Is it the program or the interpreter? Modeling the influence of program characteristics and interpreter attributes on visitor outcomes.

Powell, Robert B. ; Stern, Marc J.

Introduction

Much has been written regarding the interpretive techniques that should be employed to enhance visitor outcomes (e.g., Ham, 1992, 2013; Moscardo, 1999; Knudson, Cable, & Beck, 2003; Brochu & Merriman, 2002; Lewis, 2005). These interpretive techniques, which we refer to as program characteristics, are the focus of training efforts offered by organizations such as the National Park Service (see NPS, 2003a,b,c,d,e,f) and the National Association for Interpretation (NAI) as well as college courses offered around the world. These characteristics are believed to improve the quality of interpretive communications and to contribute to reaching desired outcomes, such as inspiring audiences to form intellectual and emotional connections with interpreted resources, influencing attitudes, and in some cases motivating behaviors. Researchers and field interpreters also recognize that there are other factors, such as the attributes of the interpreter (confidence, charisma, enthusiasm, passion, apparent knowledge, etc.) that may influence the effectiveness of an interpretive program (e.g., Ham & Weiler, 2002a). However, these attributes are often overlooked in research and interpretive training. Stern and Powell (article 1, this issue) and Powell and Stern (article 3, this issue) investigated live interpretation programs provided by the U.S. National Park Service (NPS) to examine the relationship between 56 different interpretive practices and interpreter attributes and visitor satisfaction, enjoyment and appreciation for resources, and behavioral intentions. The results suggest that only certain program characteristics and interpreter attributes were significantly related with these outcomes. This study seeks to extend these findings by modeling the relative influence of these program characteristics and interpreter attributes on visitor outcomes using structural equation modeling (SEM).

The interpretive techniques promoted by professional associations and organizations have evolved over many decades and are based on experience, expert consensus, theory, and peer-reviewed research (Skibins et al., 2012). However, the empirical support for many of these best practices is largely anecdotal, because few studies to-date have attempted to isolate the influence of particular practices on outcomes through the use of experimental (or quasi-experimental) designs or comparative approaches (Skibins et al., 2012). The isolation of the influence of particular practices is challenging even with these designs, as program outcomes inevitably emerge from a dynamic interaction between the interpreter, the audience, the content, the setting/context, and the delivery (Powell et al., 2009, 2012; Archer & Wearing, 2003; Wearing & Wearing, 2001). Accounting for all factors seems a near impossibility. In this paper, we explore the relative influence of two of these elements, interpreter characteristics and program characteristics, on visitor outcomes. We define interpreter characteristics as those that may be entirely unique to the individual interpreter in any given context. These elements might include their mood, personality, or particular style of presentation. While program characteristics may also be highly dependent upon the interpreter, they could also be incorporated by design into a pre-packaged program, such as the sequence, content, theme, or logistics of the program.

We use structural equation modeling (SEM) for two reasons. First, the models give us a sense of the relative strength of influence of interpreter and program characteristics on visitor outcomes. The models can reveal the percentage of the observed variance in each outcome that can be explained by the predictors (Byrne, 2006). Second, the models allow for an examination of the interactions between interpreter characteristics and program characteristics. SEM also reveals the most parsimonious causal models for each outcome. As such, only the most predictive combination of variables remains in the final models. Examining which variables are present in the final models and their inter-relationships allows for consideration of the relative influence of program design vs. interpreter attributes. For example, if only interpreter characteristics are present in the final models, we would consider them dominant drivers of visitor outcomes. If both interpreter and program characteristics are present, it would support a view that outcomes are produced more by the interactions between interpreter and program design rather than by one or the other.

Methods

Selection of sites

We observed 376 diverse interpretive programs provided by 24 NPS park units across six regions of the NPS that generally reflected the current makeup of the NPS system (see Stern and Powell, this issue). The criteria for selecting NPS units included:

* Annual visitation greater than 35,000

* Geographic distribution across the county

* Variable distances from urban centers (urban, urban proximate, remote)

* Resource-base (cultural, natural, mixed)

* The ability to observe multiple programs in a short period of time

* Willingness to participate

The 24 selected units varied widely in terms of visitation, resource base, and locations, providing a reasonable sample from which to make generalizations regarding interpretation provided across the NPS system.

Sampling and data collection

Four researchers collected field data. Prior to each program one researcher conducted a short interview with the interpreter to collect demographic and background information regarding the program. During the program, this same researcher monitored 56 different program and interpreter characteristics and recorded these details on standardized observation sheets. After the program, we surveyed attendees that were age 15 or older using a standardized questionnaire. For programs with fewer than 50 participants, we attempted a census of all eligible attendees. In programs with more than 50 attendees, we systematically sampled attendees. From the 376 programs, we collected 3,603 surveys from visitors (for more detail, see Stern & Powell, this issue).

Data cleaning

Post-program surveys and program audits were coded and entered into Microsoft Access Database and Microsoft Excel to facilitate data entry. Data were then transferred to SPSS and EQS v6.1 software (Bentler, 2005) for screening and analyses. The visitor survey data were first screened for cases missing more than 50% of the items per factor (Tabachnick & Fidell, 2007). A total of 118 respondents were removed as a result. Data were then screened for univariate and multivariate outliers on outcome variables following Tabachnick and Fidell (2007) using Mahalanobis Distance (MAH) and studentized deleted residuals (SDRESID). A total of 58 cases were removed for exceeding +/- 3 standard deviations, or the criterion Mahalanobis Distance value (Fox 1991). This reduced our sample to 3,427 individual surveys from 376 interpretive programs.

Next we reviewed the number of valid respondents per individual interpretive program. Prior theory and research suggest that programs with a low number of attendees may be inherently different than programs servicing a larger number of attendees (Forist, 2003; McManus, 1987, 1988; Moscardo, 1999; Stern & Powell, this issue). We observed 272 programs with five or more attendees (see Stern & Powell, article 1, this issue for more extensive description). We chose this sample for the analyses conducted herein because it is most representative of programs in general and it provides a sample large enough to conduct structural equation modeling (Byrne, 2006). Because the program was our unit of analysis, our final step in data preparation included aggregating individual data at the program level by calculating the mean score of each visitor outcome for each program. For SEM purposes, all data was then grand mean centered (Tabachnick & Fidell, 2007)

Dependent variables: outcomes

Based on extensive input from the NPS and a review of literature, we developed three dependent variables (Table 1). The first dependent variable served as a measure of visitor satisfaction with the program on a scale from 0 to 10, with 0=Terrible and 10=Excellent. Two indexes were developed from other survey items following procedures outlined by DeVellis (2003) to represent visitor experience and appreciation and behavioral intentions. The items comprising each index were measured using a five-point Likert-type scale, with answer choices: Not at all (1), A little (2), Somewhat (3), A moderate amount (4), and A great deal (5). Composite scores were created for each of the scales by taking the mean of all items (for more detail, see Stern & Powell, this issue).

Program and interpreter characteristics

The independent variables used in this SEM analyses included both interpreter and program characteristics that met two criteria. We included ordinal variables that were correlated (p < 0.01) to the particular outcome in question in any context (See Stern & Powell, article 1, and Powell and Stern, article 3, this issue). We also included categorical variables with at least "moderate" effect size in association with the particular outcome in question in any context (Cohen's d > 0.5). The program characteristics (Table 2) were originally drawn from an extensive literature review aimed at identifying best practices i n the field (see Skibins et al., 2012). The interpreter characteristics were developed from the communications and education literature, though many of these factors are also referenced in the interpretation literature (Table 3). The tables also contain descriptive statistics. For more detail, see Stern and Powell (this issue).

Structural equation modeling

We used structural regression modeling (a.k.a. path analysis), a form of SEM, to examine the influence of different program and interpreter characteristics on three outcomes. We used SEM for this analysis because it is confirmatory (as opposed to exploratory) in nature and requires the researcher to have an explicit hypothesized model; it can model measurement error, which reduces inaccuracies; it allows for the analysis of a complete multivariate model including direct and indirect effects and in this case it can assess causal relationships between independent variables and a dependent variable (Byrne, 2006; Kline, 2005). In this study, all independent variables are formative (as opposed to reflective). That is, they were observed and represent a specific practice or attribute that is thought to directly influence the dependent variables (see Kline, 2005; Diamantopoulis et al., 2008; Diamantopoulis & Winklhofer, 2001; Jarvis et al., 2003, Padsakoff et al., 2007 for further explanation).

We used the EQS v6.1 software (Bentler 2005) to perform the statistical analyses, which progressed in several stages. First, the data were screened for univariate and multivariate deviations from normality. Next, we used structural regression modeling to assess the causal relationships between independent variables and each dependent variable (three separate models). For each outcome, we began with a model that contained all interpreter and program characteristics that met the criteria described above for that outcome. The starting list of program practices and interpreter attributes used in the hypothesized models are in Table 4. To develop the final structural regression models we used an iterative process in which diagnostics (modification indices: Lagrange Multiplier Test (LM), Wald Test) indicated potential modifications, including removal of independent variables from the model, to improve fit and parsimony.

Structural regression analysis provides multiple statistics that can be used to evaluate the "fit" of a specified model (Byrne, 2006). In this paper we report the Satorra-Bentler Scaled Chi-Square (S-B [chi square]), Robust Comparative Fit Index (CFI), Standardized Root Mean Square Residual (SRMR), the Robust Root Mean Square Error of Approximation (RMSEA) and its associated 90% confidence interval (Bentler & Yuan, 1999; Byrne, 2006). The S-B [chi square], which should be interpreted like a [chi square], is reported because it corrects for the degree of kurtosis in the data (Satorra & Bentler, 1994). The Robust CFI accounts for non-normality in the data and is an "incremental or comparative fit index" that evaluates the change in fit between the hypothesized model and the "independence model" (Byrne, 2006, 97; Bentler, 1990; Kline, 2005, 140). The independence model assumes that all the variables in the model are unrelated. The CFI represents the total covariation in the data and is measured on a scale of 0 to 1 with values greater than .9 indicating an acceptable fit and values greater than .95 indicating an excellent fit (Byrne, 2006; Hu & Bentler, 1999). The SRMR statistic provides the average difference between the sample and the predicted correlation matrices and thus is not susceptible to non-normality (Byrne, 2006). The SRMR uses standardized values with the range of scores between 0 and 1; values less than .1 are considered acceptable and less than .05 are considered a good fit (Hu & Bentler, 1995; Kline, 2005). The Robust RMSEA also accounts for non-normality in the data and is based on the average lack of fit per degree of freedom; therefore, as the fit improves, the RMSEA decreases. As such, this measure is sensitive to the degrees of freedom and the complexity of the model (Byrne, 2006). Like the SRMR, the scores range between 0 and 1, with values of .05 to .08 deemed acceptable and values less than .05 considered excellent (Browne & Cudeck, 1993; Hu & Bentler, 1999).

Beta weights in structural regression models reflect the effect size of an independent variable on the dependent variable. [R.sup.2] values gauge the predictive validity of the structural model, explaining the proportion of the total observed variance in the dependent variable explained by the model. It is recommended to assess [R.sup.2] values independently of fit indices, as the latter do not pertain to predictive validity (Kline, 2005).

Results

Three models were created based on the list of variables in Table 4. All independent variables (interpreter and program characteristics) were first entered as direct predictors of each outcome. In each case, the initial fit of each model was deemed unacceptable (Byrne, 2006). Through an iterative process, we adjusted the models using diagnostics that indicate potential model changes that would improve fit and parsimony. This generally involves removing variables one at a time based on statistical indicators produced at each stage of the modeling process. As the iterative modeling continues, it also can include adding or changing the nature of relationships between variables. In the end, a single "best fit" model is produced that represents the most parsimonious and predictive model for each outcome. The resulting models are displayed in Figures 1, 2, and 3.

Figure 1 represents the final model pertaining to how the interpreter and program characteristics influenced visitor satisfaction. Fit indices for the final "satisfaction" model (SB[chi square] = 5.39, p < .07; CFI = .99; SRMR = .029; RMSEA = .08) indicated the model was an acceptable representation of the relationships present in the data (Byrne, 2006; Kline, 2005). Authentic emotion was a strong predictor of interpreter's confidence ([beta] = .388, p < .05) and a weaker predictor of visitor satisfaction ([beta] = .171, p < .05). Organization was also a strong predictor of interpreter's confidence (([beta] = .300, p < .05), but not a direct predictor of visitor satisfaction. Confidence was a strong predictor of visitor satisfaction ([beta] = .307, p < .05). Appropriate for the audience was also a significant predictor of visitor satisfaction ([beta] = .183, p < .05). The model accounted for 35% ([R.sub.2]) of the variance in confidence and 27% ([R.sub.2]) of the variance in visitor satisfaction.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

The final structural regression model for visitor experience and appreciation had the same structure as the final visitor satisfaction model (Figure 2). Fit indices for the model (SB[chi square] = 4.45, p < .1; CFI = .99; SRMR = .027; RMSEA = .069) indicated the model was an acceptable fit of the data (Byrne, 2006; Kline, 2005). The only structural differences between this model and the satisfaction model involved the relative strength of confidence and appropriateness for the audience. Appropriate for audience ([beta] = .288, p < .05) was the strongest predictor of visitor experience and appreciation, followed by authentic emotion ([beta] = .139, p < .05) and confidence ([beta] = .068, p < .05). The model accounted for 35% ([R.sub.2]) of the variance in confidence and 16% ([R.sub.2]) of the variance in visitor experience and appreciation.

[FIGURE 3 OMITTED]

The model in Figure 3 represents how interpreter and program characteristics predicted intentions to change behaviors. Fit indices for the model in Figure 3 (SB[chi square] = 7.38, p < .05; CFI = .96; SRMR = .040; RMSEA = .03) indicated the model was an acceptable representation of the relationships present in the data. Having a goal to influence behavior ([beta] = .145, p < .05), appropriate logistics ([beta] = .153, p < .05), and humor quality ([beta] = .223, p < .05) were significant positive predictors of intentions to change behaviors. Use of sarcasm ([beta] = -.170, p < .05) was a significant but negative predictor of intentions to change behaviors. The model accounted for 10% ([R.sub.2]) of the variance in intentions to change behaviors.

Discussion: Is it the interpreter or the program?

We used structural equation modeling to examine the relative influence of interpreter and program characteristics upon visitor outcomes at live interpretation programs across the U.S. National Park Service. The resulting models reveal three main lessons. First, it appears in each case that both interpreter and program characteristics influenced visitor outcomes. Second, depending on outcome, certain program practices and interpreter attributes provided the best model fit and predictive power. Third, the final models accounted for a relatively low percentage of the overall variance in visitor outcomes. We explain each finding and some important limitations in the interpretation of the analyses below.

In each model, both interpreter and program characteristics influenced outcomes. The satisfaction and the visitor experience and appreciation models each contained authentic emotion and charisma, organization, confidence, and appropriate for the audience. In each model, authentic emotion and charisma and organization were mediated by confidence. In other words, the model suggests that authentic emotion and charisma and organization each help to create interpreter confidence, which in turn enhances visitor outcomes. Authentic emotion and charisma also served as a direct causal predictor of each outcome, as did the appropriate for the audience variable.

The final structural regression model of intentions to change behaviors suggests that humor quality, appropriate logistics, and intending to influence behaviors through a program positively influenced intentions to change stewardship behaviors. The use of sarcasm was associated with weaker intentions to change stewardship behaviors. In other words, interpreters that successfully employed humor, ensured that their audience's needs were met, and explicitly intended to influence their audience's behaviors were more successful at doing so. Meanwhile, overly sarcastic interpreters were less likely to influence changes in behavioral intentions. Interestingly, only 7% of all interpreters interviewed in the study explicitly intended to influence audience behaviors (Table 3). Ham (2013) reminds interpreters that outcomes, such as behavior change, do not happen magically; instead a program should be planned and developed with an outcome in mind. When focusing on behavior change, numerous techniques may increase the likelihood of influencing specific behaviors (Ham et al., 2007; Powell & Ham, 2008; Stern & Powell, this issue).

Certain limitations in the data and analyses are important to consider when interpreting these findings. First, structural equation modeling explicitly aims to produce the most parsimonious predictive or in this case causal model for selected outcomes. As such, independent variables that may be strongly related to outcomes are commonly removed during the modeling process due to their relationships with other variables. For example, the connection variable is highly correlated with organization, authentic emotion and charisma, and confidence (see Stern et al., this issue). As a result, it may be removed from a model because it explains a redundant proportion of the variance in the outcome as the other independent variables. This is the case with many of the program practices and interpreter characteristics tested in this analysis. It would be inappropriate to assume that their absence in the final models reduces their significance in influencing more positive outcomes.

Second, the models accounted for 10% to 27% of the variance in the outcomes. The strongest model accounted for 27% of the variance in satisfaction. The weakest model accounted for 10% of the variance in behavioral intentions. This suggests that much more is at play than simply the interpreter and the program elements. Interpretive programs are complex phenomena, and audience outcomes can be influenced by characteristics of the individual audience members, the makeup of the group, and the location and context of the program, in addition to characteristics of the program and the interpreter (Powell et al., 2009). Past research into communications (see Ajzen, 1992, for more) suggests that few consistent trends emerge when attempting to examine the range of source (interpreter) factors, receiver (audience) factors, channel (program) factors, and message (content) factors that influence outcomes resulting from communications. These factors vary with each program and produce an almost unlimited number of interactions and potential combinations (Falk, 2004). We examine a small portion of these additional factors in a separate article in this issue (Powell & Stern, this issue).

Relatively low [R.sub.2] values may also be a product of the lack of variance observed in satisfaction and visitor experience and appreciation scores. We discuss this issue in greater depth in a separate article in this issue as well (Stern et al., this issue). Predictive ability may be particularly low for behavioral change for a number of reasons. As noted earlier, few programs actually targeted behavioral change as an outcome. As such, changing behavioral intentions may have been more of a side effect than an intended outcome of a program. Moreover, many interpretive program goers may already perform many of the behaviors discussed in interpretive programs, leaving little room for change (see Stern & Powell, this issue, for a more detailed discussion).

Despite the limitations, the results suggest that outcomes are influenced by both program and interpreter characteristics and that these characteristics interact and influence each other. For example, confidence may ultimately emerge from an interpreter's passion for the resource and careful planning, which leads to good organization. Because most prior research and formal training have focused on what we have categorized as "program characteristics" (Skibins et al., 2012), we urge future researchers, trainers, and practitioners to give some meaningful attention to interpreter attributes and delivery styles. Training programs might add elements that could improve interpreters' abilities to project confidence and authentic emotion. Some lessons for doing so might be found in the formal education field, where "affinity-seeking" and immediacy behaviors have garnered some attention (e.g., Finn et al., 2009). These practices involve efforts to ingratiate teachers with their students by reducing the social distance between them (see also Stern & Powell, this issue; Stern et al., in press). Interpretive organizations might also consider these findings in light of the role of the individual interpreter in program development. If organizations can provide opportunities for creating and sustaining authentic connections between interpreters and the resources they interpret, they might enhance interpreters' abilities to convey their own passions to their audiences. Finally, we urge researchers to consider how different program and interpreter characteristics may function differently in varying contexts.

References

Ajzen, I. (1992). Persuasive communication theory in social psychology: A historical perspective. In M. J. Manfredo (Ed.), Influencing human behavior (pp. 1-27). Champaign, IL: Sagamore Publishing.

Archer, D., & Wearing, S. 2003. Self, space, and interpretive experience: The interactionism of environmental interpretation. Journal of Interpretation Research, 8 (1): 7-23.

Beck, L., & Cable, T. T. (2002). Interpretation for the 21st century: Fifteen guiding principles for interpreting nature and culture (2nd ed.). Champaign: Sagamore.

Bentler, P.M. 1990. Comparative fit indexes in structural models. Psychological Bulletin, 107: 238-46.

Bentler, P.M. 2005. Eqs 6 structural equations program manual. Encino, CA: Multivariate Software (www.mvsoft.com).

Bentler, P.M. and K.H. Yuan. (1999). Structural equation modeling with small samples: Test statistics. Multivariate Behavioral Research, 34: 181-97.

Brochu, L., & Merriman, T. (2002). Personal interpretation: Connecting your audience to heritage resources: Fort Collins, CO: InterpPress.

Browne, M.W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In Testing structural equation models, ed. K.A. Bollen and J.S. Long, 445-55. Newbury Park, CA: Sage.

Byrne, B.M. (2006). Structural equation modeling with eqs: Basic concepts, applications and programming. Second ed. Mahwah, NJ: Erlbaum.

DeVellis, R.F. (2003). Scale development: Theory and applications Applied social research methods. 2nd ed. Thousand Oaks, CA: Sage Publishing.

Diamantopoulos, A., Riefler, P., & Roth, K. P. (2008). Advancing formative measurement models. Journal of Business Research, 61(12), 1203-1218.

Diamantopoulos, A., & Winklhofer, H.M. (2001). Index construction with formative indicators: An alternative to scale development. Journal of Marketing Research, 38(2); 269-277

Falk, J. (2004). The director's cut: Toward an improved understanding of learning from museums. Science Education, 88(S1), S83-S96. doi: 10.1002/sce.20014

Finn, A. N., Schrodt, P., Witt, P. L., Elledge, N., Jernberg, K. A., & Larson, L. M. (2009). A meta-analytical review of teacher credibility and its associations with teacher behaviors and student outcomes. Communication Education, 58(4), 516-537.

Forist, B. (2003). Visitor Use and Evaluation of Interpretive Media. A Report on Visitors to the National Park System. National Park Service Visitor Services Project. http:// nature.nps.gov/socialscience/ docs/Visitor_Use_and_Evaluation.pdf. Accessed May 24, 2012.

Fox, J. 1991. Regression diagnostics. Newbury Park, CA: Sage.

Frauman, E., & Norman, W.C. (2003). Managing visitors via "mindful" information services: One approach in addressing sustainability. Journal of Park and Recreation Administration, 21 (4): 87-104.

Ham, S. (2009). From interpretation to protection: Is there a theoretical basis? Journal of Interpretation Research 14(2): 49-57.

Ham, S. H. (1992). Environmental interpretation: A practical guide for people with big ideas and small budgets. Golden, CO: Fulcrum Publishing.

Ham, S. H. (2013). Interpretation: Making a difference on purpose. Golden, CO: Fulcrum Publishing.

Ham, S. H., Brown, T. J., Curtis, J., Weiler, B., Hughes, M., & Poll, M. (2007). Promoting persuasion in protected areas: A guide for managers. Developing strategic communication to influence visitor behavior. Southport, Queensland, Australia: Sustainable Tourism Cooperative Research Centre.

Ham, S.H., & Weiler, B.M. (2002a). Toward a theory of quality in cruise-based nature guiding. Journal of Interpretation Research, 7 (2): 29-49.

Ham, S H., & Weiler, B.M. (2002b). Tour guide training: a model for sustainable capacity building in developing countries. Journal of Sustainable Tourism, 10(1): 52-69.

Hu, L.-T., & P.M. Bentler. (1995). Evaluating model fit. In Structural equation modeling: Concepts, issues, and applications, ed. R. H. Hoyle, 76-99. Thousand Oaks, CA: Sage.

Hu, L.-T., & P.M. Bentler. (1999). Cutoff criteria for fit indices in covariance structure analysis: Guidelines, issues, and alternatives. Structural Equation Modeling, 6: 1-55.

Jacobson, S. K. (1999). Communication skills for conservation professionals. Washington, D.C.: Island Press.

Jarvis, C.B., MacKenzie, S.B., & Podsakoff, P.M. (2003). A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research, 30(2); 199-218.

Kline, R.B. (2005). Principles and practice of structural equation modeling. 2nd ed. New York: The Guilford Press.

Knapp, D., & Benton, G. M. (2004). Elements to successful interpretation: A multiple case study of five national parks. Journal of Interpretation Research, 9(2), 9-25.

Knapp, D. & Yang, L. (2002). A phenomenological analysis of long-term recollections of an interpretive program. Journal of Interpretation Research, 7(2), 7-17.

Knudson, D. M., Cable, T. T., & Beck, L. (2003). Interpretation of cultural and natural resources, (2nd ed.). State College: Venture Publishing.

Larsen, D. L. (2003). Meaningful Interpretation: How to Connect Hearts and Minds to Places, Objects, and Other Resources. Fort Washington: Eastern National.

Lewis, W. J. (2005). Interpreting for park visitors (9th ed.). Fort Washington: Eastern National.

Madin, E. M. P., & Fenton, D. M. (2004). Environmental interpretation in the Great Barrier Reef Marine Park: An assessment of programme effectiveness. Journal of Sustainable Tourism, 12(2), 121-137.

McCroskey, J.C., Richmond, V.P., & Stewart, R.A. (1986). One on one: The foundations of interpersonal communication. Englewood Cliffs, NJ: Prentice-Hall.

McManus, P.M. (1987). It's the company you keep ... The social determination of learning-related behaviour in a science museum. The International Journal of Museum management and Curatorship, 6: 263-270.

McManus, P. M. (1988). Good companions: More on the social determination of learning-related behaviour in a science museum. International Journal of Museum Management and Curatorship, 7(1), 37-44.

Moscardo, G. (1999). Making visitors mindful: Principles for creating quality sustainable visitor experiences through effective communication. Champaign: Sagamore.

National Park Service. (2003a). Interpretive Development Program. Module 101: Fulfilling the NPS mission: The process of interpretation.

National Park Service. (2003b). Interpretive Development Program. Module 103: Preparing and presenting an effective interpretive talk.

National Park Service. (2003c). Interpretive Development Program. Module 210: Prepare and present an effective conducted activity.

National Park Service. (2003d). Interpretive Development Program. Module 220: Prepare and present an interpretive demonstration or other illustrated program.

National Park Service. (2003e). Interpretive Development Program. Module 230: Interpretive writing.

National Park Service. (2003f). Interpretive Development Program. Module 311: Interpretive media development.

Podsakoff, P.M., MacKenzie, S.B., Lee, J.Y., & Podsakoff, N.P. (2007). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5) 879-903.

Powell, R.B., Brownlee, M.T.J., Kellert, S. R. & Ham, S.H. (2012) From awe to satisfaction: Immediate affective responses to the Antarctic tourism experience. Polar Record. 48(2): 145-156.

Powell, R.B., Kellert, S. R., & Ham, S.H. (2009). Interactional theory and the sustainable nature-based tourism experience. Society and Natural Resources, 22(8): 761-776.

Powell, R. B., & Ham, S. H. (2008). Can ecotourism interpretation really lead to pro-conservation knowledge, attitudes and behavior? Evidence from the Galapagos Islands. Journal of Sustainable Tourism, 16(4): 467-489.

Regnier, K., Gross, M., & Zimmerman, R. (1992). The Interpreter's Guidebook: Techniques for Programs and Presentations. Interpreter's Handbook Series. Stevens Point: UWSP Foundation Press.

Satorra, A., & Bentler, P. M. (1994). Corrections to test statistics and standard errors on covariance structure analysis. In A. von Eye & C. C. Clogg (Eds.), Latent variables analysis (pp. 399-419). Thousand Oaks, CA: Sage.

Skibins, J.C., Powell, R.B., & Stern, M.J. (2012). Exploring empirical support for interpretation's best practices. Journal of Interpretation Research. 17(1): 25-44.

Sharpe, G. W. (1976). Interpreting the environment. New York: John Wiley & Sons.

Stokols, D., & Altman, I. eds. (1987). Handbook of environmental psychology. New York: John Wiley & Sons.

Tabachnick, B.G. and L.S. Fidell. (2007). Using multivariate statistics (5th ed.). Needham Heights, MA: Allyn & Bacon.

Tilden, F. (1957). Interpreting our heritage (3rd ed.). Chapel Hill: The University of North Carolina Press.

Veverka, J. A. (1998). Interpretive master planning: the essential planning guide for interpretive centers, parks, self guided trails, historic sites, zoos, exhibits and programs (2nd ed.). Tustin: Acorn Naturalists.

Wearing, S., & Wearing, B. (2001). Conceptualizing the selves of tourism. Leisure Studies, 7: 11-23.

Ward, C.W., & Wilkinson, A. E. (2006). Conducting meaningful interpretation: A field guide for success. Golden: Fulcrum.

Robert B. Powell

Department of Parks, Recreation and Tourism Management and School of Agricultural

and Forest Environmental Sciences, Clemson University

Marc J. Stern

Department of Forest Resources and Environmental Conservation, Virginia Tech

Table 1. Outcome (dependent) variables with descriptive statistics.

Outcomes                                         N    Mean   St. Dev.

Satisfaction: 0 to 10 scale                     272   8.94     0.64

Visitor experience and appreciation             272   4.41     0.32
  ([alpha] = .89): 1 to 5 scale

* Made my visit: to this park more enjoyable          4.55     0.30
* Made my visit to this park more meaningful          4.49     0.32
* Enhanced my appreciation for this park              4.36     0.37
* Increased my knowledge about the                    4.45     0.34
  program's topic
* Enhanced my appreciation for the National           4.27     0.36
  Park Service

Behavioral intentions ([alpha] = .94):          272   2.92     0.64
  1 to 5 scale

* Changed the way I will behave while I'm             2.92     0.67
  in this park
* Changed the way I will behave after I               2.92     0.61
  leave this park

Table 2. Program characteristics, their definitions, and descriptive
statistics.

Program          Definition
characteristic

Organization     Equally weighted composite mean score of 6 program
([alpha] =       characteristics:
0.82)
Scale:           * Quality of the introduction (Brochu and Merriman,
1 to 5           2002; Ham, 1992; Jacobson, 1999): Degree to which
Mean: 3.34       the introduction captured the audience's attention
S.D.: 0.71       and oriented (or pre/disposed) the audience to the
                 program's content and/or message.

                 * Appropriate sequence (Beck and Cable, 2002; Ham,
                 1992; Jacobson, 1999; Larsen, 2003): Degree to
                 which the program followed a logical sequence.

                 * Effective transitions (Beck and Cable, 7742;
                 Brochu and Merriman, 7742; Ham, 3342; Jacobson,
                 1999; Larsen, 2003): Degree to which program used
                 appropriate transitions that kept the audience
                 engaged and did not detract from the program's
                 sequence.

                 * Holistic story (Beck and Cable, 2002; Larsen,
                 2003; Tilden, 1957): Degree to which the program
                 aimed to present a Holistic story (with characters
                 and a pilot) as opposed to disconnected pieces of
                 information.

                 * Clarity of theme (Beck and Cable, 2002; Brochu
                 and Merriman, 2002; Ham, 1992; Jacobson, 1999;
                 Knudson, Cable, and Beck, 2003; Larsen, 2003;
                 Lewis, 2005; Moscardo, 1999; Sharpe, 1976; Veverka,
                 1998; Ward and Wilkinson, 2006): Degree to which
                 the program  had a clearly communicated theme(s). A
                 theme is  defined as a single sentence (not
                 necessarily  explicitly stated) that links
                 tangibles,  intangibles, and universals to organize

                 and develop  ideas.

                 * Link between introduction nnd conclusion (Beck
                 and Cable, 2002; Brochu and Merriman, 2002; Larsen,
                 2003): Degree to which program connected conclusion
                 back to the introduction in an organized or
                 cohesive way (i.e., program "came full circle.")

Connection       Equally weighted composite mean score of 5 program
([alpha] =       characteristics
0.88)
Scale: 1 to 5    * Link tangibles to intangible meanings and
Mean: 2.77       universal concepts (NPS Module 101; Beck and Cable,
S.D.: 0.78       2002; Brochu and Merriman, 2002; Ham, 1992;
                 Knudson, et al, 2003; Larsen, 2003; Lewis, 2005;
                 Moscardo, 1999; Tilden, 1957; Ward and Wilkinson,
                 2006): Communication connected tangible resources
                 to intangibles and universal concepts.

                 * Cognitive engagement (Knudson, et al., 2003;
                 Moscardo, 1999; Sharpe, 1976; Tilden, 1957;
                 Veverka, 1998)): Degree to which the program
                 cognitively engaged audience members in a
                 participatory experience beyond simply listening;
                 i.e. calls to imagine something, reflect, etc.

                 * Relevance to audience (Beck and Cable, 2002;
                 Brochu and Merriman, 2002; Ham, 1992; Jacobson,
                 1999; Knapp ami Benton, 2004; Lewis, 2005;
                 Moscardo, 1999; NPS Module 101; Sharpe, 1976;
                 Tilden, 1957; Veverka, 1998): Degree to which the
                 program explicitly communicated the relevance of
                 the subject to the lives of the audience.

                 * Affective messaging (Jacobson, 1999; Lewis, 2005;
                 Madin and Fenton, 2004; Tilden, 1957; Ward and
                 Wilkinson, 2006): Degree to which the program
                 communicated emotion (in terms of quantity, not
                 quality).

                 * Provocation (Beck and Cable, 2002; Brochu and
                 Merriman, 2002; Knudson, et al., 2003; Tilden,
                 1957): Degree to which the program explicitly
                 provoked participants to personally reflect on
                 content and its deeper meanings.

Appropriate      Degree to which basic audience and program needs
logistics        were met (i.e., restrooms, weather, technology,
Scale: 1 to 4    accessibility, shade, etc). (Jacobson, 1999;
Mean: 3.71       Knudson et al., 2003)
S.D.: 0.93

Appropriate      Degree to which the program aligned with audience's
for audience;    ages, cultures, and level of knowledge, interest,
Scale: 1 to 5    and experience. (Beck and Cable, 2002; Jacobson,
Mean: 3.93       1999; Knudson et al 2003)
S.D.: 0.70

Multisensory     Degree to which the program intentionally and
Scale: 1 to 3    actively engaged more than just basic sight and
Mean: 2.39       sound. (Beck and Cable, 2002; Knudson et al., 2003;
S.D.: 0.71       Lewis, 2005; Moscardo, 1999; Tilden, 1957; Veverka,
                 1998; Ward and Wilkinson, 2006)

Physical         Degree to which the program physically engaged
engagement       audience members in a participatory experience;
Scale: 1 to 4    i.e., through touching or interacting with
Mean: 1.42       resource. (Beck and Cable, 2002; Knudson, et al,
S.D.: 0.39       2003; Lewis, 2005; Moscardo, 1999; NPS Module 101;
                 Sharpe, 1976; Tilden, 1957)

Verbal           Degree to which the program verbally engaged
engagement       audience members in a participatory experience;
Scale: 1 to 5    i.e, dialogue (a two-way discussion). (Knudson, et
Mean: 2.71       al, 2003; Moscardo, 1999; Sharpe, 1976; Tilden,
S.D.: 1.42       1957; Veverka, 1998)

Fact-based       Program communicated only fact-based information.
messaging        (Frauman and Norman, 2003; Jacobson, 1999; Lewis,
Binary: 27%      2005; Tilden, 1957; Ward and Wilkinson 2006)

Clear message    Degree to which program's message (s) was clearly
Scale: 1 to 4    communicated; i.e., the "so what?" element of the
Mean: 2.70       program. (Beck and Cable, 2002; Brochu and
S.D.: 0.94       Merriman, 2002; Ham, 1992; Jacobson, 1999)

Consistency      Degree to which the program's tone and quality were
Scale: 1 to 3    consistent throughout the program. (Beck and Cable,
Mean: 2.78       2002; Ham, 1992)
S.D.: 0.37

Table 3. Interpreter characteristics observed in the study, their
definitions, and descriptive statistics for cases analyzed in this
paper.

Interpreter      Definition
characteristic

Confidence       Equally weighted composite mean score of 3
(a = 0.70)       interpreter characteristics:
Scale: 1 to 4
Mean: 3.28       * Comfort of the Interpreter (Lewis 2005; Moscardo,
S.D.: 0.49       1999; Ward and Wilkinson, 2006): Degree to which
                 the interpreter presenting the program seems
                 comfortable with the audience and capable of
                 successfully presenting the program without
                 apparent signs of nervousness or self-doubt.

                 * Apparent knowledge (Ham and Weiler, 2002a; Lewis,
                 2005; Ward and Wilkinson, 2006): The degree to
                 which the interpreter appears to know the
                 information involved in the program, the answers to
                 visitors questions, and has local knowledge of the
                 area and its resources.

                 * Eloquence (Lewis, 2005): The extent to which the
                 interpreter spoke clearly and articulately, and did
                 not mumble or frequently use filler words such as
                 "um" or-"like"

Authentic        Equally weighted composite mean score of 3
emotion and      interpreter characteristics:
charisma
(a = 0.85)       * Passion (Beck and Cable, 2002; Ham and Weiler,
Scale: 1 to 5    2002b; Moscardo, 1999): The interpreter's apparent
Mean: 3.57       level of enthusiasm for the material, as opposed to
S.D.: 0.85       a bored or apathetic attitude toward it. The
                 overall vigor with which the material is presented.

                 * Charisma (Ward and Wilkinson, 2006): A general
                 sense of the overall likeability/charisma of the
                 interpreter, commonly recognized by seemingly
                 genuine interaction with the visitors, including
                 smiling, looking people in the eye, and having an
                 overall appealing presence.

                 * Sincerity (Ham, 2009): The degree to which the
                 interpreter seems genuinely invested in the
                 messages he or she is communicating, as opposed to
                 reciting information, and seems sincere in the
                 emotional connection they may exude to the message
                 and/or the resource. In other words, the extent to
                 which the interpretation was delivered through
                 authentic emotive communication.

Responsiveness   The extent to which the interpreter interacts with
Scale: 1 to 3    the; audience, collects information about their
Mean: 2.81       interests and backgrounds, and responds to their
S.D.: 0.41       specific questions and requests or non-verbal cues.
                 (Jacobson, 1999; Knudson et al., 2003; Lewis,
                 2005))

Humor quality    How funny is the interpreter overall? Does the
Scale: 1 to 4    audience react positively to the interpreter's use
Mean 2.08        of humor and seem to enjoy it? (Ham and Weiler,
S.D 0.73         2002b; Knapp and Yang, 2002; Regnier et al., 1992)

Sarcasm          The degree to which the interpreter used sarcasm
Scale: 1 to 3    (the use of mocking, contemptuous, or ironic
Mean: 1.23       language or tone) or self-deprecation that was not
S.D.: 0.46       meant to be serious, as a part of presenting their
                 program.

Audibility       The extent to which the interpreter could be
Scale: 1 to 3    clearly heard and understood by the audience.
Mean: 2.86
S.D.: 0.36

Impatience       Exhibition of explicit impatience toward audience
Binary: 1.8%     members.

Goal: Behavior   Intention of the interpreter for the program to
Change           influence audience's behavior. (Ham, 2013)
Binary: 7%

Table 4: Variables included in hypothesized models for each outcome.

                                             Visitor
                                            Experience    Behavioral
Variable                    Satisfaction       and        Intentions
                                           Appreciation

Interpreter
characteristics

Audibility                       X              X
Authentic emotion and            X              X             X
  charisma
Confidence                       X              X             X
False assumption about           X              X
  audience
Goal: Behavior change                                         X
Humor quality                    X              X             X
Impatience                       X
Responsiveness                   X              X
Sarcasm                                                       X

Program characteristics

Appropriate for audience         X              X             X
Appropriate logistics            X              X             X
Clear message                    X              X             X
Consistency                      X              X
Connection                       X              X
Multisensory engagement          X
Organization                     X              X
Verbal engagement                X              X             X
Fact-based messaging             X