Close this search box.

An Emotional Systems Research Methodology

As it stands today, there is virtually no formal, data-driven support for the broad claims of Bowen theory. Such a data-driven basis would be comprised of well validated explicit predictive models published as experimental research in peer-reviewed journals outside the Bowen Network. It is a primary aim of Alaska Family Systems to identify and work toward solving this problem.

This state of affairs is understandable, as the theory was developed between the 40’s to the 70’s when psychiatric research and publication was more informal. But however unfortunately, this problem took years of training and a 1000-page theoretical dissertation to identify and articulate. In terms of “a science of human behavior” (Bowen, 1988, pp. 354), this is is not an ideal state of affairs.

Terms like “science” and “research” are part of the Bowen theory vernacular. But while new trainees often ask Bowen Center faculty about the scientific validity of the theory, there is virtually no published experimental literature to refer them to. Terms like “solid” and “a theory from validated and provable facts alone” (Bowen, 1978, pp. 352) portrays the theory as scientifically established while the theory remains unvalidated and the original research rife with methodological problems in the modern era.

How then to rectify this problem?

One of the functions of the Family Diagram App Seminar is to generate principles for research of emotional systems. The seminar does so by providing a place to observe researchers toiling through real, inductive research (See article: Induction and Family Diagram). The app seminar and app itself provides the necessary technical constraints to organize such efforts.

An emerging qualitative research methodology for the family emotional system is one product of the app seminar. As with any formal research methodology, it clarifies the mundane, repeatable parts of a research method so that researchers can devote more energy to their research questions and less to sorting out how to conduct the research.

This article describes the terms and processes of the methodology such as they stand today. This is not a final product but a work in progress. The methodology does not yet have a name, and many of the terms remain provisional.

The methodology has one aim: To argue to a good faith critic that a particular emotional process has occurred in a family based on nothing but what the data indicate. This simple statement has many parts to it, all of which are covered in the sections below.

Please feel free to submit any critique or feedback to

Goal of the Methodology

“Arguing” An Opinion

First, the goal is to argue a theoretical formulation using pure induction. There is zero exception to this most basic rule of scientific proof. There can certainly be many phases of speculation and allowing biases to rein. But at the end of the day, the goal is not to merely present conclusions but to make an inductive argument, then offer that argument up to critique from the scientific community.

In other words, the researcher recognizes that their theoretical interpretation of their data is merely an opinion, and their goal is to subject that opinion to good-faith critics for their best effort to disprove it; to really sick the hounds on it.

The status quo in family research today is to present a theoretical formulation as a fact without evidence and without asking for ruthless critique from outside the Bowen Network. The underlying principle in the proposed methodology is that whatever prediction stands up to the highest level of scrutiny must be solid. This is the highest aim of scientific research.

As an example, that a child was the “focused on” will often be presented as fact without evidence. What is the exact definition for “focused on” used in the research? Where are the examples in the observational data set that do fit this definition, and where are the examples that do not? What process of critique has this term and concomitant examples been subjected to? Does this process include validation of inter-rater reliability from researchers who do not share a bias for Bowen Theory?

If these problems are not fixed in family research, then hypotheses will never tested and biases will perpetually be confirmed. This methodology is intended to correct this problem by setting a standard for rigor in family research.


Central to the ability to critique is clear measurement. Without a way to measure, or at least count, data, there can be many opinions about what is observed.

For example, it is straightforward how to count siblings. Virtually all claims of parentage can be assumed correct. But exactly how does one know that “anxiety” was occurring in a single person? What about “anxious attention”? Or “functioning”? Or “differentiation”?

Was each datum recorded as a number with a mechanical instrument? Or was some qualitative scale used? Or was it a subjective impression of an increase or decrease? Or direct eyeball observation based on some operational definition? If so, what is the definition? Is the definition appropriately generic or does the researcher change it ad-hoc to fit each case? Or is the recording of anxiety mere self report?

The above list of questions walks down a spectrum of precision from precise mechanical measurement to self report. The less the precision, the less power an argument can have. The more precision, the more a good-faith critic has to work with.

In psychometrics, the study of measuring psychological constructs, the degree of imprecision in an instrument or method of recording data is called measurement error. The more vague a term or concept is, the more measurement error there will be when different people use it to record data. The more measurement error, the lower the predictive power of the term.

A goal is to decrease the amount of measurement error in the terms and concepts of Bowen Theory.

Data Appropriate for Emotional Process

The methodology presented here is for arguing that a particular emotional process has occurred in a formal, scientific manner. It is not for merely telling a story about a family. There are basic requirements for data that can describe an “emotional process.”

To start, an emotional process occurs over time. Therefore, data that describes an emotional process will always be a time-series data set. The app enforces this through the timeline as the core of the data model. The timeline consists only of events, which always have a date, factual description, and arbitrary variables which are determined by the idiosyncrasies of the research question at hand. Example variables are: anxiety, symptom, and others. This methodology therefore requires that researchers get their data organized on a timeline.

An emotional process also consists of shifts within or between individual people. Shifts must be recorded at a particular time in the timeline or they cannot be included in conclusions about an emotional process. Shifts can be either an increase or decrease in a variable in lieu of an absolute numerical measurement for the variable.

Shifts can pertain to individuals or between individuals. Because the term “emotional process” assumes that all shifts are reciprocal, shifts that pertain to individuals are tracked as markers of functioning to be tested through correlation in time with shifts recorded with other individuals.

In Bowen theory, an emotional process consists of a chain of automatic/instinctual responses to interpersonal events. In other words, theory suggests that some automatic responses modulate anxiety, and anxiety in combination with the level of functional differentiation modulates symptoms (Havstad & Sheffield, 2018). Therefore, data that describes emotional process must capture shifts in anxiety and functional differentiation in terms of automatic/instinctual responses to interpersonal events.

Avoid Over-Interpretation

The inductive mode of arguing an opinion is data-driven by definition. An “argument” is presented by:

  • first presenting the raw data
  • then presenting the interpretation of the data
  • then asking for critique

The interpretation must only interpret the data presented. In other words, it is the responsibility of the researcher to restrict their interpretations to what the data indicate. The tendency today is to interpret, or draw conclusions, well beyond what the data indicate on their own. Example statements that over-interpret given data available today are, “this child received the majority of the parent’s focus,” “the regression in this family began with this grandfather,” and “the family is an emotional unit.”

There is nothing inherently wrong with over-interpretation. In fact, it is assumed that researchers will tend to over-interpret. But the researcher should put forth every effort to minimize over-interpretation, and to correct over-interpretation by pulling in a good-faith critic. The good-faith critic functions to correct over-interpretation by representing the bias of under-interpretation. It is the responsibility of the well-trained researcher to preemptively identify over-interpretation and label it what it really is, “speculation.”

Good-Faith Critique

What makes a good-faith critic? Such a critic should not share the biases of the researcher, and should be orientated toward discovery of the reality of the situation. This process of critique is an extremely nuanced social process that operates via the mechanisms of collective intelligence (Mann & Helbing, 2017).

Research on collective intelligence can help explain how the social aspect of science works to help converge on the truth. Humans, bees, even cells bring relative dissonance to where they would direct energy for a given common problem. This relative dissonance can be called “bias.” It is the somewhat chaotic, instinctual mixing/synthesis of these biases that gives rise to collective intelligence. A general rule is that the average of a set of disparate biases is often more in line with objective reality than a single expert bias (Lorenz, Rauhut, Schweitzher, & Helbing, 2011).

Therefore, so long as the objective really is to discover the reality of a situation as opposed to merely confirm a bias or gain social status, researchers must seek out opposing biases to their own to reveal weaknesses in their argument or model.

To be effective, such arguments based on data must be presented in clear and concise form. As of today, the status quo for Bowen Theory is to present a rather long case study. These case studies exceed the attention span for a critic that does not share the biases of Bowen Theory itself. A good target to communicate a case using minimum sufficient evidence is in less than a few minutes.

Generalized claims must include more than one sample, not a single case study. The interpretation of the data itself should take less than one minute. New technical terms, otherwise known as “jargon”, must be defined through, and confined to, specific examples representing the definitions used in the argument.

For example, if the technical term triangle is used in an argument to a person not familiar with Bowen theory, it should be defined in terms of a few single-statement examples. Such examples may be, “Person A told Person B that Person C is bad at their job and something should be done about it,” “Person B restates the claim that Person A is the best thing that happened to the company every time Person B meets with Person C,” etc.

Most case presentations in Bowen theory today include a single hypothetical emotional process in a single family. To make a generalized claim that such emotional processes exists across families, more than one family sample is required. For example, to claim that a phenomenon called an “emotional shockwave” exists “in the human family” in general, multiple examples of emotional shockwaves must be provided, properly critiqued, and the observations properly replicated by other good-faith researchers.

This methodology is aimed at solving this final point; to organize a concise, data-driven argument about an emotional process for provision to a good-faith critique. The definition of “emotional process” scattered about the Bowen Theory literature dictates a certain kind of data that is too complex to organize concisely without technology.

This is the primary function of the Family Diagram app. The methodology is aimed at the tedious, human-driven part of using that technology to record, organize, and present the data.

The Methodology

The methodology consists of two major phases, assuming a researcher is starting with information about a family scattered among disparate sources. All of these steps reflect standard principals of qualitative research ultimately aimed at testing hypotheses. There are no new research principles created here, only new terms that seem to help family researchers argue their inductions on emotional process to good-faith critics using the pure, data-driven, inductive mode.

The two phases in the methodology reflect the two major phases in all scientific research aimed at prediction; qualitative and quantitative research. Broadly speaking, qualitative research is creative and quantitative research is reductive. Qualitative research begins with minimal assumptions and generates new data from raw observations. Quantitative research then takes that data and reduces it into conclusions.

The two steps in this methodology are called Collecting and Massaging. Collecting is creative while Massaging is reductive.

Phase 1: Collecting

The first phase of the methodology is creative. This means it is creating a body of raw observations and putting them all in one place. In terms of the app, this means generating a timeline of factual observations.

Many people get hung up on this first phase. It appears difficult for many folks experienced with Bowen theory to collect all known information and put it in chronological order. Identifying and articulating this difficulty is important because it reveals the state of the art in family research and points to the work to be done.

This phase has taken more than a year for some for a small amount of information on a single family. To draw contrast, this phase might take a couple of weeks for a typical masters degree candidate working under supervision in the sciences on an existing data set similar to what most family researchers have on their own family. Nevertheless, this step is simple enough that a family researcher can begin by simply organizing their information on paper or in a spreadsheet.

In addition, it appears far more difficult for many to differentiate between fact and opinion on that chronological data set. This is important to point out when the goal is to generate a timeline of facts alone.

To compare, a typical doctoral candidate in psychology will use a quantitative methodology which by definition eliminates confusion about what is fact and opinion insofar as the statistical facts it produces are concerned (though how those statistical facts are interpreted are almost always opinion). An atypical doctoral candidate in psychology that uses a qualitative methodology will inherently blur the lines on fact and opinion somewhat because qualitative methodologies always impose some degree of opinion on raw observations through coding.

Collecting data on “emotional process” appears to blur the lines between fact and opinion well beyond typical qualitative methodologies in doctoral dissertations. How, then, can one make accurate claims using a dataset full of opinion and unsubstantiated “fact”?

The first phase of this methodology, Collecting, is intended to solve this problem. The product of Collecting is a relatively generic, factual dataset on emotional process that can be used for many purposes. Such a data set is organized to allow for random access, a term from computer science indicating that any one datum in a medium can be accessed at any time with maximum efficiency.

Collecting is separated into two steps; Gathering and Parsing. Gathering is analogous to picking all the blueberries and putting them into a pile on the ground. Parsing is analogous to cleaning the berries and organizing them into bags to be processed. At this stage the berries can be processed in one of many methods resulting in very different berry products.

In other words, Collecting is the generic phase of processing data that must always occur regardless of how the data will be used. Therefore, the product of collecting is a generic, random access dataset on emotional process. If a goal was to crowdsource general purpose data on emotional process, the crowd-sourcing effort would consist of this first phase of Collecting.

Difficulty: Red Pill or Blue Pill?

This basic first phase of Collecting has proved difficult for most who have been involved in research under Bowen theory for some time. Observations from individual consultation reveal that decades of involvement with Bowen theory does not equate to being able to systematically organize such a dataset. There are some formal research skills that need to be developed in almost all cases.

However, those who complete just this first phase with one family case demonstrate clarity about what is and is not substantiated by the evidence.

Interestingly, this clarity seems to come as a “red pill” realization. “Red pill” is a reference to the wonderful and archetypally rich scene in The Matrix where the film’s hero Neo must choose between ignorance, a blue pill, and enlightenment, a red pill. The catch is that enlightenment comes at the cost of having to deal with realities of increased insight.

This realization does not seem to be explainable to those who have not put the time and toil in to discover it on their own. At least not yet.

As a note, Collecting will likely not be straightforward for those not scientifically trained and whose professions primarily deal with people, such as therapists, clergy, and organizational consultants, etc. It will be straightforward for most not scientifically trained people whose professions deal with things, such as engineers, mathematicians, economists, etc. The latter group tends to depend on research and development that must precisely prove its ideas in order to proceed, while the former does not.

This methodology is perhaps more intended to clarify research principles for emotional process for the former group.

Sample: One’s Own Family?

The entire methodology will come easier to those who research families other than their own. There is a tradition of “researching” one’s own family in Bowen theory, and that anyone can be a scientist (M. Kerr, 1988) in this regard. Such “research” is integral to an effort at differentiation of self. It can certainly follow scientific principles in an effort to differentiate one’s thinking function from one’s feeling function in the context of one’s own family.

Further, anecdotal report, including report by this author, suggests that such “research” as part of an effort at differentiation of self can contribute increased objectivity and new thinking for formal research. One certainly has easier access to one’s own family but not other families.

However, such “research” on one’s own family has never proved productive enough for formal experimental research. If such research exists then this author would certainly like to hear about it (submit responses to Further, it appears as though research on one’s own family and formal experimental research has never been formally differentiated.

Therefore, if the goal is to draw substantial conclusions, for example for experimental publication outside the Bowen Network, the suggestion here is to use families other than one’s own.

Step 1: Gathering

The first step in Collecting is to gather all of the available information in one place. Many researchers start with information scattered around all different media, e.g. hand-written notes, audio recordings, photos, legal documents, diagrams, human memory, etc. Some researchers will go out and record new data through real-time observation.

Gathering is analogous to going into the wild and recording raw data on baboon movements or bee selection. Or retrieving existing datasets from previous researchers or archives. In conventional field work this step might be called “recording” raw observations or “collecting” raw data. The difference here is that information about family might come from many different sources. Gathering is about putting data from all sources in one place.

Gathering data into one place allows the researcher to avoid interrupting their current train of thought to go down a rabbit hole to retrieve a piece of information. For example, the only evidence of a grandmother’s memory of the negative feeling orientation between two sons may only exist in a 30-year-old letter. The fact that this grandmother reported her opinion of this relationship at that time needs to be recorded somewhere along with all other information for the case. The goal is to move beyond the only way of accessing this information being physically picking up and reading the letter itself.

Gathering involves putting all of the information from different sources in the same place, on the proverbial researcher’s desk. The standard for this is quite strict, it means literally recording them into a single medium in a single format. What format is used is not as important as it being one format.

Initial attempts at this can be on paper, spreadsheet, or word document. One way of organizing it is to get it into chronological order. Experienced or tech-savvy folks will enter it strait into the app as events. Either way, getting all the information into a single format is the goal.

Step 2: Parsing

The next step is to parse the gathered information into the app’s timeline. The product of this step is a heap of data. In computer science, a heap is an organized set of data that allows for random access. As previously discussed, random access is a term from computer science meaning efficient access to any one datum in a medium at any time.

In other words, a heap is an blank slate for scientific analysis.

For the most part, parsing means getting all the information into the app’s timeline. This means going through each datum one at a time, assigning it a date and time so far as this is possible, and converting opinion to fact via the functional fact tool. The result is a diagram and timeline of facts alone.

The first step is to organize all known facts into chronological order. Chronological order means that data are organized into events with timestamps. Timestamps are a single date and time combination.

People are added to the diagram along with what is known about their birth, adoption, and death timestamps. Pair-Bonds are added between people along with what is known about their bond, marriage, separation, and divorce timestamps. All of these dates that have specific fields under people and pair-bonds will be automatically added to the timeline.

Other items on the timeline can be added as arbitrary events. Events are always added under people or pair-bonds. Event titles are ideally less 5 words so that are readable on the timeline.

Some of the data will have precise timestamps, and some will be quite vague, for example anywhere within a 10-year range. The goal is to get as precise as possible while finding ways to mark some dates as being uncertain. One imperfect but useful convention for indicting imprecise timestamps is to set the month and day to new years day with an approximate year.

The term “parsing” comes from linguistics and computer science, and means separating information into known syntactic chunks. Here, it means either A) picking apart statements from a transcript into factual events on a timeline, or B) picking apart events on a timeline into events or symbols on the diagram. Most notably, parsing means identifying opinions and converting them to fact using the concept of a functional fact, described in the next section.

Functional Facts

The term “functional fact” was coined by Murray Bowen “mainly…to facilitate research” (Bowen, 1978, pp. 360). A functional fact is an idea that functions as a fact for a person because that person uses it to make decisions. It is that person’s current working model of something in the real world.

For example, it can be a fact that a person had a dream, but the content of the dream may not be factual (Bowen, 1982). Similarly, it can be a fact that person A thinks person B is “bad,” but whether or not person B is “bad” is not really a fact. While these ideas are common in Bowen theory vernacular, the precise application to inductive research has not been clarified.

This methodology makes a simple but important contribution to Bowen’s idea of a functional fact for research; that a functional fact occurs at a specific point in time, and that adding timestamp allows it to be added to a timeline. This allows opinions to be recorded on a timeline of facts alone and for these opinions to contribute essential information about an emotional process.

Continuing the above example, it is a fact that person A reported that person B is “bad” at a particular point in time. The report of this opinion would be recorded in the app’s timeline, clearly laying out A) when the opinion was reported, B) whose opinion it was, C) who the opinion was about, and D) who the opinion was reported to. This method captures all applicable triangles involved in the opinion, including those involving the researcher. Recording events like this one includes the “feeling orientation” (Bowen, 1978, p. 438) or emotional valence, of one person to another at that point in time.

Statements or actions indicating emotional valence from one person about another helps reveal the inside and outside positions of triangles.

The first step of parsing functional facts is identifying items in the timeline that are still only opinion. Experience in individual consultation shows that it can be exceedingly difficult for many to do this, regardless of level of experience with Bowen theory. This is simply a skill. It comes more naturally to some than others. It is not clear how to teach this skill other than by going through many examples in individual consultation. However, once a person obtains this skill, they seem not to lose it. Some never obtain it.

The product of parsing is a timeline of facts alone. Data on family, particularly behavior as emotional process, includes many feelings and opinions by definition. Parsing converts the many feelings, impressions, and opinions into facts on the timeline or symbols on the diagram with discrete timestamps or start and/or end timestamps that end up on the timeline.

Emotional Process Symbols

In the early days of the app and associated method for app consultation, virtually all users asked the same question; How do I show X, Y, Z on the diagram? Usually “X, Y, or Z” was some emotional process term from theory. This initially posed a challenge, as the app only had so many symbols and adding the requested symbols introduced a rats nest of conceptual complexity.

This problem required quite a bit of thought. The result was the most important design principle in the app; that users are expected to enter the evidence for whatever it is they want to show, not to merely provide another tool to illustrate their conclusions. This decision made the small set of relatively consensual emotional process symbols sufficient.

The entire methodology stems from this one app design principle and has revealed the status quo for research Bowen theory in a way never revealed before.

As a tool for recording emotional process, the app does provide a small set of relatively consensual emotional process symbols. It is certainly possible to use these symbols to illustrate one’s conclusions as opposed to the evidence for those conclusions. This is usually done by adding symbols without dates or notes.

However, the preferred method for adding emotional process symbols is to add the symbol along with the evidence for them. The evidence for them consists of A) either a discrete timestamp or start and/or end timestamps, and B) the associated raw observations added to the Notes field for the symbol and/or symbol’s events. The app automatically adds an emotional process symbol’s timestamps to the timeline as events. Adding the raw observations to the Notes field allows a critic to decide whether they agree with the coding for that symbol. It also allows formal definitions for those symbols to emerge as a collection of examples.

Including the raw observations being coded as a symbol is essential to the inductive mode. For example, consider that person A reported to the researcher on Jan 16th, 2020 that they had an argument about how to mix margaritas with person B around noon on May 5th, 2012. This could be coded as a conflict symbol between person A and person B with a timestamp of 05/05/2012 12:00pm, and Notes field that reads:

Argued with person B about how to mix margaritas. (Reported by person A to researcher on Jan 16th, 2020)

This note captures the raw evidence for the decision to code this as “conflict,” and includes all information about the triangles involved including the triangle with the researcher. A critic can then decide whether they agree that this should be coded as “conflict,” “repression,” “trauma,” or some other theoretical idea.

Most importantly, this practice will force researchers to confront the pervasive imprecision in technical terms from Bowen theory. What is conflict exactly? How about projection? Or distance? Or child-focus? What is the evidence that any of these are occurring? What operational definitions were used to record them? Would others agree with that definition? Would others agree with a particular code that depends on that definition?

The app forces researchers to confront these problems head on. If the decision were to simply add a symbol for every idea requested by a user, then the app would become just another genogram where the symbols had no integrative theory at all (Butler, 2008).

Thus, the step of parsing may produce few emotional process symbols. Those symbols that are added to the diagram will have timestamps and associated raw observations added to the notes field of the symbols. One exception is the hypothetical baseline emotional configuration, described below.

Hypothetical Baseline Emotional Configuration

Theory implies that every family has a baseline emotional configuration, or baseline emotional process. A baseline emotional configuration is the inside and outside positions of the family, use of the four anxiety binding mechanisms, etc.

The baseline is the rule upon which acute, episodic variations occur. Acute fluctuations hypothetically occur in response to shifts in anxiety. They consist of a relatively transient emotional process that functions to contain the anxiety.

Currently, the state of the art in clinical practice or family research is to draw the family’s baseline configuration on the diagram using somewhat consensual symbols from Chapter 10 in Micheal Kerr’s 1988 book Family Assessment (Bowen & Kerr, 1988). This is certainly a contribution to the current state of clinical practice. It is likely the only formal way to visualize a process of emotion or feeling over time in the field of psychotherapy. It is certainly the only way that is also associated with an integrative theory of human functioning.

However, in this inductive methodology this “baseline emotional configuration” is only hypothesis. In fact, this methodology implies that almost all family diagrams today are hypothetical. The evidence for this is that there is almost always insufficient evidence of a baseline configuration exists in a single family case. Further, the existence of a “baseline emotional configuration” phenomenon has never been formally collected, validated, and published.

This methodology addresses this problem by distinguishing a hypothetical baseline emotional configuration from a hypothetical episodic fluctuation over that baseline. It does so by formalizing the collection of data to support or refute the existence of either. A hypothesis would be that evidence for baseline and episodic configurations will emerge from the data if collected and analyzed systematically.

This would be one example of an induction that might be argued to a good-faith critic using this methodology.

In the meantime, a hypothetical baseline can be drawn onto the diagram by adding emotional process symbols without timestamps. A tag called “Hypothetical Baseline” can be created and added to each emotional process symbol to make it easy to show only the symbols depicting the hypothesis. As more data comes in, the hypothesis can be modified to fit.

Phase 2: Massaging

The second phase of the Methodology is the analytical phase. It involves the careful application of scientific principles to draw conclusions from the systematically organized data.

The data to be analyzed was prepared and normalized in the first phase of Collecting. Now that the data is all in the same format, it is possible to start combing through it to see what patterns emerge. At the end of the day, this phase is meant to be the fun part of family research.

The primary product of this phase is an induction or set of inductions (See the article: Induction and Family Diagram). Induction is the act of explaining a pattern. An induction is not a fact, but a best guess at a general rule for a sample of observations.

Inductions are meant to be tested against further data and modified to fit.

It is suggested to take time to clear one’s mind of the data gathering process before proceeding with the second phase. This will help remove biases developed from the collection phase to evaluate the data what the dat shows and does not show as clearly as possible.

This phase takes time and reflection. It is usually necessary to return to the same data again and again, looking at it different ways each time. Experience has shown that it is especially difficult to see past one’s subjective impressions of human behavior.

The Family Diagram app was designed to facilitate this act of returning to the same data again and again in different ways. It allows one to slice a heap of chronological data into different cross-sections (i.e. via tags) to see new relationships in it. Even the unorganized timeline can be reflected upon many times to see what was not seen before.

Now that the app removes the technological barriers to plainly laying out the data, the biggest bottleneck to making sense of it is the observational and conceptual capacity of the researcher.

Looking for patterns requires the highest level of systems thinking available to the researcher. This is a mental shift from the first phase. If the first phase was successful, then this mental shift can happen more completely because the data is completely collected and organized.

If the first phase is not successful, then the researcher will be caught up in problems of the first phase while trying to proceed to the second phase. This undifferentiated mode will be less efficient, more error-prone, and more vulnerable to confirmation bias. If this is the case, it is better to return to the first phase to ensure that all original materials and notes can be left behind in favor of the new, unified heap of data.

Step 1: Pruning

The first step in the analytical phase is to prune unnecessary data. This requires evaluating each datum in terms of its relevance to the research question. Theoretical a priori, or assumptions from theory, are imposed on the data to decide what is and is not important for the research question.

Pruning the diagram can involve deleting people that don’t seem pertinent. Folks who are used to making a single diagram for their entire family will likely have added more people than are necessary to answer the research question. Similarly, folks who are new to this methodology will likely have entered superfluous events.

Some events on the timeline may be removed if deemed impertinent. This is because a goal of the methodology is to see what one has not seen before. This requires initially adding more data than seems pertinent. But once sufficient reflection has occurred, many events bubble to the surface as more important and the rest can be removed.

Because the final goal of the methodology is to present a minimum sufficient argument, pruning must remove all superfluous information from the diagram file.


It probably makes sense to first make a copy of the original diagram file whenever data is removed. The heap diagram produced in phase 1 is a general purpose dataset and therefore should be saved in case it can be used for another research question.

Making a copy of the diagram on the hard drive for a specialized use is called forking. Forking is a term from computer science that indicates a separate copy of something has been made to be maintained separately from the original. A fork is like a creation of a new genetic lineage, never to return to its ancestral line.

Like synaptic pruning in the brain, pruning data specializes a diagram file for the project at hand. Who knows what completely different research question this heap diagram may be used for in the future?

File storage is cheap and backups are easy. Backup early, backup often. Collect, fork, then prune and massage.

Free users of the app can only edit a single diagram and so cannot copy it, so forking is not an option for them. Users with a professional subscription can fork diagrams because they can manage multiple diagram files.

Sizing People

The more people there are on a diagram, the harder it is to fit everyone into view. This can create diagrams where the people are so small that their names, ages, and diagram notes are unreadable. Removing impertinent people, making people with less significant roles smaller, and re-arranging people can allow the more important players and their details to be shown larger.

The story of most emotional processes involve a small enough number of people to facilitate a diagram where the main players are big enough for their details to be read.

Step 2: Massaging

The second step of massaging the data into a clear, concise argument requires the most scientific precision. This involves using the app’s features to clearly depict the evidence for each induction. The goal is to provide a clear, comprehensive, concise, minimally sufficient, and efficiently presentable argument.

Tagging Events

The primary way to argue an induction is through tags. Tags allow you to categorize events into meaningful groups. Tags are extremely flexible because there are virtually infinite ways to slice this kind of data. Deciding what tags to create and which events to tag with them can vary widely between research questions and requires critical thinking on the part of the researcher.

One way to use tags is to create a separate tag for each point when arguing an induction. As an example, reflection on one clinical case’s heap revealed three hypothetically significant periods on the timeline:

  • The first from 30 years prior where a sibling of the client was jailed for committing a murder
  • The second where an emotional process played out in the client’s early adulthood
  • The third where the client reestablished contact with a dying mother after years of cutoff.

These were the only clusters of events that the researcher decided were descriptive of the emotional pattern in this family. The researcher decided this by reflecting on what the data could show and nothing more.

First, one tag was created for each period. Then each event that fell within a tag’s time period was tagged appropriately. This is worthy of being repeated: Each event associated with a tag’s time period was tagged appropriately.

Experience shows that one must think of events as being tagged to understand these essential app features. That means tags are added to events, not the other way around.

This is equivalent to how hashtags are added to twitter posts, and how twitter posts are not added to hashtags. This allows you to search for all twitter posts containing a particular hashtag. Similarly, adding tags to events in Family Diagram allows you to search by events containing that tag.

In the example, the resulting list of tags looked like this:

  • #1: Murder, Jail, Start of Calm Period
  • #2: Out of Jail, Triangle Vignette
  • #3: End of Cutoff, Proximate Emotional Process

Tags can be named whatever you like. The #1, #2, #3 were added to the beginning of each name to A) keep these tags in order since tag lists are automatically sorted alphabetically, and B) serve as a reminder as to the order that they should be presented for the theoretical formulation to make sense.

Each tag represented the minimum necessary evidence to make the salient point for that tag’s time period. #1 had about 4 events, #2 had 8 events, #3 had about 12 events. It is important to keep the number of events in each tag to a bare minimum for an audience to be able to absorb it.

All in all, one should aim to spend only a few minutes arguing all the evidence. Covering a very large data set over 20-45 minutes is too much for a good-faith critic who does not share the bias of Bowen Theory to hold attention. Presentations that take this long will flood the mind with so many details that even the best systems thinker will be unable to maintain a broad, holistic view of the thesis.

If a diagram has three tags then one should probably only point out about 3-4 events per tag in a minimum sufficient presentation. If the presenter is clear on the evidence, clear on the inductions/interpretation, and has pruned and massaged the data well, then it should be possible to argue the entire thesis in under a few minutes.

Here are other example uses for tags:

  • Shifts in anxiety, shifts in a symptom
  • All birth events in one generation, all death events in another
  • Shifts up in functioning in one person and shifts down in functioning in another person.
Coding Symbols

As you reflect on the data, you may decide to convert events from the heap into emotional process symbols. For example, you may decide that one or more events constitutes evidence for use of a particular anxiety binding mechanism like conflict by the two people in a pair-bond. You might add a conflict symbol between the two people and give it a start date of the first event and an end date of the last event.

Or, if you are converting a single event for an argument that took place, you may create a single conflict symbol with a discrete timestamp. You can do this by leaving the “Is Date Range” checkbox unchecked and entering the timestamp of the event so that it appears only at that specific date and time.

Comparisons with the Graphical Timeline

The tabular timeline is more a tool for managing or counting data than for temporal analysis. The graphical timeline is intended to analyze the temporal relationship between different categories of events. It does this by A) displaying events on a timeline with accurate spacing, and B) stacking events by tag in horizontal rows.

For example, you could add tag events as shifts in symptom and tag other events that hypothetically impact the expression of symptoms by modulating anxiety. You could test this hypothesis by adding events and then tagging these events with as little thought to the relationship between symptom and anxiety as possible. Then you could pull up the visual timeline, search by “Symptom” and “Anxiety” tags, and see if there is any noticeable relationship between the two in the horizontal rows drawn for each tag.

Null Results

It is likely that research on Bowen Theory following this, or any other inductive methodology, will turn up nothing but null results for a while. A null result is one where it was not possible to answer the research question. This can happen for a number of reasons, including insufficient data, confusion in the definition of terms, and others.

Evidence for null results being a likelihood includes but is not limited to; lack of any existing explicit predictive models that have been validated, the ease in which one finds contradictory applications of technical terms such as conflict, projection, reciprocity, child-focus, anxiety, and differentiation of self, and the predictability of the “red pill” moment described above.

A null result is not the same as a negative result, which is where the hypothesis was found to be false. A null result is where the question was unanswerable. Wolfgang Pauli coined the phrase “not even wrong,” to describe a theory that is untestable. A perponderance of null results when research held to a conventional scientific standard will be evidence that Bowen Theory is “not even wrong.”

This point about null results is not to dissuade researchers. Instead, it is meant to make some realities of the situation clear so as to better focus on the more difficult and important questions in the theory itself.

Notes on Presenting

The end goal is to present the argument to a good-faith critic. This requires A) successfully completing phase 1 and 2, and B) technical proficiency with the app. Gathering and Massaging are required for effective presentations, but not sufficient for effective presentations.

Practice Presenting with the Tech

It is important to practice presenting the argument until the tech becomes second nature. A small blip in a live presentation usually causes anxiety to skyrocket, and tech and an inhibited thinking system are not a good combo. This includes operating the app, switching between tags and views, whatever screen share functionality you are using, the combination of the two, etc.

It is important to have predefined failsafes if you can’t find a file or if you get into a view in the app that you are not familiar with. In the app itself this is the reset button which clears all search and zooms out on the diagram. For your computer, be sure about where your files are stored. Always keep backups.

Time Each Point

The audience has an attention span that is far more limited than the presenter assumes. Also, it is often difficult for presenters to get clear enough about a particular emotional process to present it succinctly. Part of this is a factor of differentiation. Part of this is a function of the amount of time spent with the data and amount of time practicing.

If your diagram is mostly organized around tags then it is important to practice reciting the details of each tag while clicking through them. Include numbers in the names of each tag to indicate the order. Make sure you have no more than 2-3 points for each tag.

The same applies if your diagram is mostly organized around views. Practice clicking through them and timing how long it takes. Know where the reset button is and what it does.


Bowen, M. (1978). Family therapy in clinical practice. New York, NY: Jason Aronson.

Bowen, M. (1982). Subjectivity, Homo sapiens, and science. Family Center Report, 4(2) 1–4.

Bowen, M. (1988). Epilogue: An odyssey toward science. In M. Kerr & M. Bowen, Family evaluation: The role of the family as an emotional unit which governs individual behavior and development (pp. 339–386). New York, NY: W. W. Norton.

Butler, J. F. (2008). The family diagram and the genogram: Comparisons and contrasts. The American Journal of Family Therapy, 36, 169–180.

Havstad, L. & Sheffield, K. (2018). Study of Weight Loss as a Model for Clinical Research: Shifts in the Family System and the Course of Clinical Symptoms. Family Systems, 13(2), 9-31.

Kerr, M. (1988b). Some aspects of systems thinking. In R. R. Sagar (Ed.), Bowen Theory and Practice: Feature articles from the family center report, 1979–1996 (pp. 23–32). Washington. DC: Georgetown Family Center.

Kerr, M., & Bowen, M. (1988). Family evaluation: The role of the family as an emotional unit that governs individual behavior and development. New York, NY: W. W. Norton.

Lorenz, J., Rauhut, H., Schweitzher, F., & Helbing, D. (2011). How social influence can undermine the wisdom of crowd effect. PNAS, 108(22), 9020–9025.

Mann, R., & Helbing, D. (2017, May 16). Optimal incentives for collective intelligence. PNAS, 114(20), 5077–5082.

Sign Up For News

Stay current.