The Nature of Supportive Listening II: The Role of Verbal Person Centeredness and Nonverbal Immediacy

Abstract

This study examines an untested research assumption that a key component of supportive communication is active listening. Participants (N = 383) viewed a 5-minute conversation featuring a person who disclosed an emotionally upsetting event to a confederate who provided emotional support that varied in verbal person centeredness (VPC) and nonverbal immediacy (NVI). Participants then evaluated the extent to which the support provider was an active listener. Results showed that helpers who used higher levels of both VPC and NVI were rated as better listeners than those who used less person-centered and immediate support, although effect sizes were small. Results were also dependent on the operationalization of active listening.

Keywords:

• Active Listening
• Comforting
• Nonverbal Immediacy
• Social Support
• Stress

Acknowledgments

Preparation of this manuscript was partially supported by a Summer Research Grant awarded to Graham Bodie from the College of Arts & Sciences at LSU. A previous version of this manuscript was awarded Top Conference Paper at the 2011 meeting of the International Listening Association, Johnson City, TN.

Notes

Note. NVI = Nonverbal Immediacy; VPC = Verbal Person Centeredness.

Note. The AELS ranged from 1 to 7, whereas the ALOS ranged from 1 to 5.

Note. The AELS ranged from 1 to 7, whereas the ALOS ranged from 1 to 5.

Brant Burleson passed away in December 2010. We would like to express our deepest gratitude for having learned from and worked with Professor Burleson for many years. He was an outstanding mentor, a caring friend, and most of all, a supportive listener.

Using a sample of the total population of conversations as opposed to only a single example of each theoretically interesting cell rests on the notion of representativeness. First, within-condition differences of confederate VPC and NVI behaviors exist. For example, some confederates may have leaned in a bit more than others in high NVI conditions. Similarly, some confederates may have expressed a bit less person centeredness in low VPC conditions. On the whole, however, manipulation checks from trained coders confirm that conversations varied as expected for both person centeredness and immediacy. To capture the slight variations in immediacy cues and person-centered messages within each of the nine conditions, we wanted to use multiple conversations for each of these nine conditions. With respect to person centeredness in particular it seemed particularly important to capture the various ways with which each person-centered message level can be expressed (i.e., operationalized). We also wanted to make sure that other confounding factors, such as conversational topics, physical attractiveness, dialects/accent, and/or attires are randomly distributed and not patterned.

Each scale used as a manipulation check in this study was submitted to a confirmatory factor analysis to ensure it conformed to its theoretical measurement model. Fit statistics for each model indicated good fit (all CFIs > .90, SRMRs < .08, RMEASs < .05 with upper CIs < .08), and these individual results are available upon request. The only exception was the measurement model for the NVI manipulation check for participant observers, χ² (35) = 248.92, CFI = .92, SRMR = .04, RMSEA = .12 (.11, .14). Notably, the RMSEA is somewhat high; however, Chen, Curran, Bollen, Kirby, and Paxton (2008) recently demonstrated inflated Type 1 error rates for adhering to the conventional RMSEA cutoff and argued that the RMSEA must be interpreted more carefully in conjunction with the sample size, as well as other fit indices. In line with this reasoning, we retained the NVI model because the overall fit statistics suggested a correctly specified model.

Although evaluating omnibus effects in the traditional factorial design is more popular, “It is a relatively inefficient way to evaluate theories” (Furr & Rosenthal, 2003, p. 46) primarily because the omnibus ANOVA tests the residual explained sums of squares left over after the main effects and all lower interaction terms are removed (Rosnow & Rosenthal, 1989). In the current design, the 3 × 3 interaction term in omnibus ANOVA is defined as SStot − (SSa + SSb + SSc + SSab + SSac + SSbc + SSerror). That subtraction divided by MSerror is usually not theoretically meaningful (Rosnow & Rosenthal, 1991). Most germane to this study, the omnibus ANOVA conflates the linear and quadratic effects of the independent variables. Since our prediction relies on the strength of the linear effect, we test specifically for this effect. This allows for greater statistical power and more streamlined results (see O'Keefe, 2007). Unless otherwise noted, the deviation from the linear trend was not statistically significant.

The mean rating for the verbal and nonverbal factors for the judgement task involving “friendly” did not differ significantly (p = .74).

We assessed this with a simple question asking “What relationship do these two people appear to have?” with the following responses: They just met; they are close friends; they are distant acquaintances; they are family; they are romantic partners.