typing, not speed it up. This is why some of the most commonly used letters in typing, such as “e” and “t,” are located in relatively hard-to-reach positions, whereas less commonly used letters, such as “j” and “f,” are located in the easy-to-reach positions where the fingers normally rest. The Qwerty keyboard was highly functional when it was designed, but today it is less functional. More functional modern keyboards have been designed, keyboards that would make typing faster rather than slower, but people do not like to change, so we stick with the less efficient Qwerty keyboard. In evaluating a model, it is crucial that it be placed in its historical context and that popularity and mere existence not be equated with functionality and efficiency, lest we make the error of mistakenly promoting the status quo.
Related to this is a third criticism of models, that they “can tend to perpetuate some initial questionable, but fundamental, assumptions about the components of a model or the processes at work” (McQuail & Windahl, 1993, p. 3). This, too, critics say, produces a status quo bias, retarding theoretical development. For example, some early models treated communication essentially as a linear and unidirectional process—the movement of a message from a sender to a receiver. Lasswell's (1948) model mentioned earlier is an example of one such model that also has been highly influential. Another highly influential model of the communication process that emerged concurrently with Lasswell's is Shannon and Weaver's (1949). It, too, depicted communication as a one-way process (see Figure 7.5). Lasswell's and Shannon and Weaver's models suited their purposes well, but for other purposes some of the assumptions they make may have lingered well beyond their usefulness. Perhaps this is a criticism more of scholars than of models. When a scholar mindlessly adheres to an inadequate model, then yes, the model will almost certainly retard theoretical development. This is why it is such a bad idea to treat models as revered objects to be worshipped from afar. Models should be poked, prodded, kicked, and torn apart. One of the greatest purposes a model can serve is to be challenged and found wanting.
Other assumptions that models make, however, can be even more general and perhaps even insidious, critics say. As Wallace (1994) noted, “Typically, models are designed to handle routine situations. Assumptions of normality in a situation are inherent in most models. In an unusual or exceptional situation, the model provides no support” (p. 3). Not only can the application of a model in the wrong situation provide no support; it can be downright catastrophic. For example, financial institutions can [Page 121]amass amazing amounts of information about potential borrowers and create “profiles” of good credit risks. Slavish reliance on such techniques may make life easier for “model borrowers” but at the same time unfairly discriminate against others whose credit may be fine but whose circumstances are not typical enough to fit the standard model.
A fourth criticism is that models “tend to trap their originators and users within rather limited confines which they then become eager to defend against attack” (McQuail & Windahl, 1993, p. 3). Of course, this criticism can be made about any scientific endeavor, including theoreticians who refuse to let adverse discovery interfere with their cherished theories. One of the strengths of the scientific method is its ability to uncover human willfulness and capriciousness, but we should never forget that scientists are humans and humans tend to tout the familiar and to resist change (Cohen & Nagel, 1934). In many cases, these failings are neither deliberate nor conscious, but the effect nonetheless may be to retard scientific progress. It therefore is a failing against which we ought to be on guard, whether we are building models or using models built by others.
A fifth criticism of models is that they distract us from the primary mission of science, which is to develop definitive causal explanations. Critics say that the identification of powerful causal agents is deemphasized or even ignored in most models. Model builders counter by saying that they can use models to help them suggest predictions and explanations and that as long as model building is viewed as a means to an end and not an end in itself models can serve a useful purpose.
[Page 122]Underlying many of these criticisms runs a current of general dissatisfaction based on the simplicity of models. When scrutinizing a model, one can easily lose sight of the fact that it is simple by design. As McQuail and Windahl (1993) said of models,
They are inevitably incomplete, oversimplified and involve some concealed assumptions. There is certainly no model that is suitable for all purposes and all levels of analysis and it is important to choose the correct model for the purpose one has in mind. (p. 3)
Severin and Tankard (1997) echoed this idea:
No one model can “do it all.” Even if it could, it would defeat the purpose of a model—a simplified representation of the real world. … If none is available to do the job required, the researcher might well be forced to modify an existing model or even invent a new one. (p. 67)
According to Kress and van Leeuwen (1996), “It is never the ‘whole object’ but only ever its criterial aspects which are represented” (p. 6), and Holmes and Hundley (1997) referred to a model as “the artifact of choices between what should be included and excluded” (p. 14). If the model builder articulates the criteria used to select the elements and connections included in the model, that can go a long way toward alleviating objections of this sort.
Despite these warnings, scientists often display their models as if, in fact, they do it all. Kaplan (1964b) asserted that the notion that one can construct “a single comprehensive model” of anything “is no more than a prejudice” (p. 288). As he playfully put it, “Models are undeniably beautiful, and a man may justly be proud to be seen in their company. But they may have their hidden vices. The question is, after all, not whether they are good to look at, but whether we can live happily with them” (p. 288).
Another challenge facing the student interested in learning how to build and use models is the seemingly bewildering array of model typologies and the different labels attached to what appear to be very [Page 123]similar types, as well as the use of the same label for what appear to be different types. Earlier we alluded to a basic distinction between physical and conceptual models. Physical models, even when they are accurate in scale and precise in detail, offer little beyond simple description. However, we should not belittle such a contribution. Physical models can help us communicate our ideas more clearly, they can serve as heuristic aids, and they can keep us focused on certain elements and connections. They also can be helpful in testing theories.
McQuail and Windahl (1993) made a useful distinction between structural and functional models. Structural models “claim only to describe the structure of a phenomenon” (p. 2). A diagram of a radio receiver and a sketch of a metropolitan daily newsroom are examples of structural models. Functional models, on the other hand, “describe systems in terms of energy, forces and their direction, the relations between parts and the influence of one part on another” (p. 3). Models showing how radio stations or newsrooms operate would be functional. McQuail and Windahl maintain that because communication is in some degree dynamic and involves at least some elements that change states, nearly all models of the communication process are functional rather than structural.
Figure 7.6 is the organization chart of the Wall Street Journal as rendered by Shoemaker and Reese (1996, p. 152). It shows the structure of a complex organization in terms of newsroom personnel (boxes) and how they are related (arrows). This “bottom-up” model shows which elements in the system report to which others, from reporters at the bottom of the structure to the company chairman at the top. The model clearly describes the manifest power structure at the newspaper. Notice, for example, that the editorials department is completely separated from the news side, with the heads of both divisions reporting directly to the newspaper publisher.
Figure 7.7 is also a model that describes a news organization, but rather than focusing on the structure of the newsroom it focuses on the functional activities that lead to the production of news. This model, by Ericson, Baranak, and Chan (1987), shows how news story ideas are generated and developed. Ideas can flow from external sources, reporters, and news services. Ideas also can lead to sources, reporters, and news services. Once a story idea is identified, it is produced through a series of steps that start with story organization and end with the finished product (content page or anchor script). This model, influenced by Galtung [Page 124]and Ruge's (1965) theory of gatekeeping, shows the classic “winnowing” process through which editorial content is developed. Gatekeepers at various points in the production process let some ideas through while barring others, resulting in what the editor considers to be all the news that is fit to print or, at least, fits (Shoemaker, 1991).
Holmes and Hundley (1997) further categorized conceptual and functional communication models into three types. The “action” model [Page 125]treats communication as “something that one person does to another” (p. 8). A variety of labels have been applied to such models, including the “conduit” model, the “injection” model, the “hypodermic needle” model, the “transmission” model, and the “linear” model. Because communication in the action model is unidirectional and unilateral, it resembles what psychologists refer to as the “stimulus-response” model. An example of an action model of communication is Shannon and Weaver's (1949) model based on information theory, mentioned earlier. In this model, the focus is upon the system's capacity to [Page 126]transmit the maximum amount of information successfully—that is, without unacceptable levels of distortion or “noise.”
The “interactional” model builds upon the action model through the addition of feedback loops. Because they are more concerned about how humans communicate, Schramm's (1954) early models of communication add feedback as an important element in the process. Such models treat communication “as an alternating exchange of messages” (Holmes & Hundley, 1997, p. 9).
Figure 7.8 is an interactional model of communication originally proposed by Osgood in an unpublished paper and popularized by Schramm (1954, p. 24). Unlike the Shannon and Weaver (1949) model, this one treats message senders and receivers as essentially equivocal, with each party contributing to the sharing of meaning that occurs when communication takes place. Here, both parties encode messages: That is, they put their ideas—their mental images—into symbols they believe the other party will understand. The encoded message (m) is then decoded by the receiver: that is, translated from the symbols back into mental images. The process may be repeated until one or both parties tire of or are prevented from communicating, or both parties feel sufficiently satisfied that further communication is unnecessary.
Holmes and Hundley's (1997) third type of communication model, the “transactional model,” was intended to overcome the one-directional nature of action models and the constraints of strictly alternating message exchanges found in interactional models. “The key differences… are simultaneity of encoding and decoding and fusing [Page 127]of sending/receiving activities” (p. 10). Berko, Wolvin, and Wolvin's (1992) model of communication is an example, as are Schramm's (1954) later models of communication, which include overlapping “fields of experience” as an important ingredient in the communication process (Figure 7.9).
The notion of “fields of experience” is important because it reminds us that efficient communication can take place only to the extent that the parties involved share certain basic experiences. If two people do not speak the same language, they may have a hard time decoding each other's messages. A newsroom that does not reflect demographically and culturally the diversity of its targeted audience may communicate poorly with at least some segments of the audience.
Two points are worth making here, one fairly obvious, one not so obvious. We can see how productive it can be to build upon earlier models by adding elements and connections ignored previously but that we now consider important. However, though one might be tempted to think that the transactional model is “better” than the interactional model, which is “better” than the action model, it is important to keep in mind the purposes for which the model was designed. Shannon and Weaver's (1949) simpler model of the communication process is better at capturing what these model builders were attempting to capture than other models that added components irrelevant to Shannon and Weaver's concerns. Just because a model is more complicated than another, and just because a model adds something that an [Page 128]earlier model is lacking, does not mean that it is necessarily an improvement. A model needs to be evaluated in terms of its purposes. As Hanneman (1988) noted, whereas one social scientist may be trying to create time- and space-invariant general laws of social behavior, another may be attempting to uncover and understand the “deep structure” of everyday life (p. 16). Because social scientists theorize about a remarkably diverse range of subjects, we should not be surprised to find an equally remarkable diversity in the kinds of models they employ.
In the study of communication, McQuail (1994) identified four distinct types of models, the first of which he called the “transmission” model. However, what McQuail considered to be a transmission model is not the same as what Holmes and Hundley (1997) meant by the same label. According to McQuail, a transmission model of communication is any model that treats communication as a process of the transmission of a fixed quantity of information—a message—that is determined by a sender or a source. Thus, the removal of unidirectional activity through the addition of feedback and the addition of simultaneous exchanges are not distinctions that make for a new type of model. Therefore, McQuail essentially ignored the distinctions made by Holmes and Hundley and lumped their three types of models together as transmission models. The Westley and MacLean (1957) model, which according to McQuail is perhaps the most complete and most highly regarded early version of a model of mass communication, is thus still a transmission model, even though it contains feedback. Holmes and Hundley (1997), in contrast, would label this an interactional model (Figure 7.10).
Regardless of whether one considers the Westley and MacLean (1957) model to be a transmission model or an interactional one, if we take the time to examine it, we will see what a well-formed and insightful model it is for describing the process of mass communication. First, events (xs) impinge on a person's (B) senses. Second, some events sensed by a communicator (A) are transmitted to the person, whereas others are not. This communicator is thus the “source” of some information the person receives. Third, some events are sensed by a special type of communicator, a “nonpurposive encoder (C) acting for B,” such as a news reporter. Again, C transmits only some messages to the person. Finally, these sentinels (C) also receive information from sources (A), only some of which are transmitted to the person (B). By focusing on the “winnowing” process from the receiver's perspective, the Westley-MacLean [Page 129]model nicely complements the Ericson et al. (1987) model, which focuses on the process from the news gatherer's point of view.
At first glance, the Westley and MacLean (1957) model may appear to be quite intimidating because of its complexity. One of the authors well remembers when he first encountered this forbidding-looking model at the start of his graduate studies. After receiving a copy of the model in class, he turned to one of the other students in the seminar and whispered, “This looks like Custer's model of communication.” [Page 130]Puzzled, the other student asked, “Custer's model of communication?” to which he replied, “Yes, just look at all those @#$%*& arrows.”
Such irreverence notwithstanding, this anecdote points to the problem of attempting to construct a model that, for our purpose, is neither too unrealistically simple nor too inaccessibly complex. Generally, we want our models to be as simple as possible without being too crude. Like Goldilocks in the fairy tale, this may mean going to bed with a number of different versions until we arrive at one that feels just right.
McQuail's (1994) second type of model he called a “ritual” or “expressive” model. According to McQuail:
Ritual or expressive communication depends on shared understandings and emotions. It is celebratory, consummatory (an end in itself) and decorative rather than utilitarian.… Communication is engaged in for the pleasures of reception as much as for any useful purpose. The message of ritual communication is usually latent and ambiguous, depending on associations and symbols which are not chosen by the participants but made available in culture. (p. 51)
McQuail noted that for some media operations, such as news and advertising processes, the transmission model is useful but that for other media activities a ritual model does a better job. Carey (1975) offered a ritual model of communication that focuses on sharing, participating, and associating. What is considered important in this model is not “the extension of messages in space but the maintenance of society in time; not the act of imparting information but the representation of shared beliefs” (McQuail, 1994, p. 51). Another example is McQuail and Windahl's (1993) “Christmas spruce model of ritual communication” (p. 55), which consists simply of the iconic image of a tree decorated for the holidays. “In one culture at least, it symbolizes ideas and values shared and understood albeit vaguely and variously. There is clearly no instrumental purpose” (p. 55). A ritual model of communication also might describe communication that occurs during times of crisis, when the form of communication may be more important than its content and when communication may be distinguished more by its uniformity (“We stand united”) than by anything else.
McQuail called his third type of communication model a “publicity” model. McQuail noted that often the primary aim of mass media is [Page 131]neither to transmit particular information nor to unite a public in some expression of culture, belief, or values, but simply to catch and hold people's attention. In doing so, the media attain one direct economic goal, which is to gain audience revenue, and an indirect one, which is to sell audience attention to advertisers. One might say that the primary goal of some forms of communication is to sell eyeballs.
A publicity model can be useful when comparing the performances of different types of media systems, such as commercial television and public broadcasting stations, or national media systems that predominate in the use of one or the other of these types of media systems. For example, the United States has a predominantly commercial-based broadcasting system in which success is measured primarily in terms of audience size, whereas the Netherlands has a predominantly public-based system in which capturing a large audience is relatively less important. This means that in the United States much attention is given to finding efficient ways to grab the audience's attention. However, in the study of news programming formats and content, it is known that factors that increase attention can simultaneously decrease comprehension, and vice versa. Thus, sets of brief stories with dramatic visuals and sounds may draw audience attention but at the cost of greater comprehension. Longer stories with “talking heads” and less vivid pictures and sounds may improve comprehension but at the expense of losing audience attention. This means that media systems that seek to increase attention to their news products may at the same time inadvertently decrease comprehension of their news products. A national media system built upon a publicity model of communication may relegate the importance of audience comprehension behind that of getting a large share of the audience. A comprehensive comparison of attention and comprehension factors in news broadcasts in the United States and the Netherlands shows that these two national broadcasting systems do differ markedly in how they present the news. The publicity model of communication helps show why they differ (Klijn, 1998).
McQuail's fourth model of communication is called a “reception” model. The focus of this model is on the receiver and the power of the audience in giving meaning to messages. In this model, all messages are considered “polysemic”: that is, possessing multiple meanings. The context and the culture of the receiver dictate the meaning the message will have. Hall (1980), for example, showed the stages of transformation through which a media message passes on the way from its origins [Page 132]to its reception and finally to its interpretation. Arrows with broken lines can be used to indicate that audience members have the ability to “read between the lines” and can even reverse and subvert the intended direction of the message (Figure 7.11).
This reception model presents unique challenges to the social scientist who must grapple with the many possible readings, including oppositional ones, that an audience can give to a message. For example, an executive of the U.S. automobile giant General Motors once proclaimed that what is good for General Motors is good for the country, a comment that became a catch-phrase in the debate over corporate power in America. This occurred while the company's Chevrolet division was running a long-standing advertising campaign with the slogan “See the U.S.A. in your Chevrolet.” Though it appears that few receivers of this advertising slogan saw a connection between it and the other catch-phrase, some did, and the “message” they received encouraged them not to purchase a Chevrolet but to do the very opposite. As McQuail (1994) noted,
The essence of the “reception approach” is to locate the attribution and construction of meaning (derived from the media) with the receiver. … While early effects research recognized the fact of selective perception, this was seen as a limitation on, or a condition of, the transmission model, rather than part of a quite different perspective. (pp. 53–54)
As models become more complicated (e.g., as they add more parts and connections in an effort to cover more territory or to delve more deeply into territory already covered) and as they become more dynamic (e.g., as they add more moving parts and connections in an effort to show change or the possibility of change), they become more difficult to depict. Holmes and Hundley (1997) reviewed mass communication texts on this point and found that in those that compare the action, interactional, and transactional models of communication, “the transaction model is identified in these textbooks as the best, most complete, and most accurate” (p. 11). At the same time, they said, “The authors of these texts also describe the transactional model as the most difficult to visualize” (p. 11).
[Page 133]Adler and Towne (1990) suggested that model builders may become dissatisfied with two- and three-dimensional static representations and that “an animated version in which environments, communicators, and messages constantly change would be an even better way of capturing the process” (p. 15). Advancements in computer and communication technologies soon may make such ideas more practical. As Holmes and Hundley (1997) have stated, “Representing a process entails visualizing a story” (p. 12). The crucial point is to have a good story to tell, for then considerable effort may be worth putting into the visualizing of it. On the other hand, as Kaplan (1964b) suggested, a beautiful model may be momentarily distracting, but if underneath the glitter there is not much substance we tire and move on.
Models in the social sciences tend to follow certain general conventions, what Kress and van Leeuwen (1996) called the “grammar [Page 134]of visual design.” Holmes and Hundley (1997) noted that although this grammar is incomplete, it reveals “a surprisingly complex set of conventions governing the production of meaning” (p. 19). Models contain objects, which include participants (actors) and goals (actions). The activity engaged in is represented generally as a vector. A vector is simply something that has magnitude and direction. It is commonly represented by a directed line segment whose length represents the magnitude and whose orientation in space represents the direction. Many models indicate vectors as a line with an arrowhead. Lines with arrowheads at both ends indicate bidirectionality. Besides line length, another common way to indicate magnitude is line width. Overlapping of actors or fields that identify actors can imply a vector, as in Schramm's (1954) overlapping “fields of experience” to illustrate shared meaning in some of his communication models.
Vectors in communication models routinely are unidirectional or bidirectional and occur singly or in pairs. The direction and number of the primary message vectors are the key differences among action, interactional, and transactional models of communication. Action models (e.g., Lasswell, 1948) illustrate a one-way vector from sender to receiver, typically using a single vector with one arrowhead. Interactional models (e.g., Osgood's, cited in Schramm, 1954) illustrate a pair of unidirectional, alternating vectors, from sender to receiver and from receiver to sender (feedback response). Transactional models (e.g., Ericson et al., 1987) illustrate simultaneous bidirectionality with arrowheads at both ends of a vector.
As we have seen, a model can help a scientist stay focused by highlighting those parts of a system and the connections among them that the scientist thinks are most relevant and most worthy of study. A model also can help a scientist communicate to others the parts and connections of a system considered most relevant. For the theory builder, a model also can suggest new theoretical statements. It is this last use of models that we discuss here.
As we noted earlier, when a scientist is interested in explaining how something works or in predicting the outcome of a process, he or she often begins simply, perhaps with one independent variable believed to influence one dependent variable. Often, however, the [Page 135]scientist begins to think about how other variables might be involved. As more variables are brought into the picture, the number of possibilities as to how they can be combined greatly increases. For example, if we consider all the possible bivariate combinations with 5 variables, 20 theoretical statements can be derived. With 10 variables, the possible statements increase to 90. (The formula is (n×n) – n.)
Because a model can help suggest relationships, it also can help eliminate them. One powerful technique is to consider the approximate time ordering of the variables. Suppose, for example, that we are interested in how journalists use experts as sources in news stories and that, in particular, we want to know if the gender of the expert is a factor in the frequency of use. Gender is a characteristic that one receives early in life, and for the overwhelming majority of the population it does not change. Therefore, in a study such as this one, it is reasonable to consider gender as a possible cause of other events but not to view it as an effect. Gender therefore would be ordered in time before other variables. In a simple two-dimensional model using the standard visual grammar, gender might appear to the left of other variables. Arrows representing causation might be seen to proceed from gender, but none would proceed into it. By time-ordering the variables in a model, we eliminate from consideration theoretical statements that contradict that order. The result may be what Hage (1972) called a “flow chart of theoretical statements” (p. 56).
Working with a model in this way also may help us recognize subsets of variables that, in effect, represent chains of causes or effects. For example, as we examine our model we may recognize that variables fall into classes or types. We may be able to “bundle” these. Suppose, for example, that we are studying social control in a newsroom. We may be interested in some variables that are outside the newsroom. Other variables may represent newsroom structures, and still others may represent newsroom resources, newsroom outputs, and so forth. We may find not only that variables cluster into classes but that there is a time ordering to these blocks, and it may therefore be possible to establish a unidirectional flow.
Once this has been done, the next step may be to consider feedback. Variables and subsets of variables that move in the opposite direction of the established flow may produce important feedback effects that should not be ignored if one hopes to approximate reality. In the case of social control in the newsroom, for example, the extent to which management hears and addresses the concerns of workers may or may [Page 136]not affect the operation of the system. As Hage (1972) noted of the model-building process, “This may sound like an impossible task, but once variables are written down on a piece of paper or a blackboard, theoretical statements connecting them should quickly come to mind. Finding the variables has always been the harder task” (pp. 57–58).
What we have been describing here is essentially a form of path analysis, using a model as a guide. The product of our analysis—the theoretical statements we derive—will be only as good as the model. If the map is poor, it may lead us into quicksand or over a cliff. To put it in the parlance of computers, “garbage in, garbage out.” Computers can easily generate complex path models, but computers cannot easily evaluate them. Scientists may be interested in whether one concept is connected with another, but they usually are interested in more than that. They are interested in causes and consequences.
A productive method for addressing the question of causes and effects is to examine several related case studies. Take a variety of situations from different times and places so that general theoretical statements are derived. Suppose, for example, that we are interested in the role of the mass media in the successful overthrow of national leaders. We might select the Marcos regime in the Philippines, the French Revolution, the Spanish Civil War, Watergate in the United States, the collapse of Czarist Russia or the Soviet Union, and any of a number of other diverse cases that might differ in terms of location, time period, length, violence, and other important variables. The next step is to select these same places at times of relative stability: that is, cases in which the phenomenon exists or predominates and others in which it does not exist or is modest. Thus, we observe the Philippines or France 5 or 100 years earlier or later. We then look for commonalties across the selected cases. We look not only for the presence of factors during times of crisis but for the absence of these same factors during times of stability, and vice versa. How do the media behave across these different times and places? What are the similarities, the differences? Does the role of the media change before, during, and after a coup, and, if so, how? What patterns do we recognize? At this point, some readers might say, “This seems like an awful lot of trouble to go through just to derive interesting theoretical statements.” This is true. Constructing productive models and deriving from them interesting theoretical statements is, alas, an awful lot of trouble. On the other hand, building on existing models can be much less daunting and may be a good way to get one's feet wet.
Suppose one wants to model cultivation theory (Gerbner & Gross, 1976). This theory holds that television possesses properties that make it an especially powerful influence, such as its relatively high penetration rate (the percentage of the population watching) and saturation rate (the percentage of time people spend watching). Additionally, the theory maintains, the “TV world” varies systematically from the “real world,” transmitting constant and consonant moral messages about risk and power. The result, the theory proposes, is that viewers think that the world is scarier than it really is. Thus, for example, heavy viewers have a greater fear of crime.
Because cultivation is a communication process or, more precisely, a mass communication process, we should be able to relate it to any general model of communication, such as the barriers, Shannon and Weaver (1949), or Westley and MacLean (1957) models. However, because any model focuses selectively on a process's elements and connections, we should expect some models to relate better than others to a particular theory.
Consider, for example, how well the Lasswell (1948) model describes the cultivation process. This model begins with a source saying something. It therefore does not focus our attention on important aspects of the cultivation process that deal with why someone says something. At the same time, Lasswell's model draws attention to the channel, whereas cultivation almost ignores channel factors, treating the channel more like a constant than a variable.
Years before Gerbner and Gross (1976) proposed their theory of cultivation, Braddock (1958) suggested that Lasswell's model be expanded to consider two important elements it ignores. He proposed that in addition to Lasswell's five questions to describe an act of communication, two questions be asked: (a) For what purpose? and (b) Under what circumstances? Braddock argued that a fuller understanding of the process of communication results when we consider both the sender's intention in sending a message and the circumstances under which the message is constructed and transmitted.
Braddock's expanded model does a better job of relating to the cultivation process than does Lasswell's (1948) original model. Again, this should not be viewed as a criticism of Lasswell's model; the purpose of his model was not to depict cultivation per se. There is a certain appeal in being able to consider an act of communication as an isolated event. [Page 138]Nonetheless, there also is a certain appeal to the idea that an act of communication does not really begin with someone saying something but that there are forces at work that lead someone to say something. This latter idea is certainly relevant to the cultivation process, if not to the communication process generally. If we are willing to assume that an act of communication does not occur in a vacuum and that the conditions leading to an utterance are important to consider, then it is possible to expand the Lasswell model in other provocative ways as well. We might, for example, include the idea that reality (the “real world”) is viewed through the filter of one's culture (e.g., the “TV world”) and that this cultural filtration influences what gets said, or the idea that communication is not linear but cyclical, as in Osgood's model including feedback (cited in Schramm, 1954). If one ignores the italicized terms, Figure 7.12 is an example of a general model of communication that incorporates these ideas. The italicized terms change the figure into a model of the cultivation process. More recent advancements to that theory (e.g., mainstreaming and resonance) could be incorporated into this model as well.
Modifying an existing model to suit one's needs is a tried and true method for advancing science. A review of the models presented here will show the extent to which many have adapted elements of earlier ones. Given that they often are attempting to model the same general process, this should not be surprising. Furthermore, because models are by definition selective, even a slight change tends to have important implications. For example, DeFleur (1970) may have made only one major modification to the Shannon and Weaver (1949) model—adding feedback—but it represented a critical change.
By considering how well existing models represent new theories or theoretical developments, and by modifying these models to create more precise representations, the model builder can help others see more clearly commonalities across theories, how one theory fits into general models, how theoretical advancements make sense in terms of previous models, and other important aspects of theory building.
Because models are designed to clarify new theories and theoretical hunches, we should expect models to reflect the theoretical interests of the day. Thus, when theories at one time tend to focus upon certain [Page 139]aspects of communication while ignoring others, models should be expected to do the same. For example, when early theories of persuasion (e.g., Hovland, Janus, & Kelley, 1953) tended to focus on inputs and outcomes while treating the mind of the receiver as a “black box,” so did the models of that time. Later, when persuasion theories began to focus on such matters as the cognitive processing of information, the sequencing of steps over time (from exposure to yielding), and the varying activity of the receiver, persuasion models began to reflect such interests as well (e.g., McGuire, 1968). More recent theories of persuasion have tended to maintain that humans process information in two distinct ways. When one has the ability and motivation to process, one tends to do so thoughtfully, analytically, and carefully, but when one lacks ability or motivation, one tends to rely instead on peripheral cues and simple heuristics to process information. So-called “dual-process models” of persuasion, such as Petty and Cacioppo's (1981) [Page 140]elaboration likelihood model and Chaiken's (1987) heuristic-systematic model, have gone a long way toward reconciling many of the conflicting findings from earlier research. Dual-process models of persuasion have had two important consequences. They help us see how the same variable (e.g., source credibility) can have different effects on persuasion depending upon whether the variable is a component of an elaborate, systematic processing of information or a simple cue or heuristic (“I yield because I trust the source”). They also point out how attitude change can be enduring, resistant to attack, and predictive of behavior because it results from persuasion through the elaborate, systematic—“central”—processing route or can be relatively ephemeral, easily attacked, and not very predictive of behavior because it results from the cue-based, heuristic-based—“peripheral”—route (Figure 7.13).
Likewise, advancements in information-processing theories such as schema theory have led to new models of perception and cognition (e.g., Axelrod, 1973) that have themselves encouraged the development of more sophisticated information-processing models, such as Graber's (1984) model of news processing. In such models, new information is evaluated in terms of existing schemas (variously called “frames” or “scripts,” depending upon the model), which are cognitive structures that codify recurring patterns of experience and make communication efficient. Thus, we have a “restaurant schema” that, when invoked by a message, allows the recipient to focus only on details of the specific message to fill in important details. As long as the sender and receiver both possess a good restaurant script, many routine details can be left out of the message because the script assumes them. We need not mention, for instance, that we ordered food or that we received eating utensils. Communication breaks down when sender and receiver do not share schemas or when a schema's assumptions are violated but not mentioned. Problems also arise when new information violates an existing frame so that the frame has to be modified or discarded. Although that is troublesome, what's worse is when the information, however valuable, is ignored because it does not fit into an existing frame. Processing news thus becomes a matter of fitting information into existing scripts, which are thereby fortified; using information to tweak existing scripts, which are thereby weakened; using information to obliterate a script entirely, which is rare due to its cognitive costs (we have much invested in our scripts); or dismissing the information altogether because it does not jibe well enough with existing scripts. Perhaps the New York Times should change its motto to “All the news that's fit to script” (Figure 7.14).
[Page 141]Hanneman (1988) envisioned changes not only in the content of models but in their form: that is, in the way scientists build models. These changes in form and content go hand in hand. Hanneman noted that the observations social scientists make often “are composed of multiple simultaneous causal processes, operating along multiple dimensions, and occurring both within and between social actors. Such processes are inherently complicated” (p. 3). Because of this, he suggested that more attention be given to dynamics, that models should use more formal languages, and that greater use be made of [Page 142]computer-assisted model building and testing. Adoption of these approaches, Hanneman claimed, would also “bridge major gaps between those who study ‘structures’ and those who study ‘processes,’ [Page 143]and gaps between those who use ‘natural’ language to present their theories and those who use formal languages (e.g., mathematics or symbolic logic)” (p. 10). As he said, “The methods of representing patterns of social dynamics in formal models and understanding their implications for computer-assisted experiments have a great deal to offer” (p. 10).
If this were not opportunity enough, Baran and Davis (1995) made the model builder's life even more adventurous by recognizing recent advancements in chaos theory (Prigogine & Stengers, 1984) showing that many important systems, both natural and social, undergo fundamental transformations that cannot be predicted by simply examining past behavior (p. 259). As Gleick (1987) noted, “The traditional models are betrayed.… Nature is more complicated” (p. 315). For example, ambiguous messages may be deliberately transmitted precisely because they can serve different functions simultaneously. In many communication models, noise is viewed as undesirable because it is assumed to interfere with the efficiency of the transmission of the message. Shannon and Weaver's (1949) model of communication, for instance, explicitly treats noise this way. Indeed, that model was designed specifically to tackle the problem of minimizing the interference of noise in the telephonic transmission of messages. Shannon worked for AT&T's Bell Labs, and he wanted to pack as many phone call conversations onto a wire as possible without crossover of conversations or other interference. The Shannon and Weaver model viewed noise as a detriment to communication. However, is it always? What of the evasive speaker who is deliberately ambiguous about his or her position on an issue? An effective election campaigner may use unclear rhetoric that may be easily read two or more ways. What of the clever poet who uses words with two or more meanings to evoke multiple images? As Krippendorff (1986) noted, “Noise need not be undesirable as in creative pursuits or in political discourse, in which ambiguity may be intentional” (p. 21). Therefore, models of certain communication processes should be developed in which noise is considered functional rather than disruptive. Examples of the application of chaos theory to social science models also can be found in economics (Arthur, 1990; Grandmont, 1985), political science (Huckfeldt, 1990; Saperstein & Mayer-Kress, 1989), and sociology (Harvey & Reed, 1996; Young, 1991).
There is a bittersweet irony here. As Kiel and Elliott (1996) pointed out:
Clearly, the fundamental gap between the clear success of knowledge acquisition in the natural sciences versus the rather minimal [Page 144]successes in understanding the dynamics of the social realm is the inherent nonlinearity, instability, and uncertainty of social systems behavior.… Yet chaos theory teaches that the “gap” between the two sciences may have largely been artificial. As natural scientists more intensively investigate complex natural phenomena, they too must contend with the challenges that have long served to keep the social sciences in the position of a scientific stepchild. Chaos theory seems to represent a promising means for a convergence of the sciences that will serve to enhance understanding of both natural and social phenomena. (p. 3)
As both natural and social scientists continue to learn more about the worlds they seek to understand, the models they use will become more varied, more interesting, perhaps more challenging and complex, and, we hope, more useful. Keeping in mind the shortcomings of model building discussed here, we hope further that model builders take to heart the words of Kaplan (1964b), who knew well the value of taking the good wherever he found it: “The dangers are not in working with models, but in working with too few, and those too much alike, and above all, in belittling any efforts to work with anything else. That Euclid alone has looked on beauty bare is a romantic fiction” (p. 293).
Though it is understandable why scientists have been intimidated and confused by building models, it is clear that much can be gained from working with them. If we keep in mind the limitations of models, we may find that by adopting and adapting existing models and by constructing new ones, we may learn much about the objects and processes we study, and we may better convey what we have learned about them to others. Perhaps as a first step we might stop talking about “working with models” and talk instead of “playing with models.” Like all the other tools scientists use to gather information and convey it, models shouldn't frighten us. They should be fun.
- physical models
- models and modeling
- theories and models
- communication processes
- processing the news