Zigzag Modeling that Captures the Dialectic of Form to Process

Bateson’s Zigzag Model
A Logical Dialectic Ladder
from Form to Process

Beyond Logical Types

L. Michael Hall, Ph.D.

With Gregory Bateson’s wanton curiosity about how things work and how to create theoretical models for cultural phenomena, mental phenomena, and biological processes, he explored numerous fields C language, semantics, biology, genetics, cybernetics, mathematics, communication, and more. He began as an anthropologist but ended up as more of a cybernetican as he integrated systems thinking into everything he did. And it seems to me that it was his systems thinking more than anything else that enabled he to create new models and ways of thinking and which made (makes) his books and articles so rich with possibilities of new developments in Neuro-Semantics. I say that because with every reading, I discover ever new things in his writings.

Bateson’s Zigzag Ladder for Modeling

In Mind and Nature (1979), Bateson proposed the following ladder as a way of modeling “mental” phenomena in biology that takes into consideration the processes as systemic activities. He moves from description to process, that is from the typological labels given to something to the processes involved and then to the next level of labeling that defines that form to the processes that govern and determine it, to the next level, and so on.

In doing this, he uses the mathematical concept of logical levels from Russell and Whitehead to sort out the logical levels that arise or emerge from the processes. He first sorted these things out as his way of demonstrating the validity of his “Criterion 6” of “mind.”

The Description and Classification of the Processes of Transformation Discloses a Hierarchy of Logical Types Immanent in the Phenomena (p. 122).

That statement is loaded. It summarizes so much in Bateson’s thinking about “mind,” learning, levels, logical, types, experience, processes, etc. He begins with the coding of messages and that stimuli from one person to another are messages but that for person B to know what person A means,

“… there comes into existence another class of information … to tell B about the coding of messages or indications coming from A. Messages of this class will be, not about A or B, but about the coding of messages. They will be of a different logical type. I will call them meta-messages.” (123)

“The whole matter of messages which make some other messages intelligible by putting it in context must be considered…”

Bateson starts here in order to flush out the levels of experience. He says this explicitly.

“All this is premised on the existence of levels whose nature I am here trying to make clear.” (124)

To do that he separates out two kinds of actions. There is the primary state set of actions in a context, then there are those actions or behaviors “which defines context or makes context intelligible.” He calls the second “meta-communication” and acknowledges that he borrowed this term from Whorf (p. 124).

What does the meta-message do? How does it work? What is its function?

“A function, an effect, of the meta-message is in fact to classify the messages that occur within its context.”

In Neuro-Semantics we use this meta-level principle to talk about moving from one state to another state and using the second state to set the frame for the first. It classifies the first. Anger now becomes a member of the class of respect thereby creating a new gestalt, “respectful anger.” In this way, we transcend our first experience and include it inside of a higher state of mind-and-emotion. This allows us to texture our states and to create a new quality to our lives.

Bateson explains this in terms of how he took a mathematical concept and translated it to biology. This was the source for the term that he most frequently used, logical types. It comes from the field of mathematics.

“It is at this point that the theory offered here connects with the work of Russell and Whitehead in the first ten years of this century, finally published in 1910 as Principia Mathematica. What Russell and Whitehead were tackling was a very abstract problem. Logic, in which they believed, was to be salvaged from the tangles created when types, as Russell called them, are maltreated in mathematical presentation.” (124) logical

While Whitehead knew that we are amused and humored by “kidding around with the types” he said that he did not know if they had any idea that their work could be applied to humans.

“I doubt whether he ever made the step from enjoying this game to seeing that the game was non-trivial and would cast light on the whole of biology.”

But Bateson did. That was his genius. He did apply it to biology. He took events or actions at the primary level, and the classes of actions (meta-actions), and then meta-classification and showed how this applies to learning. We can learn what an action means, we can then learn to learn about how we do that, etc. His levels of learning resulted from this application of logical typing. In NLP and Neuro-Semantics we usually speak about this same things using the phraselogical levels rather than logical types. There’s no difference except that types comes from the formal mathematical theory and levels refers to the way we use it in biology and psychology.

Applying it linguistically to the phenomenon of “paradox,” Bateson writes,

“Epimenides was a Cretan who said, ‘Cretans always lie'” C I have presented this paradox here in the form of a quotation within a quotation, and this is precisely how the paradox is generated. The larger quotation becomes a classifier for the smaller, until the smaller quotation takes over and reclassifies the larger, to create contradiction.” (125)

Bateson’s conclusion from this is that “logic cannot model causal systems, and paradox is generated when time is ignored.”

Confuse Logical Types and Reap Neurosis

When there are levels embedded in levels we have to keep them separate. There is a discontinuity between them. We have to discriminate levels recognizing that we are dealing with a different logic at each level. If we do not, the category error or the error in logical typing leads to pathology. This happens when we confuse map and territory, when we fail to recognize one message as classifying another and so don’t know what kind of message it is.

In his text, Bateson decided to illustrate using the clinical experiment with dogs. It was the experiment wherein dogs were taught to discriminate between two stimuli, between a circle and an ellipse (p. 127). First a dog is trained by positively rewarding him with food when he makes the distinction. Once the dog can tell the difference, the experimenter then pushes the limits of his perception. He makes the ellipse somewhat fatter or the circle somewhat flatter so the contrast becomes less. The dog has to “put out more effort to discriminate between them.” Eventually however he cannot. It is this that then leads the dog to experience experimental neurosis. Some of the dogs become agitated, nervous, and will become aggressive. Other dogs turn lethargic, passive, and give up. Some even become comatose.

What happens that creates this pathology?

How does this relate to the logical typing of messages at different levels?

Once a dog had learned to see the context as one of discrimination, he had come to expect there to be differences. It was as if in his “mind” he was expecting it and organized on the inside with an attitude that there should be a distinction and that he should be able to make it. But when he was consistently made wrong in that context, then he ceases to know how to operate there, hence the emergence of the neurosis. Bateson commented,

… he has been very carefully trained to expect a context for discrimination. He now imposes this interpretation on a context that is not a context for discrimination.” (128)

He is no longer a naive dog. Naive dogs who have not been so taught don’t have the problem, they just guess and leave it at that. But the trained dog has jumped a logical level.

“The naive dog has not been taught not to guess, that is, he has not been taught that the contexts of life are such that guessing is inappropriate.” (129)

But the trained dog had. Guessing is inappropriate for the trained dog. He cannot guess. He must discriminate … but he cannot discriminate. He sees and relates to that context with a higher level learning and having made that jump he cannot go back. To get out of the double-bind, he would have to jump yet another logical level. And yet he does not (cannot) jump yet another logical level to recognize what’s going on. Bateson writes,

“The dog fails to transcend the jump in logical type from ‘context for discrimination’ to ‘context for guessing.'” (130)

It is this shift of logical types or levels that has created the problem for the dogs. They do not deal with the stimuli as mere stimuli, but as “a context for discrimination.” But they cannot make the discrimination. They must, but can’t. Nor are they able to step “out of that frame” and recognize what’s occurring.

Yet dolphins are able to do just that, and do so (pp. 130-132). The dolphins also because very agitated and disturbed.

“The experience of being in the wrong was so disturbing to the dolphin that in order to preserve the relationship between her and her training (i.e., the context of context of context), it was necessary to give many reinforcements to which the porpoise was not entitled. Unearned fish.” (131)

When the dolphins were released back into the training tank after the fourteenth and fifteenth sessions, they were very excited and “put on an elaborate performance that included eight conspicuous pieces of behavior of which four were new and never before observed in this species of animal.”

So what happened?

The discontinuity is a jump between the logical types.

“In all such cases, the step from one logical type to the next higher in a step from information about an event to information about a class of events, or from consider the class to considering the class of classes.” (132)

Adding Systemic Processes to the Modeling

In all of this, Bateson acknowledges that he has approached the matter of hierarchy in mental phenomena from the aspect of coding.

How are things encoded?

He answers that specific actions are like members of a class. This distinguishes members from the classes that we put them in. He notices that Pavlovian conditioning through reinforcement and extension works very well at the first level of actions, but not on the classes of activities. This explains why primary level anchoring will not extinguish meta-level concepts and structures like “exploration.”

The specific activity of a rat sticking its head into a particular hole can be reinforced positively or negatively. Give it pleasure, it will continue. Give it pain, it will stop. But the class of activities that come under the more abstract label, “exploration” will not be affected by either. If the rat is rewarded or punished by looking into a particular hole, in both instances it succeeds in its discovery. It’s “exploration” activity worked. This general (or meta-state) is reinforced by either result.

Figure 1

Classes: Exploration Play Threat
Sensory Level: Sticking head in hole Chase
Pretend Fight
As if Biting but
Not puncturing
Baring Teeth

This led Bateson to linguistically recognize the difference between what we would call nominalizations and sensory specific language. “Play” is at a different logical level than the games of Chase or Hide-and-Seek. It’s a classification of activities and it points to no one specific activity. So with “crime,” “exploration,” and thousands of other meta-terms. These names describe how we organize or punctuate activities. They are more general than any particular activity that is a member of that class. View it systemically, Bateson says,

“The relationship between the characteristics of a component and the characteristics of the system as a whole as it circles back on itself, is equally a matter of hierarchical organization.” (134)

This brings in the reflexivity of the mind, the central mechanism in Meta-States. As an interactive system of multiple parts, our mind-body-emotion system is just that C a system. And as a system, there are components and then there is the-system-as-a-whole. And these also operate hierarchally.

Now linguistically we describe things using nouns and verbs. Using nouns, we think about things C the form, the structure, the type, the classification. Using verbs, we think about processes. This gives us a double-description of whatever we’re describing or modeling. It gives us form and process. Bateson noticed that he used both classification and description (nouns) and processes (verbs) when he was involved in his anthropological studies of the Iatmul culture.

“This typing of persons led back into the study of the processes by which the persons got that way. These processes were then classified into types of process types in their turn, were named by me. The next step was from the typing of process to study the interactions between the classified processes. This zigzag ladder between typology on one side and the study of process on the other is mapped in figure [2].” (pp. 209-210)

In this hierarchal figure, Bateson has sorted out the levels in his own theorizing and modeling and sorted out the difference between the labels he used to create his typology at each level and his awareness of the processes involved. He described this as “the zigzag sequence of steps from form to process and back to form and says it provides …

“… a very powerful paradigm for the mapping of many phenomena.” (210)

In other words, the procedures of inquiry into a subject are punctuated by alternating between Classification and the Description of a process. We proceed up an alternating ladder from Description to Typology. He noted that this paradigm for modeling reoccurs wherever mental process predominates in the organization of the phenomena.

“In other words, when we take the notion of logical typing out of the field of abstract logic and start to map real biological events onto the hierarchies of this paradigm, we shall immediately encounter the fact that in the world of mental and biological systems, the hierarchy is not only a list of classes, classes of classes, and classes of classes of classes but has also become zigzag ladder of dialectic between form and process.” (211)

Wherever we find a Form, there’s an underlying Process, and wherever there’s a Process, we can set it apart by naming it. We do this in Meta-States (the form) as we meta-state (the process) one state to another. From process comes form and within form are processes. In these two processes Bateson recognized a systemic difference that moved him beyond logical types.

“So far we can go with Russell and Principia. But we are not now in Russell’s world of abstract logic or mathematics and cannot accept an empty hierarchy of names or classes. For the mathematician, it is all very well to speak of names of names of names or of classes of classes of classes. But for the scientist, this empty world is insufficient. We are trying to deal with an interlocking or interaction of digital (i.e., naming) and analogic steps. The process of naming is itself nameable, and this fact compels us to substitute analternation for the simple ladder of logical types that Principia would propose.

In other words, to recombine the two stochastic systems into which I have divided both evolution and mental process for the sake of analysis, I shall have to see the two as alternating. What in Principia appears as a ladder made of steps that are all alike (names of names of names and so on) will become an alternation of two species of steps. To get from the name to the name of the name, we must go through the process of naming the name. There must always be a generative process whereby the classes are created before they can be named.” (201).

Noting that this is a “very large and complex matter” Bateson here speaks about how he thought about logical typing in the process of modeling any logical level system. This moves us beyond a simple logical typing that can be conveyed by ladders or steps in a hierarchy. We are now forced to add the process of how we actually move to a higher level C the process we call going meta, meta-stating, the meta-function, etc. After using Bateson’s illustrations, I will illustrate with the phenomenon of stuttering.

The Systemic Logical Types/ Levels in Heating a House

Bateson modeled his Levels of Learning on this sort of zigzag paradigm and suggested that we can model many other things on it. To illustrate and given meaning to the terms feedback and calibration, he illustrates with a house thermostat. He begins with the primary state of temperature.

“At the lowest level, there is the temperature. This actual temperature from moment to moment (a process) affects a thermometer (a sort of sense organ) that is connected to the whole system in such a way that the temperature, or expressed by the bending of a double metal plate, will make or break an electric connection (a switch, a calibration) that controls the furnace. When the temperature rises above a certain point, the switch will be changed to the state called “OFF”; when the temperature falls below some lower point, the switch will be changed to “ON.” The house will thus oscillate around some temperature between the two threshold points. On this level, the system is a simple, servo circuit.” (213)

If we have temperature at the primary level, what do we have at a meta-level that would control or govern the temperature? Bateson says that the simple feedback circuit is …

“… controlled by a calibration housed in the same small box that contains the thermometer. On the box is a know that the householder can turn to change the setting, or bias, of the thermostat to a different temperature around which the temperature of the house will oscillate Note that two calibrations have their location in the box: There is the control of state, ON/OFF, and the control of HIGH/LOW temperature around which the system will operate.” (214)

Primary state temperatures move up and down the scale as determined by the meta-state thermostat and sets the frame or bias. At the primary state level, the householder experiences temperature. At the meta-state level, the householder decides and chooses the range of temperature that he or she wants. And that comes from what the householder intends to experience and believes would be best (in terms of comfort, or cost, or some other value).

Above the frame or bias level, there is the level of evaluation (“This is too cold;” “This is too hot”) and above that is a person’s personal threshold governed by the person’s genetics and trainings and values governed by the person’s status, situation, and so on.

Thinking about the systemic nature of processes that reflect back onto their previous states led Bateson to notice the overall effect of going meta in how the higher levels govern the system and feedback into the system greater knowledge.

“With each zigzag of the ladder, the sphere of relevance increases. There is a change in logical typing of the information collected by the sense organ at each level.” (215)

“We note an alternating ladder of calibration and feedback up to larger and larger spheres of relevance and more and more abstract information and wider decision.” (216)

“From this paradigm, it appears that the idea of ‘logical typing,’ when transplanted from the abstract realms inhabited by mathematicological philosophers to the hurly-burly of organisms, takes on a very different appearance. Instead of a hierarchy of classes, we face a hierarchy of orders of recursiveness.” (218)

All of this describes many of the key features in how we experience our meta-states, our self-reflexive states-about-states and the recursiveness of the frames that we set that establish our bias in a given context. It also identifies a way for modeling the mental-emotional phenomena that can either enrich or impoverish our lives. As we recognize the recursiveness, the reflexivity, the discontinuity, the logical typing, and the element of “time” in a system’s functioning, we are more able to see how a system operates, its biases, and its leverage points.

Changing within a Level and Between Levels

We have two major feedback-feed-forward loops in all of our experiences. We have the primary state feedback/forward loop when we encounter any stimulus. This creates our basic stimulus-response pattern. X event occurs. As we input the messages from that event this feedback provides us information about the world. To this we respond with thought, emotion, speech, and behavior C and it is this which we feed forward back to the world. When that event next responds, we have more feedback about it. And round and round it goes.

Yet above and beyond that feedback/forward loop, we have a meta-feedback/forward loop. As we input information from the event (feedback at that level), we also feed back our own ideas, conclusions, generalizations, beliefs, rules, etc. to ourselves. We do that as we experience thoughts-and-feelings (a state) about that experience. This feeds back yet another message (a higher message) to a higher level of our mind. With this second message, we set new frames. We develop new biases. We draw new conclusions or reinforce old conclusions. Then from that, we feed forward these new frames and biases through ourselves and then out into the world to the event. And round and round that goes.

Bateson illustrates this with the difference between shooting a rifle and a shotgun and so describes two sorts of methods of perfecting an adaptive act (pp. 211-219).

In shooting a rifle, we use the first feedback/forward loop. We aim, align rifle with target, adjust, notice error adjust again and so on until we press the trigger and shoot. The ongoing actions of self-correcting goes on within the single act of shooting. Feedback describes this whole genus of methods of perfecting an adaptive act (212).

In shooting a shotgun or revolver held under the table where one cannot correct its aim, we have a very different situation. There’s no error correction in the single act. To improve, we take the previous shooting into account and make an adjustment, try again, and so on. “Correction must be performed upon a large class of actions.” So it is through long practice that we adjust the setting of nerves, muscles, seeing, etc. to the critical event. “This genus of methods Mittelstaedt calls calibration.”

“It seems that, in these cases, ‘calibration’ is related to ‘feedback’ as a higher logical type is related to lower. This relation is indicated by the fact that self-correction in the use of the shotgun is necessarily possible only from information derived from practice (i.e., from a class of past, completed actions).” (212)

From Bateson draws a distinction between feedback and calibration. Calibration is what he calls form or classification and feedback is what he calls process. He then says that while form-process and calibration-feedback are synonymous, higher and lower logical type is more complex due to the factor of recursiveness in the process (213). And it is this that calls for the separating out of levels because form or classification of the state and the process that creates it differs at each higher level even though it is an oscillation between form and process.

Within a primary level experience, feedback gives us the ability for immediately adjustments using error as our signal for adapting. But when we have to take a collection of experiences, a sample of events, then “time” enters into the formula. Then the discontinuous jump to the next highest level means working with samplesquality of how we jump to the higher level that governs the bias that we set, the frames that we construct, and the mental world that we invent.

Stuttering Application

When someone stutters in speech, is that a single sensory-based action or is that a class of activities? You might think upon first consideration that would be an easy question to answer. It is not. It all depends. It depends upon whether the speaker is referring to a specific stutter in speech or to a whole class of activities that now operates as a meta-frame, “stuttering.”

If the person is referring to a classification then that’s why reinforcement of smooth speaking will not extinguish that category. The General Semanticist Dr. Wendell Johnson noted long ago that first we have to create that category and that there are some language groups and cultures which do not have that classification. And without the classification (frame), the sensory based activities of stuttering cannot be punctuated.

Once we have the category as a frame matrix of beliefs, values, pain/pleasure, expectations, identification, social definitions, etc., then we have moved up many logical levels classifying and embedding the sensory level inside of many other classifications. This creates the Stuttering Matrix that holds the activities in place.

First, I stumble over my words as I search for the words to express myself (process at the primary state level), then someone (myself or a parent) labels that action “stuttering” (form). Then predictions are made (in my head or in the descriptions of others trying to help!) about all the problems and bad things that will happen if I “stutter” again (a higher level description, description2). As I hear this, I become afraid and threatened by those meanings and label those processes as part of my personal traits and identity (higher level labeling, “I am a stutterer.”), and so on.


  • The zigzag approach to modeling neuro-linguistic processes or any system is a way of moving from Form (the labels that indicate solidified Structures) to the Processes that create those forms and which processes are within the Forms.
  • Following this zigzag approach enables us also to flush out the logical types within our reasoning that classifies and categorizes reality in a certain way and creates the logical levels of our experience.
  • This illustrates that levels are “logical” or psycho-logical (Korzybski) to the degree that they fit the definition of logical types.
  • Sorting out levels in this ways saves us from double-binds, from setting frames that lock us in closed loops, and that allow us to escape the field (or context) of any frame that creates limitations and impoverishment of life.
  • Recognizing at what level we’re working, thinking, reasoning, etc. also enables us to profile an experience and to work with it more elegantly. It gives us more clarity about how to reinforce or extinguish a behavior inasmuch as it enables us to distinguish sensory from evaluative behavior C single actions at the primary level or classes of actions at a meta-level.
  • In this presentation Bateson has drawn a critical distinction between sensory-based feedback of actions in an error-activated context when we want to perfect an adaptive act and the higher level (meta-level) calibration from a class of activities that sets a full neuro-semantic attitude in mind-and-body.


Hall, L. Michael. (2001). NLP: Going Meta C Modeling Using Meta Levels. Grand Junction, CO: Neuro-Semantic Publications.

Hall, L. Michael. (2002). Is there Any Difference Between Logical Levels and Logical Types? Website: www.neuro-semantics.com