This study was a qualitative study but also minimally used quantitative measures. Glesne and Peshkin (1992) asserted that "although some social science researchers (Lincoln & Guba, 1985; Schwandt, 1988) perceive qualitative and quantitative approaches as incompatible, others (Patton, 1990; Reichardt & Cook, 1979) believe that the skilled researcher can successfully combine both" (pp. 8,9). Glesne and Peshkin stated that the use of two seemingly incommensurable approaches is, in practice, somewhat resolved by having one methodology serve in a supportive role of the other. I employed quantitative measures only to gather profile information about participants and to aid me in the identification of reoccurring categories addressed by my interviewees. The primary qualitative data collection technique used in my research was interviewing. Multiple types of data were collected and qualitative methods of data analysis were supported through the use of the AQUADÔ qualitative analysis computer package.
 
 

Key informants were chosen because of their experience with local desegregation and civil rights issues and greatly helped me shape my research during the pilot study. Key informants represented both the black and white communities and a wide range of socioeconomic success. One key informant was a successful black teacher who has been an invited speaker to many types of gatherings to speak on the topic of racism. Another key informant was a white journalist and historian known for his knowledge of all aspects of St. Joseph’s history.

Newspaper articles, scholarly journals, the internet, state and local records, school board records, and miscellaneous papers and letters all contributed as archival sources to this study. Although one cannot interact with such documents as in an interview, they do provide a stable source of context data against which interview results can be interpreted (Erlandson, Harris, Skipper, & Allen, 1993). In addition, tape recordings of the interviews allowed for a kind of stability in that any time during the duration of the study they could be revisited and compared against original interpretations as a hedge against bias and to allow for the benefits of a prolonged engagement with the interviewees.

Participants in the interviews were chosen from several sources: (a) upon recommendation from a key informant, (b) from school attendance records, and (c) from my personal knowledge of past friends and acquaintances. Each interviewee also completed a survey. Instead of the random sampling employed by quantitative studies, I selected each interviewee by means of purposive sampling, a sampling technique which is guided by an established set of criteria toward the end that those chosen will have relevant knowledge and will be likely to contribute positively to the study. Thus, my reason for selecting each participant was not to gather a representative sample of the target population, but rather to choose participants who had a story to tell, who were reflective about their experiences, and who were articulate enough to translate their experiences and insights into a narrative form.

Interviewees in the study were promised anonymity in order to encourage their full participation in sharing of their experiences and reactions to those experiences. As a lifelong resident of St. Joseph, I knew firsthand that the black population is careful about their criticism of the dominant social group and may feel freer to talk openly under a cloak of anonymity. As it turned out, my perception was not totally accurate; many did not care if their identities were revealed.
 
 

Three broad areas were targeted by the survey: (a) experiences of the participants, (b) their perceptions of mechanisms used to effect desegregation, and (c) participants’ perceptions of the outcomes of desegregation. These three areas corresponded to the same areas addressed by the research questions and the interview questions. For most survey items, a five point Lykert scale was used with 1 being greatly disagree and 5 being greatly agree.

By way of quantitative support, the AQUADÔcomputer program aided in the coding of meaningful segments of interview data by counting the number of occurrences of selected words and topics. This allowed me to quickly identify portions of text for closer scrutiny. In addition, AQUADÔ aided in locating linkages between words and concepts by counting and identifying the number of occurrences in which a target word was found to be within a given number of words of a second target word. Such quantitative results sped up the process of assigning codes and determining relationships between blocks of coded text.
 
 

Grounded theory (Denzin & Lincoln, 1994; Erlandson et al, 1993; Glaser, 1978; Glaser & Strauss, 1978; Guba & Lincoln, 1994; Hutchinson, 1988; Strauss & Corbin, 1990; Strauss & Corbin, 1994) requires that theory emerge from a recursive process of interaction with data collection and data analysis. Analysis of initial data is used to inform and refine the data collection instrumentation. Analysis of the data is accomplished through three phases of coding, which may overlap. Open coding is the initial sorting of data into relevant categories. In axial coding the researcher refines the categories and begins to relate the categories together. Finally, in selective coding the dominant category to which all others are related is identified and its relationship to the other categories is tested. Open-ended questions are typically used to elicit unexpected findings from participants. Sampling is not random as in quantitative research where it is important to be able to generalize results. Instead, sampling is purposive in order that information-rich cases may be selected for analysis (Patton, 1990). Therefore, the researcher must select sources which will provide the most relevant information to the research topic and which aid in development and verification of theory. Early in my study, I chose participants that could aid in the identification of categories. I analyzed the data as I continued to collect new data and identify potential categories within which the data seemed to be logically clumped. As relationships between categories were discovered, I used purposive sampling to verify test and verify relationships. Finally, near the completion of data collection, I again employed purposive sampling to confirm and verify the emergent theory. Unlike quantitative research, there is no ideal sample size since the researcher is interested in quality of data over quantity (Erlandson et al, 1993).

I turned to the work of Strauss and Corbin (1990) for four criteria to ensure the quality of my emergent theory: (a) it should fit the phenomenon being studied by emerging from diverse data from the field and by remaining faithful to the lived reality of the participants, (b) it should offer understanding which can be verified by both the participants and experts in the field of study, (c) given that the data is rich enough and that the theory is broad enough, it should provide for generality to similar contexts, and (d) it should provide control by accurately stating the conditions under which the theory applies and, thereby, providing a possible foundation for action.

Strauss and Corbin (1990) termed grounded theory, "a qualitative research method that uses a systematic set of procedures to develop an inductively derived grounded theory about a phenomenon" (p. 24). Therefore, contrary to traditional positivist methodology, I did not begin my research with a theory but, through ongoing data collection and analysis, created theory which explained and linked together the principal concepts identified in my study. In addition, I found grounded theory to be especially suited to a critical perspective since it can be viewed as a form of social criticism (Hutchinson, 1988):

Strauss (1987) described data collection as the process of finding, gathering, and generating materials to be analyzed over time in order to give rise to a grounded theory. In my study, I gathered data from official records, documents, personal experiences, participant surveys, key informants, and participant interviews. These multiple sources provided for a rich triangulation of data (Glesne & Peshkin, 1993; Marshall & Rossman, 1989). I taped recorded interviews and transcribed each prior to importing them into the AQUADÔ analysis environment. In order to leave an extensive audit trail I created various record keeping forms and kept entered data after each interview or acceptance of a historical record or document. My goal for data collection was to identify categories of meaningful text, determine linkages or relationships between categories, and identify core categories which contribute strongly to an emergent theory (Strauss & Corbin, 1990). These three phases often overlapped.

I coded data according to the protocols initially offered by Strauss (1987) and later refined by Strauss and Corbin (1990). My analysis was interwoven with collection and coding as the level of the coding matured and moved to higher conceptual levels. The three levels of coding outlined by Strauss and Corbin (1990) were adhered to in my study and are addressed separately below.
 
 

Open Coding

My initial goal for open coding was to identify segments of text which made sense and which were relevant to the study and to assign a conceptual label to each. Meaningful portions of text were originally selected and coded if they fit any of the following criteria: (a) demonstrated relevance to the research questions, (b) detailed experiences related to desegregation, (c) demonstrated understanding of the mechanisms used to bring about desegregation, or (d) addressed issues of the outcomes of desegregation. In addition, it was important that I not miss any important information. Therefore, I gave special attention to segments which occurred often in any text. For instance, AQUADÔ reported that the interviewee I’ve renamed as Mr. Watkins volunteered meaningful text related to the concept of perseverance nine times in our one and one half hour interview. As an example, one of those coded segments read, "So, I got scared. But at the same time, I told myself that I’m going to make it. I got to work on that." Including this quantitative measurement allowed me to consider concepts which I may have otherwise missed.

Next, I compared and examined coded segments of text within their contexts in order to assign one conceptual label to all segments addressing the same phenomenon or concept. Many times this label was more conceptual than the labels attached to individual segments of text. Concept names sometimes came from the words of the participants, what Strauss (1987) called "in vivo" codes (p. 33). Often, the labels came from my notes as I reflected on the data immediately after gathering it. The concept labels remained tentative, however, as analysis was ongoing and could at any point require a refining of categories. I used theoretical memos in order to jot down questions and concerns about categories as both a way in which to refine them and to keep them in my field of attention. Strauss (1987) called these memos, "writing in which the researcher puts down the theoretical questions, hypotheses, and summary of codes" (p. 22). My memos were very informal and often included arrows tentatively joining categories in preparation for the next phase of coding.

Axial Coding

Axial coding was a grueling process in which categories were analyzed in relationship with other categories in order to identify linkages and patterns between them. To paraphrase Strauss and Corbin (1990), each category is analyzed to identify any conditions which caused it to occur, any other categories it generated or within which it is situated, any mechanisms which manage the category, and the results of effects of that management. What often occurs at this stage is that a set of categories may be recognized as being inter-related and a new, meta-category created. This new category then subsumes the earlier ones as subcategories. For instance, during the middle stages of data collection and analysis, I found that early categories of "grades," "like school," "graduation," "aspirations," "attendance," and "perseverance" were inter-related and were all being used to talk about a more comprehensive conceptual category which I named "achievement." This hierarchical creation of related categories became the "paradigm model" (Strauss & Corbin, 1990, p. 99) which emerged during axial coding and contributed to the products of the selective coding phase.
 
 

Selective Coding

In selective coding a core category was created which was robust enough to explain much of the behavior noted in the study. As I analyzed the linkages which emerged between categories and gave attention to categories which were critical for increasing explanatory power, I searched among the stories for a story, an overarching story which would represent as much of the phenomenon as possible. The story line conceptualized the story and provided the basis for the core category or theory. The core category exhibited three critical characteristics: (a) it occurred frequently throughout the data, (b) it served to link categories together, and (c) it explained much of the variation in the data (Sherman and Webb, 1988). This core category then guided additional theoretical sampling and data collection (Strauss, 1987). In this way, the emergent theory could be tested and either strengthened or disqualified. Therefore, I restricted additional coding to that which was relevant to the core category. Strauss and Corbin (1990) emphasized the primacy of the core category by declaring, "The core category must be the sun, standing in orderly systematic relationships to its planets" (p. 124).The core category I searched for was one that would incorporate St. Joseph desegregation experiences, and the meanings attached to those experiences, into a robust explanation of the silences surrounding those experiences and the public account of desegregation. Theoretical saturation was reached when additional interviewing did not develop the properties of the core category (Glaser & Strauss, 1967).
 
 

Many researchers consider it inappropriate to apply the standards against which a quantitative study is measured to a qualitative study (Lincoln & Guba, 1985; Smith, 1984; Strauss, 1987; Strauss & Corbin, 1998). For instance, Kvale (1995) attacked the "scientific holy trinity" (p. 19) of validity, reliability, and generalization as modernistic and inappropriate for a postmodern qualitative research. As a way of retaining the concepts of rigor inherent in these three and yet, at the same time, employing standards more appropriate to qualitative research Lincoln and Guba (1985) recommended the concept of trustworthiness. If a study is to be deemed trustworthy then it must "demonstrate its truth value, provide the basis for applying it, and allow for external judgements to be made about the consistency of its procedures and the neutrality of its findings or decisions" (Erlandson et al., 1993, p. 29). These components of trustworthiness are then distributed among four quality criteria: (a) credibility, (b) transferability, (c) dependability, and (d) conformability. I adopted these criteria for my study and discuss each in more detail below.
 
 

Credibility

Credibility was suggested by Lincoln and Guba (1985) as a reasonable substitute for internal validity, the fit between the data and the subject under scrutiny. Erlandson et al., 1993 recommend the following constructs for establishing credibility: (a) prolonged engagement, (b) persistent observation, (c) triangulation, and (d) member checking. In addition to employing these four indicators for credibility, I used three external auditors to verify the accuracy and appropriateness of both my procedures and findings. Each of the constructs is elaborated below.
 
 
 
 

Prolonged engagement. I began informal and formal investigations of this topic over two and one-half years prior to the final write up. Over this length of time, I had the opportunity to visit with many people informally and to read extensively regarding this topic. By having such a long term contact with the subject, I learned to recognize many faulty points of view and correct aspects of my thinking. What now seems to me to be an unfounded concern was once quite strong: I felt many blacks would not want to speak to a white man about that which was painful and possibly shameful to them. On the contrary, I never had anyone turn me down. All participants were gracious and generous in sharing the most personal of their possessions, their stories.
 
 

Persistent observation. In the same way that the final theory emerged, my study itself had a long gestation period in which many avenues were investigated and steps were often retraced until I settled upon the research questions and design which seemed the best fit for me and for this context. Persistent observation speaks to the process of experiment and reflection, what Erlandson et al. (1993) call "constant and tentative analysis." For instance, when I found I could get little or no information from white informants who had been a part of the school district administration, I assumed they were consciously hiding information from me. Only in time did I come to realize this was a normal phenomenon for them. They had no information to give me because desegregation had not been important enough for them to have recognized and retained any part of that history. Such insight, gained over time, added immensely to my understanding of this topic.
 
 

Triangulation. In order to increase the credibility of this research I actively sought out multiple sources of data and multiple perspectives with which to interpret that data. I used official records including school board minutes, attendance records, personnel directories, state manuals, and annual reports to the school board. Documents incorporated into my research included history books, periodicals, newspapers, reunion records, and school yearbooks. Key informants and interviewees included both black and white, male and female, older and younger, professional and working class, and former teachers and administrators from the district. In addition, I used informal interviews which were never recorded to conduct a pilot study to determine the feasibility of pursuing this topic. Interview responses were also checked against survey responses to add to the credibility of each participant.
 
 

Member checking. Member checks provided that "both data and interpretation" (Erlandson et al., 1993, p. 31) were verified as accurate by those who contributed to the study. Member checking occurred near the middle and ends of my study. When categories were identified as both robust and relevant enough to contribute to the core category, key participants were contacted to verify crucial data and interpretations. All participants were appraised of the conclusions of the study and invited to critique the findings.
 
 

Transferability

In a quantitative study, the extent to which findings can be generalized to a larger population is called external validity. For qualitative studies, to embrace external validity in its traditional form would be to admit that the site of study did not contain characteristics that make it different from every other site. For the qualitative study does not seek to establish a norm for the purpose of generalizing findings but investigates phenomena which are context and time bound (Lincoln & Guba, 985). For this reason, external validity has proven unsatisfactory to many researchers (Lincoln & Guba, 1985; Strauss, 1987; Strauss & Corbin, 1990). Instead, Lincoln and Guba (1985) suggested transferability as the qualitative measure to replace external validity. Transferability differs from external validity in that, rather than attempting to identify variables which will remain unchanged across time and distance, transferability offers a rich enough description of the problem and the context to enable "observers of other contexts to make tentative judgements about applicability of certain observations for their contexts" (Erlandson et al., 1993, p. 33). Whereas, in traditional quantitative research, the responsibility lies in the lap of the researcher to demonstrate generalizability of the findings to a larger population, transferability shifts this responsibility to those who would attempt to apply the findings to another setting (Lincoln & Guba, 1985). In addition to rich disclosure of the context, the qualitative researcher can provide the reader with information about the purposive sampling that was used to select participants who would provide relevant and diverse contributions to the study. I included thick descriptions of participant interviews and also openly revealed reasons for selecting each in order to increase the transferability of this study.
 
 

Dependability

Dependability loosely corresponds to the positivist criterion of reliability, the degree to which the study could be duplicated (Lincoln & Guba, 1985). However, since a qualitative study is highly dependent upon its unique context, it is recognized that "no theory that deals with a social/psychological phenomenon can actually be reproducible" (Strauss & Corbin, 1990, p. 250) as it would not be possible to exactly duplicate the parameters of this study. This does not mean, however, that a qualitative study should not be held accountable for similar standards for consistency or stability. Dependability assumes that responsibility by addressing the stability indicated by reliability and by accounting for explainable variance. My study incorporated an audit trail that documented "critical incidents, documents, and interview notes" (Erlandson et al., 1993) in order to provide the reader with enough information with which to perform an analogous study.

Three auditors were used in my study to examine raw data, data analysis procedures and results, documentation procedures, the process of inquiry, and to attest to the dependability of the findings. One auditor, an African-American male, was chosen for his personal experience with desegregation, his perspective of being a black educator in the St. Joseph school district, and his experience in researching and speaking on issues of race. The second auditor, an African-American male older than the first auditor, helped confirm much of the earlier material of my study surrounding the school setting prior to desegregation and on into the first year. It was important to me as a white male of European origin that I be guided by the racial and cultural perspectives of the first two auditors. The third auditor, a white professional in the legal system, was chosen for his knowledge of the political and social workings of the town. In addition to his insider view of the political structures, he has proven himself as a supporter of civil rights and is deeply read on the subject of race and desegregation. His insights greatly helped me understand the workings of the official silence in St. Joseph.
 
 

Confirmability
 
 

Finally, the fourth quality criterion suggested by Lincoln and Guba (1985) was that of confirmability, or how well the findings are the result of the workings of the research process rather than researcher bias. The traditional counterpart to confirmability is that of objectivity. However, qualitative studies readily recognize that "objectivity is an illusion and that no methodology can be totally separated from those who have created and selected it" (Erlandson et al., 1993, p. 34). Instead of objectivity, confirmability seeks to demonstrate a logical connection between data and its sources and between the conceptual parts and the whole (Guba & Lincoln, 1989). The way in which grounded theory calls for a constant comparison of the data and concepts encourages the kind of objectivity measured by confirmability. Such confirmability was attested to in my study through the audit trail examined by my auditors.
 
 

Lincoln and Guba (1985) cited Dexter (1970) to define an interview as "a conversation with a purpose" (p. 258). My purpose in using the interview was, as listed among others by Guba and Lincoln, a reconstruction of past experiences. Whereas, structured interviews have as their goal to collect "precise data of a codable nature in order to explain behavior within preestablished categories" (Fontana & Frey, 1994, p. 366), unstructured interviews are used so that the "problem of interest" (Guba & Lincoln, 1985) can arise naturally in the telling. Unstructured interviews offer the advantage of protecting the participant from the researcher’s prior conceptions of categories of responses and allowing them to freely share from their own interpretations of their experiences (Fontana & Frey, 1994).The drawback to using an unstructured interview is that the lack of structure can contribute to lengthy interviews which may generate a great deal of data for analysis, data which may not be related to the topic under investigation. As a compromise between the open and inductive requirements of qualitative research and the theory driven perspectives of critical theory, I used a semi-structured interview format in which selected questions were addressed at every interview (Erlandson et al., 1993). However, those questions were made to fit the flow of the interview and were asked in a conversational manner to fit with the relationship established with the participant, "a mixture of conversation and embedded questions" (Erlandson et al., 1993, p. 86). In addition, as addressed in chapter one, a dialectical approach was adopted, one that allowed me to oscillate between the world view of the informant, (e.g., by departing from the interview schedule to pursue an interesting line of inquiry), and the insights offered by the historical and structural analysis" (Wainwright, p. 10). In this way, I could engage my subjects from the critical theory perspective of a transformative intellectual as defined by Giroux (1988) and cited by Guba and Lincoln (1994): to "uncover and excavate those forms of historical and subjugated knowledges that point to experiences of suffering, conflict, and collective struggle . . . to link the notion of historical understanding to elements of critique and hope" (p. 110). Throughout the interview process, whether the respondent was leading the conversation, responding to specific questions on my part, or reacting to new information from critical theory, it was very important to me to encourage the telling of stories.
 
 

Rationale of Storytelling/Narrative as a Method of Inquiry

My desire in using a storytelling method of inquiry was to coax both content and context from participants with a minimum of direction (Van Manen, 1990). I believe storytelling is a powerful research tool for three reasons. First, it has been my experience that in the act of telling their stories, people remember that which they thought they had forgotten and make new connections between critical elements of their accounts. Second, stories tend to suppress the self-conscious editing which can take place in a more formal question and answer structure. Therefore, not only does more information get through, but also cultural attributes are transmitted. Third, stories allow both me and the reader to become a part of their world as they tell of experiences to which we can strongly relate, recognizing that we could have had the same experiences given the same contexts (Van Manen, 1990). Of special interest to me as a researcher was the way in which interviewees acted as co-participators in the discovery of explanatory theory.
 
 

Stories birth new discoveries. In the telling of their stories, participants experience epiphanies as their minds construct the elements of the stories. For instance, Mr. Henderson experienced a flash of insight into hegemonic practices embedded deep within the black community of St. Joseph when he said the following:

Stories carry culture and invite others in. Additional data is gained that may have been lost in a more formal structure of questions and answers for participants are also less likely to edit the various elements of their narratives when caught up in the reliving of experiences which is storytelling. Whereas answers to formal questions may be controlled and the information metered out as through a sieve, stories act as containers, not only of information but of culture. They can reveal the archetypes and mores of a culture in ways that the interviewer might never think to ask and the interviewee might never think to offer. A story is not only the offering of information about a culture but is also an invitation to the researcher to share in that community in a way that is a very personal and egalitarian manner of relating. Eliciting stories from participants is similar to the gathering of oral histories and has been found to be especially effective in reaching people groups that have been traditionally overlooked (Fontana & Frey, 1994). To this end, my interviews began with broad open-ended questions which created democratic spaces for stories. I prompted participants for more information and reflection by using their own words whenever possible. In addition, I encouraged interviewees to ignore questions they did not wish to answer and to take the conversation down any lines they wished to pursue. In this way, I sought to empower them to retain ownership of their own stories and to actively reflect and direct the conversation into virgin territories. At some point in the interview I would usually share a short story of my own in order to historically locate myself in the field of study and to encourage trust. Questions were developed from a careful study of the historical context, personal experiences, and from a limited pilot study. New questions often emerged from prior interviews and from auditors who were consulted at numerous check points throughout the study.
 
 
 
 

The Role of the Researcher

Traditional positivist research defined the role of the researcher as a detached observer who plays "a neutral role . . . casual and friendly, but, on the other hand, directive and impersonal" (Fontana & Frey, 1994, p. 364, 367). In recent years claims of objectivity have been challenged by a number of researchers (Mellon, 1990; Lincoln & Guba, 1985; Strauss & Corbin, 1990). Other researchers have gone so far as to suggest that the false distance created between the researcher and the subject is elitist and contributes to false and incomplete data (LeCompte, 1993). Feminists, in particular, have attacked the philosophical underpinnings of research paradigms which support masculine practices such as detachment while vilifying typically feminine traits such as sensitivity and emotionality (Fontana & Frey, 1994). Indeed, critical theory would cause us to highly question the authenticity of a researcher/researched relationship in which the subjects are positioned in an inferior position. LeCompte suggested the role of the researcher to be that of a mediator who helps participants find their voice concerning their understandings of their lived experiences. In particular, she recommends that critical researchers attend to the power inherent in the subject’s ability to name and describe their realities. Such a stance is strongly supported by other critical theorists as they have identified the act of naming as prerequisite to action (Freire, 1993; Greene, 1988). As addressed earlier, I employed open-ended questions to draw out stories in order to allow participants to find their voice, make new discoveries, and retain power and ownership over their stories. However, the interviews were recursive in nature between a hermeneutical listening and recording and a critical theory approach of supplying needed theory and terminology in order to further empower interviewees to extend their observations.
 
 

A Recursive Mode of Interviewing

As addressed in the introduction to this study, in order to synthesize a hermeneutical approach to interviewing with a critical perspective, I freely invited participants to share their experiences and meanings attached to those experiences but also critically situated emergent categories within economic and social contexts, often visiting with them about concepts and terminology from critical theory. I found this approach to be very similar to Paulo Freire’s (1993) approach described in Pedagogy of the Oppressed. I would always begin in a posture which Freire described as that adopted by "sympathetic observers with an attitude of understanding towards what they see" (p. 91). As an ongoing process within the interview, I would then "study the inhabitants [interviewees] level of awareness of these contradictions [the experiences of desegregation]" (p. 94). And finally, I offered an initial bit of theory and/or terminology in order to facilitate the "broadening of their perception" (p. 98). If they responded to the critical interjections in such a way as to latch upon them and elaborate and extend their prior understandings and articulations, I would continue to move recursively between a listening, understanding mode and a critical theory approach of providing more grist for a raised consciousness. If, on the other hand, the interviewee displayed annoyance over the interjection or was not able to accommodate my sharing for any reason, I always backed off and resumed a purely hermeneutical approach of careful listening and recording until another opportunity was offered, if ever.
 
 

With a grounded theory methodology, data analysis is an ongoing process which informs changes in sampling, instrumentation, coding of text segments, the creation of conceptual categories, and the relationships assigned between categories. Analysis is essentially the process of making sense of the data, a very time consuming process (Chenail & Maione, 1997). As I began analysis of early data, I found there were two aspects of this sensemaking. One part of it seemed to be rather technical; for that aspect of analysis I used the AQUADÔ computer program. As stated earlier, AQUADÔ facilitated the coding of meaningful segments of text, retrieval of coded text, checking for the frequency of key words and code, and helping to establish linkages between categories for axial coding. During this protracted part of analysis, I often felt as though I was losing touch with the participants. The fragmentation which is a part of open coding never seemed to get fixed, even as the parts were assembled again into categories and then subsumed into meta-categories. I wanted another perspective, one that would place me once again closer to my subjects. I also knew that the tool I was searching for would not be a replacement for the analysis protocols of grounded theory but would be an enhancement, another angle of the same view, but one that would help me retain a strong connection with each subject’s context. What I eventually settled upon was an adaptation of an analysis technique used by therapists, researchers, and theorists called Recursive Frame Analysis (RFA).

RFA was developed by Brad Keeney (1987, 1991) and extensively refined by Chenail (1991, 1995) for his use in analyzing patient talk in therapy sessions. It has subsequently been used by qualitative researchers, practitioners, and theorists to "get closer to the talk" (Chenail, 1995, p. 1) as method of analyzing many forms of discourse. Chenail cited (Keeney, 1987) as he described how throughout the course of a conversation, participants reveal many different perspectives from which they make decisions about how they will continue in the dialogue. In order to analyze the conversation, the researcher must "name the talk" (Chenail, 1995, p. 12) by identifying the contexts, or frames, used by the speaker. Conversation is seen as recursive since a speaker will often not only speak from a repetitive set of frames but will also adjust the frames or adopt new ones as the conversation progresses (Chenail, 1995).

In RFA, the basic unit of meaningful text is called a frame. Keeney imported this term from an earlier work by Bateson (1972) who described a frame as "a psychological concept for ways of understanding meaning in human and animal interaction" (Maione, 1997). A frame corresponds quite nicely to a coded section of text under the grounded theory methodology. For instance, of all the phrases and sentences spoken by an interviewee, only some are chosen by the researcher to be especially meaningful for the research topic, others may have meaning but not relevance. An example of a meaningful frame for my study was given by Mr. Watkins when he said, "My savior was sports." When discourse is analyzed under RFA, the researcher looks for ways in which frames that seem to be related may be clumped together according to a common theme called galleries (Rambo, Heath, & Chenail, 1993; Keeney, 1987; and Maione, 1997). A gallery, of course, corresponds to a grounded theory category. As an example, I placed the frame quoted above in the gallery which I eventually called transitions because it fit with other frames that described structures and experiences which either aided or hindered black students’ transition into the integrated schools. As data is amassed, it may be necessary to look for even larger, more comprehensive groupings of concepts which Maione (1997) termed wings. RFA, therefore, gives the qualitative researcher the ability to locate the galleries and wings which are often frequented by the subjects. In addition, RFA sensitizes the researcher to be able to identify contiguous galleries, those which are contextually connected and boast of shared traffic back and forth. In this way concepts which might not seem to the researcher to be connected are more easily identified.

So, why do RFA if you are already doing grounded theory? What RFA offers the researcher is a dynamic component in the analysis of categories and how they relate together to contribute to a core category or theory. Whereas the grounded theory use of categories and linkages between categories is static, one can use RFA in order to visualize how an interviewee moved among galleries during the course of an interview. For example, it might be interesting to note not only that the subject purposely moved back into a certain gallery many times during an interview but also how that was accomplished. Was there a certain hallway or entryway that was used to return each time to a favored gallery? As an example, Mr. Watkins moved back into the sports section of the transitions gallery nine times during the course of his interview. Every trip back to the gallery was of his own initiative, not as the result of a leading question. Once the researcher begins to view interview transcriptions from this dynamic perspective, other aspects of the discourse may become apparent. For instance, Chenail (1995) wrote of torqued talk in which a subject cannot seem to be able to leave a gallery. Although RFA was not used as a primary analysis tool, it was a very good fit with my style of learning and analysis and therefore, facilitated my identification of relevant galleries or categories and the relationships between each. Overall data analysis was conducted according to the traditional guidelines established within grounded theory.
 
 

All interview tapes and handwritten notes were word processed into separate documents and each line was numbered for reference. The text was then analyzed and broken down into meaningful segments, or frames, with each segment containing one item of information. The AQUADÔ software package was used to number lines of text and to separate meaningful text into separate frames. In addition, AQUADÔ facilitated the tagging of each frame with an identifier that related back to the originating document. The frames were then analyzed in order to develop categories, or galleries. Each gallery was given a description and a name which fit the description. Frames not finding a home within a gallery were retained and referred to periodically for their potential of creating unforeseen galleries or connections between galleries. As is common to grounded theory, frames were moved and galleries created and demolished as a result of a constant comparison of frames and galleries according to the emerging understanding and theory. Theoretical saturation was reached when no new or relevant data contributed to new frames, galleries, or wings.
 
 

The reason we gather data in a qualitative study is "to gain the ability to construct reality in ways that are consistent and compatible with the constructions of a setting’s inhabitants" (Erlandson et al, 1993). In addition, I wanted to insure that the constructed reality was consistent with archival sources of data. Triangulation of data from multiple sources generated a tremendous burden of paperwork. I realized early on that good methods of documenting and sorting data were essential.
 
 

Document Summaries After I examined each document or record I created a document summary which noted various details related to that piece of research. First, I noted the type of document, whether it was a photograph, a newspaper article, a letter from a library holding, an interviewee tape, or one of other kinds of documents examined in the course of this research. Second, I noted its bibliographic information: the source, author, date, location, etc. Third, I recorded a brief summary of its pertinent contents so the summary could be quickly consulted in order to aid in making connections with other documents. Fourth, I jotted down any ideas I had of how other documents might be related. This summary led me to the more detailed summary of the results gained from the interviews.
 
 

Field Note Summaries

Following each interview, I removed myself as quickly as possible from the interview site and wrote a contact summary sheet in which I summarized what I had observed and learned from the interview. Included was a brief description of the context; an observation of the interviewee; that which seemed most significant about the interview; and that which was the most unexpected, interesting, or problematic about the interview. Both the document summaries and field note summaries contributed to an ongoing reflective analysis of data.
 
 

Audit Trail

The document summaries, field note summaries, and miscellaneous notes contributed to an overall audit trail of the entire research process. In addition, process notes were included about data collection and analysis instruments which were considered and then either added or dropped from the study. I chose to integrate any journaling about the process into the flow of the audit trail so my questions and comments would be in proximity with procedures and actions. In this way, I was able to constantly question the questions and provide for adjustment as needed.
 
 

Erlandson, Harris, Skipper, and Allen (1993) cited Denzin (1970) and wrote that the credibility of a study is increased by triangulation that occurs in several dimensions: (a) multiple sources of data, (b) multiple methods, (c) multiple investigators, and (d) multiple perspectives of the analysis. I sought to increase credibility by addressing each of these kinds of triangulation. Multiple sources of data came from the study of historical records and documents, the results of an informal pilot study, a quantitative survey instrument, purposeful sampling of interviewees, and input from key informants and auditors. I employed multiple methods by following grounded theory as my overarching methodology and also including a recursive form of interviewing which allowed me to share theory with those I interviewed. In addition, a survey was used to gather a profile of the interviewees. Although I was the only formal investigator, I consider all those who participated in this study to be investigators in their own rights as they sought to recall past experiences and meanings and to situate those within the historical, social, and economic contexts from 1953 to 1959. Multiple perspectives were used as I actively followed the perspective of critical theory in examining power relations and critiqued social and economic contributions to the topic of my study and incorporated a value of caring and attention to emotions from critical feminist theory. In addition, feminist theory alerted me to issues of silencing and voice as they pertained to desegregation in St. Joseph. In order to track data gathered from these different sources, formal methods of recording were employed, beginning with a chain of evidence.
 
 

Chain of Evidence
 
 

After categories were identified, related, and tested, I constructed a chain of evidence used to make strong links between my research questions, raw data, and initial conclusions from data analysis. The chain of evidence provided connections between the data and the findings in order that an external observer could follow the flow of the research and test the robustness of the conclusions.
 
 

Member Checking

At various check points throughout the study I involved key informants in a review of the process and results of the research. Key informants were used to determine the feasibility of the study, to review survey and interview items, to critically analyze initial conclusions, and to test the final emergent theory.
 
 

Searching for Alternative Explanations of the Data

Key informants were also used in order to search for alternate explanations of the data. This was done by avoiding leading discussions at the various checkpoints described above. Instead of sharing my tentative conclusions, I would instead share data and ask for their ideas on what it meant. In addition, ongoing literature review related to new data consistently exposed me to alternative explanations which had to be measured against the historical background of St. Joseph and the data generated by the interviews. The writing of the report began long before the theory emerged and was constantly adjusted according to new and relevant data.
 
 

The major blocks of the study were written in chronological order and represent the way in which the research unfolded. For instance, as I considered this thesis I was drawn back in my memories to the episodes that I related in the preface. Next, I began to wrestle with the conceptualization of the study as represented in the introduction chapter. A review of the historical literature informed the third chapter and so on until my theory emerged and could be addressed in the final chapter. It was not until the theory emerged, however, that I felt a sense of coherence to the experience and could go back to fill in gaps and address inconsistencies. In the next chapter I write of the results I found from interviewing former white administrators, former black teachers and students. It was in the organizing of the writing of the next chapter that I began to feel comfortable that a substantive theory had, indeed, emerged using the grounded theory methodology.