Event Structure Analysis: a Qualitative Model
of Quantitative Research

David R. Heise

Chapter 9 in Nigel Fielding and Raymond Lee (Eds.), Using Computers in Qualitative Research (Newbury Park, CA: Sage)
Consult printed volume for reference list.

In the spirit of reflexivity (Woolgar, 1988), I begin by quoting myself, and later I illustrate arguments by analyzing my past work. Whether this is indeed Escher-like reflexivity (a hand drawing the hand which is drawing) or mere egotism, it does provide materials that are useful.

First the quote--a call for papers which I published on beginning an editorship.

Sociological Methodology will continue to publish important papers in the established areas of quantitative analysis. In addition, however, I'll be making an active effort to expand the literature on methods of social taxonomy and on techniques for analyzing sequences of social events. Methodologies for abstracting rules of organization will be sought especially, though they must be empirically oriented. Also, I'll welcome papers presenting advanced developments in 'qualitative' areas like historical methods, content analysis, and ethnography. (Heise. 1974: ix-x)

This hortation sounds my current theme a decade and a half later as I offer something like what the editor ordered: a methodology for modeling sequences of social events, which facilitates abstracting rules of organization, and which applies specifically to qualitative data.

Event structure analysis (Heise, 1989) materializes expert understandings about processes that might be impenetrable to the uninformed. Dealing with recorded incidents, an analyst defines events, defines logical relations among the events, and defines how each event enables and expends other events. The result is a grammar of action accounting for recorded incidents, and this model can be displayed graphically, employed for simulations, and compared with related grammars for purposes of contrast or generalization. This methodology is most appropriate in the later stages of a qualitative research project rather than at the beginning, because it depends on having incidents recorded in transcripts, End p. 136 narratives or sound-image recordings and also depends on intensive labor by an analyst with expertise concerning the incidents.

In the following section I discuss the presumptions which circumscribe the methodology's domain of application and from which the methodology gains its power. The subsequent section employs historical data to demonstrate the kinds of analyses which can be done and to illustrate an end product. In the final section, I reflect on the generalizability, utility and future of the methodology.

Required data and assumptions

Like every methodology, event structure analysis allies with certain kinds of data and a set of assumptions in order to obtain analytic power.

Event descriptions

The focus is on social events--more precisely, on verbalized renditions of events by a culture expert providing indigenous readings of social activity (Geertz, 1973). The culture expert might be a native member of the culture or an outside researcher who has achieved verstehen. Whichever--the requirement is that the expert phrases descriptions not only so that any comparable expert can recognize the events, but also with reality constraints embedded semantically so that other culture experts can comprehend readily how the events are inherently structured. The assumption here is that event descriptions represent expert understandings of what can and cannot happen in reality, and uncovering the tacit logic in the descriptions reveals the structuring of reality as perceived by experts.

Implication relations

Event structure analysis does not actually extract logical structures through semantic analysis of the event descriptions--that would be too demanding and risky. Rather event descriptions provided by an expert are fed back to the expert in order to obtain judgments about what implies what. Pair by pair the expert reports whether this event is required for that event. Experts themselves are the instruments for recovering the logic of their descriptions.

This elicitation process is facilitated through the application of syllogistic reasoning. For example, once an expert has reported that event Y requires event X, and event Z requires event Y, then there is no need for the expert to consider whether Z requires X: Z must require X, because Z requires Y and Y requires X. Drawing such inferences is essential once the number of events exceeds ten or so End p. 137 because then the number of pair-wise questions about event relations explodes to burdensome magnitudes if pairings are made mindlessly without taking advantage of prior answers.

Syllogistic reasoning with a large set of propositions is not all that easy for humans to do, but computers can obtain such derivations accurately and fast, and thus these kinds of elicitations end up being computer-assisted. Indeed, once a computer is brought into the process, it is used as well to record data (the event descriptions and their logical relations) and to draw diagrams of the logical structure which is uncovered.

Event Series

Yet even with computer assistance, it really is not practical to have an expert specify events involved in a process and describe their logical connections. For one thing, the expert may neglect crucial events no matter how often the computer prompts for something else related to the topic of interest. Moreover, the pair-wise consideration of event relations creates an overwhelming labor despite syllogistic reasoning. So the domain of application has to be more constrained in order to achieve practicality.

Event structure analysis deals with sequences of events. That is, rather than conjuring event descriptions out of context, the expert describes events which happened in specific incidents and considers those events in the order of their occurrence. One advantage of this approach is that the expert is much less likely to forget crucial events, especially if urged to give detailed descriptions of incidents. A second advantage is that the pair-wise questions about prerequisites can be reduced drastically in number--essentially by half--when considering events in serial order. The expert never needs to consider if later events were required for earlier events because the present is not influenced by the future according to a widely accepted metatheoretical assumption.

Focusing the elicitation on serially ordered events in actual incidents makes it fairly easy to obtain a model, and in the process one also acquires data for testing and improving the accuracy of the model.

Dynamic assumptions

A structure in which events are verbally defined and logically linked to each other turns into a production-system model for generating event sequences with the addition of three assumptions about how events condition each other. A production-system model is an End p. 138 action grammar that puts constraints on the strings of events that can occur over time.

The most obvious assumption is that an event cannot occur until all of its prerequisites are fulfilled. Thus in temporal sequences, the first occurrence of an event should be preceded by occurrences of all events which it logically implies.

A second assumption is that occurrence of an event depletes conditions produced by prerequisite events so, if the event is to happen again, the prerequisite events must happen again also. Thus, two different occurrences of the same event should be separated by occurrences of all the event's prerequisite events.

The third assumption is that an event ordinarily does not repeat unless conditions which it produced were depleted by some consequence--by an event that has the focal event as a prerequisite. Thus repetitions of an event should be separated by occurrence of a consequence.

A corollary of assumption two yields a further inference about intervening consequences. The corollary is this: when an event occurs it depletes its prerequisite events, and thereafter any other event with the same prerequisites cannot occur until the prerequisite events are repeated. Now shift focus to one of the prerequisites. The corollary means that it is impossible for more than one of that prerequisite event's consequences to occur without the prerequisite repeating.

Beyond these general assumptions, we also allow that some pairs of events (which have to be identified ad hoc) have a peculiar relation in which each primes and depletes the other. Entering your office and leaving your office is an example of such a pair: once you enter your office, you have to leave before you can enter again; and after you leave your office, you must enter once more before you can leave again. Clearly, if an event is in such a commuting pair, then repetitions of the event have to be separated by occurrences of the other event in the pair.

Adjusting assumptions

Typically, an implicational structure and the assumptions about event ordering constrain event sequencing so stringently that the model cannot account for sequences of events in actual incidents. Then the assumptions may have to be weakened for some events.

The first assumption--that occurrences of events require prior occurrence of all prerequisites--can be relaxed by allowing that an event may be primed by occurrence of any of its immediate prerequisites instead of all of them. In other words, required events End p. 139 can be treated as disjunctive prerequisites rather than as conjunctive. Then the first occurrence of the focal event does not have to be preceded by occurrence of all prerequisites, only by occurrence of one prerequisite (and its prerequisites). Similarly when applying assumption two we do not expect every possible prerequisite to occur between repetitions of the focal event but only enough to prime the focal event again.

The second assumption--that events use up the conditions which permit them--can be weakened by allowing that occurrence of some particular event does not deplete the conditions produced by a specific prerequisite. Thereby we can expect a repetition of the focal event without necessarily having a repetition of the event whose conditions still remain in force.

Weakening assumption two also weakens the corollary of assumption two. Thus, if we allow that an event is not depleted by one of its consequences, then a single occurrence of the event might be followed by occurrences of several different consequences.

The third assumption--that the conditions created by an event have to be undone by other happenings before the event repeats again--can be eliminated for some particular event in a model. Doing so allows the focal event to repeat without being depleted by occurrences of a consequence in between times.

In the current implementation of this methodology, commutative event pairs ordinarily are not specified at the outset, so weakening assumptions about commutation is not a general issue. Rather, commutation typically is brought into a model after having weakened assumption two to the point that we no longer have interesting constraints on the orderings of some event and its consequences. At that point a commutation may be identified so as to require that repetitions of the event always are separated by an occurrence of one particular consequence.

Such adjustments in dynamic principles are the usual way of shaping a model so that observed event series are consistent with sequences of events which the model can generate. Sometimes, though, no such adjustment will correct an inconsistency, or a required adjustment would be clumsy or would not make sense at all. Then attention turns to adjusting the logic structure, or to looking for errors in data, or to the possibility that the model is inadequate.

Adjusting logical relations

A logic structure constrains event sequencing by specifying which events have to precede occurrence of a focal event and which events might ensue from occurrence of a focal event. Thus the generative End p. 140 ramifications of a model can be changed drastically by deleting a logical connection between two events or by adding one. Such changes are not to be done casually because they conflict with the best judgment of the expert who created the logic structure. However, sometimes even the original expert can be convinced that he was wrong in judging that one event is required for another when he sees the problems his judgment creates in interpreting sequences, or that he was wrong in judging one event as unrelated to another when he sees the opportunities for explanation which such a linkage provides. Then the logic structure reasonably can be changed to make the model more consistent with an observed series of events.

Changing the logic structure can change the meaning of at least one event, and therefore a corresponding change in descriptive phrasing may be required, too. Moreover, once an event is reconceptualized, there could be ramifications in judging how the event logically relates to still other events.

Adjusting serial data

Occasions do arise when no acceptable changes in assumptions or in logic structure will fix an inconsistency between a model and the serial record of events. There still is one more way to achieve consistency: change the record of events. Such changes are adopted only within the usual criteria of historical analysis (McCullagh, 1984) that the weight of the evidence, including the expectation provided by the model itself, supports the modification. This theory of error in qualitative data parallels error theory in quantitative research: observations arc fallible, and reality corresponds to expectations generated by a conceptual model, providing that the model accounts for most observations.

Rejecting a model

The last resort is to scrap the model entirely, a most unlikely outcome if definitions of events and their interrelations were obtained from a competent expert, but a probable denouement if data were obtained from novices or outsiders. For example, my own attempts to make models on topics where I am unlearned have yielded formalizations of ignorance which are intractably unadjustable so as to be consistent with series of observed events.


The medium for performing the intricate analyses which are involved in event structure modeling is a microcomputer program called ETHNO (Heise and Lewis, 1988). The program can be End p. 141 employed directly by culture experts who are literate, or it can be used in computer-assisted interviews, or the program can be used by researchers who claim expert competencies (verstehen) concerning their topics.

Two of the major options in ETHNO are: 'Create a structure', which carries out computerized elicitation of events and logical connections; and 'Analyze a series', which tests consistency of a model with event series and which allows a model and serial data to be adjusted for better convergence. These routines will be examined in detail below.

The end product of ETHNO analyses is a model specified in terms of event definitions, a diagram of the logical relations among events, and delineation of the assumptions which turn the logic structure into a generative system.


Having already illustrated this methodology's applications in ethnography (Corsaro and Heise, 1990), content analysis (Heise, 1988) and the study of careers (Heise, 1990), I now want to illustrate the procedure as a form of historiography. I happen to know no body of historical events (never mind historic) so well as my own work, and that is why I ruminate about my past. The product of analyzing my professional activities and those of my associates will be a model of scientific activity, and it may be of some use in the sociology of science.

A time-ordered list of events related to the development of Affect Control Theory (ACT) constitute the data for analysis. Affect Control Theory (Heise, 1979; Smith-Lovin and Hcise, 1988) elaborates the idea that people avoid events which create 'tension' in affective associations. Selecting low-tension behaviors yields normative action for people in specified roles. Selecting roles (instead of behaviors) to minimize tension corresponds to social labeling processes in which identities are assigned to people on the basis of their actions. Emotion reflects the amount and kind of tension produced by an experience.

ACT'S mathematical model applies the theoretic principle, operating on databases from empirical studies in order to provide computer simulations of social interaction. In the simulations, actors are verbally characterized by identities alone or by modifier-identity combinations, and settings may be specified. Results of simulations include verbal predictions about the behavior of interactants, their emotions, and the social labels and traits which they would attribute to each other on the basis of normal behavior or disruptive events. End p. 142

I assembled the list of ACT-related events from my own vita, from chronicles in ACT reports and publications, and from records in my own files. This provided a framework of career events (degrees, employments, publications, external fundings, editorships) and of activities by key associates (co-authors, students, correspondents). I then added events which in my view were critical parts of the research process--events .like collecting data, collating data, analyzing data, estimating equations, etc. The preliminary list of 160 events was sent to two other highly active ACT researchers, Lynn Smith-Lovin at the University of Arizona and Neil MacKinnon at the University of Guelph, Canada, and each provided corrections and expansions related to their own participation in the research program.

Exploratory analyses with ETHNO revealed that additional events (for example, within-university fundings) had to be added in order to create a logical structure, and the extra details were recalled by searching files when possible. Also, as I continued working on this project, I recalled events with low salience (like publication rejections, grant denials, and work by students who had left the program long ago). Indeed, the pool of related events seems bottomless, and I have purposely included only events which were directly or indirectly consequential in some public way.

Considerable effort was given to obtaining a correct chronological order not only by year but within years. However, records often did not provide such fine grain, and I had to resort to reasoning and reminiscing in order to reconstruct the stream of events within busy years.

The final list of 298 events is too long to provide here, but Appendix 9.1 shows the beginning and the end of the data. Note that the events are characterized abstractly enough to reveal the repetitiveness of research activities.

Mechanics of elicitation

Elicitation was begun by selecting Create a Structure from ETHNO'S main menu, selecting a framework appropriate for processing historical data, and naming the domain of events 'ACT'.

Events were entered in sequence. An event repetition was positioned in the serial record by entering the event description exactly the same as the first time or by entering the event's abbreviation.

In response to each non-repetitive entry, ETHNO asked questions in order to determine the logical relations between the last entered event (call it Event L) and prior events (for example, Event P). The question always had the same general form: Does Event L require End p. 143 Event P (or a similar event)? An expanded phrasing of this question would be: Does an initial occurrence of Event L require the prior occurrence of Event P or of some other event which can serve as the functional equivalent of Event P? A 'yes' answer established a logical link between Event L and Event P. The program immediately assessed implications of this logical connection in order to reduce the number of subsequent questions.

An updated diagram showing the current logic structure appeared on the computer's screen as each event was processed.

After the logical relations of every distinct event were defined, ETHNO filed the event descriptions, the logical linkages and the event series for future use.

Elicitation concerns

Appendix 9.2 illustrates dialogue with the computer during the elicitation phase and the kinds of thinking I did in order to answer questions about prerequisites; it also includes comments about ETHNO'S operations. I purposely show a portion of my first effort (rather than the final elicitation which led to the model displayed later) in order to demonstrate how a model evolves during its construction: the 'Revision' notes focus on changes.

Entries into ETHNO were the phrases in the 'Action' column of Appendix 9.1 preceded by R (for researcher). For example, the first three events were:

R read
R funded locally
R contributed funding.

The terse phrasings, which were convenient during ETHNO analyses, are expanded to the semantic forms that actually influenced my judgments in Table 9.1.

Some of my initial phrasings of events were superseded because systematic processing of the data refined my conceptualizations about research (Appendix 9.2 gives some examples). In other cases, phrasing stayed the same but interpretations changed: for instance, initially university support for computer usage was included as an aspect of local funding, but that interpretation was dropped when it became evident that computer funding occurred automatically whenever needed and so it need not be noticed.

The final list of events includes some events that were not in the initial list at all. These eventually were added in order to account for complexities in incidents that were not obvious at first. For example, 'R wrote report' eventually had to be added in order to deal with intricacies of co-authorships. End p. 144

I ultimately deleted a few events entirely; for example: 'R affiliated with ACT research program'. I never could decide when this event happened for other people, and I could not figure out how the event applied to myself. I finally decided that this was a projection I made onto other people when they initiated projects related to ACT rather than a decision others themselves made, and I removed the event from the corpus.

When answering questions about prerequisites, I typically recalled the specific incident and wondered what would have happened if the focal event had not been preceded by the specific prior event, and I decided that the prior event was not required for the focal event if I easily could imagine the focal event occurring in the circumstances even without the prior event. When unsure about the necessity of the prior event, I recalled other experiences in which the focal event occurred--including the communicated experiences of other researchers--to see if the prior event always preceded the focal event. A single instance of the focal event occurring without the prior event (or a functional equivalent) led me to discard the prior event as a prerequisite for the focal event. Thus my own experience in operating as an expert suggests that experts judge logical relations by applying the method of analytic induction (Znaniecki, 1934; Robinson, 1951) to a corpus of experiences stored in personal memory.

Thinking about prerequisites was rapid and easy when events were phrased well. When phrasing was 'off', I frequently had to ponder whether prerequisites were functionally equivalent to each other. Phrasing also was important in keeping me focused on certain actors (for example, in reminding me that I was considering only researchers) and in differentiating contrasting objects (for example, in reminding me that there is a difference between a dataset collected for a single project and a multi-use database).

Mechanics of series analysis

I presumed that a series of actions by a specific researcher was explainable by a general model, and therefore I concatenated the event series for different researchers. Thus the event series which I analyzed was not a single progression but rather 16 blocks of time-ordered events, one for each different researcher, ranging in length from 145 events down to 2 events. On reaching the marker for a new series (the name ACT instead of an event. description), ETHNO automatically started a new analysis, while retaining any changes in the model which were made while analyzing previous blocks.

Testing congruence between the model and the series data was begun by selecting Analyze a Series from ETHNO'S main menu and End p. 145 then recalling the file containing data for the ACT research program. The logic diagram for the model was displayed on the screen with the abbreviation of the first event blinking.

Pressing the ENTER key visually marked the first event as accomplished and lit up paths to all events which became possible once the first event was done. (ETHNO's graphic representation of completed events and possible events was inspired by Clarke's, 1983, idea of penciling over completed portions of a behavior plan.) Meanwhile the abbreviation for the second event started blinking. Repeated pressing of the ENTER key stepped through the events in sequence, and, as additional events were accomplished, other abbreviations were marked as done, other lines to newly possible events were brightened, and visual marking was removed from events which had been used up by consequences.

Eventually an event was encountered which could not be explained in terms of the logical structure and the default assumptions about how events enable and deplete one another. In a superimposed window the program explained the problem--either an unfulfilled prerequisite or an event repeating without an intervening consequence-- and began offering suggestions on how the problem could be solved.

After one of the suggestions was adopted, the program implemented the change and automatically reanalyzed all prior events to make sure the solution did not create problems earlier in the series. Then it stepped through more of the events in the event series.

On reaching the last event the program filed for future use the event descriptions, logical relations, assumptions applying for each event, and the complete event series. The file contained all modifications incorporated during the series analysis.

Modifications from series analysis

Appendix 9.3 illustrates the dialogue with the computer which occurs during a series analysis. Though the illustration is limited to the same events as were treated in Appendix 9.2, an example of each of the major types of problem appears in Appendix 9.3, and some (but not all) of ETHNO's different kinds of suggested solutions are exemplified.

In the course of series analyses I made one change in logical structure, added commutation for one pair of events, gave disjunctive prerequisites to 13 events, eliminated depletion along 37 (of 58) linkages between events, and allowed 30 of the 42 events to be repeatable without depletion.

In all, 30 items in the serial record of 298 events were added, moved or changed from one classification to another as a result of End p. 146 series analysis. Mostly these were cases in which I was reminded to add an event that was recalled easily when logic demanded. Once I had an event six positions higher in serial ranking than it should have been (though still in the right year). In two cases, series analysis uncovered the fact that I had recorded erroneously the kind of event which occurred: external versus local funding in one case; database measurements versus single-study measurements in another case.

ETHNO suggested the solutions which were adopted in all cases except one. In the one instance where ETHNO could offer no suggestions, I had glossed over two events in sequence (collating and analyzing a database) which are required for publishing a database study. The missing events had to be added to the event series using an ETHNO editing routine in order to proceed.

Final model

The event structure model for the ACT program of quantitative research consists of event descriptions, a logical structure, and specifications concerning how each event's logical relations constrain event orderings.

Table 9.1 lists the events in the model, showing ETHNO abbreviations, short names used in analyses, and detailed descriptions which gave an appropriate semantic basis to the logic structure.

Table 9.1 Event descriptions

Ana analyzed database. R (a researcher) extracted new information from a database through transformations of measurements or by focusing on a subset of cases.
Cer certified in profession. R obtained a doctorate or other professional credential while not involved in the research program.
Col collated data. R organized the stimuli used to obtain measurements from respondents, organized and verified the measurements themselves, and then perhaps computed descriptive statistics (e.g. means) which could be the basis for further analyses.
Con contributed funding. R employed personal funds to foster research activities.
Cln contracted with publisher. R obtained a written contract assuring that a publisher would accept a book manuscript for publication.
Den denied funding. R received a letter from an external funding agency in which the agency declined the opportunity to support proposed research activities.
D1n denied monograph publication. R received a letter from a publisher in which the publisher declined the opportunity to publish a book manuscript reporting research activities.
Edi edited journal issue. R negotiated with the official editor of a journal and End p. 147 thereby gained editorial control over one issue of the journal for the purpose of promoting a topic or a research program.
Est estimated equations. R concretized algebraic portrayals of relations between various measurements by estimating equation parameters as numerical values through statistical analyses of a sample of measurements. Some collation of data is presumed to be part of equation estimation.
Fun funded locally. R received funds administered within a university to free a researcher from remunerative activities like teaching or to buy research materials and services.
F1n funded externally. R received funds from a source beyond the researcher's own university to free one or more researchers from remunerative activities like teaching or to buy research materials and services.
Gat gathered database measurements. R measured people's subjective responses to verbal stimuli in order to create a database. (ACT databases were created solely through survey methods.)
Gav gave invited talk. R was invited to speak at a conference or a colloquium outside of the researcher's own university, and the talk led to a publication about the research program by the researcher or by someone in the audience.
Imp improved simulation system. R made the output of a simulation system more realistic by refining the computer program. by incorporating more refined equations and rules, or by refining usage of a database.
Iss issued simulation system. R found a way of distributing a simulation program, with databases and instructions, so that people could operate the system on accessible computers.
Joi joined faculty. R obtained a professorship permitting pursuit of intellectual interests and enlistment of student researchers into the research program.
Mat mathematized formulation. R constructed a mathematical derivation which transformed assumptions about reality along with empirically based equations describing a process into additional equations describing another process.
M1t mathematized methodology. R constructed a mathematical derivation which resulted in the definition of a complex methodological procedure.
Mea measured responses. R obtained measurements of people's responses to a number of verbal stimuli (presented in a questionnaire or by a computer) in order to conduct a specific analysis.
Per performed simulation. R employed a simulation system to enter information about social situations and obtain a computer report about theoretical predictions.
Pro programmed simulation system. R programmed a computer in order to implement an empirically grounded mathematical model (plus additional rules) while making use of a database such that a variety of problems could be set up easily and theoretical predictions examined readily.
Pub published database study. R published a description of methods and of the results of processing a database in order to address some issue.
P1b published equation estimations. R published an article describing how some process can be given an algebraic formulation and how numbers were found to make the equations concrete and descriptive of reality.
P2b published methodology. R published a report describing a generalized research procedure and discussing its benefits and limitations. End p. 148
P3b published theory formulation. R published a statement claiming that some abstracted aspects of reality are interrelated in a principled way.
P4b published simulator results. R published illustrative simulation results in order to communicate a theory's capacity for portraying reality.
P5b published research monograph. R published a lengthy systematic exposition describing activities and outcomes in a research program.
P6b published math derivations. R published a report describing how a mathematical derivation was obtained and how the results are to be interpreted.
P7b published research overview. R published an exposition outlining the claims, activities and products of a research program.
P8b published test of theory. R published a report defining a theoretical assumption or prediction, how the claim was examined empirically, what the results were, and how the results favor or undermine the focal theoretical formulation as well as other theoretical formulations.
P9b published edited book. R published a collection of writings by various authors on a particular topic or research program.
Ran ran (a social psychology) experiment. R constructed real social situations representing distinctive circumstances and assessed some aspects of people's responses to the different circumstances.
Rea read. R consumed reports and publications regarding theory, research or research methods from within the researcher's own research program or from other research programs.
R1a reanalyzed prior study. R performed new analyses on measurements which were collected and analyzed previously, and the new analysis addressed the same issue as the prior work.
Rec received doctorate. R was awarded a doctorate degree certifying the person as a competent researcher.
Req requested funding. R sought external funding from a government agency or from a foundation or from an outside research institution through submission of a proposal outlining a research plan and a budget for specific research activities.
Sol solicited paper. R requested preparation of a report by another scholar or researcher, with assurance that the report would be published in an edited book.
Sub submitted research monograph. R sent a book-size manuscript reporting theory and research to a publisher for possible publication.
Tes tested theory. R used empirical data to examine the accuracy of a theoretical assumption or prediction. Some collation and analysis of data is presumed to be part of testing a theory.
Uti utilized existing database. R made use of a collated database in order to conduct some kind of research.
Vis visited other faculty. R visited another faculty during a sabbatical leave from his or her own faculty.
Wro wrote research report. R prepared a report interpreting literature, describing how an experiment was conducted or how measurements were made, how statistical or other kinds of analyses were done, or how mathematical solutions were derived. End p. 149

Logical connections among the events (with each event represented by its abbreviation) are shown in Figure 9.1. The topmost entry is simply the name of the system; the tier below the topmost entry shows events which have no specified prerequisites within the system.

Lines traced downward from an event define the event's potential consequences. (Gaps in lines occur when two unrelated paths cross each other.) For example, Contributed funding (Con) can lead to Measured responses (Mea) or to Gathered database measurements (Gat). Tracing further reveals indirect consequences; e.g., Contributed funding may lead indirectly through Measured responses to Tested theory (Tes) or Estimated equations (Est) or Wrote research report (Wro).

Lines ascending from an event define the event's prerequisites, which are disjunctive if the event abbreviation is printed all in capitals, conjunctive otherwise. For example, the prerequisites for Improved simulation system (Imp) are Programmed simulation system (Pro) AND Performed simulation (Per); the prerequisites of Measured responses (Mea) are Contributed funding (Con) OR Funded locally (Fun). Only two events in the system (improving and issuing a simulation system--Imp and Iss) have conjunctive prerequisites rather than disjunctive.

ETHNO's diagrams ordinarily do not show a line between two elements if that line can be reproduced by tracing direct paths (for instance, there is no line between Con and Tes because their connection can be determined by tracing through Mea). I forced two exceptions in this model--direct lines indicate that gathering database measurements (Gat) can be directly supported by personal contributions (Con) or by local institutions (Fun), as well as being indirectly dependent on past contributions or past local funding. Dynamic assumptions which cannot be represented in the diagram are shown in Table 9.2.

Thirty of the 42 events had to be declared repeatable without depletion. Mostly the repeatable events have permanent products (like publications) so the designation of being repeatable without End p.151 depletion is natural. In a few instances the designation results from a current limitation in ETHNO--the program does not stack event occurrences and count depletions; thus an event like 'analyzed database' has to be repeatable because researchers who are pursuing several projects simultaneously may do several analyses before writing up any of them.

One pair of events was put into a commutative relation: submission of a research monograph to a publisher (Sub) and having the manuscript rejected (D1n). The commutation represents the process whereby submission is made to one publisher at a time, and rejection primes another submission. Submitting manuscripts always is essential for getting the manuscripts published, and this is the case whether they are articles or books, but the model amalgamates submission and publication in the case of articles because there are so many forms of article publication that listing component steps would clutter the model substantially, and, in any case, most ACT article manuscripts were not rejected, so the detail would be of little value. Note that a commutative structure does not cover submission and denial of grant requests, first, because multiple grant requests often are made at once, and, second, because the intellectual framework and the personnel which justify a request may vanish before a re-submission can be made.

Overview of the model

The model presented in Table 9.1, Figure 9.1, and Table 9.2 defines 42 research-related events and specifies how ordering of those events is constrained, thereby forming a grammar of action which generates sensible sequences of action in a quantitative research program. The model accounts for event series containing 298 events by 12 associates of the program and by four outside scholars.

Five events are customary happenings in academic careers: Gave invited talk, Joined faculty, Read, Received doctorate, Visited other faculty. Certified in profession is a convenience construct which simplified inclusion of researchers who obtained their PhDs before becoming involved with ACT (and in one case was used to represent certification in a non-sociology profession which permitted applying for special funds). Thus, professional participation in the academic world is an important aspect of the kind of research which I modeled.

Another five events deal with economic aspects of research: Contributed funding, Denied funding, Funded externally, Funded locally, Requested funding. A quantitative research program based on empirical data has to attend in one way or another to research financing. End p. 152

Many events in the research program relate to publishing. Nine events, Published database study, Published equation estimations, Published moth derivations, Published methodology, Published research overview, Published simulator results, Published test of theory, Published theory formulation, Issued simulation system, define types of publication. Note that these events refer to kinds of statements rather than to publication of standard units like articles or chapters. It is the content of publications which relates most directly to research activities, and a single publication may contain several kinds of content, therefore amount to several events. Seven other events also relate to publishing, but their focus is on processes involved in creating books and special issues of journals: Contracted with publisher, Denied monograph publication, Edited journal issue, Published edited book, Published research monograph, Solicited paper, Submitted research monograph.

Events which have no further consequences within a system amount to system outputs, and publications and public presentations are the outputs of a research program, these intellectual products being exchanged with other research programs. Denials of publication also are kinds of outputs, and in a sense these foregoings are exchanged with other research programs, too. Soliciting papers for an edited book involves control of resources which also may be offered to other research programs.

Finally, 15 of the events are specifically focused on research. Three of these, Gathered database measurements, Measured responses and Ran experiment, are intrinsically social in social science research. (Measured responses is distinguished from Ran experiment in that stimuli in experiments are real social situations rather than printed presentations; a database study is distinct from Measured responses in that the sample of people or the sample of stimuli is sufficiently comprehensive that the data may be broken into different categories for various kinds of analyses.) Otherwise--in contrast to the academic, economic and publishing aspects of research--these technical activities might be implemented without engaging in much social interaction or written correspondence: Analyzed database, Collated data, Estimated equations, Improved simulation system, Mathematized formulation, Mathematized methodology, Performed simulation, Programmed simulation system, Reanalyzed prior study, Tested theory, Utilized existing database, Wrote research report.

The core research activities (in the center of the diagram) essentially consist of three lines of development: testing theory, developing simulation systems, and developing mathematical formulations. Because most of the events in these sequences End p. 153 become possible with occurrence of a single prior event (either because that event is the sole prerequisite or because it is one of the several prerequisites in a disjunctive set), the core research activities typically are 'straight-line' ventures in which one step naturally leads to another rather than occasional opportunities which suddenly open on completion of a variety of immediate prerequisites.

Many research-related events do not deplete their prerequisites, so the conditions created by events in a research program often last long enough to support several events. For example, receiving a doctorate provides a continuing resource; reading is practically non­depletable in the sense of providing permanent intellectual resources; and collation of a database, estimation or derivation of equations, or writing a research report similarly provide long-lasting resources for a research program. Thus a quantitative research program cumulates research capital over time, enhancing its durability and productivity.


The model presented here does not deal with every interesting sociological aspect of research programs. For example, the model shows that funding is essential for some research events, but it does not show that funding accelerates research events--an effect which is evident in the historical record and which might be formalized through a different methodology (for example, Tuma et al., 1979). Also, the model shows how research products are generated, but it does not show how different research programs are linked through the exchange of those products, though that is a common theme in the sociology of science and readily studied through application of network methodology (for example, Burt, 1990).

The model presented here inevitably is conditioned by social institutions which were in the background during the research activities. For example, publication events are important in the model, and that is because the historical record contained numerous publishing events, deemed meaningful and faithfully recorded by researchers because publishing is demanded by the academic institutions that employ researchers. Quite a different picture of research might emerge in other supporting institutions (like corporations) which promote the recording of different events, thereby changing the recorded history from which the model is derived.

The model is to some degree a product of the unique experiences of the expert constructing the model. For example, events related to End p. 154 simulation might not appear in the models of expert representatives from other quantitative research programs, and other researchers might distinguish more events relating to experiments and testing theory than I did. Indeed, ten years from now I might provide a somewhat different model of the ACT research program because I will have accumulated more experiences by then.

Limitations notwithstanding, the model has several kinds of utilities. First, as a scientific product, it is an object that can be compared and classified with similar objects (like models of other academic research programs and of corporate research) in order to identify differences or to abstract general features of the domain. Second, as a dynamic simulator of research events, the model might be used to socialize neophytes and to disseminate knowledge that could foster wider institutionalization of quantitative research programs. Third, the model permits 'action at a distance' (Latour, 1987) for administrators in academic, funding and publishing institutions by facilitating the anticipation and direction of events which the authorities themselves do not produce.

Event structure analysis has a close affinity with Abell's (1987; 1988) method of comparative narratives, but whereas event structure analysis is focused primarily on data analysis procedures, Abell's work on comparative narratives is focused primarily on solving related conceptual problems. Exchange might be expected between these two research programs. For example, ETHNO's abstraction routine might incorporate Abell's idea that abstraction amounts to a homomorphic reduction. Meanwhile empirical projects like this one clearly indicate that Abell's preferred homomorphism has to be weakened in order to deal with cycling, because repetition of events is ubiquitous in human action once one goes beyond the most concrete level of description.

Note: The following researchers are represented in the data: C. Averett, W. Douglas, D. Heise, L. Keating, L. Lazowski, N. MacKinnon, R. Morgan, B. Smith, L. Smith­Lovin, L. Thomas, B. Wiggins, D. Willigan. Also included are some activities by four outside scholars: T. Kemper, H. Smith, S. Stryker, P. Thoits. Not included are events by graduate students whose work yielded no publications as of February, 1989: C. Cassel, M. Young, L. Wood, R. Sands and S. Lerner at the University of North Carolina; I. Okuyama and M. Brondino at the University of South Carolina; T. Leowinata at the University of Guelph; M. Malone and D. Barrett at Indiana University; D. Robinson-Reeve at Cornell University.



Create a Structure was chosen from ETHNO's main menu. Asked for a framework file, I entered D (for dynamic), in order to process historical data. ETHNO provided a screen of instructions and asked for the name of the structure I was going to create.

I began entering data by typing ACT, an acronym for the title, 'development of Affect Control Theory'. ETHNO asked for the first event. I typed: R read, which is the first event given in Appendix 9.1, ignoring specific details.

At that point ETHNO drew a diagram on the screen, Act at the top, Rea below it, connected by a vertical line. (Each entry on an ETHNO diagram is represented by a three-letter abbreviation formed from the second word of the entry.)

In a separate window at the bottom of screen ETHNO asked, 'Next happening?'

Henceforth I show the event which I entered and ETHNO's abbreviation for it. Then I show the questions ETHNO asked and the answers which I gave, followed by an explanation of my answers. I also provide commentaries to indicate how definitions and interpretations were revised in the course of analyses and to explain ETHNO's operations.

R funded locally Fun

Does 'R funded locally' require 'R read' (or a similar event)'? N

ETHNO needed to know the logical relation between a researcher reading and a researcher being locally funded. Prior reading of research reports is not really required in order to receive funding which is administered within a university, only that a researcher be affiliated with the university (and all characters in the ACT research program had university affiliations as students or post-doctoral fellows or faculty). Thus I answered 'no'.

ETHNO notes:

ETHNO did not ask whether funding is required for reading because funding occurred later than reading.

R contributed funding Con

Does 'R contributed funding' require 'R funded locally' (or a similar event)? N
Does 'R contributed funding' require 'R read' (or a similar event)?

Why do researchers use personal funds for research projects? Few prior events are available to serve as explanation in this case. I assumed that funding from a university is not essential. I said that reading is required, thereby letting the act of reading take on the extra meaning of developing enthusiasm for a researchable idea.

ETHNO notes:

At this point the ETHNO chart which appeared on the screen looked as follows.

This shows that events of receiving funding and reading are part of the ACT event structure. Additionally it shows that contributing personal funds for research implies reading (that is, enthusiasm for an idea). Note that reading is a necessary but not End p. 157 sufficient condition for personal contributions: ETHNO diagrams display developmental relations, not causal relations.

R collected database Col

Does 'R collected database' require 'R contributed funding' (or a simUar event)? Y
Does 'R collected database' require 'R funded locally' (or a similar event)?

Assembling a sizeable database of quantitative measurements requires money for questionnaires, computer materials, etc. So funding from some source is required. I answered 'yes' to both questions because in this particular case funding from both sources was required (I collected data on 1,000 words, and I could not have done so without my fellowship plus some personal expenditures).


The name of this event eventually changed to 'R gathered database measurements' (Gat) to emphasize that this event refers only to the fieldwork aspect of assembling a database.

R funded locally Fun

At this point in the original event series there was a repetition of the event 'R funded locally'.


The local funding in this case was support for computer usage, and eventually all instances of computer support were deleted from the corpus because obtaining university funding to support computer usage never was a contingent factor in research dynamics--such funding always was available (even though one had to 'apply' for it).

ETHNO notes:

Since this was a repetition, I entered the event's ETHNO abbreviation, whereupon ETHNO incorporated a repetition of the event into the event series but asked no questions since the logical relations of this event were already defined.

R programmed database Pro

Does 'R programmed database system' require 'R collected database' (or a similar event)? Y

I reasoned that social researchers do not develop database systems in the abstract, that a researcher has to collect a database before he would bother entering a program of instructions to a computer in order to process data, and thus I answered 'yes' to the question.


The description of this event changed to 'R collated data' (Col) in the course of analyses, purposely eliminating reference to computer programming because there were instances where the same operations were conducted without computer assistance.

ETHNO notes:

No additional questions were asked about this event because ETHNO computed answers on its own. Programming implies collecting data, and collecting data implies two kinds of funding, so programming also implies these kinds of funding. Moreover, personal funding implies reading, so programming implies earlier reading.

R analyzed database Ana

Does 'R analyzed database' require 'R programmed database system' (or a similar event)? Y End p. 158

Changing 'R programmed database system' to 'R collated data' modifies some interpretations dramatically, and so it is best to show my original reasoning in quotation marks.

'Durkheim, Sorokin, and other early sociologists analyzed huge databases without computer assistance, but the discipline changed just prior to my entrance into the field, and I am among the sociologists who employ a computer in order to cope with a large set of quantitative data. I answered "yes" to this question to reflect the technology and the economics of the field currently.'


Since 'programmed' ended up simply as 'collated', the original reasoning is irrelevant and somewhat misleading: current research (my own included) still involves many of the same operations as classic research, though these operations can now be done more efficiently with computers. However, the logic structure remains the same for these events because analyzing data requires collating data.

Though not in the series that I analyzed initially, 'R wrote research report' (Wro) eventually was added as the next event. 'Wrote research report' deals with unpublished manuscripts that got incorporated into co-authored publications (or which had other kinds of consequences) and had to be included wherever appropriate once identified as part of the system.

R received doctorate Rec

Does 'R received doctorate' require 'R analyzed database' (or a similar event)? N
Does 'R received doctorate' require 'R programmed database system' (or a similar event)?
Does 'R received doctorate' require 'R collected database' (or a similar event)? N
Does 'R received doctorate' require 'R contributed funding' (or a similar event)? N
Does 'R received doctorate' require 'R funded locally' (or a similar event)? N
Does 'R received doctorate' require 'R read' (or a similar event)? Y

Original reasoning:

'Obtaining a doctorate in the social sciences usually involves collecting and analyzing data, but that is not essential. Funding events of any kind have no logical relation to granting of a degree. It is inconceivable, though, that a PhD would be awarded without the candidate knowing some of the literature in his or her field, so I answered that receiving a doctorate does imply reading.'


Inclusion of 'R wrote research report' in later analyses led to quite a different structure. Receiving a doctorate requires writing a research report (a dissertation), and research reports of quantitative researchers may involve significant confrontations with data. Essentially, later analyses forced me to recognize that I was dealing not with anyone in 'social science' but only with people in quantitative research programs.

The event series in Appendix 9.1 next shows an event which was glossed in my initial analysis: 'R requested funding' (Req). Ultimately, I had to include 'R requested funding' in the system in order to explain how denials of funding could happen, and, once included, this event had to be placed everywhere it happened.

R funded externally F1n

Does 'R funded external1y' require 'R analyzed database' (or a similar event)? N
Does 'R funded externally' require 'R programmed database system' (or a similar event)?
Does 'R funded externally' require 'R collected database' (or a similar event)? N End p. 159
Does 'R funded externally' require 'R received doctorate' (or a similar event)?
Does 'R funded externally' require 'R contributed funding' (or a similar event)?
Does 'R funded externally' require 'R funded locally' (or a similar event)?

Original reasoning:

'Empirical analyses are not essential in order to receive a financial award from outside of one's university (though having done some empirical analyses might improve one's chances). Nor are prior funding experiences necessarily relevant. The implicit prerequisite is a PhD. (While some national awards are given to pre-doctoral students, they invariably require strong faculty sponsorship of the student which can be treated as a "similar event".)'


In the final logic structure, external funding implies receipt of a PhD, and a PhD among quantitative researchers implies past empirical research, so external funding does imply empirical research.

ETHNO notes:

The abbreviation for 'R funded externally' is formed from the second word as usual, but ETHNO substitutes a 1 for the second letter of the abbreviation in order to prevent a duplication.

Appendix 9.1 next shows an event which was not included in the series at this point: 'R certified in profession' (Cer). This event involves a different researcher, so the event was not entered until later when the block of events involving Scholar 1 were under consideration.

R joined faculty Joi

Does 'R joined faculty' require 'R analyzed database' (or a similar event)? N
Does 'R joined faculty' require 'R programmed database system' (or a similar event)?
Does 'R joined faculty' require 'R funded externally' (or a similar event)?
Does 'R joined faculty' require 'R collected database' (or a similar event)?
Does 'R joined faculty' require 'R received doctorate' (or a similar event)? Y
Does 'R joined faculty' require 'R contributed funding' (or a similar event)?
Does 'R Joined faculty' require 'R funded locally' (or a similar event)?

Original interpretation:

'Joining a faculty requires a PhD, with only rare exceptions (covered by the "similar event" phrase). No specific kinds of research experience are universally required, and prior funding is not essential.'


Anything that requires a PhD implies past empirical research experience for quantitative research.

R read Rea

This repeats an earlier event.

ETHNO notes:

The abbreviation rea is entered, whereupon ETHNO logs the event into the event series but asks no questions since the event is already positioned in the logic structure.

R published database study Pub

Does 'R published database study' require 'R analyzed database' (or a similar event)? Y End p. 160
Does 'R published database study' require 'R Joined faculty' (or a similar event)?
Does 'R published database study' require 'R funded externally' (or a similar event)?
Does 'R published database study' require 'R received doctorate' (or a similar event)?

A researcher (or someone under the researcher's supervision--a similar event) has to analyze a database in order to publish statistics and interpretations based on the dataset. Publications are not dependent on other career events of any kind since both students and non-academic researchers publish.

ETHNO notes:

Observe that ETHNO asked no questions about the relation of 'R published database study' to 'R read', 'R funded locally', 'R contributed funding', 'R collected database', or 'R programmed database'. The answers to the relational questions concerning these events were derived syllogistically from the fact that publishing a database study requires analyzing a database.

R contributed funding Con

ETHNO notes:

This repetition of a prior event was entered by typing its abbreviation in order to place it in the event series. The program asked no questions since the structural position of the event was already known.

Appendix 9.3: ETHNO series analysis

After I selected the Analyze a Series option from ETHNO's menu and entered the name of the file, the program displayed the logic-diagram on the screen with the abbreviation for the first event blinking to indicate that the event was 'occurring'. I pressed the ENTER key, and the blinking item changed to the abbreviation for the second event. I proceeded through the first seven events this way without ETHNO noting problems.

R read
R funded locally
R contributed funding
R collected database
R funded locally
R programmed database system
R analyzed database

The eighth event, though, caused an interruption.

R received doctorate

Problem! Conditions for this event are not fulfilled.

Getting a doctorate requires reading, and ETHNO assumed that the researcher's reading got used up when he contributed funds for research because making a contribution also depends on reading. The default assumption is that events deplete their prerequisites.

Having stated the problem, ETHNO began possible solutions.

Is 'R read' not required for:
Rec 'R received doctorate'
End p. 161
Con 'R contributed funding'
Enter abbreviation of non-dependent event.

ETHNO suggested two ways to change the logical structure so that a problem would not exist here. If reading is not required for getting a doctorate, then it would not matter that the reading was used up--getting a doctorate does not depend on it. On the other hand, if reading is not required for a funding contribution, then the contribution would not have depleted the reading, and the reading still would be available for getting a doctorate.

I decided that I made no errors in specifying the consequences of reading, so neither of these solutions was acceptable, and I skipped to ETHNO's next offered solution.

Can 'R contributed funding' happen without depleting 'R read' (y or n)? Y

This indeed is the solution, and the rest of the implementation follows below. First, though, consider ETHNO's final suggestion.

Rea 'R read' might have happened--unrecorded--just before 'R received doctorate' (y or n)? Skipped.

Maybe the original reading got used up by making a contribution, but then some more reading was done in order to get a doctorate. If so, then ETHNO could insert another reading event to correct the data record. This is not the desired solution in this case, but the idea is applicable in some cases.

Now, returning to the actual solution: after I typed Y ETHNO wanted to know what event does deplete reading, and it offered all the other consequences of reading as possibilities.

'R read' Is depleted by 'R received doctorate' (y or n)? N
'R read' Is depleted by 'R verbalized formulation' (y or n)?
'R read' Is depleted by 'R published methodology' (y or n)?

I answered 'no' to all of these questions because in my opinion reading is depleted only by senility or death--events which are not in the model.

To allow that some indirect consequence of reading might be what depletes reading, ETHNO next asked for the abbreviation of the depleting event.

Identify event that DOES deplete 'R read' (For none, enter Act.) Enter the abbreviation: act

ETHNO allows for the possibility that none of the recorded events acts as a depleter and tells how to escape from specifying a depleting event, and that was the option I chose.

Even if no recorded event depletes reading, ETHNO allows that you may want to change the structure at this point and incorporate a new event that does deplete reading.

Add a depleting event (y or n)? N

I declined the offer. Thereupon ETHNO accepted the idea that reading is a non­depletable event.

The program started over and ran through all of the prior events in the series (just to make sure this solution did not create problems elsewhere). Then ETHNO proceeded to the next event, and the next interruption occurred. End p. 162

R funded externally

Problem! Conditions for this event are not fulfilled.

This is the same kind of problem as before with the same kind of solution, so I simply list the ETHNO questions and my answers without further commentary.

Is 'R received doctorate' not required for:
Joi 'R joined faculty'
F1n 'R funded externally'
Enter abbreviation of non-dependent event.

Can 'R funded externally' happen without depleting 'R received doctorate' (y or n)? Y
'R received doctorate' is depleted by 'R joined faculty' (y or n)?
Identify event that DOES deplete 'R received doctorate' (For none, enter Act.) Enter the abbreviation:
Add a depleting event (y or n)? N

The next event caused no problems.

R joined faculty

The event after that was a repetition of  'R read', and it caused another interruption.

R read
Problem! This event undepleted since last occurrence. Can this event be repeated without depletion (y or n)?

ETHNO assumes by default that events are not repeated unless used up by consequences, but in this case we have specified that reading is not used up by any research events. ETHNO's suggested solution was that reading might be repeated whether it is depleted or not; and that was ETHNO's only offered solution in this case. I accepted the ETHNO suggestion, and the program continued with the analysis.

No additional problems were encountered with ACT events prior to 1966, the arbitrary stopping point for this illustration.

R published database study
R contributed funding