Competing and complementary explanations of communication media use.

ABSTRACT

This study examines how well an integrative model explains media use, the relationships among the causal elements in the model and the replicability and generalizability of the model on two new communication media, electronic mail and voice mail. The findings suggest that a better understanding of new media use can be attained by considering different explanatory constructs simultaneously. By replicating the model for multiple new communication media, this study demonstrates that, a general model of new media choice and usage may indeed be attainable and that the subtle differences between the causal models warrant future research.

Keywords: Social influence, Critical mass, Media Richness, Social perspectives, Rational perspectives, Media Choice, Media use, Electronic Mail, Voice Mail, Structural equations modeling

1. INTRODUCTION

With abundant communication options, the nature of organizational communication has changed in fundamental ways (Canessa and Riolo, 2003). Appropriate media choice and use can have significant effects on communication effectiveness (Lind and Zmud, 1995). As such, research explaining how and why individuals use the new media is of increasing importance to both researchers and practitioners alike. Researchers in this area are currently debating between different theories as appropriate explanatory constructs for explaining new media use. The predominant perspective takes a rational and technologically deterministic view of how and why individuals choose to communicate using a certain medium. Theories such as Media Richness Theory (Daft and Lengel 1986) are within this rational perspective. In attempting to explain media usage for new media, research has been plagued with mixed and inconsistent findings (Dennis and Kinney, 1998; Panteli 2002) especially for explanations of new media such as electronic mail and voice mail. This has lead researchers to investigate other explanations that take into account the interactive nature of the new media.

Some researchers have suggested extending existing theories such as Media Richness Theory (Sheer and Chen, 2004; Carlson and Zmud, 1999) to take into account social factors (Kock, 2001), social influences (Fulk and Boyd, 1991; Fulk et al., 1990) and critical mass issues (Markus, 1987). These researchers contend that how and why individuals choose to use these newer media may be rooted in more social and structural considerations rather than a rational examination of the characteristics and features of a communication medium. These social perspectives may be seen as legitimate explanations complementing and not necessarily substituting for rational approaches to explaining media use.

Unfortunately, researchers have stressed the differences rather than the complementary nature of these theoretical perspectives. In doing so, researchers have claimed that one perspective may be superior at providing explanations of media usage in organizations. Other researchers have illustrated that it may be premature to try to claim this dichotomy. With the evidence in support of each perspective, it is hard to discredit or embrace any one of these perspectives as superior to the others. Recently, researchers have been calling for a more integrative approach that would examine both the competing and complementary nature of these perspectives as well as interrelationships among their constructs. However, empirical evidence is still lacking. The purpose of this paper is to employ an integrative approach that would capture the influence of each of these perspectives on new media usage separately and in combination. The model proposed studies these perspectives as multiple and interrelated determinants of new media usage behavior in organizations.

The empirical research on new media choice and usage has also been restricted to comparisons of a new medium with traditional ones (Marginson et al. 2000; Zack, 1994). The basic limitation of this line of research is that it assumes that there is an "ideal" medium (Reder and Schwab, 1988) usually face-to-face communication. New communication media are then assessed on how much they deviate from this "ideal".

This approach leads researchers to focus on the common set of capabilities shared by the ideal and the new media and overlook the additional capabilities of the new media, not found in the ideal. Not surprisingly, such a biased comparison results in a tendency to regard the new media as similar to one another and yet both different and deficient in comparison to their traditional counterparts.

The objective of this paper is to provide a plausible explanation of how and why individuals use the new media. The study is guided by a model that proposes multiple influences on individual communication media choice and usage. The paper tests the proposed model on one communication medium - electronic mail--and then proceeds to replicate and test its robustness for another communication medium--voice mail. This provides for a strong test of the replicability and thus the generalizability of the model to two new media.

2. THEORETICAL BACKGROUND

The Rational Perspective is the most prominent of the three perspectives considered in this study and has generated the most empirical research. According to this perspective individuals evaluate the characteristics of the medium and its communication capabilities to determine whether the communication medium will be appropriate for the activity to be performed. For effective communication, the medium should neither provide more nor less communication capability than that required.

Much of the work on media choice and usage has focused on Media Richness Theory. Empirical investigation of the theory have yielded mixed results. (ElShinnawy and Markus, 1997; Panteli, 2002). Researchers have recently voiced the concern that this research stream's sole focus on media richness may have limited our understanding (Sheer and Chen, 2004). Appropriated features of the technology may be the most likely to influence and determine usage (Griffith, 1999). Three general features that are important in understanding media choice and usage are functionality, usability and substitutability.

In traditional media, functionality is both limited and fairly constant across media. However, the range of functions in the new media are much more extensive and vary widely from one medium to another (Huber, 1990). Functionality of the medium is defined in terms of its technical features, i.e. what the technology is capable of doing, and how easy it is to use its technical features to perform the task at hand (Goodwin, 1987, Reder and Schwab, 1988; Davis, 1989; Adams et al. 1992). Functionality refers to the ability of the medium to provide such features as: documentation, editing, broadcasting, storage and retrieval, advanced sending functions such as forwarding, carbon copying and receiving abilities such as scanning and prioritizing incoming mail. The more cumbersome it is to perform any of these functions, the less the functionality of the medium. Usability refers to how effective the medium is for performing the task at hand. The appropriateness and utilizability of the medium are important elements of usability. Usability also refers to the convenience of using the medium to both the sender and receiver of the medium. Substitutability refers to the appropriateness of the medium to accomplish the communication task when compared with other media (Grudin, 1988).

Within the rational perspective, work has been grounded on the assumption that the technological characteristics of a medium are fixed and are not mediated through individual perceptions. This has led to inadequate attention to the social context in which usage decisions are made. Such issues as the skill and experience of users with the medium, the social influence of peers and superiors who use the medium (Kock, 2001) and the availability of a critical mass of users with whom to communicate (Markus, 1987; 1990), have been ignored. Within the rational perspective, the behavior of others is considered irrelevant to the choice to use or derive benefit from the medium.

Within the social perspective, researchers alert us to the notion that how and why individuals use a communication medium may be rooted in social factors rather than objective considerations of the functionality, usability and substitutability of the communication medium. In addition, these social perspectives introduce the idea that the objective characteristics of a medium may not be objective at all. Our perceptions of the characteristics of a medium may be a function of what others tell us, our own observations of the behavior of others, or simply whether anyone else is using this medium i.e. the availability of a critical mass of users.

The Critical Mass Perspective was developed explicitly for the study of interactive media (Markus, 1987; 1990). According to the Critical Mass Perspective, the decision to use a medium is a rational, but interdependent one. While the new media provide efficiency benefits in terms of communication, these benefits cannot be realized independently. Because of the interdependent nature of the new media, the benefit derived increases with the number of communication partners (Rohlfs, 1974).

The Critical Mass Perspective adds two important elements to theories based on the Rational Perspective. First, it recognizes that interactive media choice is an interdependent, rather than an independent, decision. Second, it recognizes that unless there is a critical mass of users for a medium, using the medium may be "costly", and thus even appropriate media may not be used. In summary, the Critical Mass Perspective proposes that a medium can provide maximum benefits to its users only when it has achieved universal access. The benefits accruing to a communication media user increase with the proportion of others in the community who are also using the medium (Markus, 1987; Fulk et al., 1990). Certain resources are required in order to achieve or approach universal access. The fewer resources individuals are required to contribute, the more the likelihood of universal access. The higher the skill and effort and the equipment costs required of an interactive medium, the less the likelihood of universal access, and the less likely is the decision to use the medium.

Like the Critical Mass Perspective, the Social Influence Perspective (Fulk, 1993; Fulk et al., 1990)deals with the social context of the decision to use a medium. The Social Influence Perspective contends that the social environment has direct effects on media usage. The Social Influence Perspective focuses on media usage as the result of influence from superiors, peers or highly regarded others. Superiors and co-workers exert influence on individuals' media choice and usage by making overt statements about the characteristics of the medium and the task (Salancik and Pfeffer, 1978). This increases the salience of these features. Superiors and coworkers also provide cues about the nature of the task and the appropriate media options for the task. Social influence can also take the form of vicarious learning from observing the experiences of others (Bandura, 1977). If positive experiences result, individuals are more likely to use the medium in the same manner as the one observed. Choices that lead to undesirable effects are avoided.

The Social Influence Perspective differs from both critical mass and the rational perspective in one important aspect, assumptions about rationality. In contrast to the other two perspectives, the Social Influence Perspective holds that, the decision to select and use a medium is subjective, retrospective, and influenced by peers, superiors, and others. The Social Influence Perspective adopts the view that sense-making is retrospective. Rather than prospectively directing choice, individuals interpret their past behavior and develop explanations for it (Weik, 1979). Social influence by peers and superiors provides certain norms and expectations that define and constrain retrospective justification and rationalization of decisions (Salancik and Pfeffer, 1978).

This perspective also differs from the Critical Mass Perspective. The Social Influence Perspective (Fulk et al., 1990) emphasizes the behavior and choices of superiors and peers as direct influences on individual media choice and usage behavior. On the other hand, the Critical Mass Perspective is more concerned with the proportion of individuals who use the interactive communication medium. Media choice and usage need not meet efficiency criteria to be considered rational in a social context. Indeed, individuals make choices that may be rational from their perspective but do not meet efficiency criteria (March, 1974). This perspective subsumes rational decision-making as being merely a single possible outcome of the social influence process. The effect of social influence is proposed to be systematic variation that is linked to social context.

3. A MODEL OF NEW MEDIA CHOICE AND USAGE

Different studies employing the same theoretical perspective have often yielded mixed or conflicting results. However, it is hard to discredit or adopt a single one of these perspectives without reservation. We are beyond the stage of focusing on validating a single perspective. We need to study these perspectives in combination as part of a model that explains individual media choice and use of the new media in organizations. While no model will explain the whole process, a model based on these three perspectives takes into account the characteristics of the media, the social environment in which media usage occurs, the structural constraints involved in new media use and contextual determinants such as the nature of the task performed. The model hypothesizes that media use is determined by direct effects of constructs derived from the three theoretical perspectives and indirect effects via relationships among the explanatory constructs in the model (Figure 1). The causal model is proposed to be applicable for the study of the new media in general. Its applicability and generality will be tested on two new media, namely voice mail and electronic mail. Below we discuss each of the proposed relationships. First, we discuss the proposed direct relationships derived from the rational, social influence and critical mass perspective. Then we discuss the relationships among the constructs and their indirect relationship to media use behavior.

[FIGURE 1 OMITTED]

Building on the rational choice perspective discussed in the previous section, the model proposes that media use will depend on the characteristics or features of the medium such as its functionality, usability and substitutability and that media characteristics will be direct determinants of media usage (proposition 1).

In addition, organizational theorists have long recognized that the kind of work individuals perform affects their communication behavior in organization. For example, organizational theorists have found that the nature of the task is a good predictor of communication media usage (Conrath, 1978). This assumes that individuals make a conscious and rational decision to match the appropriate medium to the communication task at hand. Following the rational choice perspective, it is hypothesized that the type of task performed will have direct effects on media usage (proposition 2).

According to the Critical Mass Perspective, the decision to use a medium is an interdependent one. While the new media provide efficiency benefits in terms of communication, a single user of an interdependent medium cannot derive any benefit. Because of the interdependent nature of the new media, the benefit derived from use of a computer-mediated communication medium increases with the number of communication partners using it (Rohlfs, 1974; Steinfield, 1986). The higher the percentage of individuals in the community using the medium, the more likely it is that individuals will use the medium. This perspective concedes that unless there is a critical mass of users for a medium, there is a cost to using the medium that influences the decision to use the medium. Media usage is proposed to depend on the availability of a critical mass of users or the potential for achieving such a critical mass (proposition 3).

The Social Influence Perspective deals with the social context of the decision to use a medium and recognizes that media usage occurs in a social environment that has direct effects on it. Superiors and coworkers influence individuals' media choice and usage by increasing the saliency of media features through overt statements about the characteristics of the medium and the task (Salancik and Pfeffer, 1978). In addition, peers and superiors provide others with vicarious learning experiences. The Social Influence Perspective emphasizes the behavior and choices of superiors and peers as direct influences on individual media choice and usage behavior. According to this perspective, the volume and patterns of use are a function of pressure from and observation of peers, superiors and others. A user is likely to use a medium frequently if his or her superior or peers do so. The model proposes that media usage will be a function of social influence from peers and superiors and others that are highly regarded (proposition 4).

Despite the clear conceptual differences between the Critical Mass Perspective and Social Influence Perspective, they are also very much related due to their emphasis on social context. A critical mass of users may be necessary in order for individuals to use a medium (Markus, 1987). However, it may be that this critical mass of users was only attained through the influence of a single powerful supervisor. Similarly, social influence may be an important predictor of media use, yet a supervisor's use of a medium may be due to an available critical mass of users. The model thus proposes that the availability of a critical mass is related to the championship of a single or a set of dominant influential others and that the degree of social influence for media use may be a function of the availability of a critical mass of users (proposition 5).

An important contribution of this model is that it proposes relationships among the explanatory elements originating from the different perspectives. While some work within the rational perspective acknowledges that the social environment may have an effect on media use (Trevino et al. 1990), most of the research in this area has largely been indifferent to the social context within which media use occurs. The social influence perspective suggests that an individual's perceptions of the characteristics of a medium may be socially constructed. The model presented in this paper proposes that the social environment may indirectly affect media use by influencing people's perceptions of the characteristics of a medium. This influence occurs through observations of others behavior or statements made by other users of the medium regarding their experiences with using the medium. The model postulates that while characteristics of a medium are important determinants of media use, these characteristics may be neither objective nor absolute across individuals. We propose that media use is indirectly determined by social factors that influence an individual's perceptions of the functionality, usability and substitutability of a medium (proposition 6).

Additionally, it is postulated that perceptions of the characteristics of a medium are only meaningful after a critical mass of users is available. Moreover, when there is an available critical mass of users, more of the characteristics of the medium can be used. For example, if a critical mass of users does not exist for electronic mail, the benefit of setting up a distribution list would not be worth the effort. To use voice mail asynchronously and send messages directly to voice mail boxes of a group of people would also be pointless if members of the groups do not have voice mail. Some users who do not operate in an environment where a critical mass exists may not even be aware of the existence of some of a medium's advanced features. As such, perceptions of the characteristics of a medium will differ depending on the availability of a critical mass of users. We thus propose that perceptions of the characteristics of a medium will be determined by the availability of a critical mass of users (proposition 7). This relationship suggests that media use will be indirectly affected by the availability of a critical mass of users through its effect on individuals' perceptions of the characteristics of a medium.

4. RESEARCH METHOD

Can media choice and usage of the new media be explained by a model that proposes multiple influences from three different theoretical perspectives? Is this model applicable across the new media? To investigate the proposed model, this study makes use of data collected from a survey-based field study that used multiple data collection techniques in two large organizations that had access to both electronic mail and voice mail. The data is analyzed using structural equations modeling with latent variables.

The principal data-gathering instrument was a standardized questionnaire. With the help of organizational coordinators, 468 questionnaires and pre-paid envelopes were distributed to all individuals in the subunits selected in both organizations. 211 usable questionnaires were returned for electronic mail and 202 were returned for voice mail. Participation was both voluntary and confidential. The questionnaire provides data for the outcome and explanatory variables of interest in this study. It also provides background information about work group, skill and experience with each medium, temporal and geographic distance, and the type of work performed.

This study uses a causal modeling method-- structural equation modeling with latent variables--to represent the constructs and test hypotheses (Bagozzi, 1982). The goodness of fit of the measurement model was determined by analyzing the squared multiple correlation and the standard errors for each item and the coefficient of determination for the explanatory (exogenous) variables and for the outcome (endogenous) variable. In order to develop measures for both the outcome and explanatory variables, single item variables as well as composite variables were created. The composite variables were developed using confirmatory factor analysis. All of the variables are manifest, endogenous or exogenous, variables that are used in the model development and testing. Below we discuss the constructs used in the study. For each construct there are two sets of items one for measuring electronic mail and the other for measuring voice mail. In some cases for simpler presentation the items are presented generically for both media.

Media usage is measured as the number of messages sent by an individual on an average day. Messages sent are used as a measure of amount of usage rather than messages received because it provides a more accurate reflection of the amount of media use. To measure critical mass a set of items were designed to capture the degree to which individuals perceive that the communication medium has attained or has the potential of achieving a critical mass status in the organization. The measures used included the importance and availability of the access device--networked computer for electronic mail and telephone for voice mail as well as the percentage of those inside the organization, outside the organization as well as regular communication partners who use the communication medium in question. For voice mail there are 4 items: 1. Teluse: This refers to the availability of the necessary infrastructure to support communication on voice mail. Teluse is a single item variable that measures the degree of importance of telephone use in an individual's work. 2. Mostuse: This is a single item variable that measures the degree of importance of having most other people in the organizations use voice mail in promoting the respondent's voice mail usage. 3. Outuse: This is a single item variable that measures the degree of importance of having individuals outside the organization access voice mail to promoting voice mail usage. 4. Partner: This is a single item measure of the importance attached to having your regular communication partners use vmail to the respondent's vmail usage. For electronic mail there are 3 items: 1) Mostuse: This is a single item measure of the extent to which most people in the organization use electronic mail. 2) Partner: This is a single item measure of the extent to which the respondents' regular communication partners use electronic mail. 3) Compwork: This is a single item measure of the importance of computer use in the respondents' work. This refers to the availability of the necessary infrastructure to support communication on electronic mail.

Social Influence was operationalized as the degree of social influence peers and superiors have on vmail usage behavior by determining the degree of importance individuals assign to the influence of pressure from superiors and peers on their voice mail usage behavior (Fulk et al. 1990). It consists of two elements: 1) the degree of importance pressure from superiors has on promoting medium use and 2) the degree of importance pressure from peers and others that are highly regarded has on promoting media use.

Perceived Media Characteristics: There are three measures of this concept that relate to whether individuals perceive the characteristics of the medium as appropriate for the tasks they perform (Daft and Lengel, 1986). 1) Usability: This variable measures how usable the medium is in terms of convenience to senders and message recipients. It includes items that question the importance of a) convenience to the user b) convenience to others and c) appropriateness for the messages the user sends in promoting medium use. 2) Functionality: This encompasses factors that deal with operational issues in using the medium (Grudin, 1988; Yates and Orlikowski, 1992). For electronic mail it includes a) multiple, different electronic mail systems, b) problems getting and remembering email addresses c) dealing with the cumbersome nature of an communication system d) dealing with unwanted messages such as junk mail e) handling the need to interface with the system using a keyboard. For voice mail it includes a) System is cumbersome to use, b) Limited capacity; mailbox overflows, c) Limited editing: have to start over to change message, d) Too much junk mail. 3) Substitutability: It is operationalized as the degree to which the medium possesses the technical features to support a user's communication needs and the ability of other media to fill a user's communication needs.

Task Characteristics: This refers to the type of work and tasks individuals perform and are involved in. It includes: managerial tasks or work that involves bargaining, negotiating, persuading, and managing people, support tasks or work that is of a supporting nature such as scheduling, assembling and distributing documents, and charting data, and professional tasks that involve, searching for or gathering information, and writing reports or documents. Individual Differences: Individual differences variables suggested in prior research such as skill, experience, geographical and temporal distance as well as organizational membership were measured to capture and observe their impact on the proposed model. Skill refers to the level of knowledge as a user of the medium. It is measured on a 0-1 scale ranging from 0=novice, just beginning to know the system, 1= expert, knows all the basic features plus more complex and specialized ones. Experience is calculated as the number of years the respondent has been a user of the medium. Steinfield and Fulk (1986) found that geographic and temporal distance can determine choice and usage. These two variables are measured as number of days per month spent traveling out of the office, and number of days per month spent working at least part of the day at home. Organizational membership may be an important variable to control for since it may have an indirect effect on media choice and usage through the social influence and critical mass derived concepts. Organizational membership is a two level categorical variable.

Rigorous testing of the adequacy of the measurement model was conducted using structural equations modeling. We were assured of a good fit in the measurement model (see the results section for additional details). It was thus possible to proceed with testing the strength of the causal relations among the latent constructs within the context of the structural model specified. For a causal model to be useful and statistically testable, it must be over-identified. When a model is over-identified it has the potential to be rejected by the data. This model has 16(16+1)/2=136 variances and covariances and (5 [THETA][epsilon], 11 [THETA][delta], 4 [gamma], 1 [zeta], 3 [PHI], 4 [lambda] Y, and 11 [lambda][chi]), 39 parameters to be estimated. The model is thus over-identified. Each equation in the model is also over-identified.

Chi-squared values, the goodness of fit index (Joreskog and Sorbom, 1988) and the root mean square residual were used to evaluate the goodness of fit of the model to the data. In addition, incremental fit indices were used to evaluate differences between models (Bentler and Bonett, 1980). Standard errors were used to provide an indication of the importance of the parameter to the model as a whole. Q plots were examined to determine the distribution of standardized residuals. Parameters whose estimates were small compared to their standard errors were eliminated from the model and the resulting model was reestimated (Hayduk, 1987). The t-values for each estimated parameter were examined. The structural model was tested and refined on electronic mail. The new model and hypotheses derived from it were then tested on voice mail for replication purposes. This provided cross-validation of the resulting model. It provided valuable information about whether the empirically based modifications represent genuinely valuable information about the model (Bentler, 1980). Moderate replication was conducted where the critical theoretical concepts were held constant to determine the robustness of the empirically generated model.

5. RESEARCH FINDINGS

How well does the hypothesized model explain the causal process underlying new media usage in organizations? The hypothesized model was tested on electronic mail using LISREL and then replicated on voice mail. LISREL estimation of the model indicates that the original hypothesized theoretical model (p<.01) does not accurately reflect the causal process that generated the data (Figure 2). A specification search was conducted guided by theoretical and empirical procedures to determine a model with a good fit. The specification search involved: 1) Eliminating non-significant parameters to provide additional degrees of freedom, and 2) Employing incremental fit indices using a difference in chi-square test to compare incremental models (Bentler and Bonett, 1980) and modification indices (Joreskog and Sorbom, 1988). The latter step involved migrating from the simplest, most restrictive model to less restrictive models.

[FIGURE 2 OMITTED]

Table 1 summarizes the LISREL estimation of the original and revised model. The revised model (Figure 3) provides a good fit with (p = 0.093, Chi-Square [chi square] = 84.98, df = 69) and a goodness of fit index of 0.947 and a root mean square residual of 0.051. The total coefficient of determination for structural equations is 0.968. The standardized residuals for the model are all smaller than the 2.58 value suggested by Joreskog and Sorbom (1988). Total coefficient of determination for X Variables is 0.874 and 0.852 for Y Variables.

[FIGURE 3 OMITTED]

Evaluation of the measurement model indicated that it had adequate reliability and validity. In evaluating the measurement model the magnitude of the standardized parameter estimates, their significance, the squared multiple correlations for each manifest variable in relation to its latent construct, and total coefficients of determination were computed and investigated. The squared multiple correlation (SMC) is the percentage of variation in the observed variable that is explained by the common factor it is intended to measure. It is a measure of the reliability of the manifest variable. Table 2 summarizes the parameter estimates for relations between manifest variables [[lambda].sub.[chi]], [[lambda].sub.Y], and their latent constructs [zeta] and [eta]. All parameter estimates are significant.

The convergent validity of the measurement model was assessed using composite reliabilities, similar to Cronbach's alpha for multiple indicator models (Bagozzi, 1981; Lawrence 1990). The results show that composite reliabilities are high (range = 0.874 -0.968). Table 3 summarizes the results of the structural model (see figure 3). Perceived media characteristics has a significant direct effect on email usage ([[beta].sub.21] = .527, t = 5.013 > 1.96, p =.05). In turn, Electronic mail usage has a reciprocal effect on individuals' perceptions of the characteristics of a medium ([[beta].sub.12], t = .431, t = 2.618, p = .05). The availability of a critical mass of users does not have a significant direct effect on electronic mail usage. However, the availability of a critical mass of users has a strong and significant effect on perceived media characteristics ([[gamma].sub.12] = .678, t = 5.475, p = .05) Task characteristics does not have any significant effect on perceived media characteristics and thus, no indirect effect on electronic mail usage. Social influence, pressure from superiors and others, did not have any effect on either electronic mail usage or perceived media characteristics. Moreover, there was no significant relationship ([[gamma].sub.21] = .472, t = 4.438, p = .05). Moreover, there was no significant relationship between social influence and either of the critical mass or task characteristics constructs.

The model was cross-validated by replicating it on voice mail. It is hypothesized that the structural model will provide a plausible representation of the causal process that generated the voice mail data. In other words, we expect that the same model should provide a plausible explanation for voice mail usage. While the measurement model is allowed to vary, the critical theoretical concepts and the relationships among them are held constant to determine the replicability of the model. The following model is proposed: H1: Voice mail usage is determined by perceived media characteristics. H2: Voice mail usage is determined by the characteristics of the task performed. H3: Perceived media characteristics are determined by critical mass of users. H4: Perceived media characteristics are determined by voice mail usage.

Table 4 summarizes the LISREL estimation of the hypothesized voice mail model. Replicating the model on voice mail provides a good fit with (p = 0.031, [chi square] = 142.74, df = 113) and a goodness of fit index of 0.924 and a root mean square residual of 0.090 (Figure 4). The total coefficient of determination for structural equations is 0.963. The standardized residuals for the model are all smaller than the 2.58 value suggested by Joreskog and Sorbom (1988). The total coefficient of determination for the manifest variables is high (range = 0.771 to 0.963).

[FIGURE 4 OMITTED]

The model provides a good fit to the data. Hence, we cannot reject the null hypothesis that the covariance matrix was generated by the hypothesized model (Bentler and Weeks, 1980; Long, 1983). Furthermore, we cannot reject the hypothesis that the model is replicable. These results provide cross-validation for the model and provide assurance that the empirically generated modifications to the model have value, beyond the specific email data, from which they were generated (Bentler, 1980). Table 5 provides a summary of the results of the hypothesis testing. It presents a list of the hypothesized relationships, LISREL's parameter estimates and whether or not the hypothesis was supported.

Three of the four hypotheses were supported. Hypothesis 1 predicted that voice mail usage will be a function of perceived media characteristics. This hypothesis is supported ([[beta].sub.21] = .548, t = 4.611 > 1.96, p = .05). Hypothesis 2 predicted that voice mail usage will depend on the kind of tasks individuals perform. This hypothesis was also supported ([[gamma].sub.21] = .561, t = 4.364 > 1.96, p = .05). H3 predicted that critical mass will have an effect on perceived media characteristics. The results show support for this hypothesis ([[gamma].sub.21] = .834, t=4.205, p=.05). Finally, H4 predicted that voice mail usage will influence individuals' perceptions of the characteristics of voice mail. This hypothesis was not supported. ([[beta].sub.12] = .196, t = .755 < 1.96, p = .05).

6. DISCUSSION

This study has examined individual media usage behavior in organizations. The study was guided by a model that proposed multiple influences on individual communication media choice and usage based on three theoretical perspectives and relationships among them. In this section we discuss the implications of our findings for each of the theoretical perspectives, the relationships among the explanatory constructs. The implications of replicating the model are also discussed.

The revised model suggests that electronic mail usage is directly affected by perceived media characteristics. The technological characteristics of the medium and particularly how appropriate and usable the medium is for the task performed, have the greatest direct influence on electronic mail use. On the surface, these results support the traditional rational view of media usage (Daft and Lengel, 1986; Yates and Orlikowski, 1992; Grudin, 1988) and suggest that the characteristics of a medium are a strong indicator of its use.

Interestingly, using electronic mail has a reciprocal, positive and significant effect on how its characteristics are perceived. This implies that using electronic mail influences individuals' perceptions about electronic mail and in turn, how they use it. The more electronic mail is used, and the greater the variability in how it is used, the greater the individual's knowledge of both the basic and advanced features of the medium.

The characteristics of the task performed also have a direct effect on media usage. Specifically, the kind of work performed affects electronic mail usage. These results are congruent with traditional organizational theorists' findings that the kind of work individuals perform affects their communication behavior (Kotter, 1982; Mintzberg, 1973). These results are also in line with media richness theory (Daft et al. 1987), which suggests that media use is a function of both the characteristics of the medium and the characteristics of the task. In other words, individuals evaluate both the characteristics of the medium and the task in their decision to use the medium. Other contextual determinants: skill, experience, distance, or organizational membership had no significant influence on the model.

As predicted, critical mass has a strong positive effect on perceived media characteristics. The relationship between critical mass and media characteristics is the strongest one in the model. The availability of a critical mass of users is a determinant of perceived media characteristics for electronic mail. Individuals' perceptions of the characteristics of a medium depend on whether or not there is a critical mass of users. For example, the usability of electronic mail system is a direct function of whether or not users believe that a critical mass of users is available. In addition, perceptions of the functionality and substitutability of electronic mail depend on whether the user perceives others to be users of the medium.

In order to fully understand the relationship between critical mass and perceived media characteristics, the indirect relationship between critical mass and electronic mail usage must be taken into account. The results suggest that having the necessary characteristics to perform a task is a necessary, but not a sufficient, determinant of electronic mail usage. The indirect effect of critical mass on electronic mail usage can be explained in a number of ways. First, the greater the number of communication partners, the easier it is to justify learning advanced features such as marking a message as urgent or attaching files, documents or presentations to an electronic mail message. In addition, the group usage functionality of electronic mail can be fully utilized only if a critical mass of group members are electronic mail users. Moreover, individuals obviously send more messages when they know that other organizational members are users of electronic mail.

Missing from these results is any support for the social influence perspective. While it was predicted that pressure from superiors, peers or others that are highly regarded will result in a critical mass of users and vice versa, no such relationship was found. In addition, social pressure has no effect on perceptions of media characteristics or on email usage. Moreover, there is no relationship between social pressure and the kind of work individuals perform. The lack of support for the social influence perspective is surprising since the measurement model indicates that the social influence manifest variables are both reliable and have significant and high parameter estimates. It is interesting to note, however, that historically researchers have found it troublesome to capture effects that are based on social influence. This is explained by the fact that social influence is often covert and thus problematic to try to capture as an explanatory construct. In addition, individuals like to believe that their choices are a result of conscious, purposive decisions and not the result of external pressure from superiors or peers (Allison, 1971). This may explain the lack of any significant effects with regard to the Social Influence Perspective.

On the surface, the finding that new media usage is directly influenced by individuals' perceptions of both the characteristics of the medium and the characteristics of the task, provides support for rational perspectives of media choice (Daft and Lengel, 1986; Yates and Orlikowski, 1992; Grudin, 1988). Individuals evaluate characteristics of the technology and the task to determine how and how much the medium will be used. This significant result is in line with research findings following this perspective.

However, when the effect of perceived media characteristics on usage is placed in the context of the full model, it is clear that while perceived media characteristics has a direct effect on usage, it is not an independent construct. Media characteristics is a consequence of one or more other explanatory factors. The availability of a critical mass of users has a direct effect on perceived media characteristics. How many others are using the medium influences individual's perceptions of the characteristics of a medium and how usable and appropriate it is for the activities they are performing. This finding also reconciles some of the mixed results of research in this area. It suggests that while perceived media characteristics are an important determinant of new media usage behavior, arguments regarding media usage are only complete when the effects of critical mass are taken into consideration. The critical mass effect has explanatory value for new media usage due to the nature of the new media. The new media are interdependent in nature. A single user of a medium cannot realize any utility from using electronic mail. This is in sharp contrast to other forms of communication where no critical mass is necessary for successful communication to ensue.

This study makes a number of contributions. First, the findings confirm our hypothesis that a better understanding of new media usage can be attained by considering different explanatory constructs simultaneously. New media usage is a complex phenomenon that depends on more than a single explanatory construct. The mixed results of previous research are due to approaching the study of media usage from a reductionist, single dimensional point of view. The different explanatory constructs of media usage are neither contradictory nor competing. They are complementary and interrelated. Hence, understanding media usage behavior requires considering its complex and multi-dimensional nature. This study shows that new media usage is a function of a diverse set of interrelated factors. It steers us away from the belief that one theoretical model is necessarily better or worse than the others by revealing the interrelations among the different causal elements.

Second, there are interactions among the causal determinants of media usage. Any examination of new media usage must consider that while the characteristics of a medium may be a necessary factor for explaining use, media characteristics are neither objective, absolute, nor independent. The characteristics of a medium are subjective, and are determined by other factors that include the availability of a critical mass of users, actual media usage and the kind of task performed. It is clear that while perceived media characteristics have a direct effect on usage, it is not an independent construct. Media characteristics is a consequence of one or more other explanatory factors. Perceptions of the characteristics of a medium are influenced by the availability of a critical mass of users, the general social context within which media use occurs, and actual media usage behavior.

Third, a fascinating finding is that the assumption of the unidirectional nature of causality is fallible. Media choice theorists have worked under the assumption that perceptions lead to behavior and not vice versa. The findings in this study suggest that people's perceptions influence how they behave, AND the outcomes of their behavior change their perceptions. Interestingly, historically, organizational theorists have argued that as organizational researchers we should investigate the bi-directional nature of this relationship (Weick, 1979; Bandura, 1977). In studying media choice and usage we need to be sensitive to this reciprocal relationship and allow for the possibility that individuals' behaviors and the consequences of their actions are important determinants of their perceptions.

Fourth, this study confirms that our neglect of the context within which media usage occurs is problematic. The process of media choice is not one that can be abstracted from other structural considerations. A potent force in explaining media usage is the availability of a critical mass of users. Without consideration of whether there is a critical mass of users of the medium, explanations of media usage based on the Media Characteristics Perspective are incomplete. This might explain the conflicting results of studies employing the Media Characteristics Perspective. Attempts to explain media choice and usage behavior need to incorporate or control for the influential effects of the availability of a critical mass of users.

Fifth, the Social Influence Perspective emphasizes the effects of the attitudes and the behavior of superiors and peers on an individuals' own attitudes and behaviors. This perspective assumes that social influence effects are so influential that they can be analyzed directly. This study suggests that the social influence effect is hard to discern. These results are in line with numerous other empirical work which suggest that social influence plays almost no role in media usage decisions (Rice and Aydin, 1991). Previous work in this area has, in general, lead to inconclusive findings (Pease, 1988; Davis, Bagozzi and Warshaw, 1989). The lack of support for this perspective is usually blamed on Social Influence Theories which fail to provide explicit guidance regarding how to operationalize social influence (Rice and Aydin, 1991).

The purpose of replicating a model using a different electronic communication medium was twofold: First, to determine whether the same model could explain different new communication media use, second, to determine whether the revised model has explanatory power beyond the electronic mail data from which it was created. The model did fit the data for both electronic mail and voice mail, providing generalizabilty across two new media. One interesting difference between the two models is noteworthy. The reciprocal relation between electronic mail usage and perceived media characteristics did not extend to the voice mail model.

Why is this relationship not significant for voice mail? One explanation is that electronic mail offers a wider range of functionality than does voice mail. We speculate that as voice mail technology advances and provides greater functionality, a reciprocal relationship between usage and perceived media characteristics may emerge. The lack of a bi-directional relationship in the voice mail model may also be explained by the superficial similarity between voice mail and telephone use. Users have a pre-conceived notion of how to use voice mail that is transferred from their knowledge of telephone use. For electronic mail, however, individuals have no superficial understanding of how to use it. Their perceptions of its characteristics are thus formed and shaped as it is used. As users become more familiar with electronic mail, this reciprocal relationship may disappear. On the other hand, as users start to use the advanced features of voice mail and realize the functionality it offers beyond the telephone, this reciprocal relationship may emerge.

7. CONCLUSIONS AND FUTURE RESEARCH

This study contributes to our understanding of the causal mechanism underlying new media usage in organizations. In so doing it takes into account rational and social perspectives as complementary forces that can affect media usage, to develop a theoretically based model of new media usage. It confirms our hypothesis that a better understanding of new media usage can be attained by considering different explanatory constructs simultaneously.

The study suggests that the differing explanatory constructs of media usage are neither contradictory nor competing. They are complementary and interrelated. Hence, understanding of media usage behavior necessitates considering its complex and multi-dimensional nature. This study emphasizes that new media usage is a function of a diverse set of interrelated factors. It steers us away from the belief that one theoretical model is necessarily better or worse than the others by revealing the interrelations among the causal factors. This study also shows that any examination of media choice and usage of the new media is incomplete without an investigation of the perceived characteristics of the medium and the task, and external situational constraints such as the availability of a critical mass of users. In addition, this study suggests that an examination of media choice and usage as an emergent process that unfolds with media usage may be a promising avenue of research. By replicating the model on other new communication media, this study has also shown that a general model of new media choice and usage may indeed be attainable. However, by also uncovering the existence of some differences between the models, this study shows that there are differences even among the new media that warrant consideration and future research.

The results of this study provide some promising avenues for future research. The first such avenue is to analyze the differing role that social determinants play for different new electronic communication media while accounting or controlling for technological features of the media. Another exciting avenue for future research is the role of emergent theoretical perspectives (March, 1974; Weick, 1979) in understanding media choice and usage in organizations. Because communication researchers have been inclined to study well established communication media, the tendency has been to examine media choice and usage as a function of preferences and attitudes and not the reverse. It would be informative to challenge this static view and examine preferences as both emerging from usage as well as guiding usage. Future research on media choice and usage should consider both, media choice as well as, the adaptation process that results from actual media usage.

REFERENCES

Adams, D.A., Nelson, R.R., Todd, P.A.(1992). Perceived Usefulness, Ease of Use, and Usage of Information Technology: A Replication. Management Information Systems Quarterly

Allison, G.T. (1971). Essence of Decision. Boston: Littel, Brown.

Bagozzi, R (1981) An Examination of Two Models of Attitude Multivariate Behavioral Research, 16, p323-359.

Bagozzi, R.P., (1982). A Field Investigation of Causal Relations among Cognitions, Affect, Intentions and Behavior. Journal of Marketing Research, November, 19, 562-584.

Bandura, A. (1977). Social Learning Theory. Englewood Cliffs, NJ: Prentice Hall.

Bentler, P.M. (1980). Multivariate analysis with latent variables: Causal modeling. Annual Review of Psychology, 31, pp. 419-456.

Bentler, P.M. and Bonett, D.G. (1980). Significance tests and goodness-of-fit in the analysis of covariance structures. Physical Bulletin, 88. pp. 588-600.

Bentler, P.M. and Weeks, D.G. (1980). Multivariate analysis with latent variables. In P.R. Krishnaiah and L. Kanal, eds., Handbook of Statistics. Vol. 2. Amsterdam: North Holland, pp., 747-771.

Canessa, E. and Riolo, R. (2003) The effect of organizational communication media on organizational culture and performace. Computational and Mathematical Organization Theory. Boston.

Carlson, J. and Zmud, R. (1999) Channel Expansion Theory and the Experiential Nature of Media Richness Perceptions. Academy of Management Journal. Briarcliff Manor: Vol 42, Iss 2.

Conrath, D.W. (1978). Organizational communication behavior: Description and prediction. In M.Elton, W. Lucas, and D. Conrath (Eds.), Evaluating New Telecommunications Services. New York: Plenum.

Daft, R.L., and Lengel, R.J. (1986). Organizational Information Requirements, Media Richness and Structural Design. Management Science, 32, 554-571.

Daft, R.L., Lengel, R.H. and Trevino, L.K. (1987). Message equivocality, Media Selection, and Manager Performance: Implications for Information Support Systems. Management Information Systems Quarterly, 11, 355-366.

Davis, F.D. (1989). Perceived Usefulness, Perceived ease of Use, and User Acceptance of IT, Management Information Systems Quarterly, September, 13,3.

Davis, F., Bagozzi, R., and Warshaw, P. (1989). User Acceptance of Computer Technology: A Comparison of Two Theoretical Models. Management Science, 35, pp. 982-1103.

Dennis, A. and Kinney, S. (1998). Testing Media Richness Theory in the New Media: The effect of Cues, Feedback and Task Equivocality. Information Systems Research, vol 9, no. 3.

ElShinnawy, M. and Markus, L. (1997). The Poverty of Media Richness Theory: Explaining People's Choices of Electronic Mail vs. Voice Mail. International Journal of Human Computer Studies, 46, 443-467.

Fulk, J. (1993). Social Construction of Communication Technology. AMJ, 36, 921-950.

Fulk, and Boyd B. (1991) Emerging theories of communication in organizations. Journal of Management, 17.

Fulk, J., Schmitz, J., and Steinfield, C.W. (1990). A Social Influence Model of Technology Use. In Fulk and Steinfield (eds). Organizations and Communication Technology pp 117-140.

Goodwin, N (1987). Functionality and Usability. Communications of the ACM, 30, 3, pp. 229-233.

Griffith, T.L. (1999). Technology features as triggers for sensemaking. AMR, 24, 3, 472-498

Grudin, Jonathan (1988). Why CSCW applications fail: problems in the design and evaluation of organizational interfaces. Proceedings of the Conference on Computer-Supported Cooperative Work, pp 85-93.

Hayduk, L.A. (1987). Structural Equations Modeling with LISREL. Baltimore: Johns Hopkins University.

Huber, G. P. (1990). A theory of the effects of advanced information technologies on organizational design, intelligence, and decision making. Academy of Management Review 15 (1), 47-71.

Joreskog, K.G. and Sorbom, D. (1988). LISREL 7: A Guide to the Program. Chicago: SPSS, Inc.

Kotter, J.P., (1982). The General Managers. Free Press, New York.

Lawrence, B.S., (1990). At the Crossroads: A Multiple-Level of Individual Attainment. Organization Science, 1,1, pp. 65-87.

Lind, M.R. and Zmud, R.W. (1995) Improving Interorganizational Effectiveness Through Voice Mail Facilitation of Peer To Peer Relationships. Organization Science, 6, 4, July-August

Long, J.S. (1983). Confirmatory Factor Analysis: A Preface to LISREL. Sage University Paper series on Quantitative application in the Social Sciences, 07-033. Beverly Hills and London: Sage Publications.

March, J.G. (1974). Bounded Rationality, Ambiguity, and the Engineering of Choice. Bell Journal of Economics, 9, 338-353.

Marginson, D., King, M. and Mcaulay, L. (2000). Executives' use of information technology: comparison of electronic mail and an accounting information system. Journal of Information Technology, 15

Markus, M.L. (1987). Toward a "Critical Mass" Theory of Interactive Media: Universal Access, Interdependence and Diffusion. Communication Research, 14,5, 491-511.

Markus, M.L. (1990). Toward a "Critical Mass" Theory of Interactive Media. In Fulk and Steinfield (eds). Organizations and Communication Technology pp 194-218

Mintzberg, H. (1973). The Nature of Managerial Work. New York: Harper and Row

Panteli, N. (2002). Richness, Power Cues and Email Text. Information and Management: Vol. 40, issue 2.

Pease, P. (1988). Factors influencing the actual use of video conferencing managers for organizational communication. Unpublished Ph.D. dissertation. University of Southern California

Reder, S., and Schwab, R.G. (1988). The communicative economy of the workgroup: multi-channel genres of communication. Proceedings of the CSCW, Portland OR, pp 354-368.

Rice, R.E. and Aydin, C. (1991). Attitudes toward New Organizational Technology: A Network Proximity As a Mechanism for Social Information Processing. Administrative Sciences Quarterly, 36, pp. 219-244.

Rohlfs, J. (1974). A Theory of Interdependent Demand For a Communication Service. The Bell Journal of Economics and Management Science, 5,1, pp 16-37.

Salancik, G.R. and Pfeffer, J. (1978). A Social Information Processing Approach to Job Attitudes and Task Design. Administrative Science Quarterly, 23, 224-243.

Sheer, V. and Chen, L. (2004). Improving Media Richness Theory: A Study of Interaction Goals, Message Valence and Task complexity. Management Communication Quarterly, Vol no????

Steinfield, C.W. (1986). Computer mediated communication in an organizational setting: Explaining task-related and socioemotional uses. In M.L. McGlaughlin (Ed.), Communication Yearbook 9 (pp. 777-804). Beverly Hills: Sage.

Steinfield, C.W. and Fulk, J. (1986). Task Demands and Managers' Use of Communication Media: An Information Processing View. Paper presented at the Academy of Management, Chicago.

Trevino, L.K., Daft, R.L. and Lengel, R. (1990). Understanding Managers' Media Choices: A Symbolic Interactionist Perspective. In Fulk and Steinfield (eds). Organizations and Communication Technology, pp 71-94.

Weick, K. (1979). Cognitive Processes in Organizations. In Staw, B.M. (ed.) Research in Organizational Behavior, Vol.1: 41-74. Greenwich, CT:JAI Press.

Yates, J., and Orlikowski, W.J. Genres of Organizational Communication: A Structurational Approach to Studying Communication and Media. Academy of Management Review. Vol. 17, no. 2, 1992

Zack, M.H. (1994) Interactivity and Communication Mode Choice in Ongoing Management Groups. Information Systems Research 4,3, 207-239

Maha ElShinnawy, The American University in Cairo, EGYPT

Dr. Maha ElShinnawy earned her Ph.D. at the University of California, Los Angeles (UCLA) in 1993. Currently she is Associate Professor of Management in the Department of Management of the School of Business, Economics and Communication at The American University in Cairo. Previously, she was Associate Professor at the Lowry Mays College and Graduate School of Business at Texas A&M University.

TABLE 1: ELECTRONIC MAIL MODEL FIT

                    [chi square]  Degrees of Freedom  p     Model Fits?

Hypothesized Model  168.31        97                  .00   NO
Revised Model       84.93         69                  .10   YES

TABLE 2: ELECTRONIC MAIL MEASUREMENT MODEL

PARAMETER        DESCRIPT.  ESTIMATE  T-VALUE   SMC

[lambda][chi]11  MGR TASK    0.685     6.328   0.408
[lambda][chi]21  PROF TASK   0.237     2.718   0.061
[lambda][chi]31  SUPP TASK   0.447     5.439   0.232
[lambda][chi]42  MOST USE    0.830    10.835   0.553
[lambda][chi]52  PARTNER     0.920    11.242   0.593
[lambda][chi]62  COMPWRK     0.585     6.439   0.218
[lambda]Yll      USABILITY   0.515     *****   0.383
[lambda]Y21      FUNCTION    0.183     3.246   0.063
[lambda]Y31      SUBSTITUT   0.416    4.7597   0.144
[lambda]Y82      EMOUT       0.564    7.6390   0.568

TABLE 3: ELECTRONIC MAIL STRUCTURAL MODEL

PARAMETER  DESCRIPTION  PARM. ESTIMATE  T-VALUE  SIGNIFICANT

[gamma]21  TC->EMUSAGE      0.472        4.438       YES
[gamma]12  CM->MC           0.678        5.475       YES
[beta]21   MC->EMUSAGE      0.527        5.013       YES
[beta]12   EMUSAGE->MC      0.431        2.618       YES

TABLE 4: VOICE MAIL MODEL FIT

                     Chi-Square   Degrees of
                    [chi square]    Freedom     p    Model Fits?

Hypothesized model     142.74         113      .031      YES
Revised Model          124.68         112      .194      YES

TABLE 5: REPLICATED VOICE MAIL MODEL

Hypotheses  Description  Par. Estimate  T-Value  Supported?

    H1       MC->VMUSE       0.548       4.611      YES
    H2      TASK->VMUSE      0.561       4.363      YES
    H3        CM->MC         0.834       4.205      YES
    H4       VMUSE->MC       0.196       0.755       NO