In the field of qualitative research, a very famous method is the "grounded theory" proposed by Glaser and Strauss (1967) . Grounded theory is a qualitative research method whose main purpose is to establish theory based on empirical data (Strauss, 1987: 5). Researchers generally have no theoretical assumptions before starting research. They start directly from actual observations, summarize empirical generalizations from original data, and then move up to theory. This is a method of building substantive theory from the bottom up, that is, searching for core concepts that reflect social phenomena on the basis of systematic collection of data, and then constructing relevant social theories through the connections between these concepts. Grounded theory must be supported by empirical evidence, but its main feature is not its experiential nature, but that it abstracts new concepts and ideas from empirical facts. In terms of philosophical thought, the grounded theory method is based on the paradigm of post-positivism, which emphasizes the falsification of constructed theories.

The method of grounded theory originated from a field observation of medical staff handling dying patients in a hospital in the 1960s by Glass and Strauss (1965, 1968). The formation of this aspect is related to two aspects of theoretical thought, which come from philosophy and sociology respectively: First, American pragmatism, especially the thoughts of Dewey, G. Mead and Peirce, who emphasize the importance of action and focus on the importance of action. The situation of the problem is dealt with and methods are generated in problem solving; another influence comes from the Chicago School of Sociology, which extensively uses field observation and in-depth interview methods to collect data, emphasizing understanding social interaction, social processes and social changes from the perspective of actors. .

1. The basic idea of ??grounded theory

The basic idea of ??grounded theory mainly includes the following aspects.

1. Generating theories from data

Grounded theory places special emphasis on improving theories from data, believing that only through in-depth analysis of data can a theoretical framework be gradually formed. This is an inductive process, condensing the data from bottom to top. Different from general grand theories, grounded theory does not conduct logical deductions from the researcher's own pre-set hypotheses, but instead conducts inductive analysis based on the data. The theory must be traceable to the original data from which it was generated and must be based on empirical facts. This is because grounded theorists believe that only theories generated from data are viable. If the theory is consistent with the data, the theory has practical uses and can be used to guide people's specific life practices.

The first task of grounded theory is to establish a substantive theory between grand theory and micro operational hypothesis (that is, a theory applicable to a specific time and space), but it does not exclude the establishment of a formal theory with universal applicability. construction. However, the formal theory must be based on the substantive theory. Only after the substantive theory is established on the basis of data, the formal theory can be established on various related substantive theories. This is because grounded theory believes that knowledge is accumulated and is a process of continuous evolution from facts to substantive theory, and then to formal theory. Constructing formal theory requires a large number of data sources and the intermediary of substantive theory. If a formal theory is constructed directly from a data source, the jump is too large and many loopholes may arise. In addition, formal theory does not have to have a single constituent form, but can cover many different substantive theories, integrating, condensing, and generating many different concepts and perspectives into a whole. This intensive formal theory has richer connotations than those of single formal theory and can provide meaningful explanations for a wider range of phenomena.

2. Be sensitive to theory

Since the main purpose of grounded theory is to construct theory, it particularly emphasizes that researchers should remain highly sensitive to theory. Whether in the design stage or when collecting and analyzing data, researchers should remain sensitive to their own existing theories, previous theories, and the theories presented in the data, and pay attention to capturing clues to new theories.

Maintaining theoretical sensitivity can not only help us have a certain focus and direction when collecting data, but also pay attention to looking for concepts that can express the content of the data in a more concentrated and condensed way when analyzing the data, especially when the data content itself is relatively loose.

Usually, qualitative researchers are better at conducting detailed descriptive analysis of the phenomena under study, but are not particularly sensitive to or particularly interested in theoretical construction. Grounded theory, out of its own special concern, believes that theory has stronger explanatory power than pure description, so it emphasizes that researchers should be sensitive to theory.

3. The method of constant comparison

The main analytical idea of ??grounded theory is comparison. Constant comparison between data and data, theory and theory, and then based on the data and theory The correlation between them extracts the relevant categories and their attributes. There are usually four steps in comparison: 1) Compare data according to concept categories: After coding the data and classifying the data under as many concept categories as possible, compare the coded data under the same and different concept categories. Compare them and find attributes for each concept category. 2) Integrate relevant conceptual categories and their attributes, compare these conceptual categories, consider the relationships between them, and connect these relationships in some way. 3) Outline the initially presented theory, determine the connotation and extension of the theory, return the preliminary theory to the original data for verification, and at the same time continuously optimize the existing theory to make it more refined. 4) State the theory, and describe the acquired data, conceptual categories, characteristics of the categories, and the relationships between the conceptual categories layer by layer as answers to the research questions.

4. Theoretical sampling method

When analyzing data, researchers can use the theory initially generated from the data as the standard for the next step of data sampling. These theories can guide the next step of data collection and analysis, such as selecting data, coding, and establishing coding and filing systems. Each theory currently presented has a guiding role for researchers and can limit where and how researchers should go next. Therefore, data analysis should not just stop at mechanical language coding, but should carry out theoretical coding. Researchers should continually develop hypotheses about the content of the data, generate theories through round-robin comparisons between data and hypotheses, and then use these theories to code the data.

5. Use literature flexibly

Using relevant literature can broaden our horizons and provide new concepts and theoretical frameworks for data analysis, but at the same time, we must also be careful not to Too much use of previous theories. Otherwise, the thoughts of the predecessors may constrain our thinking, causing us to intentionally or unintentionally apply other people's theories to our own data, or in other words, apply our own data to other people's theories, which is what people call it. "Cutting your feet to fit your shoes" rather than "tailoring your clothes to suit your needs".

While appropriately using previous theories, grounded theory believes that the researcher's personal interpretation can also play an important role in constructing the theory. The reason why a researcher can "understand" the data is because the researcher brings his own experiential knowledge. The theory generated from the data is actually the result of the continuous interaction and integration between the data and the researcher's personal interpretation. There is actually a triangular interaction between the original data, the researcher's personal pre-understanding and previous research results. The researcher must combine the original data and his or her own personal judgment when using the literature. The researcher himself should develop the habit of asking himself and being asked, listen to the multiple voices in the text, and understand the interactive relationship between himself and the original data and documents.

6. Theoretical evaluation

Grounded theory has its own standards for checking and evaluating theories, which can be summarized as the following four: 1) Concepts must be derived from original data; After the theory is established, you should be able to return to the original data at any time and find rich data content as the basis for argumentation. 2) The concepts in the theory themselves should be fully developed and the density should be relatively high, that is, there are many complex concepts and their meaning relationships within the theory, and these concepts are located in dense theoretical situations. What is different from what Geertz (1973) called "thick description" is that grounded theory pays more attention to the density of concepts, while "thick description" mainly provides a dense description of the research phenomenon at the descriptive level.

3) Each concept in the theory should have a systematic connection with other concepts. “Theory is a reasonable connection between concepts and sets of concepts” (Strauss & Corbin, 1994: 278). Each concept should be closely related to each other. intertwined together to form a unified, intrinsically linked whole. 4) A theory connected by a set of concepts should have strong application value, be applicable to a relatively broad range, have strong explanatory power, and be theoretically sensitive to the subtleties of the parties' behavior, and can make suggestions on these phenomena. related theoretical issues.

2. The operating procedures of grounded theory

The operating procedures of grounded theory generally include: 1) Generating concepts from the data and registering the data step by step: 2) Continuously reviewing the data Compare with concepts and systematically ask generative theoretical questions related to concepts; 3) develop theoretical concepts and establish connections between concepts; 4) theoretical sampling and systematically code data; 5) construct theory , and strive to obtain the density, variability and high degree of integration of theoretical concepts. Coding the data step by step is the most important step in grounded theory, which includes three levels of coding.

1. First-level coding (open login)

In first-level coding (open login), researchers are required to keep an open mind and try to "suspend" Personal "prejudices" and the "stereotypes" of the research community will register all data as they present themselves. This is an operational process of breaking down the collected data, giving it concepts, and then reassembling it in new ways. The purpose of registration is to discover conceptual categories from the data, name the categories, determine the attributes and dimensions of the categories, and then name and categorize the researched phenomena. The process of open registration is similar to a funnel. The registration range is relatively wide at the beginning, and then continuously narrows the range until the code number is saturated. When registering data, researchers should ask specific and conceptually related questions about the content of the data. When asking questions, keep your original research purpose in mind, and leave room for unforeseen goals to emerge from the data. An important principle that researchers should abide by at this stage is: believe everything and believe nothing (Strauss, 1987: 29).

In order to deepen their analysis, researchers should frequently stop and write analytical memos while conducting open access to data. This is an effective means of analyzing data, which can prompt researchers to think about theoretical issues arising in the data, and gradually deepen the preliminary theories they have constructed through writing. The main purpose of this round of registration is to open up the exploration of the data, and all interpretations are preliminary and undecided. The researcher's primary concern is not what concepts are contained in the text at hand, but how it can further the inquiry.

When conducting open login, you can consider the following basic principles: 1) Register the information carefully and do not miss any important information; the more detailed the login, the better until it is saturated; if you find If a new code number is obtained, further raw data should be collected in the next round. 2) Pay attention to the words used by the parties, especially those original words that can be used as code numbers. 3) Give each code number a preliminary name. The naming can use the original words of the person involved, or the researcher's own language. Don't worry about whether the naming is appropriate. 4) When analyzing the data line by line, ask specific questions about relevant words, phrases, sentences, actions, meanings, and events, such as: What does this data have to do with the research? What categories can this event produce? What specific information does this information provide? Why do these things happen? 5) Quickly analyze the dimensions of some concepts related to the data. These dimensions should be able to evoke cases for comparison; if no cases are generated, they should be looked for immediately. 6) Pay attention to the relevant items in the login paradigm listed.

2. Secondary coding (associative coding)

The main task of secondary coding (also called associative coding or axial coding) is to discover and establish relationships between concept categories Various connections to show the organic connection between various parts of the data.

These relationships can be causal relationships, temporal relationships, semantic relationships, situational relationships, similarity relationships, difference relationships, equivalence relationships, type relationships, structural relationships, functional relationships, process relationships, strategic relationships, etc. In axis registration, researchers only conduct in-depth analysis of one category at a time and look for relevant relationships around this category, so it is called an "axis". As the analysis continues, the various connections between the various categories should become more and more specific. When conducting correlation analysis on conceptual categories, researchers must not only consider the correlation between the conceptual categories themselves, but also explore the intentions and motivations of the researchers who express these conceptual categories, and put their words into take into account the context of the time and the socio-cultural background in which they live.

After the relationship between each group of conceptual categories is established, researchers still need to distinguish which are the main categories and which are the secondary categories. After these different levels of categories are identified, researchers can connect the relationships between them through comparison. When all master-slave relationships are established, researchers can also recombine the original data in new ways. In order to discover whether these analysis methods have practical significance, researchers can also establish a prototype of theoretical construction guided by action orientation or interaction orientation after exploring various categorical relationships. This theoretical prototype puts the focus of analysis on handling and solving practical problems, and its theoretical basis is the practical rationality of the parties involved.

3. Three-level coding (core coding)

Three-level coding (also known as core coding or selective coding) refers to: in all discovered concept categories After systematic analysis, a "core category" is selected, and the analysis is continuously focused on those code numbers related to the core category. The core category must be repeatedly proven to be dominant in comparison with other categories and be able to encompass the greatest number of research results within a relatively broad theoretical scope. Just like the strings of a fishing net, the core category can pull up all other categories as a whole and play the role of "outline and lead". To sum up, core categories should have the following characteristics: 1) Core categories must occupy a central position among all categories, be more concentrated than all other categories, be meaningfully related to the largest number of categories, and be most relevant. Strength becomes the core of the material. 2) The core category must appear frequently in the data, or the indicators that express this category must appear in the data with the greatest frequency; it should represent a relatively stable phenomenon that appears repeatedly in the data. 3) Core categories should be easily related to other categories. These relationships should not be forced, should be established quickly, and the content of the relationships between them should be very rich. 4) In a substantive theory, a core category can easily develop into a more general theory; before developing into a formal theory, the relevant data need to be carefully reviewed and tested in as many substantive theoretical fields as possible . 5) As the core categories are analyzed, the theory will naturally develop forward. 6) Since core categories are constantly logged in dimensions, attributes, conditions, consequences, strategies, etc., their subordinate categories may become very rich and complex. Finding internal variations is a characteristic of grounded theory.

In the core entry stage, researchers should always ask: "At what general level can this conceptual category(s) belong to a larger social analysis category? Is it possible among these conceptual categories? Summarize a more important core? How can I string together these conceptual categories to form a systematic theoretical framework? "The memo written by the researcher during this period should be more focused and analyze the density of theoretical integration of the core categories. Integrate theories until theoretical saturation and completeness are achieved. After the core categories are found, it can provide directions for the next step of theoretical sampling and data collection.

The specific steps of core login are: 1) clarify the story line of the data; 2) describe the main categories, sub-categories and their attributes and dimensions; 3) test the initial hypotheses that have been established, Fill in the concept categories that need to be supplemented or developed; 4) select core concept categories; 5) establish systematic connections between core categories and other categories. If we find more than one core category at the beginning of the analysis, we can connect related categories through continuous comparison and eliminate those categories that are not closely related.

Let me give an example to illustrate the above three-level coding process. When I was researching the cross-cultural interpersonal communication activities and meaning interpretations of some Chinese students studying in the United States (1998), I logged in the data step by step. First, in the open login, I found many "local concepts" used by the interviewees, such as "interest, desire, coming and going, prepared, often, in-depth, caring for others, taking care of others, managing, saving face" , lose face, reserved, considerate, tolerant, business-like, emotional communication, enthusiasm, warmth, close buddies, closeness, return, wandering outside, circle, unstable, unsafe, overwhelmed, big child, inferior to others, national self-esteem, no Comfortable" etc. Then, in relational login, I found some connections between the above concepts, connecting them under seven main categories: "Interaction, people, emotional exchange, friendship, outsider, self-esteem, change". Under each main category, there are related categories. For example, under "human feelings" there are "caring and taking care of others, being considerate and tolerant, saving face and being reserved", etc.; under "outsider" there are "freedom from outside, circle" , overwhelmed, unsettled, insecure, lonely, homesick, free and at ease” etc. Finally, after all the categories and generic relationships were established, I defined the core category as "culture's construction of the self and the relationship between others" during the core login process. After further analyzing the original data under this theoretical framework, I established two grounded theories: 1) Culture has a directional effect on individuals' concepts of self and self and interpersonal behavior; 2) Cross-cultural interpersonal communication has a directional effect on individuals' Self-cultural identity has the function of reconstruction.

4. An example of analysis

In the following, let me borrow from Strauss's book "Qualitative Analysis for Social Scientists" (1987: 12-7) An example is provided to demonstrate the process of analyzing data using the grounded theory method.

The hypothetical research question is: "Will (and how) the use of machines and equipment in hospitals affect the interaction between medical staff and patients?" We see many machines and equipment connected to patients in the ward. On the body, we can form a preliminary category - "machine-body connection" - to represent this phenomenon. Based on the observation results, we initially decided to divide the machines into two categories: machines that connect to the outside of the patient's body (such as the skin); and machines that connect to the inside of the patient's body (such as the nose, mouth). This distinction leads to two dimensions of the category "machine-body": internal connections and external connections. We can then further break down these dimensions, such as "connections within," and we can continue to ask: "Do these machines cause pain to patients? Are they safe for patients? Are they comfortable? Are they scary?" Ask these questions We can use a dichotomy: "yes" or "no", or we can use a continuum, from "strong" to "weak". Of course, this classification comes not only from the original data we collected in the field, but also from our own empirical data (for example, these internal organs of the human body are very sensitive, and machine connections may cause pain in these parts; the one protruding from the patient's belly The pipe looks terrible, so this pipe is probably unsafe).

The above questions involve the consequences of actions or things: "If this thing looks like this, is it likely to have life-threatening consequences?" At this time, we can also add some other specific conditions , such as: if the patient moves too fast, or he turns over while sleeping at night, or this tube falls out, his body becomes inflamed, and his life will be threatened in such a situation. We can also ask questions about the strategies used by medical staff: "Why did they put the tube in this way and not that way?" or about the strategies used by the patient: "Did he negotiate with the nurse to use another method?" We can also ask questions about the strategies used by the patient. Ask questions about the interaction: "What happened between him and the nurse when the machine was connected to him? Did they tell him beforehand, did they give him some warning? Did they just go ahead without telling him?" After doing this, he felt frightened?” (The last question also involves the results of the interaction between the two parties.)

After preliminary answers to these questions, we can begin to form some hypotheses. . Some hypotheses still need to be further tested through observation or interviews, but we can conduct observations and interviews in a more targeted manner.

We may find that a tube connected to the patient's nose is uncomfortable but safe. So we can do interviews on that. If we wish to continue probing for "conditions that lead to unsafe conditions," we can ask the nurse, "At what point do these connections become unsafe for the patient?" We can also watch for when the patient's nose becomes unsafe for the machine to connect to. When, some conditions occur, such as the connection is suddenly disconnected, or there is a problem with the connection method.

This thinking clue can further guide us to subdivide the dimensions, ask more questions, and form more preliminary hypotheses. For example, for machine connections that are more likely to come off, we can ask: "How did they come off? Was it an accident, negligence, or intentional (such as when the patient was annoyed, uncomfortable, or scared)?" Nurses use What strategies and techniques are used to avoid or prevent detachment? Give special care to the patient? Emphasize that the individual's safety depends on not moving or breaking the connection no matter how painful it is. Or through a 'cooperative' approach? Connected for a few hours? Or remove the machine periodically to let them relax? "These questions, assumptions and distinctions are not necessarily "true", but if they are "true", we can explore it further and find out the "yes-no" —maybe” and “why”. Obviously we always end up asking more questions about the conditions and consequences, not only about the patient himself, but also about the patient's relatives, the nurses, the different staff, the functioning of the ward, and perhaps Redesign of certain machine parts.

The above-mentioned more targeted inquiry will naturally lead us to ask: "Where can I find evidence of 'X' or 'Y'?" This question raises the issue of "theoretical sampling". Through the previous investigation, we began to look for relevant people, events and actions as a basis for sampling for our initial (perhaps very primitive) theory. For novice researchers, this sampling is usually done covertly in comparative activities, mainly comparing different sub-dimensions.

Guided by the above theory, we can also conduct sampling more widely. For example, sampling the safety and comfort of other machines that are attached to the human body, such as X-ray equipment, airplanes, toasters, lawn mowers, or those employed in the hands of workers who break concrete pavements in the streets. The shock to the body when the machine vibrates. This comparison is not intended to formulate a general theory about all machines or safe/hazardous machines, but rather to provide theoretical sensitivities regarding the use of medical equipment in hospital settings. Our external sampling is closely linked to internal sampling. Of course, these comparisons can also be drawn from other experiential data of our own (so-called "anecdotal comparisons"), such as our own personal experiences with machines, watching others use machines, reading novels, autobiographies or reports about machines ourselves Literature etc.

This article focuses on the grounded theory method in qualitative research, especially its basic ideas and operating procedures. There are many other different ways of constructing theories in qualitative research. Researchers may adopt different ways of treating and processing theories if they are trained in different schools, have different ways of looking at problems, and have different research situations. Therefore, readers need to adopt an open and flexible attitude when considering issues of theory construction.

Chen Xiangming (1998): "Sojourners and "Foreigners"—A Study on Cross-Cultural Interpersonal Communication of Chinese Students Studying in the United States", Changsha: Hunan Education Press.

Geertz,C.(Ed.).(1973).The Interpretation of Cultures. NewYork: Basic Books.

Glaser,B.(1978).Theoretical Sensitivity. Mill Valleyociology Press.

Glaser, B. (1982). Generating Formal Theory. In R. G. Burgess (Ed.), Field Research: A Source Book and Field Manual. London: George Allen & Unwin (Publishers) Ltd .

Glaser,B.& Strauss,A.(1965).Awareness of Dying. Chicago:Aldine.

Glaser,B.& Strauss,A.(1967). The Discovery of GroundedTheorytrategies for Qualitative Research. Chicago: Aldine.

Glaser,B.& Strauss,A.(1968). Time for Dying. Chicago:Aldine. Strauss, A. (1987). Qualitative Analysis for SocialScientists. Cambridge: Cambridge University Press. Strauss, A. & Cor bin, J.(1990).Basics of Qualitative Research: GroundedTheory Procedures and Techniques, Newbury Park: Sage.