A CAUSAL MODEL OF FACTORS INFLUENCING FACULTY USE OF TECHNOLOGY

Based on earlier studies using the 1999 and 2004 National Study of Postsecondary Faculty (NSOPF) data [1, 2], a causal model explaining faculty technology use was constructed. Path analysis was used to test the causal effects of age, gender, highest degree, discipline (health science or not), recent research productivity, and teaching load on faculty use of websites in teaching. Two models, one for faculty from Research I institutions and the other for faculty from Community Colleges, were tested and both models fit the data with satisfying indices. Results confirmed that age, highest degree, and teaching loads influenced technology use directly, but indicated the lack of relationship between research productivity and technology use in teaching. An additional connection is suggested from discipline to teaching load. One important difference between the two models is that the impact of gender and teaching load on research productivity is significant for faculty at Research institutions, but not for faculty at community colleges. The models confirm the consistent and relatively strong relationship of teaching load to faculty technology use.


I. INTRODUCTION
Why is there such interest in whether and to what extent faculty use technology in higher education?Certainly, the interest may simply be because students come to college with an interest in the latest technologies [3] or that employers insist that students graduate with appropriate technical skills [4].But it is also possible that technology has become, in the minds of some state and higher education leaders, a "silver bullet" that can solve higher education's problems of low productivity or poor performance [5,6].Clearly, higher education institutions are interested in seeing faculty use of technology increase as evidenced by a focus on faculty development, growth of distance education, and e-learning among the issues monitored in the annual survey of the top ten issues of Chief Information Officers compiled by Educause.While this research cannot answer definitively whether technology is that silver bullet, it does attempt to model the factors which affect faculty technology use.In other words, it investigates how technology comes to be used by faculty, but not whether it achieves all that has been ascribed to it.

Data on Faculty Use of Technology
Current data on faculty technology use indicates that faculty are increasing their use of technology.In a 1998 survey by the National Education Association [7], 70% of faculty had a computer at home, 25% had been involved with distance education, and 27% had a web site for their classes.It is very likely that these figures have increased to higher levels as use of course management software (CMS)-such as WebCT, Blackboard, or others that provide course materials and activities online-increased from 14.7% of the institutions answering the Campus Computing survey in 2000 to 33.6% of the institutions doing so in 2003 [8, p. 7).Two-thirds of the faculty in the University of Wisconsin system reported increased use of a CMS [9].But other data paint a picture of faculty who are less involved in technology.For example, while 80% of public four-year institutions had course management systems (e.g., WebCT, Blackboard) in 2001, only 20% of courses taught by faculty used these systems [10].Contrast these data with more recent data from the 2004 NSOPF that indicates 82% of faculty use email but only 50% used websites in their teaching.These data indicate an increasing, although not universal, use of various technologies by faculty.If faculty use of technology is to be encouraged, it is necessary to know what factors influence their use of websites in teaching.In other words, more research is needed to clarify their use of these new tools in professional activities.

Demographics and Workload
Because faculty technology use has been promoted as a "silver bullet," it is therefore important to understand which faculty come to use technology and the role of workload.Age and gender are widely thought to influence the use of technology.In national studies of the U.S. population, the Pew Internet and American Life Project have documented continuing differences in Internet use, but also a narrowing or elimination of those differences.Younger individuals use the Internet more frequently and more of them log in than older Americans [11].However, for some specific activities (use email, get news, online purchasing), the different age groups are not that different.For example, 88% and 94% of Gen Y (18-28 year olds) and Leading Boomers (51-59) use email, 72% and 74% (respectively) get news online, and 68% and 67% (respectively) make online purchases.Males use online resources more than females, although black women and women under 30 exceeded their male peers in Internet use [12].Women, however, are more "enthusiastic" [12, p. 1] online communicators, while men are more likely to perform transactions online, such as purchases and banking.In any case, it is clear that the genders are closing the earlier gap in Internet use.Given these research results, perhaps male and female higher education faculty are also becoming more alike in their use of technologies.
The impact of technology on workload has been amply documented [13,14].Workload increases through a number of processes, including learning new skills [15], spreading work time over a larger portion of the day because of increased use of email and course management systems [16]; keeping online courses up-to-date through continuous upgrading [15]; and increasing development time as learning goals ascend Bloom's learning taxonomy from knowledge to create [17].In fact, for one individual, workload doubled [18] as preparation time increased as did one-to-one instruction; Brown [19] claimed a 40% to 50% workload increase; in another study, 76% of faculty surveyed spent more time preparing and delivering courses [20].These results are duplicated on a larger scale: the National Center for Education Statistics [21] found that faculty teaching distance education courses actually had a higher teaching load (which may capture the higher teaching loads of community colleges where much distance education has occurred).In addition, faculty of distance education courses had higher average office hours and hours spent on student email per week [21].
An interesting insight into faculty perceptions can be drawn from studies on wholly online courses.Geith and Vignare [22] and Geith and Cometa [23] found that all of the surveyed faculty felt they spent more time when teaching online.However, when the researchers investigated the faculty's self-reported hours, only three of the nine online sections actually consumed more faculty time per student than more traditional courses.Five sections consumed about the same amount of faculty time and one section consumed less time.In a similar effort to understand why faculty believe technology increases their workload is Hislop's [24] study.Four faculty completed time logs while teaching two courses each-one online and one face-to-face-where both courses had the same curriculum.Instructional time did not differ much (333 hours online versus 347 hours face-to-face), but the online courses did take more time per student and interactions with students were spread out over more days in the week, rather than on the days when the class met [24].It is interesting to see that the Geith et al. studies [22,23] would indicate that faculty perceptions of how much time it takes to teach using technology are not reliable, but Hislop [24] would claim that faculty perceptions of time may be more credible when the analysis breaks down total time spent into per-student or other analyses.Or it may be that faculty perceptions about time spent online is different because they are also dealing with patterns of behavior or interaction that are different from earlier face-to-face experiences.Clearly, this is an area that requires further study.
Although the increases to workload seem to be sharp in the early stages (when learning new skills and developing new online resources occur), little research has been done on whether workloads decrease as faculty improve their skills and take advantage of a "learning curve."In this vein, Bartolic-Zlomislic and Bates [25] identified a rapid learning curve for faculty in three case studies of online programs that included time to learn the software to be used, how to design courses, and provide online instruction.
What is intriguing in most of the research on the relationship of technology and workload is that it is in one direction: in other words, technology increases faculty workload.Therefore, this study will focus on the causal relationship from teaching workload to technology use.

Models of Faculty Technology Use
The National Survey of Postsecondary Faculty (NSOPF:04) presents an opportunity to model faculty technology use by using the variables included in this national database.This dataset includes demographic variables of interest as well as several variables capturing workload, which seemed important for analyzing faculty technology use based on the research literature already reviewed.In other words, it is the best national database for this research and it has the further advantage of moving beyond the one-institution studies so prevalent in earlier research.
In two previous studies, Meyer and Xu [1] and Xu and Meyer [2] used NSOPF:99 and NSOPF:04 data to identify the factors that influenced faculty use of technology in their teaching.In the first study [1], hierarchical multiple regression analysis found that individual factors with a statistically significant relationship to faculty technology use included gender, age, highest degree type, teaching load, and research productivity [1].Younger faculty, faculty with doctorates, and faculty with higher teaching loads were found to use email and the web relatively more in their teaching.Faculty with higher research productivity were found using email more often than websites.One limitation of the study was that interactions among variables were not investigated.
In the second study [2], institutional factors such as the institutional expenses, student full-time equivalent enrollments, faculty/student ratios were investigated along with individual factors.By including both sets of variables (a total of 40 separate variables) and using Bayesian networks to examine relationships among variables, the analysis resulted in a finding that differences in technology use were apparent for faculty from different academic disciplines and for faculty employed by institutions of different Carnegie types.The most interesting results indicated that several variables capturing teaching (teaching as the individual's principal activity, student contact hours, percent time teaching, student credit hours taught) had significant and independent influences on technology use.
One consistent finding from both of these studies was the significant relationship between faculty teaching load and faculty technology use.With Bayesian analysis suggesting that teaching loads impact technology use, could it be that higher teaching loads send faculty members looking for technologies to use in order to do their job?This seems inconsistent with the literature on how technologies increase faculty workload, but both the importance and strength of this relationship indicate that in-depth investigation is needed into the causal relationships between workload and technology used in light of other related factors.Hence, this study tests a model that establishes the causal relationships between factors, including workload, influencing faculty technology use.

Proposed Causal Model
The causal model clarifies how individual and institutional factors-already found to be influential in Xu and Meyer [1] and Meyer and Xu [2]-interact and impact faculty use of technology in teaching.The proposed model is based primarily on these two studies.Rather than being a full-blown theory, the model (shown in Figure 1) is an abstraction that describes the probable causal relationships among significant factors that determine faculty use of technology.The plausibility and validity of the proposed model will be evaluated with the 2004 NSOPF data.

Res. productivity
Using technology in teaching through teaching load and research productivity, given that many studies have found that female faculty have heavier teaching loads, which leave them with less time to conduct research [26,27].Academic disciplines are related to faculty technology use [1], since some disciplines (e.g., health fields) had distinctive patterns of technology use in comparison to other disciplines (e.g., business, education).The model recognizes the impact of teaching load on research productivity, but not the reverse.The reverse relationship (that research productivity impacts teaching load) is theoretically viable, but so far has not been borne out by earlier analyses.The model also recognizes the critical organizing role of teaching load in its effect on technology use as an intermediary of gender and education level as well as having an independent influence.
Although a number of institutional influences on technology use (i.e., appropriations, enrollment size) were evaluated in [2], only the Carnegie classification of institutions was found to explain variations in technology use in teaching.However, rather than incorporate all Carnegie types into this model, the proposed model was evaluated separately for faculty in Research I institutions and compared to faculty in community colleges.In the NSOPF data, the 1994 and 2000 Carnegie classifications are available.The Research I institutions are classified in terms of the level of research activity, which is measured by factors including, among others, research and development expenditures and number of doctoral programs.Community colleges are two-year institutions that mainly confer certificates or associate's degrees (For more information, please visit http://www.carnegiefoundation.org/classifications/).Because of the distinct academic missions, faculty responsibilities are somewhat different.Research I institutions emphasize faculty research as well as teaching and service, and community colleges predominantly stress teaching [28,29].These two institutional types were chosen to focus on whether different expectations for faculty would play a role in faculty use of technology in teaching.How the model fits the two different faculty samples can shed more light on the causal effects between factors as well as the validity of the proposed model across institution types.

A. Data Source
For this study, faculty members were selected if their principal activity was either teaching or research from the NSOPF:04 data.NSOPF:04 is the fourth national survey of postsecondary faculty sponsored by the National Center of Educational Statistics (NCES) since 1988.Even though the questions stay fairly consistent with minor modifications across surveys, a different sample of faculty was selected from the national pool in each survey.In the 2004 implementation of the survey, the original sample consisted of approximately 1,070 postsecondary institutions and more than 34,300 individuals employed in those institutions.The weighted response rate is 76% for the faculty survey.NSOPF surveys used complex sampling procedures including multi-level stratification and unequal probability selection of individuals.Following the discussion on data weighting of complex survey samples by Thomas and Heck [30], the data were weighted in both descriptive and inferential analysis.After data preprocessing, 2,748 faculty members from Research I institutions and 3,112 faculty from community colleges were included in the analysis.

B. Variables
The research objective is to propose and validate a causal model of factors determining faculty use of technology in teaching.Because the NSOPF:04 data showed that 82% of faculty members used email in their teaching [2], it is fair to conclude that email use has become the norm among higher education faculty.Thus, email use is not considered as an indicator of faculty adaptation of technology in this study.
Rather, the outcome variable capturing technology use will focus on the variable titled "website for any instructional duties" (coded as 0 for No, 1 for Yes).The original item in NSOPF:04 is worded in this way: Did you have one or more web sites for any of your teaching, advising, or other instructional duties?(Web sites used for instructional duties might include the syllabus, readings, assignments, and practice exams for classes; might enable communication with students via listservs or online forums; and might provide real-time computer-based instruction) [31, p. 23].
Table 1 presents the sample's distribution of faculty use of websites in teaching by gender and highest degree types.It shows that 61.6% of faculty members in Research I institutions used websites in their teaching, while only 39.1% of faculty in community colleges did so.Age is faculty members' biological age at the time of data collection (2003-04); gender is coded as 1 for male and 2 for female.To keep a parsimonious model structure, disciplines are simplified into a binary variable, health science disciplines (coded as 1) versus non-health-sciences disciplines (coded as 0), because faculty in health science disciplines were found to have distinct patterns from those in other academic areas when it comes to using technology in teaching [1,2].Educational level is measured by highest degree type, which includes doctoral, master's, first professional, and bachelor's degrees (coded as 4, 3, 2, and 1, respectively).A few measures of teaching load were available in the NSOPF data, but "total student credit hours" was used in this study because increases in either the number of classes or the class sizes generally translate into an increase in teaching loads.
Last, because the causal model is a snapshot of faculty technology use at a certain point in time, it was decided to use research productivity in the last two years rather than career total research productivity.Measures of research productivity include the number of publications in juried journals, non-juried journals, books, book reviews, and presentations.It is difficult to justify eliminating any of these measures because different disciplines place different importance on them.Thus, an exploratory factor analysis (EFA) procedure was done to combine these variables of scholarly productivity into one more reliable measure of productivity.To be specific, one underlying factor was extracted that has an eigenvalue greater than 1 using a principal-components method.The scree plot also confirmed the singlefactor structure because the extracted factor was separated by a clear "elbow" from the others that had eigenvalues lower than 1.No rotation was needed to simplify the structure because only one factor was retained.Thus, the scores of the extracted factor, accounted for 36.4% of the total variance, was used as the measure of research productivity in the study.

C. Analytical Methods
The data were prepared SPSS first and imported into LISREL 8.

A. Overview of Results
Figure 2 shows the final path model of faculty technology use.In the process of path modeling, the direct effect from research productivity to technology use was removed due to non-significance.Also, an indirect effect was added from health science disciplines to faculty use of technology through teaching load as a result of analysis and further consideration of how teaching in the health science disciplines at research institutions may depend on technology to a lesser extent than other disciplines.This is an interpretation of the results that must be followed up with more detailed research.Statistically, the added path is supported by the single degree LaGrange-Modifier-based modification index that shows that the model Ȥ 2 decreased by 8.40 to 11.43 with a reduction in the degrees of freedom by 1.In Figure 2, each path is marked with two numbers: the path coefficient (beta weight) for Research institution faculty before the slash (/) and the path coefficient for community college faculty after the slash.The standardized direct, indirect, and total effects of each variable on faculty use of websites are summarized in Table 2.In this section, the two models are presented separately.For each model, the fit is measured by four indices: Ȥ 2 coefficient, root-mean-square error (RMSEA), standardized root mean square residual (SRMR), and comparative fit index (CFI).Note that dozens of model fit indices have been made available in LISREL.In general, they fall in three categories.First, absolute indices evaluate the overall discrepancy between observed and model-implied variance/covariance matrices; Ȥ 2 and SRMR fall under this category.For good model fit, Ȥ 2 is expected to be non-significant, although it may not be the case in many studies because it is a statistic that is very sensitive to sample size.SRMR needs to be lower than .08 for acceptable models.Second, incremental indices evaluate a model's absolute or parsimonious fit relative to the null model.CFI is one of them and must have a value greater than .95for good model fit.Third, parsimonious indices evaluate the overall discrepancy between observed and model-implied variance/covariance matrices while taking into account the simplicity of the model structure.RMSEA belongs to this group and a value lower than .06 is desired [32].The four indices reported in this article are selected based on a discussion in McDonald and Ho [33] and followed examples in other published SEM studies in higher education research.In addition to fit evaluation of the models for Research I institutions and community colleges separately, at the end of this section, the similarities and differences of the two models are also briefly discussed based on the multi-group analysis.

Figure 2. Path Analysis Model of Factors Influencing Faculty Technology Use
Notes.

B. Path Analysis Model for Faculty in Research I Universities
The path model of faculty use of websites in teaching has a RMSEA of approximately 0.026, which is nonsignificant (p = 0.99).The value of this index suggests an excellent fit of the proposed causal model to the observed data.The Ȥ 2 coefficient is 11.43 with df = 4 and p = 0.022; given the large sample size, this is another positive sign of good model fit.The SRMR of this model is 0.012, which further confirms the model fit from the perspective of average magnitude of residuals.The fourth fit index, CFI = 0.99, also means the models fit the data quite well.
With the model goodness of fit confirmed, the path coefficients between the predictor and predicted variables show that the strongest predictor of faculty technology use is the highest degree type (0.154), although the indirect effect of highest degree on technology use through teaching load is not significant.Increase in teaching load (0.151) contributes significantly to greater use of websites in teaching.Age is also a significant predictor for technology use (-0.099), suggesting that older faculty are less likely to use the web while teaching.Faculty in the health sciences have a relatively higher teaching load (.056), but the indirect effect of discipline on technology use is weak (0.008).Given that discipline had a direct effect on website use (-.056), the total effect shows that faculty in health science disciplines use websites for teaching purposes significantly less than those in non-health-science disciplines.Gender is not a significant predictor of technology use (-.006).It is worth nothing that this model suggests that female faculty members had significant lower teaching loads (-0.065) and fewer number of publications than their male colleagues in the last two years (-0.134), while increased teaching load is found to lower research productivity (-0.066).Overall, 3.8% of the variance in faculty use of websites is accounted for by this path model.

C. Path Analysis Model for Faculty in Community Colleges
The same model structure explains community college faculty use of websites given that the model goodness of fit indices are RMSEA = 0.024 (p = 1.00),Ȥ 2 = 11.43 (df = 4, p = 0.027); SRMR = 0.011, and CFI = 0.99.However, the parameter estimates reveal that the magnitude of impact exerted by individual factors on faculty use of technology in this model is different from the model for faculty in Research I institutions.The strongest predictor of faculty technology use is teaching load (0.159): heavier teaching loads appear to drive professors to increase use of websites in teaching.The second strongest predictor is highest degree (0.101), indicating that faculty with higher education levels use technology more often.Older faculty use websites less in teaching (-0.044).Health science disciplines have a nonsignificant direct effect on technology use.However, combined with an indirect effect through teaching load, being in health science disciplines means community college faculty actually use more technology than their colleagues in other disciplines (0.044).Again, gender is not a significant predictor of technology use (0.019).Contradictory to earlier studies, female faculty actually had lighter teaching loads (-0.050); though, unlike faculty in Research I institutions, recent research productivity is comparable for both males and females for community college faculty.This may be due to the community college's lower emphasis on research for faculty, which equalizes the demand for research productivity for all faculty.Finally, possessing a higher degree meant heavier teaching loads (0.213) and more research productivity (0.086), but there is not a significant relationship between teaching load and research productivity.Overall, 4.4% of the variance in faculty use of websites is accounted for by this path analysis model.

D. Multi-Group Path Analysis
The same model structure appears to explain faculty technology use quite well in the two different institution types.Nonetheless, differences in path coefficients and explained effects warrant further examination of the cross-group model fit.Thus, a multi-group path analysis was run, in which crossgroup equality constraints were imposed on the path estimates to determine whether the direct effects in the unconstrained models, as previously discussed, differ by Carnegie type (Research vs. Community College).The analysis shows that the goodness-of-fit for the constrained models is substantially worsened (RMSEA = 0.257; SRMR = 0.21) when the two unconstrained models are compared.The poor fit of the cross-group model provides further evidence that observed differences in the parameter estimates and effects between the two unconstrained models are significant and that the factors driving faculty use of technology in teaching interact differently for faculty in Research I institutions and for faculty in community colleges as observed in the two unconstrained models.

V. DISCUSSION
This analysis has produced some intriguing results.First, the model as proposed and then refined appears reliable and sound.Age, degree level, discipline, and teaching load influence faculty use of technology, with teaching load being the greatest and most consistent influence.While the original model hypothesized an impact from research productivity to technology use in teaching, it was not maintained in the final model.This is difficult to explain without additional research being done that uses data collected to specifically explore this finding.It is possible that time spent on conducting research has little connection with using technology in teaching.Or, the relationship between research productivity and using the web in teaching could be obscured by contradicting forces.For example, experienced researchers may be able to put less effort into their teaching (assuming it takes some effort to employ websites in teaching) or perhaps the proliferation of course management systems make the use of the web in teaching easier and less time-consuming, therefore making it easier for busy researchers to use the web.
Another modification to the proposed model involved adding a link between academic discipline (health science or not) and teaching load.This is easy to understand considering teaching loads are different by discipline and health science faculty often work additional clinical or field rotations.It also adds to the importance of teaching load, which mediates all of the variables in the model except age.
There are differences between Research and Community College faculty and their use of websites in teaching.Discipline (health science or not) does not have significant direct effect on technology use for community college faculty, nor are gender and teaching load significant influences on research productivity, as befitting the community college's focus on teaching rather than research.Faculty in Research institutions produced significant relationships for disciplines (health science or not) on the use of websites in teaching as well as gender and teaching load on research productivity.In other words, the two types of institutions are different in their missions (i.e., their focus on teaching for community colleges versus teaching and research for Research institutions), and this difference is captured by the model.One of the reasons for retaining recent research productivity in the model is to capture its continuing importance for faculty in Research institutions, even if its impact on technology use is unproven.
There are many similarities between the two models, perhaps capturing some basic or foundational qualities of the faculty job that are true whether the faculty person is working at a Research institution or Community College.Age matters for technology use, with older faculty perhaps finding it more difficult to keep up with new technologies (which does not mean they are not willing or not able to learn).Highest degree also matters, as does teaching load.The more likely that faculty possess the doctorate and have a high teaching load, the more likely they will use websites in their teaching.But perhaps in contrast to early expectations that females used technologies less than males, gender did not have significant influence on faculty use of websites in teaching in either type of institution.
In this study, as the faculty person's teaching workload increased, the use of websites in teaching also increased.This may be capturing a need of faculty to seek new ways of teaching as demands on them grow.This finding of the connection of teaching load to technology use should not be construed to imply we advocate increasing teaching loads to encourage greater technology use.This would be a dangerous misinterpretation.It is not clear if faculty with higher teaching loads choose to adopt the use of the web in order to achieve some efficiencies in the use of their time or to lighten their load, so to speak.Or, this relationship may capture institutional pressures to teach more and use more technology, perhaps in response to budget problems, a need to address enrollment growth, or pressure from institutional leaders.There may also be a natural limit to how much web use can be incorporated into a busy teaching load; while course management systems might alleviate this problem, there is a practical limit to how much any faculty can accomplish within a 24-hour day.In other words, without paying attention to teaching faculty how and when to apply technology so that technology can help relieve pressures on their time, realizing greater teaching productivity (in terms of increased courses and/or enrollments) may not occur [34].
But what should institutions do based on this model?Should they increase teaching loads and only hire young Ph.D.s? Clearly not.However, they may decide to research the causal relationships of teaching load or research productivity on technology use and explore the conditions that could affect this relationship.Perhaps there are differences based on institutional type (e.g., small private liberal arts institutions) or disciplines which have different expectations for using technologies to teach.Second, institutions may want to consider exploring whether policies on faculty workload are affecting technology use or if there is a limit to this relationship.
As for future research, this model is based on both literature and previous studies conducted on the NSOPF 1999 and 2004 datasets.It explains the causal relationships among the variables, and does so with good model fit indices.But because the variance in faculty technology use explained by the models is not satisfactory, we need to continue to explore other hypotheses that may fit the data equally well or better.
It is important to remember that the current model was developed and confirmed using a single database (NSOPF:04).In order to confirm, modify, or refute this model, future research needs to use more recent data and/or collect additional data elements that could fill important holes in the model.
It will be important to explore other mitigating or preexisting variables that may better explain the relationships in the model.There is literature on the roles of barriers to faculty using technology, including a loss of face-to-face, live interaction with students [35,36] and a lack of support or assistance to learn and implement the technologies [13,35,36,37].Because the NSOPF:04 database does not include variables that capture these barriers, they could not be included in the model.
Future research ought to also explore how faculty motivations to use technology modify relationships in the model.Faculty are motivated to use technology in order to reach new audiences [35,36,38], to pursue an interest in learning new skills [36,39], to enjoy the flexibility of the new approach [13,35,36,39], and to stay up-to-date [35] and relevant [40].Betts [41], Schifter [42,43], and Rockwell et al. [14] explored the role of faculty's intrinsic and extrinsic motivations to use technology, and these motivations should be included in further research on the model in this research.
Furthermore, future research ought to explore the differences between Carnegie types and whether the individual's preference for teaching (or research) affects the institutional type they choose to work in and their willingness to use technology for teaching.
Finally, it is important for readers to keep in mind the limitations of this study when evaluating the findings.First, the causal model was generated using data from a national survey.Because the survey relied on self-reported information, the research questions were answered with limited depth and breadth.Second, the model proposed in this study is based on the authors' understanding of the literature and knowledge about faculty work.Nonetheless, readers need to be aware that, even though the models fit the data with good indices, there could be other competing models that statistically fit the data equally well or better.Further research is needed in this area.

VI. CONCLUSIONS
So is faculty use of technology a silver bullet or not?Technology use is, as of 2004, widespread and almost universal for email.It is easy to understand why younger faculty use technology more than their counterparts, but it is less clear why teaching load influences technology use although it does so consistently despite different analytical approaches and statistical tools.The merits of this model lie in the extended understanding of factors influencing faculty technology use from personal level to institutional level (which includes differences resulting from discipline or institutional type), and statistically modeling their causal relationships that have not be done previously.It is worth noting that the final models are only one of the models that explained the data structure well, and there are possible alternatives that may explain faculty technology use equally well or even better thereby resulting in a higher R 2 .
If technology was ever perceived as a silver bullet, it is a bullet whose trajectory is unknown, whose velocity is unknown, and whose arc and distance is unknown.In other words, while our understanding of the benefits of technology is emerging and we can state that technologies are certainly worthy tools, technology hardly qualifies for the status of a silver bullet that solves everything magically.Some problems it can help solve, although doing so will take effort, but solving all things will have to await the next silver bullet, if it exists.

Figure 1 .
Figure 1.Proposed Causal Model of Faculty Technology Use in Teaching

Table 1 . Faculty Use of Website in Teaching by Gender and Highest Degree Type (Weighted)
7to perform path analysis by reconstructing the covariance matrices through maximum likelihood estimation and validate the proposed causal structure of faculty technology use.Separate path analyses were run for faculty in Research I universities and in community colleges; fit indices were examined for each model and parameter estimates studied.Then, a multi-group path analysis was conducted by imposing a cross-group equality constraint on the path estimates to determine whether the model direct effects differ significantly for faculty in Research I or Community College institutions.Maximum likelihood function was used to estimate the model parameters.Multiple model fit indices are discussed below.Since the available faculty samples are relatively large, Į = .01 is used for tests of significance.

Table 2 . Direct and Indirect Effects
Note.Numbers in bold indicate statistical significance at Į = .01.