Subjective experience

The term ‘perception’ has three kinds of meaning in PCT discussions: a perceptual signal, a rate of firing in the nervous system; a numerical variable in a model of a behaving system that may be implemented as a computer program; and the experience of perceiving. An aim of PCT is to justify our belief that the former two correlate not only with each other (a hoped-for outcome of neuroscience research), but also with our experience of perceiving, and indeed nerve firings and model variables would be a big shrug without that correlation to subjective experience.

In neuroscience terms, we talk about a perceptual signal as some sort of aggregate or average of rates of firing in a bundle of nerves.

Neural spike trains present analytical challenges due to their noisy, spiking nature. Many studies of neuroscientific and neural prosthetic importance rely on a smoothed, denoised estimate of a spike train’s underlying firing rate.
Methods for Estimating Neural Firing Rates, and Their Application to Brain-Machine Interfaces

“Firing rate” in the neurophysiological sense is an idealized theoretical concept.

(That said, many neuroscience researchers look instead for temporal patterning in neural spike chains, in preference to firing rates. Peter Cariani at Mass Eye and Ear never did come around from that conviction, and dropped away from CSGnet. One explanation for the preference is that they imagine it might account for pattern recognition, which living organisms do well but computers don’t. HPCT ought to provide a good account of pattern recognition; I am not aware that it does.)

In a PCT model of a kind of behavior, perceptual signals are quantitative variables which are claimed, sometimes explicitly, to correspond to perceptual signals in the neurophysiology sense. These variables in the model are theoretical constructs. We can derive them from instrumental measures of physical phenomena that demonstrably impinge upon the relevant sensors. In principle, we can even measure rates of firing from those

In practice, and in talk, we always revert to perception as experience, or as we imagine it to be experienced. Even if we don’t acknowledge this explicitly, 'fess up—it’s always the foundation of which the other two are theoretical analogs. When we talk about perceptions, almost always we mean aspects of what we experience, even when we use the language of model variables and neural firing rates.

A virtue of triangulating subjective experience to model variables and imputed neural activity is objectification. Physical science looks for quantifiable analogs or correlates of ordinary experience. But the direct investigation of subjective experience is actually a well established discipline with a long history. These disciplines (plural, actually) specify practices and conditions and indicate experiences that can result and how to recognize them. Experimental method: establish these conditions, do these things, and if your experiment is successful you will see these results as others have done before.

In their instructional material, many of the diverse schools of Vedanta, Buddhism, etc. carry culture-specific imaginings. This is to be expected: a great deal of the furniture and traffic of subjective mental experience is the product of memory and imagination. The disciplines teach discrimination, and this they have in common with Western science, which in its history has all too often been misled by imagining.

Prominent Buddhist teachers have joined with scientists in exploring the relation between these two ways, objectification by modeling and direct, disciplined investigation of subjective experience. The Dalai Lama wrote The Universe in a Single Atom: The Convergence of Science and Spirituality (2005), and there are many other books, videos, etc. on the relation of Buddhism and physics. PCT is eminently relevant.

So how much can we do on just the basis of perception as experience?

1 Like

I recently encountered a difference between me and someone else thinking about PCT, that relates to this post (only loosely, forgive me if I get it wrong). In describing the perceptual control hierarchy, I tend to explain perception as experience. The experience of planning, of following a path, of feeling whether something is true or not, of recognizing a certain plant species.

I call these experiences perception. However, I also think that all these kinds of experiences are controlled perceptions (not always controlled well, but under control). Even the perceptions at the lowest levels are controlled. The other person was surprised by my idea that all perceptions in our experience are controlled perceptions, thinking that the controlled perceptions mainly relate to our conscious experience (such as conflict resolution at higher levels).

So since then I’m thinking: what are perceptions that are not controlled? Can we experience these perceptions or does experiencing mean that they are brought under our control? Of course we’ve got a perceptual signal that would be a ‘pure’ perception, but then, can we ever experience just a perceptual signal seperate of the control process that contains that signal?


With your main interest being MoL, of course (perception=firing rate in neurons) and (perception = quantitative variable in theory, model, and computer simulation) take a back seat to (perception = experience). PCT requires a balanced grasp of all three, and each has explanatory power for the other two.

what are perceptions that are not controlled? Can we experience these perceptions or does experiencing mean that they are brought under our control? Of course we’ve got a perceptual signal that would be a ‘pure’ perception, but then, can we ever experience just a perceptual signal separate from the control process that contains that signal?

Look at what Bill says about ‘passive observation mode’ (B:CP p. 222). Don’t get too caught up in his theory/model notion of switches. However neurons and neurochemicals do it, the perceptual input function assembles lower-level signals into a perception at that level and passes it up to higher levels without comparing it to any reference value for it. In observation mode, the perception does whatever it does and you don’t care. “Observation” does imply awareness, but it’s obvious that Bill intended this ‘mode’ to cover the general case, with or without awareness.

This is easy to confuse with absence of disturbance. So long as variability of the perception is close enough to the reference variable the perception does whatever it does just as though there were no reference value for it, and no error signal results from the comparison, but as soon as it changes in a way that is significant for you your resistance to that change demonstrates that you have been controlling that perception all along. Bill proposed (and I agree) that this is true even if you lack effective means of resisting the disturbance.

Remembering that discussion with Bill was the basis of saying (as Bill did) that we control the perception of the sun rising in the east as distinct from rising in some other location. A more obvious and acceptable case would be the sunrise not appearing at all, e.g. if pyroclastic clouds darken the sky. In that case, we might slip into proposing that our perception of those clouds was the controlled perception. I don’t think that would be correct. They are the source of the disturbance. We usually don’t perceive disturbances and their sources, but when we do they may become controlled perceptions in their own right, after the primary disturbance.

Scientific observation is controlled perception in which the controlling systems form memory of the observed perception, or prosthetic extensions of memory (record-keeping). Does some other higher-level system nullify or damp any systems which may be asserting a preferred value for that perception, i.e. any sources of reference input for it? We call this ‘objectivity’. Actual practice involves (or should involve) cross-checks because we know that we control without awareness, and that not all of the sources of our reference values are acceptable in science.

One pitfall of grounding your understanding in the experiential sense of ‘perception’ is confused talk about experiencing error. I have seen quite a bit of this. Error signals never enter the perceptual input side directly. Reference input signals are constructed many-one in a reference input function just as perceptual input signals are constructed many-one in a perceptual input function, and error output from one system branches to many reference input functions. What you may experience is perceived consequences of poor control or loss of control. I think it’s important to keep the language of MoL straight about this, not only because talk of experiencing error muddles the model and therefore muddles its explanatory power, but more immediately for the client because the consequences of error are perceptions which may be observed and the origins of those perceptions may be perceived and brought under control, whereas error itself cannot.

Eva, Bruce,

two very quick and very incomplete comments.

I have had a good discussion with Martin about uncontrolled perceptions, but I have now no time to recollect it. However, we have much, much more sensor organs and afferent nerves than effector organs and efferent neurons. Thus it is natural to think that we cannot control all our perceptions (as such). But still we can utilize the uncontrolled perceptions when controlling other perceptions. For (not so elaborated) example, I control for tasting ice-cream but I do not control for hearing the irritating sound of an ice-cream van. Still I can go to the ice-cream van and get relief for my ice-cream-controlling-error if I hear the sound and there are no suitable shops near.

Similarly I still disagree that we do not usually control for sun rise. Technically it is not even possible. It is possible as “mythical action” like giving sacrificial offerings to the god of sun. But today at least I do not control for sun rise. However, the sun rise is one of the countless important stabilities I need for controlling my other daily perceptions. If it happened that in some morning the pyroclastic clouds darken the sky, I would not first start to pray for sun to rise, but tried to click the lights on. Then later I would start to control for removing the clouds and returning the sun, but not today.

Hi Eva, it could be really helpful to discuss this. I have always thought that just because behaviour is the control of perception, it doesn’t mean that all perception is controlled or even controlled by behaviour.
That said, I can see how selective attention - merely focusing on an element of the environment - requires the control of perception (but not necessarily the control of the perception of that element in anyone else). It may involve behaviour (e.g. eye movement) or it may not (maybe lowering the threshold of specific input functions). I would tend to describe this form of ‘control’ as either ‘passive’ observation or imagination mode, or a combination of the two. But it is different from controlling the variable using behaviour via a feedback function in the environment - the default controlled mode.
It’s certainly an important question Eva and opens up many more…

Thanks all for replying - and I’m happy the site is back up so I can reply!

@bnhpct: An interesting point about the experience of error. If I get you right, error in the pct sense cannot be experienced. It is the consequences that can be felt. I do like to explain the mechanism of control by using a video like this: The Most Unsatisfying Video in the World ever made - YouTube. You obviously control many perceptions, and only notice once something is wrong and that perception disturbed. The moment the sun rises in the west, for example. Thanks for the notion to be careful about discussing the experience of error - that is a shortcut that messes up the model since the error itself is a signal that is not experienced. Right?

I think this question of whether we do experience uncontrolled perceptions lays bare different views of what control means in our experience. If I get you right, @EetuP, you believe we only control for sunrise at certain specific moments, when you need the sun to be there in order to fulfill other goals.

Might there be a difference in understanding what it means to control? For example the sun rise (this discussion has been done before, I realise). If today the sun in reality hasn’t risen, but instead my house has been enveloped by a gigantic film studio and they simulated the daylight out of my office window, I might not have noticed the difference at all. All I control is my perception of the sun rising, not the actual rising or the movement of the earth.

@wmansell I agree with “I have always thought that just because behaviour is the control of perception, it doesn’t mean that all perception is controlled or even controlled by behaviour.” For example: If I have to teach in a room with poor ventilation, I will either open the windows myself, or someone else may open the windows for me. My behaviour doesn’t matter, it’s the perception of ventilation that matters. Whether I take action or not, I control the level of ventilation in the room.

So wrapping up: I don’t think I’m writing something silly if I discuss control in the broader sense (not just the active default controlled mode). However, I might need to pay attention that not everyone (in our outside of the PCT realm) thinks of control in the same way.

Hi Eva,

thanks for interesting thoughts – and an amusing video…

I am not so sure about the impossibility to experience the error. Martin has somewhere suggested a different kind of control diagram where also the error signal was echoed to the input function of a higher control unit but I can’t remember the details now. But quite phenomenologically we do not experience perceptions as such but always as evaluated against our references, either as good or bad things. This would mean that we experience perceptions as already compared i.e., as error signals. Of course this does not fit to standard control diagram, but neither does the conscious expressions fit there.

Are all perceptions controlled – either by behaviors or else? Perhaps an application of a dimension I have learned from semiotics could be useful here: in the other end are virtually controlled perceptions, then actually controlled and still after that the actively controlled. Virtually or possibly controlled perceptions are those which we could in some situation control but at the moment there is no active reference signal for them. Actually controlled are those where there is a reference signal but no error signal and no output because, for example, the perception is already inside the tolerance limits. Actively controlled are those perceptions to which we try to affect by output.

(Note that the same perception, for example the sunlight, can be controlled with different reference values, as either good or bad. If you are sunbathing your reference is high and when you try to sleep in a tent your reference is low. If you are developing photographs in the dark room, you may have no reference for perceiving sunlight.

That there is no reference, nether low or high, requires that there is some kind of switch, which puts the control offline but not the perception: passive observation state.)

1 Like

I believe this is a proposal that an error signal can add its influence to a perceptual signal. It does not become the entirety of a perceptual signal.

It is often the case that we pay attention where control falters or fails. It also frequently happens that we notice perceptions that we were experiencing without awareness. You would deny that this is possible, if you say that experience is limited to conscious awareness of perceptions to which you are paying attention.

And you omit the possibility of passive observation (B:CP p. 222) as an experience.

We do not have consensual definitions of consciousness, awareness, and attention. Superficially, attention is the directing of sensory organs to one or another part of the environment. We sense pressures by bringing a part of the body with pressure sensors into contact with that which is sensed. In the visual modality, attention seems limited to the field of vision, as determined by rotating the eyeballs and head. In the auditory modality, we can likewise orient the ears and use a hand to extend the pinna (cup the ear). However, as I look at these characters on this screen I can direct my attention to neighboring parts of the visual field to either side, above, and below. I can direct my attention to the auditory modality or other sensory modalities. I can direct my attention within the auditory field without reorienting my head or ears. I can direct my attention to the NSF proposal that I need to write or to a remembered photo on my phone of my newest grandson and his mother. These do not involve any sensory organs that I know of. In your terms, they involve an intention to experience this perception rather than that perception. This is close to the problem Warren is wrestling with.


BN: We do not have consensual definitions of consciousness, awareness, and attention.

That is so true! And in additions we seem to have still less a consensual definition of experience. This makes discussion quite futile. But I try to continue once I have more time.

This whole thread seems to equate conscious “experience” with “perception” in the PCT sense. I think that is entirely wrong, and have thought so for years before this forum existed. The non-conscious perceptions such as perceptions of the tensions in a muscle fibre bundle (as Bill explained somewhere early in B:CP) are the ones we reorganize to be able to control, as are higher level ones to which a sports pro or a teacher musician guide our consciousness.

If a golf pro says you should feel your belly button at the end of your swing facing in the direction you wanted the ball to go, how many of their students would have had the conscious experience of feeling where their belly button was facing when (as usual) the ball had gone astray? Most of what we do every day consists of actions in control of perceptions of which we are not conscious,and of which we can become conscious only with training and effort.

On the other hand, as has been mooted many times over the three decades of my involvement with PCT, we are conscious of (experience) more complex perceptions, such as that we need to swing a club if our golf ball is to progress toward the hole. Doubt about what to do has been claimed to be a characteristic of consciously experienced perceptions, the question being about controlling actions about which one must think.

Thinking is controlling variables in imagination. The variables you control when you think are categorical, even if they are labels such as “pi” for particular values of continuously variable perceptions. When thinking about what to do about an experienced perception, the solution arrived at requires a selection among all the non-consciously controllable perceptions. “Should we walk or take the bus” deals with actions we know how to do, because we have controlled many other perceptions either by walking or by taking a bus. But we have not used them much, if ever, to solve this particular problem.

Usually when we walk we are not conscious of controlling all the hip, knee, and ankle joint angles by deciding how and when to tense the appropriate muscles in each leg. We just control the relevant perceptions non-consciously, not experiencing them at all unless we try to, and have the skill to succeed.

For years I have tried to get people to distinguish between perceptions controlled non-consciously in the reorganized hierarchy, and conscious perceptions that by definition are experienced whether they are controlled or not.

Eva (Message 3) asks about perceiving error. I have two answers for that. In Figure 7.2 of PPC, I show a possible kind of inter-level connection that is mathematically equivalent to the Powers connection of perception of level N being an input to the perceptual input function of some controller at Level N+1. This circuit takes only the level N error value and adds it to the level N+1 reference value to produce the equivalent of the Level N perceptual value. This circuit connection has many benefits, and I prefer it to the Powers connection. If it is anywhere near correct, then the error variable has the same status as any perceptual variable.

My second answer relates to experience. We certainly can experience what we are trying to achieve when there’s a problem achieving it, and we can experience the current state and what is wrong with that state (the error). Consciously, then, we can indeed perceive reference and error values.