Visitors: This page is out of date. It has been kept in place for archival purposes only. For the most up-to-date information please see the new index page.

Look: Finding patterns

When you look at stories and answers to questions about them, you can find patterns that provide insights. Note that I do not say you can find answers that provide solutions. Stories are not laboratory instruments and they cannot be perfectly controlled or measured. Any act of looking at stories involves interpretation which is guaranteed to vary from person to person. If you are looking for answers in stories you will not find them. However, what you will find, and lots of it, is things that make you think, things that surprise you, things that bring new perspectives to your exploration, and things you may not want to hear but need to hear.

There are three general ways of looking at the information you gather in a story project: looking at stories alone, looking at answers alone, and looking at stories and answers together.

Looking at stories

pine needle pattern

Groups of stories can form emergent patterns

The method I've used for looking at stories and found useful is inspired by grounded theory. The basic idea of grounded theory is that instead of constructing a hypothesis and then collecting data to test it, for some topics (and especially when dealing with interpretation of texts) it is better to start with the data and let the theory emerge from it, grounded in it so to speak. I'm not very concerned with the debates on whether the method really generates "theory" or is scientifically rigorous, but I've found the general ideas very useful for looking at stories. If you know enough about grounded theory to know there are two schools, you should know I like the Glaserian or inductive/emergent side of things for narrative work.

What I suggest to look at stories is the following. If it is at all possible, I suggest having two or more people go through this process independently and share their results to maximize the diversity of interpretation.

Soak up stories

First, read all the stories you have (and here I am assuming you haven't collected thousands). I like to think of it as "soaking up" stories like a sponge. As you read, highlight everything that "jumps out" at you. What jumping out means I hesitate to define, but it's sort of like things people said that seem important to them or to you or to the project's goals. I usually try to highlight phrases or clauses or sentences rather than single words. I call the things you select "elements", though grounded theory calls them "codes" (which to me, being a programmer, sounds strange).

You can select elements in a few different ways:

  1. by printing the stories and circling or highlighting words
  2. by opening a document and bolding words (this is the equivalent of circling, on the computer)
  3. by copying words from one document to another

Marking things by circling them on paper or bolding them on the computer is faster and more intuitive (less distracting), but it requires you to do a second step of copying what you've circled or bolded to the computer or other pieces of paper after selection is complete. Copying and pasting can feel artificial and goes more slowly, but then again you've got the text right there when you are ready to move on. I've done it both ways. For smaller bodies of text I think the circling/bolding method is far superior; but when I have a lot of text to get through or not much time to work with I give up and use the copying method.

Feel free to move back and forth through the stories and change what you've selected as you go. Sometimes you get halfway through the stories and then realize something you hadn't been selecting has been growing in importance in your mind; or conversely, you realize that something you have been selecting is not really very important after all. It's perfectly okay to go back and make changes. What you want to do is reach a saturation point where you feel that the things you have selected, when extracted from the whole texts, will capture enough of what is important about what people said that you can leave the whole texts behind without losing very much essential meaning.

Cluster

When you have finished reading the stories and highlighting elements, you should have a nice pile of short pieces of text: things like "he never came back" and "I felt abandoned" and "the winter was so long" and so on. Next you just take these things and cluster them, merging and splitting clusters as you go. I've done this clustering in three ways:

  1. by writing or printing elements on sticky notes or pieces of paper and moving them around on a table or wall
  2. by copying elements onto shapes in a presentation program and moving them around on the screen
  3. by pasting elements into columns in a spreadsheet

Which of these you do depends on how much time you have and how many elements you have. If you have a lot of time or few elements, you can painstakingly copy the texts to pieces of paper or bubbles in a presentation program. Having the freedom to move things around fluidly improves your ability to think intuitively, and physical space is better than virtual space for this sort of emergent thinking. But if you have a lot of elements or not much time, the spreadsheet method, though less intuitive, works. Just make a new column whenever you think you have a new cluster to place items in.

Remember to cluster based on unspecified similarity, meaning that you should put things together if they seem like they belong together without thinking about why you are doing it. It's important to stop yourself from categorizing things on one dimension like "strength of emotion" or "negativity", because that will give you a narrow interpretation of what people said (and it gives bias purchase). This is one place where I think grounded theory falls down a bit, because it calls the groups "categories" and the activity "sorting", which seems to me too directed for true emergence to take place. The act of clustering is one of using your intuitive connecting mind, not your analytical separating mind, to see relationships that you cannot easily articulate but that may be important nonetheless.

When you have placed all of your items into clusters, review the clusters to see if you want to change anything. When you don't feel any tensions in the group of clusters (i.e., nothing wants to split or join), stop.

Describe the clusters

Once your clusters are complete, describe them to yourself. What does each cluster say to you about what the people who told the stories feel? Also, look at how the clusters interact, as though they were members of a family or players in a game. Are there tensions or open conflicts between members? Are there some that reinforce each other? Are there coalitions? What does the whole body of clusters say?

Looking at answers to questions

ice triangles

Answers to questions can connect in interesting ways

If you asked questions about the stories people told, you will want to look at patterns in the answers. Here are some suggestions for doing that.

Single questions

The first and most obvious thing to do is simply to look at how people answered each question. What proportion of people said their story made them feel sad and what proportion said it made them feel encouraged? What is the distribution of answers across the five age ranges? And so on. Are the answers what you expected? What surprises do you see?

Pairs of questions

If you asked, say, five questions about each story, you will have twenty-five pairwise combinations between them. You can ask things like, when the storyteller was over fifty, were they more or less likely than otherwise to say their story was uplifting? Or, when the storyteller said their story was about trust, were they more or less likely than otherwise to say their story ended badly? And so on. There are many possible combinations to explore. If you have few stories or few questions you may be able to explore them all, but if you can't, choose pairs you think may be interesting together.

Larger patterns

Above the level of pairs of questions, you can ask broader questions. Here are a few ways to explore larger trends:

  1. Pretend the data is an omniscient observer of whom you can ask any question you like and get an accurate answer. If you came up with some ideal questions when you started planning the project, you can go back to those now. Also, reading some of the stories can help you think of questions you'd like to ask your omniscient friend. For each ideal question, see what sort of patterns you can find in your data that answer the question, even if only partially.
  2. Think about what your assumptions are about what you will find in the data, then see if you can surprise yourself. For example, you might see the question "How long have you been doing this sort of work?" and think "I'll bet the old-timers have more stories about how things used to be better." Then go and test that assumption, if you can.
  3. Take the single-question and question-pairs observations you've already made and start rearranging them. Do they say anything synergistic if you put them together? That might lead you to explore more patterns.
  4. When you make an observation, think about what questions it leads to in turn. For example, if you find out that most of the stories that made people feel "glad" took place over a year ago, you might want to look at all of the answers regarding stories that took place in the past year to see if you can see any other patterns that can explain the loss of "gladness". And so on.

Looking at stories and answers

Looking at stories and data together can arise out of either end of the partnership -- from looking at stories alone or from looking at questions alone. Say you are looking at stories and want to find out whether stories in which you noticed a trend towards fatalism happened in younger or older veterans, so you look at the answers to the age question for those stories. Or say you notice a group of five stories in which people said the story was "too dangerous to tell in public" yet made them feel "enthused" about their work, so you read those stories to find out what happened to trigger that interesting combination of answers. As you develop observations from either side of the equation you should be able to tack back and forth. The stories and the answers to the questions are really just two ways of finding out the same things, and they should complement your understandings of what the people are saying to you.

Who is listening to whom?

multicolored leaves

Different groups interact in story projects

The people who might be involved in a story project fall into three groups:

  • people who are being listened to: people in the "group of interest"
  • people who are listening: those who are running the story project (this probably includes you)
  • people who are listened for: usually the project sponsors, or people who have responsibility for what is being asked about and may be in a position of authority over it

These three groups might be all the same group, for example if you are helping your own community with its own project:

  • listened to: the community
  • listening: the community (and you in particular helping them do that)
  • listened for: the community

Or they may be three separate groups, for example if you have been asked by friends who own a coffee shop to ask their customers to tell stories about the shop:

  • listened to: customers of the shop
  • listening: you
  • listened for: shop owners

Or two of the three groups might be the same (e.g., you are the owner of the shop; or you are a customer of the shop).

Now you may recall that in the "Why work with stories?" section I said:

The main difference between this approach and many others that collect stories is this: a person who is working with stories does not tell or interpret or change or even select stories, ever. All of these things are done only by the people in the group of interest.

When you plan how you will work with stories in your story project, it is important to think about these three groups and how you will be asking them to help you with the project. The best story projects are those in which the people who are listened to participate in all phases of the project, meaning they do at least some of the looking at, thinking about, and talking about stories. (In the coffee-shop example, you might ask some customers to do a sensemaking exercise.) However, in some cases this is not possible, usually because the group of interest is unavailable, unable or uninterested.

When the group of interest cannot look at, think about, or talk about stories, it is best to follow these guidelines to avoid biasing the outcome of the project.

Rule 1: Separate statements

First, the listening and listened-for groups (i.e., those not in the group of interest) should get into the habit of separating statements about stories, answers and patterns into three categories:

  • Observations are things that anyone could be expected to see and agree on. An observation is something like "People younger than 20 were more likely than people 20 or older to say their story reflected peer pressure." Anyone in any of the three listening groups can make an observation.
  • Interpretations are opinions about what an observation or story or answer or pattern means. Most of the answers people give to questions about their stories will be interpretations. Only people in the group of interest should make interpretations unless the second rule is followed (see below).
  • Implications are opinions about what could or should be done about the issues raised. Answers to questions about stories sometimes involve implications, such as when the question is something like "What do you think should be done about the problem described in this story?" As with interpretations, only people in the group of interest should make implications unless the second rule is followed.

Rule 2: Provide provoking perspectives

The purpose of interpretations and implications is to provoke thought and discussion, not to provide answers or solutions. To support this, people in the listening or listened-for groups should always generate at least two differing interpretations or implications per observation. Try to make the interpretations compete as though people with opposite perspectives said them. You can even make some of the interpretations and implications deliberately naive and extreme in order to provoke reactions that jar yourself and others out of habitual thought patterns. Following this rule can be difficult but it greatly reduces the trap of confusing interpretations with answers, and helps any group of people broaden their reflections and discussions on the topic and consider fresh perspectives.

You can school yourself in creating competing interpretations and implications. For example, you can think:

  • you could look at it this way or that way
  • this sort of person might say this and this sort of person might say that
  • it could mean this or it could mean that
  • one person might say this and another person might say that
  • you could take it this way or that way
  • one way to look at it might be and another might be

And so on. Sometimes it can be helpful to think of people you know (or can imagine) who would not be expected to agree on the topic and imagine what they might say, like "my mother would say this, and my college friend would say that" or, for a group, "our happiest customer would say this, and our angriest customer would say that." These habits of thought can become accepted practice in your group.

A note on technology

You may have noticed that I have said nothing about what technology you should use to look at stories. There are three reasons for this.

  1. If you know how to use a spreadsheet to count and sort and graph results I don't need to tell you anything (and if you don't know how to use a spreadsheet, there are plenty of resources on the web to help you). Using a statistical package and writing your own scripts to process data are also options you may have available which I don't need to explain.
  2. The folks over at Cognitive Edge sell software specifically designed to support looking at patterns in narrative data, and I don't want to step on their toes.
  3. I have no idea what you need or have to work with, and it's impossible for me to make recommendations for the wide range of things people might possibly want to do in a story project. There is no one best way to support storytelling with technology; there are many possible ways.

For further reading

Reader comments, tips and advice can be found on the Look: Finding patterns Google Group page.

Next: Think: Making sense