In the research study , Crisis Negotiations: a Preliminary Investigation of Facework in Naturalistic Conflict Discourse, Randall G. Rogan and Mitchell R. Hammer analyze the concept of face in high tension situations.  Rogan and Hammer chose crisis negotiations as the representative subject for tense situation analysis.  They conceptualized crisis negotiations as “unique communicative situations in which one or more individuals use themselves or other individuals to bargain with authorities” (p 216.)  They suggest crisis negotiations follow a general sequential pattern beginning with perpetrator’s motives followed by a police response resulting in emotional excitation.  In these situations, they say, the authorities’ primary objective centers around shifting communication from crisis interaction to normative interaction.

            The cross sectional subjects of Rogan and Hammer’s study were three FBI transcripts of actual crisis negotiations conducted in the field.  The subjects include a person ready to commit suicide, an emotionally unstable person and a hostage incident involving a father and his children.  These subjects seem to offer a good range for examining the concept of crisis negotiations.  The object of this content analysis was to independently code these transcripts for facework features and then analyze the results of that coding scheme.  They stress that not much literature exists around this type of study.

            Rogan and Hammer borrowed face theory- whose origins they attribute to the 1967 work of Goffman- to aid them in operationalizing their research.  The area of interest primarily borrowed from Goffman’s work is concern for self projected image.  Rogan and Hammer conceptualize facework as an immediate phenomenon taking place spontaneously within the context of communication and dynamic social interactions.  Their concept of a facework system consists of a three dimensional model deconstructed into three areas of interest.  These three areas of interest include: the communicator’s locus of interest (i.e. either self directed or other directed); a face valence scale ranging from threat to honor; and finally, a temporal feature which determines whether a statement restores or defends face.  With their three dimensional operating model, Rogan and Hammer initially distinguish a seven code scheme with which to perform content analysis.  These codes include: Defend Self Face, Attack Self Face, Restore Self Face, Neutral Face, Defend Other Face, Attack Other Face and Restore Other Face.

            One area of concern I noticed with this study is that Rogan and Hammer do a good job illustrating what they mean by the initial seven codes listed; but then they add two more codes to the scheme with little explanation as to why and less reference for what the codes mean in this study.  Positive Backchannel and Negative Backchannel were the two codes added to the study.  The authors only suggest that these additional codes give a broader understanding of the facework taking place in crisis negotiations.  As stated in the research questions, Rogan and Hammer are attempting to validate their three dimensional model of facework; positive and negative backchannel seem to fall outside of their three dimensional operating model and therefore seem problematic.  The addition of these two codes would seem to challenge both the validity of the three dimensional model itself, but especially the reliability of the study.  In order to reproduce these results, a researcher would need to know what positive and negative backchannel mean.  The authors never really clarify these terms.

The three FBI transcripts of interactions between perpetrators and negotiators form the sample selected for this study.  There’s no mention of how these transcripts were sampled, so it is assumed they utilized purposive sampling which may lead to concerns with the representativeness of their results.  Rogan and Hammer trained research assistants in the use of their nine item coding scheme and this presents a couple mores areas of concern with their research.  One, they never mentioned how many coders were involved in this study and the second concern is about content analysis in general.  What affect does code training have in manufacturing false reliability between coders?  It is difficult at times to determine what exactly is being measured.  Is it the crisis negotiations taking place in real life, or is it really just the artificial codes being measured?  If the latter is true then the research may lose credibility based upon lack of applicability in the field.  To the authors credit though, they do not offer much in the way of recommendations since they acknowledge the field of study is so new.  They also stated that since their research was unique, they would not attempt to offer any hypothesis.

Rogan and Hammer chose uninterrupted speaking turns as the unit of analysis for this research.  From the FBI transcripts, more than 1800 units of analysis were produced.  Interestingly, about half of the units of analysis were attributed to the perpetrator and the other half to the negotiator.  Due to ambiguity in some of the units of analysis, coders were instructed to utilize a Gestalt impression of the unit or to determine the “dominant level of facework conveyed by message” (p. 221.) 

To pre-test coding reliability after training, coders were given 150 units of analysis for independent coding.  The results from this pre-test suggest an 81% level of agreement on coding resulting in a kappa of .79.  With these results, Rogan and Hammer cautioned readers that their findings were tentative.  This portion of the research also offers another area of concern.  The coding scheme seems to suffer from a certain lack of clarity in accounting for ambiguous facework statements which generates the need to add two new codes as well as utilizing Gestalt interpretation methods further eroding future reliability of the study.

Rogan and Hammer posed three research questions- with no hypothesis- to focus their study.

1. What facework behaviors are present in perpetrator and negotiator interactions in crisis negotiations? 
2. What is the relationship between perpetrator-facework behavior and negotiator-facework behavior? 
3. To what degree is the three-dimensional coding scheme used in the study a reliable measure of facework behavior? (220)

The kappa result of coding 1800 units of analysis within the three crisis negotiation transcripts was .70.  Individually, each negotiation scenario resulted in a relatively minor kappa differential varying between .73 for crisis negotiation scenario three and .65 in scenario two. 

Construct validity for results was examined in the post test.  After completing the research study, two trained coders were selected to independently code the same 85 facework expressions developed by Lim and Bower.  Construct validity was determined to be positive because these two coders reached 92% agreement on the 85 expressions.  However, the authors fail to mention how these two coders were sampled from the original pool of coders.  It is also unclear whether these two coders were even involved in the original study, although it is to be assumed.

Analysis of the data results from the coding exercise took the form of a frequency analysis and a lag sequential analysis.  Some of the more interesting results from the frequency analysis were an absence of Attack Other Face codes.  The researchers also noted a high percentage of negotiator Restore Other Face codes and a high percentage of perpetrator Restore Self Face codes.  The lag sequential analysis produced interesting suggestions, such as, a seemingly dualistic nature surrounding the code for Restore Self Face.  In the crisis interactions it was determined that Restore Self Face behavior exhibited by negotiators seemed to result in Restore Self Face behavior by perpetrators.  The researchers suggest elements of Attack Other Face may exist within the coding of Restore Self Face behavior.  This could be a communication phenomenon unique to crisis negotiations; however, this could also be a phenomenon unique to the coding scheme.

In conclusion, the researchers reiterated their intent was to conduct an exploratory investigation to produce descriptive statistics.  They felt their three dimensional model possessed construct validity and provided fairly reliable coding for speaking turns.  They felt their data was more informative when analyzed by incident type (i.e. scenario one vs. scenario two) rather than forming a sweeping analysis of crisis negotiation in general.  The data could be valuable to crisis negotiators by providing insight into communication cues signaling a perpetrator’s intent to become violent or destructive.  Rogan and Hammer hypothesize, at the end of their project, that crisis negotiators should focus on expressive needs more than instrumental needs through a form of communicative social support.  They felt this could best be achieved by focusing on the strategic content of the negotiator’s message behavior.

Rogan and Hammer’s admitted areas of weakness in this study involve the unit of analysis and coder discretion.  There was observed difficulty on the part of the coders to distinguish between proactive and retroactive facework.  In the midst of a dialogue, they observed the temporal function becoming ambiguous and subject to interpretation.  They also expressed a concern over the uninterrupted talking turn unit of analysis possibly being too large to code reliably.  Their suggestions for future research in this area include: coding smaller units of analysis (e.g. thought units) and allowing multiple codes on the same unit of analysis.  They acknowledge a need for more data analysis before making field recommendations as well as a need to study the effects cultural background has in crisis negotiations.