Thomas G. Cummings University of Southern California Christopher G. Worley University of Southern California Pepperdine University INVESTICn DO ROZVOJE VZDĚLÁVÁNI ?\ SOUTH-WESTERN t » CENGAGE Learning- Australia ■ Brazil • Canada • Mexico • Singapore • Spain ■ UnitedKingdom ■ UnitedStates 4240769147 Collecting and Analyzing Diagnostic Information Organization development is vitally dependent on organization diagnosis: the process of collecting information that will be shared with the client in jointly assessing how the organization is functioning and determining the best change intervention. The quality of the information gathered, therefore, is a critical part of the OD process. In this chapter, we discuss several key issues associated with collecting and analyzing diagnostic data on how an organization or department functions. Data collection involves gathering information on specific organizational features, such as the inputs, design components, and outputs presented in Chapters 5 and 6. The process begins by establishing an effective relationship between the OD practitioner and those from whom data will be collected and then choosing data collection techniques. Four methods can be used to collect data: questionnaires, interviews, observations, and unobtrusive measures. Data analysis organizes and examines the information to make clear the underlying causes of an organizational problem or to identify areas for future development. The next step in the cyclical OD process is the feedback of data to the client system, an important process described in Chapter 8. The overall process of data collection, analysis, and feedback is shown in Figure 7.1. THE DIAGNOSTIC RELATIONSHIP In most cases of planned change, OD practitioners play an active role in gathering data from organization members for diagnostic purposes. For example, they might interview members of a work team about causes of conflict among members; they might survey employees at a large industrial plain about factors contributing to poor product quality. Before collecting diagnostic information, practitioners need to establish a relationship with those who will provide and subsequently use it. Because the nature of that relationship affects the quality and usefulness of the data collected, it is vital that OD practitioners clarify for organization members who they are, why the data are being collected, what the data gathering will involve, and how the data will be used.1 That information can help allay people's natural fears that the data might be used against them and gain members' participation and support, which are essential to developing successful interventions, Establishing the diagnostic relationship between the consultant and relevant organization members is similar to forming a contract. It is meant to clarify expectations and to specify the conditions of the relationship. In those cases where members have been directly involved in the entering and contracting process described in Chapter 4, the diagnostic contract will typically be part of the initial contracting step. In situations when1 data will be collected from members who have not been directly involved in entering and contracting, however, OD practitioners will need to establish a diagnostic 122 PART 2 The Process of Organization Development [Figure 7.1]- The Data-Collection and Feedback Cycle Core Activities Planning to Collect Data J 7\ Collecting Data Li Analyzing Data Feeding I— Back Data r~jv Following Up SOURCE: FEEDBACK AND ORGANIZATON DEVELOPMENT by Nadler, D. © 1977. Reprinted by permission of Pearson Education, Inc., Upper Saddle River, NJ. contract as a prelude to diagnosis. The answers to the following questions provide the substance of the diagnostic contract:" 1. Who am I? The answer to this question introduces the OD practitioner to the organization, particularly to those members who do not know the consultant and yet will be asked to provide diagnostic data. 2. Why am I here, and what am I doing? These answers are aimed at defining the goals of the diagnosis and data-gathering activities. The consultant needs to present the objectives of the action research process and to describe how the diagnostic activities fit into the overall developmental strategy. 3. Who do I work for? This answer clarifies who has hired the consultant, whether il be a manager, a group of managers, or a group of employees and managers. One way to build trust and support for the diagnosis is to have those people directly involved in establishing the diagnostic contract. Thus, for example, if the consultant works for a joint labor-management committee, representatives from both sides of that group could help the consultant build the proper relationship with those from whom data will be gathered. 4. What do I want from you, and why? Here, the consultant needs to specify how much time and effort people will need to give to provide valid data and subsequently to work with these data in solving problems. Because some people may not want to participate in the diagnosis, it is important to specify that such involvement is voluntary. 5. How will 1 protect your confidentiality? This answer addresses member concerns about who will see their responses and in what form. This is especially critical when employees are asked to provide information about their attitudes or perceptions. OD practitioners can either ensure confidentiality or slate that full participation in the change process requires open information sharing. In the first case, employees are frequently concerned about privacy and the possibility of being punished for their responses. To alleviate concern and to increase the likelihood of obtaining honest responses, the consultant may need to assure employees of the confidentiality of their information, perhaps through explicit guarantees ol response anonymity. In the second case, full involvement of the participants in their own diagnosis may be a vital ingredient of the change process. If sensitive issues arise, assurances ol confidentiality can co-opt the OD practitioner and thwart meaningful diagnosis. The consultant is bound to keep confidential the issues that are most critical for the group or organization to understand.1 OD practitioners must think carefully about how they want to handle confidentiality issues. CHAPTER 7 Collecting and Analyzing Diagnostic Information 123 6. Who will have access to the data? Respondents typically want lo know whether they will have access to their data and who else in the organization will have similar access. The OD practitioner needs to clarify access issues and, in most cases, should agree to provide respondents with their own results. Indeed, the collaborative nature of diagnosis means that organization members will work with their own data to discover causes of problems and to devise relevant interventions. 7. What's in it for you? This answer is aimed at providing organization members with a clear delineation of the benefits they can expect from the diagnosis. This usually entails describing the feedback process and how they can use the data to improve the organization. 8. Can I be trusted? The diagnostic relationship ultimately rests on the trust established between the consultant and those providing the data. An open and honest exchange of information depends on such trust, and the practitioner should provide ample time and face-to-face contact during the contracting process to build this trust. This requires the consultant to listen actively and discuss openly all questions raised by participants. Careful attention lo establishing the diagnostic relationship helps to promote the three goals of data collection.4 The first and most immediate objective is to obtain valid information about organizational functioning. Building a data collection contract can ensure that organization members provide honest, reliable, and complete information. Data collection also can rally energy lor constructive organizational change. A good diagnostic relationship helps organization members start thinking about issues that concern them, and it creates expectations that change is possible. When members trust the consultant, they are likely to participate in the diagnostic process and to generate energy and commitment for organizational change. Finally, data collection helps to develop the collaborative relationship necessary for effecting organizational change. The diagnostic stage of action research is probably the first time that most organization members meet the OD practitioner, and it can be the basis for building a longer-term relationship. The data collection contract and subsequent data-gathering and feedback activities provide members with opportunities for seeing the consultant in action and for knowing her or him personally. II the consultant can show employees that he or she is trustworthy, is willing to work with them, and is able to help improve the organization, then the data collection process will contribute to the longer-term collaborative relationship so necessary for carrying out organizational changes. METHODS FOR COLLECTING DATA The four major techniques for gathering diagnostic data are questionnaires, interviews, observations, and unobtrusive measures. Table 7,1 briefly compares the methods and lists their major advantages and problems. No single method can fully measure the kinds of variables important to OD because each has certain strengths and weaknesses.^ For example, perceptual measures, such as questionnaires and surveys, are open lo self-report biases, such as respondents' tendency to give socially desirable answers rather than honest opinions. Observations, on the other hand, are susceptible to observer biases, such as seeing what one wants to see rather than what is really there. Because of the biases inherent in any data collection method, more than one method should be used when collecting diagnostic data. If data from the different methods are compared and lound lo be consistent, it is likely that the variables are being measured validly. For example, questionnaire measures of job discretion could be supplemented with observations of the number and kinds of decisions employees are making. If the two kinds of data support each other, job discretion is probably being assessed accurately. If the two 124 PART 2 The Process of Organization Development [Table 7.1]- A Comparison of Different Methods of Data Collection METHOD MAJOR ADVANTAGES MAJOR POTENTIAL PROBLEMS Questionnaires 1. Responses can be quantified and 1. Nonempathy easily summarized 2. Predetermined questions/missing 2. Easy to use with large samples issues 3. Relatively inexpensive 3. Overinterpretation of data 4. Can obtain large volume of data 4. Response bias Interviews 1. Adaptive—allows data collection 1. Expense on a range of possible subjects 2. Bias in interviewer responses 2. Source of "rich" data 3. Coding and interpretation 3. Empathic difficulties 4. Process of interviewing can build 4. Self-report bias rapport Observations Unobtrusive measures Collects data on behavior, rather than reports of behavior Real time, not retrospective Adaptive 1. Coding and interpretation difficulties 2. Sampling inconsistencies 3. Observer bias and questionable reliability 4. Expense 1. Nonreactive—no response bias 2. High face validity 3. Easily quantified 1. Access and retrieval difficulties 2. Validity concerns 3. Coding and interpretation difficulties SOURCE: FEEDBACK AND ORGAN IZATON DEVELOPMENT by Nadler, D. © 1977. Reprinted by permission of Pearson Education, Inc., Upper Saddle River, NJ. kinds of data conflict, the validity of the measures should be examined further-—perhaps by using a third method, such as interviews. Questionnaires One ol the most efficient ways to collect data is through questionnaires. Because they typically contain fixed-response queries about various features of an organization, these paper-and-pencil measures can be administered to large numbers ol people simultaneously. Also, they can be analyzed quickly, especially with the use of computers, thus permitting quantitative comparison and evaluation. As a result, data can easily be led back to employees. Numerous basic resource books on survey methodology and questionnaire development are available.'' Questionnaires can vary in scope, some measuring selected aspects of organizations and others assessing more comprehensive organizational characteristics. They also can vary in the extent to which they are either standardized or tailored to a specific organization. Standardized instruments generally are based on an explicit model of organization, group, or individual effectiveness and contain a predetermined set of questions that have been developed and refined over time, For example. Table 7.2 presents a standardized questionnaire for measuring the job design dimensions identified in Chapter 6: skill variety, task identity, task significance, autonomy, and feedback about results. The questionnaire includes three items or questions lor each CHAPTER 7 Collecting and Analyzing Diagnostic Information 125 Job Design Questionnaire Here are some statements about your job. How much do you agree or disagree with each? -[Table 7.2] MY JOB: STRONGLY SLIGHTLY SLIGHTLY STRONGLY DISAGREE DISAGREE DISAGREE UNDECIDED AGREE AGREE AGREE 1. provides much variety..... [1] [2] [3] [4] [5] [6] [71 2. permits me to be left on my own to do my own work.......... !-l [2] [3] [4] i'=: Iii [7, 3. is arranged so that I often have the opportunity to see jobs or projects through to completion......... [1] [2] [3] 14] [5] [6] [7] 4. provides feedback on how well I am doing as I am working . , , [1] [2] [3] [4] [5] (6. [7] 5. is relatively significant in our organization............ :ii [2] [3] [4] [51 [6] [7] 6. gives me considerable opportunity for independence and freedom in how I do my work............. tu [2] [3] [4] [5] [6] [7] 7. gives me the opportunity to do a number of different things...... n; !2l [3] [4] [5] [6] [71 8. provides me an opportunity to find out how well I am doing......... in [/; [3] [4] [5] [6] [7] \ 9. is very significant or important in the broader scheme of things . . . [1] [2] :3i [4] [5] K>; [7] 10. provides an opportunity for independent thought and action . . . tu [2] Pi [4] [51 [6] [/: 11. provides me with a great deal of variety at work.............. :u [21 [3] [4] 15. [6] \n 12. is arranged so that 1 have the opportunity to complete the work 1 start...................... m [2] [3] [4] [5] [6] [/] 13. provides me with the feeling that I know whether 1 am performing well or poorly..... [1] [2] [3] [4] ''a [6] [7] 14. is arranged so that I have the chance to do a job from the beginning to the end (i.e., a chance to do the whole job).................... in [2] [3] [41 ibl [6] [7] 15. is one where a lot of other people can be affected by how well the work gets done............... in [2] [3] 'a: [5] 6] [7] Scoring: Skill variety ................... Task identity ................ Task significance .......... Autonomy .................... Feedback about results . questions 1,7,11 questions 3, 12, 14 . questions 5,9, 15 . questions 2, 6, 10 . questions 4, 8, 13 SOURCE: Reproduced by permission of E. Lawler, S. Mohrman, and T. Cummings, Center for Effective Organizations, University of Southern California. 126 PART 2 The Process of Organization Development dimension, and a total score for each job dimension is computed simply by adding the responses for the three relevant items and arriving at a total score from 3 (low) to 21 (high). The questionnaire has wide applicability. Ii has been used in a variety ol organizations with employees in both blue-collar and white-collar jobs. Several research organizations have been highly instrumental in developing and refining surveys. The Institute for Social Research at the University ol Michigan (http:// vvvvvv.isr.iunich.edu) and the Center for Effective Organizations al the University of Southern California (http://ceo-niarshall.usc.edu) are two prominent examples. Two of the Institute's most popular measures of organizational dimensions are the Survey of Organizations and the Michigan Organizational Assessment Questionnaire. Few other instruments are supported by such substantial reliability and validity data.7 Oilier examples of packaged instruments include Weisbord's Organizational Diagnostic Questionnaire, Dyer's Team Development Survey, Cameron and Qtiinn's Organizational Culture Assessment Instrument, and Hackman and Oldham's Job Diagnostic Survey* In fact, so many questionnaires are available that rarely would an organization have to create a totally new one. However, because every organization has unique problems and special jargon for referring to them, almost any standardized instrument will need to have orga nidation -specilic additions, modifications, or omissions. On the other hand, customized questionnaires are tailored to the needs of a particular client. Typically, they include questions composed by consultants or organization members, receive limited use, and do not undergo longer-term development. They can be combined with standardized instruments to provide valid and reliable data focused toward the particular issues lacing an organization. Questionnaires, however, have a number of drawbacks that need to lie taken into account in choosing whether lo employ them lor data collection. First, responses are limited to the questions asked in the instrument. They provide little opportunity to probe for additional data or to ask for points of clarification. Second, questionnaires tend to be impersonal, and employees may not be willing to provide honest answers. Third, questionnaires often elicit response biases, such as (he tendency to answer questions in a socially acceptable manner. This makes il difficult to draw valid conclusions from employees' sell-reports. Interviews A second important measurement technique is the individual or group interview. Interviews are probably the mosl widely used technique lor collecting data in OD. They permit the interviewer lo ask the respondent direct questions. Further probing and clarification is, therefore, possible as the interview proceeds. This flexibility is invaluable lor gaining private views and feelings about the organization and for exploring new issues that emerge during the interview. Interviews may be highly structured—resembling questionnaires—or highly unstructured—starling with general questions that allow the respondent to lead the way. Structured interviews typically derive from a conceptual model of organization funclion-ing; the model guides the types of questions that are asked. For example, a structured interview based on the organization-level design components identified in Chapter 5 would ask managers specilic questions about technology, strategy, organization structure, measurement systems, human resources systems, and organization culture. Unstructured interviews are more general and include the following broad questions about organizational functioning: What are the major goals or objectives of the organization or department? How does the organization currently perform with respect to these purposes? What ate the strengths and weaknesses of the organization or department? What barriers stand in the way of good performance? CHAPTER 7 Collecting and Analyzing Diagnostic Information 127 Although interviewing typically involves one-to-one interaction between an OD practitioner and an employee, it can be carried out in a group context. Group interviews save time and allow people to build on others' responses. A major drawback, however, is that group sellings may inhibit some people from responding freely, A popular type of group interview is the focus group or sensing meeting" These are unstructured meetings conducted by a manager or a consultant. A small group of 10 to 15 employees is selected to represent a cross section ol lunciional areas and hierarchical levels or a homogeneous grouping, such as minorities or engineers. Group discussion is frequently started by asking general questions about organizational features and functioning, an intervention's progress, or current performance. Group members are then encouraged in discuss their answers more fully. Consequently, focus groups and sensing meetings are an economical way to obtain interview data and are especially effective in understanding particular issues in greater depth. The richness and validity of the information gathered will depend on the extent to which the manager or the consultant develops a trust relationship with the group and listens to member opinions. Another popular unstructured group interview involves assessing the current state of an intact work group. The manager or the consultant generally directs a question to the group, calling its attention to some part of group functioning. For example, group members may be asked how they feel the group is progressing on its stated task. The group might respond and then come up with its own series ol questions about barriers to task performance. This unstructured interview is a fast, simple way to collect data about group behavior. It allows members to discuss issues of immediate concern and to engage actively in the questioning and answering process. This technique is limited, however, to relatively small groups and to settings where there is trust among employees and managers and a commitment to assessing group processes. Interviews are an effective method lor collecting data in OD. They are adaptive, allowing the interviewer to modify questions and to probe emergent issues during the interview process. They also permit the interviewer to develop an cmpalhetic relationship with employees, frequently resulting in frank disclosure of pertinent information. A major drawback ol interviews is the amount of lime required to conduct and analyze them. Interviews can consume a great deal ni time, especially if interviewers take full advantage ol the opportunity to hear respondents out and change their questions accordingly. Personal biases also can distort the data. Like questionnaires, interviews are subject to the self-report biases of respondents and, perhaps more important, to the biases of the interviewer. For example, the nature of the questions and the interactions between the interviewer and the respondent may discourage or encourage certain kinds of responses. These problems suggest that interviewing takes considerable skill to gather valid data. Interviewers must be able to understand their own biases, to listen and establish empathy with respondents, and to change questions to pursue issues that develop during the course of the interview. Observations One of the more direct ways of collecting data is simply to observe organizational behaviors in their functional sellings. The OD practitioner may do this by walking casually through a work area and looking around or by simply counting the occurrences of specific kinds of behaviors (lor example, the number of limes a phone call is answered alter three rings in a service department). Observation can range from complete participant observation, in which the OD practitioner becomes a member of the group under study, to more detached observation, in which the observer is clearly not part of the group or situation itself and may use film, videotape, and other methods to record behaviors. Observations have a number ol advantages. They are free of the biases inherent in self-report daia. They put the practitioner directly in touch with the behaviors in question, without having to rely on others' perceptions. Observations also involve PART 2 The Process of Organization Development real-time data, describing behavior occurring in the present rather than the past. This avoids the distortions that invariably arise when people are asked to recollect their behaviors. Finally, observations are adaptive in that the consultant can modify what he or she chooses to observe, depending on the circumstances. Among the problems with observations are difficulties interpreting the meaning underlying the observations. Practitioners may need to devise a coding scheme to make sense out of observations, and this can be expensive, take time, and introduce biases into the data. Because the observer is the data collection instrument, personal bias and subjectivity can distort the data unless the observer is trained and skilled in knowing what lo look for; how, where, and when to observe; and how to record data systematically. Another problem concerns sampling: Observers not only must decide which people to observe, they also must choose the lime periods, territory, and events in which to make those observations. Failure lo attend to these sampling issues can result in highly biased samples of observational data. When used correctly, observations provide insightful data about organization and group functioning, intervention success, and performance, For example, observations are particularly helpful in diagnosing the interpersonal relations of members of work groups. As discussed in Chapter 6, interpersonal relationships are a key component of work groups; observing member interactions in a group setting can provide direct information about the nature of those relationships. Unobtrusive Measures Unobtrusive data are not collected directly from respondents but from secondary sources, such as company records and archives. These data are generally available in organizations and include records of absenteeism or tardiness; grievances; quantity and quality of production or service; financial performance; meeting minutes; and correspondence with key customers, suppliers, or governmental agencies. Unobtrusive measures are especially helpful in diagnosing the organization, group, and individual outputs presented in Chapters 5 and 6. At the organization level, lor example, market share and return on investment usually can be obtained from company reports. Similarly, organizations typically measure the quantity and quality of the outputs of work groups and individual employees. Unobtrusive measures also can help to diagnose organization-level design components—structure, work systems, control systems, and human resources systems. A company's organization chart, lor example, can provide useful information about organization structure. Information about control systems usually can be obtained by examining the firm's management information system, operating procedures, and accounting practices. Data about human resources systems often are included in a company's personnel manual. Unobtrusive measures provide a relatively objective view of organizational functioning. They are free from respondent and consultant biases and ate perceived as being "real" by many organization members. Moreover, unobtrusive measures tend to be quantified and reported at periodic intervals, permitting statistical analysis of behaviors occurring over time. Examining monthly absenteeism rates, for example, might reveal trends in employee withdrawal behavior. The major problems with unobtrusive measures occur in collecting such information and drawing valid conclusions from it. Company records may not include data in a form that is usable by the consultant. II, lor example, individual performance data are needed, the consultant may find that many firms only record production information at the group or departmental level. Unobtrusive data also may have their own built-in biases. Changes in accounting procedures and in methods of recording data are common in organizations, and such changes can affect company records independently of what is actually happening in the organization. For example, observed changes in productivity over time might be caused by modifications CHAPTER 7 Collecting and Analyzing Diagnostic Information in methods of recording production rather than hy actual changes in organizational functioning. Despite these drawbacks, unobtrusive data serve as a valuable adjunct to other diagnostic measures, such as interviews and questionnaires. Archival data can be used in preliminary diagnosis, identifying those organizational units with absenteeism, grievance, or production problems. Then, interviews might be conducted or observations made in those units to discover the underlying causes of the problems. Conversely, Unobtrusive data can be used to cross-check other forms of information. For example, if questionnaires reveal thai employees in a department are dissatisfied with their jobs, company records might show whether that discontent is manifested in heightened withdrawal behaviors, in lowered quality work, or in similar counterproductive behaviors. SAMPLING Before discussing how to analyze data, the issue of sampling needs to be emphasized. Application of the different data collection techniques invariably raises the following questions: "How many people should be interviewed and who should they be?" "What events should be observed and how many?" "How many records should be inspected and which ones?"10 Sampling is not an issue in many OD cases. Because OD practitioners collect interview or questionnaire data from all members Of the organization or department in question, ihey do not have to worry about whether the information is representative of the organization or unit. Sampling becomes an issue in OD, however, when data are collected from selected members, behaviors, or records. This is often the case when diagnosing organization-level issues or large systems. In these cases, it may be important to ensure that the sample of people, behaviors, or records adequately represents the characteristics of the total population. For example, a sample of 50 employees might be tised to assess the perceptions of all 300 members of a department. A sample of production data might be used to evaluate the total production of a work group. OD practitioners often find that it is more economical and quicker to gather a sampling of diagnostic data than to collect all possible information. If done correctly, the sample can provide useful and valid information about the entire organization or unit. Sampling design involves considerable technical detail, and consultants may need to become familiar with basic references in this area or to obtain professional help." The first issue to address is sample size, or how many people, events, or records arc needed in cany out the diagnosis or evaluation. This question has no simple answer: The necessary sample size is a function of population size, the confidence desired in the quality of the data, and the resources (money and time) available for data collection. First, the larger the population (lor example, the number of organization members or total number of work outcomes) or the more complex the client system (for example, the number of salary levels that must be sampled or the number of different functions), the more difficult it is to establish a "right" sample size. As the population increases in size and complexity, the less meaning one can attach to simple measures, such as an overall average score on a questionnaire item. Because the population comprises such different types of people or events, more data are needed to ensure an accurate representation of the potentially different subgroups. Second, the larger the proportion of the population that is selected, the more confidence one can have about I fie quality of the sample, il the diagnosis concerns an issue of great importance to the organization, then extreme confidence may be needed, indicative of a very large sample size. Third, limited resources constrain sample size. If resources are limited but the required confidence is high, then questionnaires will be preferred over interviews because more information can be collected per member per dollar. 130 PART 2 The Process of Organization Development Tht: second issue to address is sample selection. Probably the most common approach to sampling diagnostic data in C)D is a simple random sample, in which each member, behavior, or record has an equal chance of being selected. For example, assume that an OD practitioner would like to select 50 people randomly out of the 300 employees at a manufacturing plant. Using a complete list of all 300 employees, the consultant can generate a random sample in one of two ways. The first method is to use a random number table printed in the back of almost any statistics text; the consultant would pick out the employees corresponding to the first 50 numbers under 300 beginning anywhere in the table. The second method is to pick every sixth name (300/50 = 6) starting anywhere in the list. If the population is complex, or many subgroups need to be represented in the sample, a stratified sample may be more appropriate than a random one. In a stratified sample, the population ol members, events, or records is segregated into a number of mutually exclusive subpopulations and a random sample is taken from each sub-population. For example, members of an organization might be divided into three groups (managers, while-collar workers, and blue-collar workers), and a random sample of members, behaviors, or records could lie selected from each grouping to reach diagnostic conclusions about each of the groups. Adequate sampling is critical to gathering valid diagnostic data, and the OD literature has paid little attention to this issue. OD practitioners should gain rudimentary knowledge in this area and use professional help if necessary. TECHNIQUES FOR ANALYZING DATA Data analysis techniques fall into two broad classes: qualitative and quantitative. Qualitative techniques generally are easier to use because they do not rely on numerical data. That fact also makes them more open to subjective biases but also easier to understand and interpret. Quantitative techniques, on the other hand, can provide more accurate readings of the organizational problem. Qualitative Tools Of the several methods for summarizing diagnostic data in qualitative terms, two of the most important are content analysis and lorce-field analysis. Content Analysis A popular technique for assessing qualitative data, especially interview data, is content analysis, which attempts to summarize comments into meaningful categories. When done well, a content analysis can reduce hundreds ol interview comments into a few themes that effectively summarize the issues or altitudes of a group of respondents. The process of content analysis can be quite formal, and specialized references describe this technique in detail.1' In general, however, the process can be broken down into three major steps. First, responses to a particular question are read to gain familiarity with the range of comments made and to determine whether some answers are occurring over and over again. Second, based on this sampling ol comments, themes are generated that capture recurring comments. Themes consolidate different responses that say essentially the same thing. For example, in answering the question "What do you like most about your job?" different respondents might list their coworkers, their supervisors, the new machinery, and a good supply of tools. The first two answers concern the social aspects of work, and the second two address the resources available lor doing the work. Third, the respondents' answers to a question are then placed into one of the categories. The categories with the most responses represent those themes that are most often mentioned. Force-Field Analysis A second method for analyzing qualitative data in OD derives Irom Kurt Lewin's three-step model of change. Called force-field analysis, this method organizes