Tree in California
Tree in California (Photo credit: Wikipedia)

This article is about my research into the structure of what we say when we write or speak. Many years ago, while teaching a course on technology and public policy at Carnegie Mellon, I discovered a logical form that underlies the discussion of complex issues. This logical form had a branching structure so I called it the “issue tree.”

Since then I have done a lot of research on issue trees, as well as using diagrams of them to help my clients deal with complex issues. Knowing that issue trees exist whenever we talk or write can be helpful so I will explain the basics here. For those who want to know more I have a crude little textbook on it that I wrote for my own classroom use, entitled “Issue Analysis: an introduction to the use of issue trees and the nature of complex reasoning.” Also, I have already presented the issue tree in a Kitchen article as a model of scientific progress and looking at that graphic may help one understand what I am about to say.

The basic pattern of the issue tree is quite simple. An issue begins when a statement is made (or a question is asked). There can be several immediate responses. The discovery was that statements made in response are answering questions of the initial statement, questions that are typically unspoken. These unspoken questions define precise relations between the statements.

As discussion continues more statements are made, each answering an unspoken question of some specific prior statement. Thus there is a precise branching structure underlying everything that is said. Much follows from this. Note that these component statements may be made by a single person or document or by many, all at once or spread over a long period of time.

By way of example, here is a link to the issue tree underlying the fine print in an automobile loan contract. The textbook has more examples, however it is mostly about how to draw issue tree diagrams. Note that the issue tree and the issue tree diagram are two different things, just as roads and roadmaps are. The issue tree is there when we talk and write, individually or collectively over time, whether we see it or not. This is not just a diagramming technique; rather it is a basic fact about human discourse.

There are some well known indicators that expressed thought has a branching structure. The outline, for example, and the nesting of blog comments are both tree structures. There are also various so-called “mind mapping” techniques. My discovery is that this structure is precise, well defined and universal.

Actually it is not quite this simple because sometimes a statement will refer to a body of prior statements rather that a single one, but that is an advanced topic. The real question is, how is this useful to know?

First and foremost is the fact that when we talk and write we are articulating complex and sophisticated structures, even in the most mundane cases. We are constantly deciding which sentences to respond to and which unspoken questions of those sentences to answer. I start my textbook with a seemingly silly minor argument just to make this point. We are not just piling on statements; rather we are creating systems of ideas.

While the combination of unspoken questions being answered can be a complex structure, the questions themselves are typically very simple. They frequently include the famous who, what, where, when, why and how, as well as “such as” (which is asking for an example) and “what evidence”. If there is disagreement, they will also include objections and replies. The questions are not really there, mind you, unless they are expressly asked; they just represent the specific relations between the statements being made.

Then there is what I call the “jumping problem,” which is a major source of confusion in human communication. Speaking and writing are both linear in the sense that one sentence comes after another, but they are expressing a branching tree structure. The problem is that there is no way to put each sentence directly after the sentence it is linked to and building on. Graphically this means that we are forced to jump around in the issue tree and sometimes these jumps are quite large. The reader or listener must then make the connection back to the prior sentence and this can be difficult, hence the potential for confusion. The jumping problem is discussed in the textbook.

Knowing the issue tree is there creates a new science of sorts because there are a lot of structural features that can be counted and measured in human discourse. These include rates of branching, the allocation of attention to different sub-issues, or to different question types, as well as patterns of jumping. For example meetings tend to have very low branching rates, because it is hard to get back to prior statements. Thus meetings may not be a good way to explore complex issues. Many of the items in my taxonomy of confusions are based on issue tree features.

Issue tree diagrams can be useful in articulating or dealing with complex issues. They have been used to deal with various sorts of issues including strategic planning, decision making, policy and regulatory analysis, litigation, contracting, and system design. But it is important just to know the structure is there whenever you talk, write, read or listen.

On an historical note I first presented the issue tree in a conference on philosophy of technology at the University of Illinois in Chicago Circle in 1973. The proceedings were finally published in 1979 as “The History and Philosophy of Technology,” edited by Bugliarello and Doner, University of Illinois Press. Mine is Chapter 14 entitled “The Structure of Technological Revolutions.” By that time I had left academia for the world of issue analysis and management consulting, but I have continued the research as well.


13 Thoughts on "The Issue Tree Structure of Expressed Thought"

I found this very interesting and went back and looked at your textbook. I was wondering if you know the work of Walter Kintsch, who analyzes propositions ( which I’ll loosely define as words that can take on a predicate as opposed to say prepositions) in a text in order to determine the text’s complexity. To me your issue trees and his diagrams resemble one another. In any case, even if you never heard of him, this was very interesting.

I am familiar with his latent semantic analysis because I use term vector similarity to map the structure of science. But these methods are more about words than about propositions. Propositions are the topic of mathematical logic. They can be thought of as the shortest sentences a text can be translated into, which are called the atomic propositions, plus how these are combined in the actual sentences.

The most detailed issue tree one can build is that in which each node is an atomic proposition. This is seldom necessary but in the case of the auto loan note we approached it. This was a project to translate consumer credit contracts into plain language. The attorneys were concerned that no legal provisions were lost in the translation and the issue tree was the tracking system that made sure this did not happen.

Most of my work has involved understanding complex issiues not detailed text analysis. But there have been cases where the text was the source of the issue, such as with major federal regulations.

It strikes me that your work coincides nicely with the branch of philosophy known as informal logic. Also, of course, it has much in common with the old practice of sentence diagramming to reveal the underlying grammatical structure; one might call your issue tree diagramming a form of analyzing the grammatical structure of reasoning.

Yes, mathematical logic is a big part of it. What I am diagramming is how our sentences fit together. Classical logic mostly looks at deduction and induction but I am looking at understanding. After all, 99% of decision making is understanding the situation.

But there is nothing informal about the issue tree, except the fact that all us folks do it. Informal suggests vagueness but what we say is typically very precise. To steal a point from Chomsky, the number of sentences that meaningfully answer a question posed to another sentence is vanishingly small compared to all the sentences in the language. We are actually extremely good at what we do when we speak and write. Whether what we say is true or not is another matter, not part of logic.

Informal logic is NOT mathematical logic, which is the same as formal logic. And “informal” does not mean that there is anything vague about it. A good account of the field is provided here: Note that “diagramming” is explicitly mentioned as one of the modes of informal logic.

You are quite right that my work falls within the domain of informal logic, but I think it abrogates the distinction between formal and informal logic. That is, the issue tree relations are extremely formal in the mathematical sense of that term. It is like algebra word problems, which were still algebra even though they came before the formalism. Formalism is an attempt to describe reality.

On the other hand the issue tree relations are relations of understanding not of inference. I was trained by mathematical logicians but they had a lot of trouble with the issue tree. The question is what is the scope of mathematical logic? I claim that it is human reasoning not just human inference. Inference is but a small part of reasoning, compared to understanding. The logicians did not see it that way. So I left.

Informal logic scholars would agree with you that “inference is but a small part of reasoning.” I learned formal or mathematical logic from one of the giants of the field, Alonzo Church, at Princeton as an undergraduate and learned more later from Charles Parsons in grad school at Columbia. But then, as a publisher at Penn State, I built a list of books in informal logic, including five books by Douglas Walton.. So I have an appreciation of the contributions of both types of approach.

  • Sandy Thatcher
  • Jul 11, 2013, 12:32 AM

This brings back some memories from my linguistics undergrad. It was an enjoyable read! I wish I had time to read your text book, you might have covered the following:

There are some tree’s which propagate through a root system. I would suggest that to overcome some of the jumping issues and the complexity of larger or groups of previous issues feeding into a current issue you could expand the model to a forest of trees interconnected via a root system. You know the adage: you can’t see the forest for the trees? It would apply when participants in the discourse jump in mid stream and focus on a single issue loosing sight of the bigger one or the initial one.

You would create an issue eco-system if you will…

Do the branches and twigs in an issue tree ever connect with each other, grafting as it were? Which is another way of asking: does the tree ever become a network? Or is it essential to this concept that it is specifically a tree and not a network? I’m guessing your answer is that the tree structure–which you’d have to traverse in reverse to get to another branch–is an essential feature of the concept. Or not?

This mention of twigs makes me think of the “twigging phenomenon” adopted as a way of visualizing the ever increasing specialization of knowledge as it develops new branches and then twigs on those branches.

Indeed Sandy, my issue tree model of scientific progress makes precisely this point.

However a twig can become a very large part of science so the twig metaphor may not be a good one. Major sciences like atomic theory and quantum mechanics began as twigs, as it were. Also, some great advances involve going way back up into the accepted tree of knowledge and starting a new line of reasoning. Relativity is a prominent example. The point is that (1) growth does not necessarily occur at the extremities and (2) when it does occur it can swamp the origin, sometimes explosively. Exploding swamps? I may have to work on that metaphor.

Correct Bill, the issue tree does not become a network, except in the case of ambiguity. The path to a sentence defines the role that sentence is playing, which is part of its meaning, so convergence would mean ambiguity. There is a network structure if you look at where the ideas come from, via citations or references for example, but that is a separate relation distinct from the issue tree relation.

Traversing in reverse is also correct. In fact one can roughly measure the distance between sentences by the number of nodes between them, first backing up from one node then going down to the other. Trees have only one path connecting any two nodes, unlike networks which have multiple paths.

But the pattern of jumping around in the issue tree, even while building it, can be a network and often is. Thought is a tree while attention is a network. One has to return repeatedly to the same node in order to initiate new branches. This returning of attention is what is hard to do in meetings, by the way, compared to the case where the sentences are sitting there in a fixed medium waiting to be thought about. Our research found meetings having branching rates of say 1.2 to 1.4 while issues thrashed out in newspapers had branching rates of 2 to 4.

To me the topology is astounding. It is amazing that we do this when we simply talk and write.

Agreed, very cool! Thanks for the further explanation. It strikes me that while it might seem that creative thinkers “jump around” in the tree, what they really do (if they’re good) is _traverse_ the tree, and even (if only instinctively) appreciate it as a tree. What might seem to be simple “thinking outside the box” might instead be an ability to do that traverse-in-reverse that is key to questioning assumptions, appreciating context, grasping that there are alternative paths, etc.

Leave a Comment