• Michael I. Eizenberg, Founder of the American Council for International Studies (ACIS). At the time of the interview, Mr. Eizenberg was ACIS president and CEO, a post he held from 1978-1997. Mr. Eizenberg began his career as an overseas "courier" (tour guide) for ALSG (above), for which he later served as an associate director before breaking away to form ACIS.

• Peter Jones. At the time of the pilot interview, Mr. Jones was Vice President in charge of operations for ACIS, where he had served in an executive capacity since the company's founding in 1978. In 1997, Jones was named president and CEO of ACIS. Jones also began his career as an ALSG courier, and served as an assistant director at ALSG for several years before joining Eizenberg to form ACIS.

• Dr. Theodore Voelkel currently serves as academic director for ACIS. More significant for this study, Voelkel co-founded ALSG with Markle in 1965, where he was employed variously as associate director of marketing and public relations and/or as associate executive director until ALSG's demise in 1993.

It was recognized that these four informants were initially "enculturated" into the industry by the same entity (ALSG), and were thus more likely to share cultural assumptions than might other informants who may have been enculturated elsewhere. However, these informants were chosen for the pilot study for three primary reasons: (1) taken together, they represented more longevity in the industry than any other combination of informants; (2) as a result, they possessed considerable depth and breadth of knowledge about the industry as a whole; and (3) all were accessible and willing to be interviewed not only for the pilot study, but also for the primary study.

To assure the informants' comfort level while they became acquainted with the nature of the project and to accommodate two informants' requests, pilot interviews were not taped. Instead, notes were taken as scrupulously and in as much detail as possible. Immediately following, these notes were transcribed via word processor into a format as closely resembling the actual interview as was possible for the researcher to reconstruct. Here, it should be noted that an adequate comfort level was apparently achieved; all informants indicated that taping subsequent interviews would be permissible.

Results of the pilot study indicated the following:

1. The original interview guide was determined to be too open-ended and flexible to yield a sufficient depth, breadth, and variety of data. As a result, it was here decided to adhere more closely to Phillips' original, more specific interview guide, rather than to the more loosely constructed one that was originally conceived. Appropriate modifications were made, the results of which may be seen in Appendix C.
2. It became readily apparent that the initial intention to conduct repeated face-to-face interviews, viz. Schein (1985), was not a reasonable expectation. Indeed, as Phillips points out, traditional qualitative protocols, when applied to organizational research, present a "variety of problems . . . . [they are] time consuming, costly to and intrusive upon the organization, and costly to the researcher both financially and in terms of the normal progression of an academic career . . . . In a study of industry culture rather than a single organization's culture, these problems are multiplied by the number of participating companies" (1990, 61-62). In short, it was simply not possible to secure from elites the amount of time necessary to conduct repeated interviews, at least not within a reasonable period of time. Hence, it was determined that (1) each elite who participated in the pilot study would, if at all possible, be interviewed a second time, using the revised and better-focused guidelines; and (2) other informants would be interviewed only once. However, given the rich data gleaned from the pilot study itself, it seemed reasonable to assume that "redundancy" (sufficient data) could be achieved in this manner.
3. When coded for thematic content, a surprising number of common concepts emerged from the relatively small body of interview data, an indication that the study was potentially feasible and fruitful. In fact, cultural data that emerged from the pilot interviews were both plentiful and striking.
4. The researcher's familiarity with the industry — albeit very dated familiarity — was perceived to be somewhat a negative factor in the first two interviews conducted. Specifically, the researcher failed, in these instances, to follow up several of the "taken-for-granteds" (e.g. assumptions) that came up in the interview, a situation that likely is attributable to assumptions (in common with informants or not) that the researcher herself made, as an erstwhile member of the culture. Once noted, this potential pitfall remained uppermost in the researcher's awareness, and every attempt was made to prevent its unduly biasing the research. In the end, given the continuity of results (discussed below), it was felt that this potential problem did not materialize, at least, not to the extent that it appeared to bias the study's findings.
5. Although most interviews were scheduled to last approximately one hour, most were extended by half an hour or more, which intuitively seemed to be a result of the participants' growing interest in the project, as the interview progressed. This tendency to "go overtime" was noted and future interview schedules for data collection were arranged accordingly.

Data Collection

Once the pilot study was completed, cooperation of additional informants was solicited by letter and personal contact and secured by the informant's written permission (Appendix A). Then, dates and times for face-to-face interviews were scheduled; in conjunction with the interviews, on-site visits were scheduled as well. In general, research was completed at one site before moving on to the next. Due to circumstances beyond the control of the researcher, three sites were not visited; therefore, data was gathered from these organizations via telephone interviewing. Only one organization of the eight members identified did not participate in the interviews.

Interview Protocol

Each informant was interviewed "in depth" — defined by Kahn & Cannell (1957) as "a conversation with a purpose" (149). For purposes of qualitative research, in-depth interviewing occurs when "the researcher explores a few general topics to help uncover the participant's meaning perspective but otherwise respects how the participant frames and structures the responses" (Marshall & Rossman, 1995, 80). This technique is consistent with uncovering "meaning" from a participant's point of view, a goal that is " fundamental to qualitative research — [that] the participant's perspective . . . should unfold as the participant views it, not as the researcher views it" (80).

As Marshall & Rossman (1995) also note, in-depth interviewing enables the researcher to "get large amounts of data quickly" (80) as well as to "understand the meanings people hold for their everyday activities" (81). Indeed, McCracken (1988) declares the long interview to be "one of the most powerful methods in the qualitative armory" (9) primarily because it "gives us access to individuals without violating their privacy or testing their patience" (11) as would extended observation, thereby fostering the achievement of "crucial qualitative objectives within a manageable methodological context" (11).

Interviewing is widely regarded to be a key component of the case study (Yin, 1994; Marshall & Rossman, 1995; McCracken, 1988); indeed, Yin declares that interviews are "one of the most important sources of case study information" (84), and moreover, that "key informants are often critical to the success of a case study [because they] not only provide . . . insights . . . but also can suggest sources of corroboratory evidence — and initiate access to such sources" (84).

Yin (1994) notes that the most common interview protocol is "open-ended" (84), although a more "focused" interview utilizing "a certain set of questions [albeit open-ended ones] derived from case study protocol" (85) may also be employed. It is this slightly more "focused" (85) yet flexible type of open-ended interview, based on the Phillips (1994) guide, that this study eventually employed. Originally, an even more open-ended guide was constructed; however, by the fourth interview of the pilot study (discussed above), it was determined that a guide more closely in line with Phillips' question set would prove more fruitful. Thus, a "fifth pilot" interview — which also served as the first interview of the actual study — tested and confirmed the modified interview guide. Here, it should be noted that as each interview progressed, most tended to gravitate naturally to areas that emerged as especially significant or cogent for each participant. As a result, it was not unusual for the "final" interview guide to be altered, in medias res, as each interview progressed. Generally, however, questions and answers/discussion centered around the topic are those indicated by the interview guide. Appendix B contains the interview guide for the pilot study; and Appendix C contains the "final" interview guide used for this study.

All interviews except two were audio taped, with the permission of the informants. Two informants preferred not to be taped, so copious notes of these informant's responses were taken by hand and transcribed on computer immediately following the interview, akin to procedures used in the pilot interviews.

In all, a total of 18 informants representing seven of the eight member companies identified participated in the study. Five of the seven participating companies were represented by a minimum of two informants; the other two were represented by the owner/CEO interview.

Each informant's identity was coded to indicate (a) personal identity, (b) number of interview (e.g. 1 of 1, 2 of 2, and the like) and (c) page of transcript. Although the key to the coding scheme shall remain unpublished and unknown to anyone but the researcher, this identification was deemed necessary for purposes of data analysis. In addition, it became necessary to abandon the anonymity rule in two cases. It became apparent that both Dr. Markle and Dr. Voelkel, because of their longevity in the industry and primary influences upon the culture, would need to be identified in the reporting of data. Thus, their permission to be identified was sought, granted, and verified in writing (Appendix D).

Once the interviewing process was completed, the primary data set comprised some 22 interviews. Although most informants were interviewed only once, for reasons discussed above, some were interviewed twice, including the pilot study participants, who were interviewed once during the pilot and once during the main investigation. Two pilot informants also corresponded regularly with the researcher via e-mail throughout the data-gathering stage of the study (in addition to having been interviewed), offering supplementary information that was pertinent to the study. Taken together, this correspondence was perceived to be tantamount to separate interviews because it was topically congruent with issues raised in/by the interview process. Thus, data gathered in this manner was transcribed and coded as additional interviews of these participants.

On the average, interviews lasted slightly over an hour. Taped interviews produced an average of 29 pages of transcribed text, whereas noted interviews produced only about half that, on the average. However, one informant was a particularly rich source of data whose two interviews resulted in some 108 pages of transcribed text.

Although time and financial constraints experienced by both the researcher and the participating entities precluded a more populous sample, the richness of the interview data was such that "redundancy" was clearly achieved, as will be discussed below. The "redundancy test" simply means that interviewing (data-gathering) should continue until such time that "no new information appears to be forthcoming . . . [which signals that] the researcher has finished the collection task" (Taylor, 1994, 269). As Taylor notes, "how long this takes depends . . . on the thoroughness of the research, the scope of the research study, and the shared patterns among the participants" (269). In this study, although "redundancy" was achieved fairly early in the process, the necessity of collecting data from as many member entities as possible altered the traditional conception of the redundancy test. That is to say, redundancy was perceived as attached to the entire set of informants, not to just the informants within each entity. Nevertheless, the important point to note is that although interview data may not have been totally comprehensive (no data ever are), sufficient redundancy was nonetheless achieved, and the data gathered was therefore sufficient to support the study's findings, and beyond.

Documents

Selected available documents were used as a secondary source of data, primarily to aid in corroboration and triangulation of interview data. Although several different types of documents (including memos, letters, position papers, historical data, financial records, and miscellaneous documents of historical interest) were available from two of the organizations, the only type of documentation available to the researcher in equal measure, representing all entities in the study, were standard marketing tools, including "early bird flyers," annual catalogues, teacher handbooks, and a variety of other related marketing pieces, which the researcher categorized as "miscellaneous promotionals." Of these, individual company catalogues proved the most helpful in corroborating interview data, as well as in providing additional insights and information.

Documents were secured in one of two ways. First, the researcher requested from each entity a standard information packet, one normally sent to potential customers/clients. All but one entity responded with packets that contained, de minimus, a company "catalogue" and some version of a "teacher handbook." In addition, if a company visit occurred, the researcher requested to see "any documents that might enhance the study" and suggested a list of possible inclusions. Responses to this request were mixed. Although the researcher asked to see the same set of documents in each case, some companies were more forthcoming than others, especially with historical documents and/or documents other than those published for public use.

Two notable exceptions to the tendency towards "closed files" proved to be a virtual treasure trove of historical documentation, however. For example, one informant provided the researcher with access to a near-complete set of industry catalogues dating as far back as 1971. Another informant with considerable longevity in the industry opened personal archives to the researcher. Taken together, these serendipitously acquired data enabled the researcher to "open a window to the past" (as it were) that might otherwise have remained closed.

Finally, the definition of "documentation" was extended to include member entities' web sites, which all but one of the companies maintained. Taken together, these sites provided a wealth of information not only about the industry itself, but about its perceived environment.

Yin notes that the purpose of document analysis is primarily "to corroborate and augment evidence" (Yin, 1995, 81) from the interviews — indeed, that is the "most important use" of documents, which themselves "play an explicit role in any data collection in doing case studies" (81). Marshall & Rossman (1995) commend the review of documents as "an unobtrusive method [for gathering data] . . . one rich in portraying the values and beliefs of participants in the setting" (85). Documents analyzed for this study proved useful in both these areas.

Transcripts and documents together comprised some 743 pages of text.

Methods of Analysis

All data (e.g. text), regardless of source, was analyzed, first by using a thematic coding, then by conducting an interpretive analysis of themes that emerged. "Although there are few descriptions of this process in the literature . . . [it generally] involves noting regularities in the [unit of analysis] chosen for study" (Marshall & Rossman, 1994, 114).

Interviews

First, the researcher transcribed and numbered all lines of interview text, using Microsoft Word, in order to facilitate the coding process.

The actual coding process was somewhat complicated, and perhaps may be most effectively visualized as a "2X2" procedure whereby interviews and then documents were each analyzed in two contexts: (1) Phillips' cultural assumption categories and (2) Pacanowsky and O'Donnell-Trujillo's list of cultural performance types.

As suggested by and extrapolated from Mohan (1993, 171) and Strauss & Corbin (1990, 73), each interview was read holistically, to "grasp the dominant

thematic content" (171). These holistic themes were recorded on notecards and set aside for possible use as categorical labels later in the process. Next, each interview was coded for "categories, relationships, and assumptions that inform the respondent's view of the world in general and the topic in particular" (McCracken, 1988, 42) — a procedure that began with "open coding" (Strauss & Corbin, 1990, 61) of the data. "Open coding" is a process wherein "data are broken down into discrete parts, closely examined, compared for similarities and differences, and questions are asked about the phenomena as reflected in the data" (62). Open coding begins when phenomena are identified and labeled, which may result in "dozens, even hundreds of conceptual labels" (65). In this study, the initial unit of analysis was set as each disparate "idea" or "event" — which usually required a line-by-line dissection of text. This initial coding step surfaced approximately 178 conceptually-labeled phenomena.

Following this step, the concepts were grouped into logical categories "that seem[ed] to pertain to the same phenomena" (65). In keeping with the study's first research question, categories were based generally on Phillips' (1990) categories of cultural assumptions in industries, and more specifically, on the subcategories previously found to exist within the assumptions — in short, conceptual labels were grouped according to appropriate Phillips' subcategories. As this process continued, all data that did not seem to "fit" within this schemata (e.g. data that might eventually suggest modifications of the Phillips' categories) were highlighted and recorded in separate, additional, and descriptively labeled categories or subcategories. Moreover, and as suggested by Phillips (1990), "to assure that themes not directly categorizable within the typology were not overlooked . . . all issues emphasized during the course of each informant interview were recorded separately" (119). In other words, in addition to categorization, separate notations were made of the issues/codes deemed to be most salient within each interview. This was done primarily to mitigate the effects of any researcher or design bias.

Thus, even though some may argue that in using Phillips' categories and subcategories, the researcher effectively imposed her own reality on the data, the important points to note are these: (1) The study's attempt to partially replicate Phillips' study required at least an initial employment of her typology; however, (2) the special care taken to note and code any relevant data that fell outside the perimeters of the Phillips typology should tend to mitigate researcher or design bias.

Once the text was coded, categorized, and examined in light of the first research goal, interview transcripts were completely re-coded in the context of the second goal. This time, however, since specific communicative processes were sought, these processes were identified and classified (coded) as they occurred in the transcripts, relying upon the "heuristic listing" of performance categories and subcategories suggested by Pacanowsky and O'Donnell-Trujillo (1983). For example, when a "metaphor" was discovered, its content was written down and coded under the category labeled Performances of Passion; then, it was further categorized in terms of its performance "type" (e.g. metaphor, story, vocabulary, jargon, and construct). Early in this process, it became apparent that, regardless of their content or theme, all performances relevant to this study were in fact Performances of Passion (e.g. storytelling and repartee) in that they were enacted through the use of language. Moreover, it became obvious that in order to arrive at a first-hand understanding of the performances of Sociality, Politics, and in most cases, Enculturation, one would need to become a participant-observer within each of the entity's various industries, and for a significant period of time, at that. This was not possible, given the limitations imposed by time, expense, and relative brevity of access. Likewise, although Performances of Ritual were discussed by each informant, data gathered within the interview format precluded the actual observation of that performance, relying instead upon the verbal description of that behavior. Thus, it was determined that Performances of Passion &# stories and repartee, including metaphor, jargon, vocabulary, and relevant constructs &# would alone comprise this study's analysis. Because it reveals informants' unique "way[s] of talking about" what they do (Pacanowsky & O'Donnell-Trujillo, 1983, 128), an analysis of the language-reliant performances in this category was adjudged to be sufficient for meeting the instant study's goals. In short, and to paraphrase Hall (1959), culture is manifested in informants' use of language, and their use of language is the culture &# at least, in the scope of this study's data.

After these "performances" were noted and categorized throughout the data set, then each performance category (e.g. each metaphor, each use of specialized language, each story, and the like) was re-coded for thematic content, using the open coding process described above. During this stage of coding, it became apparent that the Phillips' (1990) typology of cultural assumptions might be overlapped with Pacanowsky & O'Donnell-Trujillo's various Performances of Passion. Once completed, this grid presented a very interesting and arguably effective means of framing the discussion of how cultural assumptions are manifested in communicative performances. Thus, it was decided to proceed with this cross-referencing format, in hopes of illuminating more fully the thematic nature of the performances themselves, and thereby of describing more precisely the role that communicative processes play in the "spinning" of cultural assumptions. Figure 2-1 shows how the grid for cross-referencing was initially conceptualized and constructed.

Table 2-1: Performance Categories Cross-Referenced with Assumptions

Documents

Documents were photocopied and their lines were numbered by hand, line-by-line and page-by-page. Thereafter, document analysis was conducted along the same lines as analysis of interview transcripts; that is, each document was subjected to the open-coding, categorizing, and cross-referencing processes previously described. However, as explained above, because of marked dissimilarities of documents accessed across organizations, comparative text was limited primarily to the standard marketing materials published for public consumption by each entity (including those found on company websites). As a result, document analysis served the somewhat constrained — albeit important — purposes of corroborating and augmenting (Yin, 1995) data generated in the interviews.

Finally, historical documents gleaned from the two "serendipitous" archival searches were coded, using the open coding and categorization processes described above; however, this data set was kept separate from the "main" set of contemporary documents, so as to mitigate any historical bias that might have contaminated more current data.

Validity

Triangulation — "the act of bringing more than one source of data to bear on a single point" — (Marshall & Rossman, 1995, 144) has gained wide acceptance as a strategy to enhance the general validity of the qualitative study. In the instant study, triangulation occurs in two ways. First of all, interviews of the informants triangulate each other. As well, document types common to each "entity participant" were analyzed, primarily to triangulate interview data. As Marshall & Rossman (1995) point out, "a study in which multiple cases, multiple informants, or more than one data-gathering method are used can greatly strengthen the study's usefulness for other settings" (144). Moreover, as Taylor (1995) notes, "good qualitative researchers always worry about . . . whether a pattern has been discovered or if one has been imposed" (276). To ensure discovery, as opposed to imposition, Taylor goes on to suggest that participants be allowed to "check" the researcher's interpretations (276). To that end, a brief summary of the cultural assumptions surfaced by the data was prepared and given to each participant (Appendix F). Participants were asked to comment on the summary as they saw fit, especially with regard to their general feelings about its accuracy and its applicability to their own company and/or situation. Feedback from these summaries indicated that although there was not complete agreement with the researcher's findings, the majority of responses did, in fact, confirm that common assumptions exist across this industry, and that the researcher's articulation of those assumptions is generally perceived to be accurate.

Findings will be presented in Chapter 3 and discussed in Chapter 4.

---

All dissertation material copyright © Linda Gayle Lyle, 1998. All rights reserved.
E-mail: llyle@utk.edu