Emergent Complexity and the
Role Attributes of Media

Davis Foulger
CUNY Brooklyn College, Montclair State University, and Evolutionary Media
May, 2005
Presented at the Spring, 2005 Meeting of the International Communication Association


This paper explores the "roles in media" as a consequence of the processes of structuration. It treats the generic roles that people play in enabling the smooth operation of a medium as generically useful solutions that solve recurrent problems across a variety of media. The processes of structuration are explored through the development of a typology, based on a dataset that codes 18 generic roles across 167 distinct media. Seven distinct role-based clusters of media are identified within a two dimensional solution. When the clusters are viewed as endpoints within these dimensions, several emergent solutions to role complexity in media are observed, each of which appears to be a consequence of a different kind of complexity. One of these solutions appears to parallel the hierarchical subassemblies suggested by Simon (1969) as the single emergent solution to complexity. The other emergent solutions to role complexity, serial complexity and floor contention, appear entail very different role profiles. Four research questions are satisfied by the results. The typology successfully groups structurally similar media. The structures reflect general solutions to problems that are encountered in the operation of media. A fundamental set of problems, different kinds of complexity, engenders the solutions. Finally, the typology is found to have practical value in its suggestion that that while technology may enable the convergence of user interface devices for very different media, role structures may make it difficult to merge companies that manage different kinds of media.


A medium of communication is, in its essense, a system that has been invented and structured by people for the purpose of enabling and supporting the exchange of messages. This is true for all media, even the simplest forms of intimate communication (which are by no means simple when the rules of such interaction are taken into account). It is more obvious as media grow more complex, however, such that there is a tendency to think of complex media that reach audiences of millions as "the media" and to treat "simpler" systems like face-to-face interaction as contexts. It remains that there are a wide range of generic characteristics by which any medium of communication, including such diverse media as the telephone, oil paintings, films, small group interaction, legislative assemblies, newsletters, live musical performances, books, broadcast radio, and many other media can be usefully compared.

One such set of characteristics are the roles that people play in the routine functioning of media. Giddens (1986) directly identifies the roles that people play in social structures, including communication systems, as an emergent, yet enduring, social structure. Social roles are, from Giddens perspective, building blocks with which we build the structures and processes of conjoint activity. Taylor et al (2001) extend this notion in two ways. First they assert that communication is the "primary modality" of Giddens' (1986) structuration (e.g. the primary means by which we negotiate social structures). Second, they examine the impact that various communication roles and structures played in the success and failure of new media in various organizations. Foulger (2002a) further extends this view of communication roles as structural building blocks by asserting a set of generic roles that people play in enabling the smooth functioning of individual media. These "roles in media" recur across a broad range of very different media, but combine in very different ways in different media.

If media are, following the examples of Taylor et al (2001), social structures that are themselves subject to the processes of structuration, then it follows that media are human inventions, a view that has been expressed previously (Levinson, 1979, 1997; Foulger, 1992b). If Giddens (1986) is correct in asserting that social roles occur in all social structures as an emergent consequence of purposeful activity, it should certainly follow that the same is true in media. Hence it can be expected that media will change both in response to problems, as would be the case for any tool (Petrowski, 1992, 1994), and as the unintended consequence of otherwise intentional actions (Giddens, 1986). Hence when structures emerge in media, whether the the result of intention or accident, it can be reasonably expected that those changes will be generically useful solutions that make the medium more suitable for a particular set of purposes. This should be as true for the emergence of specific roles in a medium as it is for any other attribute of the medium. It seems likely, then that a recurrence of roles across media reflects both problems that recur across a variety of media and the effectiveness of particular roles and role combinations in solving those problems. .This is probably particularly true for media when they compete for the same uses (Foulger, 1990, 2002b).

This paper examines the distribution of eighteen generic roles in media, as identified in Foulger (2002a), across 167 discrete media, which each role operationalized and each medium assessed for each role. The 167 media, listed within emergent categories in Table 3, are largely drawn from those used in prior typologies of media (Bretz, 1971; Ciampa, 1989; Foulger, 1990, 1992; Hoffman and Novak, 1996). The eighteen roles, including transcriber/recorder, selector/gatekeeper, publisher/producer, director, performer, content editor, advertiser, content integrator, reproducer, distributor/carrier, retailer, representatives/advocate, critic, regulator, investor, financial manager, and general production support, are operationalized based on the role descriptions offered by Foulger (2002a).

It is hoped that exploration of the resulting data set will prove useful useful in the creation of generic typologies of media which:

  1. Organize media into structurally similar groups.
  2. Reveal a systematic set of generic solutions that are associated with the various groupings.
  3. Expose at least some of the generic problems that media encounter in the course of their invention and evolution.
  4. Provide practical guidance to individuals and social collectives engaged in building and evolving media.

This study will treat these four kinds of value as research questions within the scope of an exploratory data analysis that combines factor analysis, cluster analysis, and multiple regression.


The data set takes the form of eighteen "dummy coded" variables. Each variable describes the presence (1) or absence (0) of each of the eighteen roles for each of 167 media.. For all but three roles, the associated role variable was coded as yes (1) if the role was typically (e.g. often or usually) associated with the medium. These coding rules were relaxed somewhat for the regulator, collector, and critic roles, where a "reasonable possibility" coded as yes (1).

The analysis combines factor analysis, cluster analysis, stepwise multiple regression, and visualization techniques. Factor analysis (principle factor, oblique rotation) was used to extract a set of dimensions that describe the distribution of roles across media. The number of dimensions was selected via a scree test. Cluster analysis is used to meaningfully group media based on the distribution of roles based on factor scores. These groupings will be displayed as a scattergram within the dimensions of the cluster analysis. A subset of the resulting factors subjected to multiple regression that included both the variables that loaded on the factors and a composite variable. The composite variable, "sum of roles", was developed for use in the multiple regression based on the results of the factor and cluster analyses.

Preliminary statistics support the use of factor analysis. The Kaiser-Meyer-Olkin Measure of Sampling Adequacy is a "meritorious" .848, easily sufficient to suggest that factor analysis is appropriate. Bartlett's Test of Sphericity is sufficient (X2=.1358, df=153, p<.001) to reject an identity matrix. An examination of the scree plot (viewable at http://www.evolutionarymedia.com/papers/roles/screeplot.htm) suggests between one and three meaningful factors. The first eigenvalue is extremely strong, but a single factor solution results in difficult to interpret clusters. The fourth and fifth eigenvalues, although suggested as factors by little jiffy (eigenvalue>1) reside in the scree. The second and third factors barely rise barely above it, but the uninterpretable cluster structure of a one factor solution and its extremely strong eigenvalue suggests multiple strongly correlated factors, making an oblique rotation advisable. Although one, two, and three factor solutions were evaluated (see Table 1), the discussion of results will focus almost entirely on the two factor solution.

Role-Based Dimensions of Media

Almost all the variables in the data set load on the single dimension of the one factor solution. Thirteen load strongly with pattern coefficients of .496 or higher. Another three can be regarded as weak loadings with pattern coefficients of .316 or higher. Only two of the eighteen variables fail to load on this dimension. The pattern of the loadings suggest labeling this dimension as production roles. The consistent strength of the loadings suggest that simply summing the number of roles associated with a medium might prove useful. Hence a "sum of roles" variable was created by simply summing the number of roles assocaited with each medium.. By itself, "sum of roles" is not illuminating. It accounts for only 14% of the variance in the dimension that suggests it, and an examination of the distribution of media along the dimension wildly mixes very different media in a way that suggests the collapse of at least two underlying dimensions.

This suggestion is strongly supported in the two factor solution, where the two dimensions are strongly correlated (r=.392). Factor scores correlations are even stronger (r=.469, p<.001, df=1/166). Overall the two factors appear to share between 15% and 22% of their variance. The first and stronger of these dimensions picks off almost all of the strongest loadings of the one factor solution. The only obviously significant change in the structure of these loadings is a much stronger director role, which dominates the factor with its .902 loading. This change in the relationship of director to what are otherwise production-related roles leads to its labeling as a directed production factor. The second dimension adds two of the variables that did not load strongly in the one factor solution to two variables that did. While the purposes associated with these roles, including selector/gatekeeper, critic, collector, and retailer, differ, they share a an orientation to content discrimination and selection. This commonality justifies applying the label content selection to this factor. Interestingly, the director role also loads on this factor, albeit weakly (-.35) and in the opposite direction. This result is worth noting, as the director role will prove to be pivotal to our understanding of both of these dimensions.

The dimensions of the three factor solution preserve the general structure of the two factor solution. A third factor, Direction versus Collection, separates away from both the Production Roles and Content Selection. These dimensions differ from those of their counterparts in the two factor solution primarily in a weakened role of Director on each. It turns out, however, the this third dimension, Direction versus Collection, adds very little to a careful analysis of the two factor solution. Hence the three factor solution will not be discussed further here beyond noting that a third dimension may be in play.

Sum of Roles and the Director Role

Table 2 shows the primary results of two stepwise multiple regressions. Each regression tests the variables that loaded on the two dimensions of the two factor solution. It should be no surprise that a substantial portion of the variance in these two factors can be accounted for by their loadings. Factor scores are, by definition, the sum of their weighted loadings. The only good reason why one might normally perform a multiple regression of a factor's components against the factor itself is to generate a simpler explanation of the factor. That is, however, exactly what is accomplished in these regressions (see http://www.evolutionarymedia.com/papers/roles/regressions.htm for the complete stepwise regressions). Just two variables, "sum of roles" and the director role, account for most of the variance in the factor scores. The two variables are the first two selected in each of the the stepwise multiple regressions. Together they account for over 85% of the variance in the two factors (over 95% of the directed production dimension and over 85% of the content selection dimension).

The "sum of roles" variable is the more powerful. Alone it accounts for 87% of the variance in the directed production dimension and 55% of the variance in the content selection dimension. This result supports the inference (described above) that the one factor solution collapses at least two underlying dimensions. Both dimensions entail an increasing number of roles. They are simply characterized by different role distributions. This is an important observation, as these differing role distributions will be important in the remaining analysis. Indeed, they make a strong argument for doing a cluster analysis.

The director role proves to be at least as interesting. Its selection for the directed production factor, where it is the dominant variable on the factor is not surprising. The role of the director is obvious when one looks at the distribution of media on the factor. What is surprising is that it loads on the content selection dimension at all. Indeed, its direct correlation with the factor scores for content selection is effectively zero (r=.004, p<.958, n=167). Once "sum of roles" has been regressed against this dimension, however, the partial correlation of the Director role with the residuals soars -.827 (p<.001), allowing it to account for 68% of the remaining variance in the content selection dimension. The effect of sum of roles in both analyses is additive and consistent with the variance the dimensions share. The effect the director role reverses, however. It is additive for the directed production dimension and subtractive on the content selection dimension. What is exposed, in effect, is a statistical interaction effect involving two deeper dimensions, role complexity and coordination and control. This "third factor" is not obvious in the two factor solution. It will be better understood through a discussion of the cluster analysis.

Role-Based Clusters of Media

The cluster analysis of this two factor solution yields seven clusters of media. Figure 1 projects the 167 media within the two dimensions using their factor scores. The cluster arrangement of the figure forms a clockface. Working clockwise around the Figure from about 8:00, the six clusters include personal interactive media, personal art media, production art media, role intensive media, moderated interactive performances, and group interactive media. A seventh cluster, intermediate scale media, resides in the center. The names of these clusters are broadly descriptive of the media within the clusters. Three vectors are clearly visible in this clockface, and none exactly follow the left to right path of the directed production vector. Note that while these media are not labeled in Figure 1, they can be viewed, within clusters, in Table 3.

A vector from the lower left (8:00) to upper right hand (2:00) corner draws what initially appears to be the traditional distinction of interpersonal (personal interactive media) and mass media (role intensive media) as a continuum through intermediate scale media. It is better described, however, as a role complexity continuum. The personal interactive media cluster that occupies the lower left corner of the figure are generally relatively role free, with only two roles, distributor/carriers and retailers, occurring for as many as half the media in the cluster. The role intensive media cluster, by contrast, encompass all of the roles in media. Five roles are associated with all media within the cluster and another ten associate with most media in the cluster. The difference in role distribution at the extreme ends of this role complexity vector is a true difference between interpersonal and mass media that is not strictly a function of audience size.

A vector from the upper left (10:00) to the lower right hard (4:00) corner describes a path from personal art media (in which a usually solitary artist creates a work of art that are viewed by others at another point in time), to moderated interactive performances (in which a performance moderator such as a judge, chairperson, or referee manages the real time interaction of multiple participants). All of the media in the personal art media cluster involve highly individual creations that generally take a while to complete. Other roles associated with this cluster, including selector/gatekeeper, collector, and critic (all of which are associated with most of the media in this cluster), are invoked only after the creative process is complete. At the other end of the continuum, moderated interactive performances is fairly role heavy. All of the eighteen roles occur in at least some media in the cluster, eleven occur at least a third of media, and five are associated with most or all of the media in the cluster. The three roles that occur in all media in this cluster are key production roles, including publisher/producer, director, and production support. The continuum proceeds, then, from individualistic creation to coordinated performance, and might reasonably be regarded as a coordination and control continuum. One end of the continuum, personal art media, requires neither coordination nor control in any large measure. The other end is heavy on both.

A vector from production art media at the top (12:00) to group interactive media at the bottom (6:00) corresponds directly to the content selection dimension of the factor analysis. The continuum from group interactive media to production art media corresponds fairly directly to the content selection dimension, and it also entails an overall increase in in the number of roles typically associated with media. While no single role is associated with all production art media, nine are associated with at least two thirds of the media in the cluster and seventeen occur for at least a third of the media in the cluster. Production art media may not be as role heavy as role intensive media, but the cluster still entails many roles. Group interactive media, by contrast, is comparatively role light. Thirteen roles occur in no more than a third of the media in the cluster. Seventeen occur in no more than two thirds. It is obvious, given this variation, that "sum of media" gives a reasonable, if less extreme, description of this continuum. What is most interesting in this dimension, however, is the eighteenth role. A single role occurs in every medium in the group interactive media cluster. That same role, director, is absent in every medium in the production art media cluster.

The Management of Complexity in Media

What emerges, then, are three role patterns that intermix, in interesting ways, through an interaction effect. One pattern centers around the exigencies of complex and highly parallelized productions. Another centers around the selective, and generally highly serial, processes associated with the production and distribution of singular works of art. A third centers around the need for a director who provides, in effect, traffic control. These patterns can be regarded as emergent solutions to different kinds of complexity that occur in media. Indeed, these three solutions to the problem of managing complexity in media appear to respond to three distinct sources of complexity.

The dominant method, at least in terms of the variance it explains in this study, is the parallel complexity of many people doing many separate tasks at the same time. Almost all of the media in the role intensive media cluster entail tens or hundreds of people working in parallel on independent tasks that must be completed before a particular moment in time (a performance, publication deadline, competition, etc.). Success, where a medium entails parallel complexity, is often dependent on the efforts of publishers/producers, directors, financial managers, and others who act to coordinate tasks and resources.

Floor control or contention-based complexity, by contrast, inevitably arises when a group or larger number of people come together to jointly participate in an medium. Whether the activity is a committee meeting, league football game, town meeting, classroom discussion, or a courtroom, there is a critical mass (number of people) at which control of the floor can no longer be managed by interpersonal means alone. When this critical mass is reached there is a natural tendency for people to seek leadership to provide floor control, rules enforcement, and other services (possibly including scheduling and resource coordination, record keeping, and other production related roles).

A third method, serial complexity, dominates in media in which singular creations (oil paintings, quilts, photographs, blown glass, etc.) proceed from creation to consumption. Even when the original artist makes the initial sale to an end consumer (as will often be the case for oil paintings and pottery), the creations persist and may be collected, sold and resold. Indeed, critics, selector/gatekeepers, retailers, and collectors may continue to process a work serially long after the original artist is dead. When such creations are reproduced (as is often the case for prints, ceramics, and blown glass), additional roles come into play, with few operating in parallel. Selection for a print proceeds to production, investment, reproduction, distribution, retailing, criticism, collection, and perhaps resale. Production and financial management often proceed in parallel with these steps, but most steps are singular, involving only one or a few people, and have only a serial dependency on previous steps.

Figure 2 illustrates the ways in which these three solutions to different kinds of complexity intermix to create the clusters of media. The number of roles is higher at the labeled end of the continuum, which has been shaded with colors that intermix to form six facets. Contention is highlighted in a red that extends from the group interactive media cluster (the pink facet) to the role intensive media cluster (purple). Serial complexity is highlighted in a blue that extends from the personal art media cluster (the blue facet ) to role intensive media. Finally, a green highlight marks the role intensive end of the parallel complexity continuum, extending from production art media (blue-green) to moderated interactive performances (khaki green).

These combinations of color show that while serial complexity and contention can occur in relative isolation from the other forms of complexity, parallel complexity does not. The role pattern associated with group interactive media emerges as a result of contention for the floor. The role pattern associated with personal art media emerges as a result of the serial handling of singular creations by people in different roles. The role pattern associated with production art media reflects a limited parallelism of production overlaying a serial creation, selection, and distribution process. The role pattern associated with moderated interactive performance resolves the more complex floor control problems of larger scale and inherently contentious processes with formalized and paralleled mechanisms of floor control, rules enforcement, and production. Role intensive media, which will be examined via exemplar in the next section, combine all three.


The results of this analysis suggest that there are three different sources of, and a corresponding set of generic solutions to, role complexity in media. One variant of complexity arises from the problems of floor control when many people contend for the floor in media that involve live coordinated performances or high degrees of face to face interaction. A second variant arises from the problems of distribution associated with highly creative efforts that involve one or a few creators, but which over time may reach many consumers. A third variant arises from creating complex productions that involve bringing the coordinated efforts of large numbers of people to a conclusion on a schedule.

The seminal study of organizational complexity, Simon's (1969) The Architecture of Complexity, does not observe such variations. It proposes instead a single general emergent solution to complexity in hierarchies of subassemblies. Simon's suggests that this solution is universally applicable, and the examples cited range from the structure of matter, biological systems, organizational systems, and manufacturing, among others. While there have recently been an increasing expression of concern that this single solution is inadequate to explaining the range of emergent complexity (Thornton, 1997; Agre, In Press), Simon's single solution has yet to be seriously challenged. Indeed, it is widely reinforced by the considerable body of organizational and management theory, going back to Henri Fayol and Frederick Taylor, that describes organizations in terms of hierarchies of subassemblies. Indeed, Taylor et al's (2001) "situated groups" are, for all intents and purposes, hierarchical subassemblies.

These results suggest, at least for media, that hierarchical subassemblies are not the only solution to complexity. The three forms of emergent complexity that emerge in this study and their combinations, as described in Table 2, suggest that there are at least six major variations in the way media organize themselves, simply in terms of the mix of specialized roles involved. Only one of these clusters, role intensive media, can be reasonably be described with hierarchies of subassemblies alone. This doesn't necessarily negate the importance of subassemblies. Personal art media, to the extent they are organized at all, might reasonably be described as organizing themselves through chains of subassemblies. Group interactive media, to the extent they are organized at all, might reasonably be described as organizing themselves in minimally hierarchical parallel subassemblies. Moderated interactive performance media, while often entailing some degree of hierarchy and some degree of subassembly, don't necessarily entail a high degree of either. Production art media are generally characterized by chains of subassemblies, only some of which are likely to be hierarchically managed.

While these results don't challenge the notion of subassemblies per se, they do challenge the necessity of hierarchy. There are other ways in which subassemblies can be usefully organized, at least in terms of the organizational roles associated with human communication media. This study makes no larger claim than this: there appear to be at least three fundamental ways in which role complexity is managed in human communication media. There may be value in asking if these solutions to emergent complexity in media are sufficiently general as to provide a useful elaboration of Simon's model. Such consideration will have to wait for another paper.


Communication media are both social constructions and human inventions. We use media as a tool of structuration and structure media through the processes of structuration. The term metacommunication reflects this dualism: it specificially refers to conversations we have about our communication, including the ways in which we use media. In this paper we have look at one limited sphere of structuration in media, the different generic roles that people play in enabling the smooth functioning of a medium. It was expected that creation of a typology of media based on these generic roles would have several benefits:

  1. It would organize media into structurally-similar groups. This can be regarded as being at least trivially satisfied insofar as we were able to give sensible names to the seven clusters of media. This value is more than trivially satisfied, however, as the groupings of media are not only structurally sensible, but structurally continuous. This is made particularly evident by the complex continuities of the "sum of media" composite variable and its interaction with the "director" role.
  2. It would reveal a systematic set of generic solutions that are associated with the various groupings. This can be regarded as satisfied by the three distinct role patterns that intermix to create the different clusters. There are systematic differences in the role requirements of complex and highly parallelized productions, the production and distribution of singular works of art, and the real time management of interaction in large groups. The differing role distributions of the seven clusters, and in particular the interaction effect of "sum of media" and the director role, are systematic solutions to the very different problems of these different kinds of media. Three distinct solutions emerge in this analysis: distribution chains, hierarchical subassemblies, and contention management. These solutions are mixed systematically in the different clusters of media such that distribution chains may entail hierarchy in some subassemblies, and contention management may parallel subassemblies and a limited hierarchy.
  3. It would expose at least some of the generic problems that media encouter in the course of their invention and evolution. This can be regarded as satisfied by the identification of the the identification of three distinct kinds of complexity, serial complexity, parallel complexity, and floor contention, that these solutions appear to solve. This outcome may be of particular value, if only for its identification of several distinct forms of complexity.
  4. It would provide practical guidance to individuals and social collectives engaged in building and evolving media. It might be argued that even the suggestion that there are different kinds of complexity and that there are systematic solutions to each that emerge naturally in the course of a medium's evolution has immediate practical implications. Foulger (2002a), based solely on the identification of the generic roles, suggests several such practical applications in such areas as communication pedagogy (e.g. role-centered instruction) and media reenginnering (e.g. generic media design patterns and generic media workflows). These results reinforce the likely value of such practical applications, and suggest a starting point for organizing a limited number of media design templates. We will close this paper, however, with a brief exploration of one potentially useful application of these results: their implications for "media convergence".

Beyond Conclusions: Applying Typology to Practice

Media convergence, the use of "one technology or resource for many different things;" Janischigg, 2001), is primary a technology construct (Patrick, 2003). In practical terms, it describes the use of computer technology (in the form of desktop PC's, laptops, video game consoles, PDA's, cellular phones, tablets, information appliances, etc) to provide multiple user interfaces to a broad range of media content, including music, movies, books, games, and more through a single device (Friedman, 2003). In the media industry, however, convergence has another meaning, gaining control of the largest possible inventory of content such that it can be repurposed for a converged media environment. In practical terms, this meaning has translated to a series of corporate consolidations and, in several cases, the merger of large media industry companies. Disney, Sony, Universal, Viacom, GE, and Time Warner (until recently AOL Time Warner) are just a few of the large players that have attempted, with varying levels of success, to leverage convergence into larger media empires.

Where this convergence has only involved mass media (e.g. media in the role intensive media cluster), these combinations appear to have often gone well. Where, however, there has been an attempt to converge an Internet company with a traditional mass media company (broadcasting, movies, music, publishing, etc.) things appear to have gone badly, and sometimes disastrously so. Disney pulled out of its brief flirtation with the Internet portal business only a year after buying into a successful search portal company. The recent de-emphasis of the AOL brand from the newly renamed Time Warner corporate name, moreover, is symptomatic of the substantial problems that have attended the merger of the largest Internet Service provider with a publishing and mass media powerhouse.

At first glance, the merger of AOL and Time Warner made sense from two perspectives. First, AOL was and remains one of the dominant presences on the Internet, with a huge hit rate against its site. To the extent that Time Warner wanted to move itself into what is emerging as the primary channel of media convergence, the Internet, bringing its resources into the service of that huge hit rate made sense. Second, AOL was and remains the dominant player in dial-up Internet access, with a huge subscriber base, while Time Warner is one of the two biggest players in cable, the technology best suited to bringing broadband Internet access to the largest number of subscribers. That combination, particularly at a time when regulators had yet to rule on whether cable companies would have to act as common carriers for broadband resellers, made AOL Time Warner the single largest player in Internet services.

What may have been overlooked, however, were the very different operating models of the media the two organizations managed. Much (perhaps most) of AOL's success follows from the support it provides for interactive media, including chat rooms, computer conferences, e-mail, instant messaging, and conference rooms. All of these media are associated with the personal interactive media cluster in this analysis and involve very little role complexity. Second, while AOL had a huge web hit rate, very little of that hit rate was directly related to AOL sourced content. Most of its mass content business, prior to its merger with Time-Warner, was sourced from other companies that partnered with AOL to provide content. AOL supplied pipes, storage, and an audience. The partners supplied the content. Another huge portion of its formidable hit rate, moreover, derived from personal and online community web pages, almost all of them supported by one or a few people with little or no coordinated support from others. From the perspective of role complexity, personal hypermedia is associated with the personal interactive media cluster. As for the role complexity AOL managed in the web portals and other mass audience web sites AOL hosted, there wasn't any. AOL was, in effect, a newsstand that sold newspapers created by others. Management of role complexity, to the extent there was any, was isolated back to the content provider.

Time Warner, by contrast, was an old line publisher that had branched out into a variety of different publishing niches. In expanding from magazine publishing into books, records, cable television networks, movies, and cable infrastructure they repeated expanded into industries with relatively compatible management structures. If one ignored its cable subscriber system, it was simply a content provider for a range of media, all of which entailed a full range of media roles and a fairly high level of complex role interrelationships. Combining the newsstand and the content provider is not, of course, inherently difficult. One simply has to recognize the very different role structures associated with the media each supported and keep those businesses seperated. Multinational conglomerates do this all the time in ways that allow each of their divisions to structure itself effectively.

The anecdotal evidence suggests that this seperation was not retained in the combination of AOL and Time Warner. The combination was viewed as a synergistic one in which the capabilities of the two companies would result in a sum that was greater than its parts, but the path to that synergy was in combination; in leveraging the content of Time Warner to expand AOL's brand on the Internet; in leveraging the subscriber base of AOL to expand the reach of Time Warners brands and content. The result, from all reports, was a war in which many AOL veterans, after an initial honeymoon phase, left; in which the management structures of Time Warner were extended into AOL, almost certainly driving up AOL's costs just as broadband was starting to eat away at AOL's subscription base.

Bottom line, in trying to build a company that could leverage media convergence via the Internet, AOL and Time Warner mixed oil and water; the oil of lower-speed role-intensive publishing infrastructures with the water of faster-paced and considerably less role-complex personal interactive media. It doesn't necessarily follow that the merger had to fail, but one can reasonably expect that the management that came from each company had no concept of how different the role structure of the other company really was. In order to succeed, the combined management of AOL Time Warner had to treat the two merging entities as different businesses with fundamentally different role hierarchies and attendant cost structures. AOL needed to treat Time Warner as a featured content provider. Time Warner needed to treat AOL as a vertically integrated distribution network. This didn't happen, and the results have been dissapointing to both management and investors.



Three Factor Solution

Two Factor Solution

1 Factor

1. Production Roles

2. Direction versus Collection

3. Content Selection

1. Directed Production

2. Content Selection

Production Roles

Generic Roles in Media

Financial management
Content Integrator
Production support
Content Editing


3 .113





Interfactor Correlations

Factor 1
Factor 2







Factor Score Correlations

Factor 1 of 1
Factor 1 of 2
Factor 2 of 2
Factor 1 of 3
Factor 2 of 3








Table 1: Three principle axis factor analyses showing the underlying dimensionalities underlying eighteen roles as applied to 167 media. The two and three factor solutions are oblimin rotated with Kaiser Normalization. Primary loadings and significant interfactor Correlations are shown in bold italic. Highlighted interfactor and factor score Correlations are significant beyond the .001 level.




Factor 1 of 2

Factor 2 of 2

Model Summary



R Sq.

Adj. R Sq.



R Sq.

Adj. R Sq.












Model 2

Sum of Sq.


Mean Square



Sum of Sq.


Mean Square















Model 2











Sum of Roles











Table 2: Stepwise Multiple Regressions of "Sum of Roles" and the Director Role against the two dimensions of the two factor solution. Each dimension was tested using factor scores generated during the factor analysis. Notice the differential treatment of the Director role in the two analyses, with the role counting as a supplement on top of the number of roles associated with each medium for the first dimension, but as a subtraction from the number of roles for the second.

Figure 1: Seven clusters of media expressed within the two dimensions of the two factor solution. While 167 unlabeled media are depicted here, some are identical in their role profiles such that two or more media are sometimes depicted by the same point. The clusters are labeled. Note that while the factor scores are presented orthogonally in this table, they are actually correlated with each other. The three vectors depicted as lines through the space are informal representations three continua that run through the space.


Figure 2: The general relationship of the three forms of complexity, contention management, serial complexity, and parallel complexity, to the clusters of media. The forms of complexity are represented as three vectors resembling those of Figure 1.



Production Art Media
Publisher/producers, reproducers, production support, distributor/carriers, retailers, collectors, critics, financial management, and selector/gatekeepers occur within most media. Representatives, Content Editors, advertisers, content integrators, and investors may or may not occur with media. Performers, transcriber/recorder, and regulators are sometimes associated with media.

Media in this cluster include artistic/professional photographs, tapestry, publication photographs, blown glass, holographic recordings (3D pictures), ceramics (plates, figurines, etc.), card and board games, postage stamps, post cards, prints, money, books, iconic toys, and teletext.

Personal Art Media
Selector/gatekeepers, collectors, and critics are associated with all media. Reproducers and production support are sometimes associated with media.

Media in this cluster include oil paintings, batik, watercolor paintings, pen and ink drawings, charcoal drawings, sculpture, quilts, and pottery.

Role Intensive Media
Publisher/producers, directors, production support, financial management, and investors are associated with all media. Selector/gatekeepers, representatives, performers, content editors, advertisers, content integrators, reproducers, distributor/carriers, retailers, and critics are associated with most media. Transcriber/Recorders, collectors, and regulators may or may not be associated with media.

Media associated with this cluster include beeping filmstrips, filmstrips, broadcast television, cable television, digital TV, satellite TV, radio, talk radio, silent film, Film with Subtitles, Motion Pictures, Video Recordings (Videotapes, DVD Video), recordings (Records, CD's, Cassettes), video games, comic books, maps, architecture, performance games, physical performance (Dance, Mime), restaurant meals, professional live musical performance, professional performance sporting events, theatrical performances, virtual reality, commercial hypermedia, video hypermedia, video on demand, interactive television, merchandise packaging, magazines, daily newspapers, weekly newspapers, multimedia documents (VideoText), direct mail advertising, on-line information, political rallies

Personal Interactive Media
Distributor/carriers and retailers may or may not be associated with media. Publisher/producers, selector/gatekeepers, directors, representatives, performers, transcriber/recorders, content editors, advertisers, content integrator, reproducers, production support, collectors, critics, financial management, investors, and regulators are sometimes associated with media.

Media associated with this cluster include telephones, cellular telephones, videophones, memo tubes, telewriting, recorded telewriting, Telex, personal electronic publications, beepers, two way beepers, telegrams, text beepers, mobs, graffiti, cave paintings, ham radio, C.B. radio, family radio, walky talkys, personal/family photographs, barter, business cards, memos, photocopier (informal) publishing, letters, decoupled board games, form submission, tape letters, voice mail, personal video, personal hypermedia, compositions/papers, facsimile, distribution list distributions, bas relief, building public address systems, contracts, news performance (wandering minstrels, purchasing, directed (e.g. lynch) mobs, oral messengers, voice pipe, broadcast signals (horns/smoke), targeted signaling (non-broadcast), string telephone, face-to-face interaction, intimacy, notes, whispering, personal decoration (makeup, clothing, jewelry, accessories), intercoms, social dancing, meals - prepared for companions, electronic mail, instant messenger, Internet telephone (CU See Me), cooperative composition, computer conferencing (newsgroups, listserves, etc.), oral history (Griot).

Intermediate Scale Media
Publisher/producer, production support, distributor/carriers, and financial management occur within most media. Selector/gatekeepers, content editors, advertisers, content integrators, reproducer, retailer, and investors may or may not be associated with media. Directors, representatives, performers, transcriber/recorders, collectors, critics, and regulators are sometimes associated with media.

Media associated with the cluster include streaming audio and video, form distribution, clipping services, road signs, store and forward facsimile, electronic clipping services, newsletters, journals, polling systems, electronic journals, lectures, multiplayer simulation games, location resolution systems (GPS), lightboards, electronic bulletin boards.

Moderated Interactive Performances
Publisher/producers, directors, and production support are associated with all media. Content integrators and financial management are associated with most media. Selector/gatekeepers, performers, transcriber/recorders, content editors, advertisers, and reproducers may or may not be associated with media. Representatives, distributor/carriers, retailers, collectors, critics, investors, and regulators are sometimes associated with media.

Media associated with this cluster include classroom discussion, store signs, town meetings, animatons, talking animatons, online databases, distributed group decision systems, microforms, information kiosks, public board meetings, judicial proceedings, legislative assemblies, street theater, religious services, amateur performance sporting events, bonfires (formal get-togethers), amateur live musical performance, and billboards.

Group Interactive Media
Directors are associated with all media. Publisher/producers, transcriber/recorders, content integrators, and production support may or may not be associated with media. Selector/gatekeepers, representatives, performers, content editing, advertisers, reproducers, distributor/carriers, retailers, collectors, critics, financial management, investors, and regulators are sometimes associated with media.

Media associated with this cluster include teleconferencing, video-conferencing, committee/group meetings, ritual ceremonies, campfires (informal get-togethers), league play sporting events, puppet shows, family interaction, informal small group interaction, interview/head-complement dyadic interaction, participatory games and sports, brainstorming, group screen sharing, bulletin boards, dioramas, group decision rooms, and speeches.


Table 3: A summary of the media clusters depicted in Figure 2 and the roles and media that are associated with them. The table has been laid out in roughly the same scheme as the figure, with media associated with the cluster listed at the bottom of each cell and the roles associated with the cluster ranked, from the top, according to the frequency with which they occur in the cluster. A role that is associated with all media (these roles are boldfaced) within a cluster occurs generally occurs in every medium in the cluster. Roles that occur "within most" media in a cluster are associated with over two thirds of the media in the cluster. Roles that "may or may not" occur are associated with between one third and two thirds of the media in the cluster. Roles that "sometimes" occur are associated with less that one third of he media in the cluster (these roles are italicized).