Hi Richard,

At the root of my comments on time standing still is a concern with using memory tests to understand the bottleneck. The theory behind the bottleneck is that the conscious mind can only absorb new information at a fairly low rate. This is measured by the memory tests and your hyposthesis is that a bottleneck in bandwidth exists somewhere between the external senses and the conscious mind.

I also think there is a bottleneck, but I am not convinced that it is static. I am not quite sure where it is, but I suspect it exists between the conscious mind, which we consider gives us our sense of existence, and the the rest of ourselves as an organism. In other words our experience of being is largely that which results from us living in the world of the model. Our experience of being is linked to the real world we live in through the bottleneck.

This theory raises a number of interesting questions, but one of them is how we experience time. A second question is the nature of the bottleneck, and whether it is static or not. By static I mean that it has a fixed bandwidth that is set by evolution. To my thinking it would be possible for the bottleneck to have dynamic bandwidth. In this theory the bandwidth increases or decreases about a mean according to the nature of the activity the human is carrying out within the real environment. In other words the human consciousness has evolved the ability to detach from its immediate envirnonment and keep only a narrow bottleneck open for normal use. This is what you are measuring with the memory tests.

When the senses detect life threatening or otherwise extremely important real world circumstanses the bottleneck opens and the consciousness couples more closely to the immediate real world environment. This leads to the perceived change in the experience of the passage of time. In the model time is completely malleable. Each of us will perceive its rate of passage in a totally individual way. This leads to anectodal ways of "whiling away the hours" in some activity or day dream unique to each of us.

However, when a serious thing happens, such as sliding off the bike in a fit of overexuberance on a dodgy surface, the bottleneck bandwidth opens. Usually the amount of information linking the model to the outside world is comaritively slow. Now, when a lot of information is coming in per unit time, the amount of information arriving through the bottleneck would be equivalent to a much longer period of time were the bottleneck still to be narrow. This has the effect of perceiving a short period of time like a much longer period of time. In other words time stands still.

In summary when the bandwidth of the bottleneck is low we perceive time passing in the environment as normal. When the bandwidth of the bottleneck opens up more information per unit time has the effect of perceiving the normal rate of time as slowing down.

If this train of thought is disproved the memory tests are viable, but if not it illustrates how the memory tests are valid only for measurement of the background bottleneck bandwidth. A further more interesting aspect of this not directly related to the passage of time thread is whether the detachment of the model from the immediate world around us is unique to the human and what evolutionary advantage this would have brought us. The latter of these could be what we cherish as free thought, and it is conceivable other primates in particular have this capability.

Jim

Jim, my bottleneck refers to the peak rate at which novel information can be integrated with the internal model. The actual rate at which novel information is absorbed at any moment will depend on the task being addressed, and more specifically the degree of concentration on external stimuli. During our lifetime, this might fall significantly on average (we know more than we want to learn), but probably is similar for the peak values (some of the memory experts are not young).
So I agree with you that in practice the bottleneck "has a dynamic bandwidth". Though I am not sure whether you are saying it can be much faster than my estimates?

As for what happens while "time stands still" during a critical experience, I think that this is when our internal model has insufficient experience to provide a rapid update, and that we efectivly are waiting for updates via our senses. We are having to create a new model "on the fly".

I recognise that the memory tests are not the best, as the task is quite unusual, and complex in terms of the number of things to be remembered. The simple maths records of adding a sequence of digits and simply remembering the running total gives higer bit rates. However apart from these tests, there are very few tasks that can be accurately quantified.

Does it have a fixed bandwidth set by evolution? you ask.
I think that the bandwidth wil be lower for more complex internal models, so a simple creature might have a higher bandwidth. It is possible that humans have a lower rate than simpler creatures when exposed to the same experience. This is because the human has a bigger world view, a bigger context in which to integrate the simple experience. A black dot passes by in the sky. The dragonfly quickly registers it as potential food, (an insect ), and responds, while the human may consider is it an airliner?, a UFO?, a bird?, what type, an insect?, a floater in his eye? etc. before responding. I realise that I am talking delays rather than bit rate, but you probably get the idea.

Richard

Hi Richard,

There are two interesting aspects arising here. I agree that the rate of integrating new information into the model is relatively slow. The slow rate of this I believe is apparent from the long childhood, etc observed in humans. When a model is up and running it provides the human with a sense of self, which we call consiousness. This blog thread is about how the running model interacts with the outside world. Previously I had speculated that the bottleneck observed in the slow rate of learning was related to the bandwidth of the communications channel between the senses and the running model. Now I am speculating the bandwidth of the communication link is variable and can go much faster than the measured rates here discussed.

This means that irrespective of how much information is presented to the conscious mind it will only build new model routines at a relatively slow rate. The bottleneck may therefore be the maximum rate of model build rather than the maximum rate of information into the model.

The second interesting thing is how we have both interpreted the perceived slowing down of time in opposite ways. In your description the model has no previous routing to cover falling off the bike. (Not true in my case I am sorry to say!). The model then begins to operate in a non predictive manner and time seems to slow down as the model must operate at the speed of information coming into it. In my interpretation the model is detached from direct interaction with the immediate environment it is connected to via the senses. When something important takes place the senses get the models attention and the communication link bandwidth increases. In normal use the BW is low, so the number of bits of information received per unit time that link the model to the real world are low. When the bandwidth of the comminication link increases, the number of bits per unit time increases. This give the perception of a lot of model time passing compared to when quiescent in the same real time interval. Hence perceived time slows down.

I think we both agree there is a bottleneck and a predictive model. The bottleneck causes a limitation to how fast the human can absorb new information. The next question is what is the nature of the bottleneck, is it a limitation of adding to the model, or a limitation of the communication bandwidth between the senses and the model?

Jim

Jim, Re "The next question is what is the nature of the bottleneck, is it a limitation of adding to the model, or a limitation of the communication bandwidth between the senses and the model? "

I believe that it is the former. However it is possible that we may have evolved some limitations in "communication bandwidth between the senses and the model" where it gained no performance advantage. Where speed of response is critical the processing is minimised and we have crude reflexes, as possessed by many simple creatures.

For example a "focussed" 2D visual observation must be incorporated into our 3D conceptual model if we are to develop our model. This is a complex processing task, and likely to be slow. A reflex to a sudden shadow in our peripheral vision, may rapidly bypass all that processing. richard

Hi Richard,

Another thing occurred to me that could be interesting following on from the train of thought in my previous reply. Your experiments with memory have indicated the present of a buffer. It seems possible that model construction could be carried out offline. In other words not in real time. If the communications bottleneck was variable and recorded a lot of interesting information into memory, this could be used later to increase the model capacity, but not in real time. This seems an intuitively probable occurance to me and may have connections to dreaming.

Jim

Jim Re Buffering
I am interested in the sustained information rate, as this will determine the constraints around lifetime learning.

"model construction could be carried out offline"
I think we need to be careful not to confuse with the creation of information internally as a result of constructing ever more complex models based on a lot of what has entered before.
there is a difference between model refinement, and the addition of new information from outside. There will be lots to be gained by comparing different internal models.

My fractal program on my PC can spend a lifetime creating ever more detailed internal fractal landscapes, once I have input the trivially small program instruction.

The high speeds achieved when multiple digits are briefly flashed up, do not indicate that high sustained rates are possible, as the senses are effectively shut down while the observer recalls what was seen. Richard

Jim, you mention of dreaming reminded me of the experience of just how fast I can dream.
I have woken, looked at the clock, seen that I don't have to get up for another 5 minutes, turned over and then had a long complex dream (seeming hours long) before being woken by the alarm 5 minutes later.

I think that there is brain scan evidence that similar brain regions participate whether we act, visualise or dream. So maybe the switching off all the information through the bottleneck (sleeping), allows the simulation to run really fast! I wonder if a problem solving dream would run slower? Richard

Hi Richard,

The rate at which we can dream, whether awake or asleep is another pointer. Model time is internal and only connects to real time through the external senses, which must go through the bottle neck as you describe it. Our perception of the passage of time is therefore set by how much attention we are paying to the environment around us, or in some cases how much demand the envirnment is placing on us. If the environment is causing us to pay too much attention to small amounts of input per unit time we find it tedious and time goes slowly. If the environment causes us to pay attention to a large amount of data per unit time we find it exciting or scary and perceivced time slows down. I suggest the reference point for perceived normal time is the bottleneck bandwidth. The model couples itself to the real world through this. When decoupled, through some form of dreaming, perceived time is model time, which is totally unreferenced to anything measurable externally. Therefore the communications link between the senses and the model is capable of a range of bandwidths but has an average which causes us to perceive the passage of time. The communications link bandwidth rate is also controlled. We pay more attention to important things, and this in turn sets the amount of information per unit time we take in, whether the model can operate predictively or not. We often re-live rapid events afterwards and make sense of them. However this is an imperfect process which often results in something useful to the individual, but not necessarily objectively accurate in terms of event interpretation and recall. I'm not sure what could be concluded by that if anything.

Jim


Jim

"I suggest the reference point for perceived normal time is the bottleneck bandwidth."
Jim, I don't agree. In normal time I am experiencing a mixture of my internal simulation and some sensory input. The bottleneck constraint is surely only experienced when all my attention is on the "incoming", the hard learning activity.
Richard

Hi Richard,

Yes and no! I propose that the perceived experience of real time is caused by the coupling of model time to real time over the bottleeck. So if the average bottleneck bandidth provides input per unit time this is the average sensation of the passage of time. Model time is detached from this. So if I am doing some thinking and also taking some environmental input, I experience time passing according to the number of bits per unit time coming into my consciousness from the environment. My model time experience can be extremely rich, but I know that this is unreal and only regard the part of my model that is paying attention to my environment as my reference. When the amount of information from the environment per unit time goes up or down, the bottleneck effect keeps the average linkage to the model fairly constant using prediction to reduce necessary input. When the prediction is unable to cope with the envirnmental circumstance more information needs to be processed per unit time. This has the effect of perceiving time as slowing down, as I experience a lot more model to real world synch events per unit time which makes me think a lot more time has passed during the same real time interval - time slows down. The opposite is true when I experience something that forces me to take in less information per unit time from the environment than I normall would. This speeds time up, in that the number of model synch events per unit real time becomes less. I think less time has passed, time slows down and it takes hours of real time to achieve minutes worth of model progression which makes me frustrated at the tediousness of having to pay a lot of attention to a small amount of input.

The difficulty in this one is the perception of time. This is how I experience the passage of time and cannot be measured.

Jim

I think that's exactly opposite to what happens in those moments - when there is a true "time slows down" experience, thye information is hugely rich. And rather than there being a rolling storyline, there is just moment-to-moment experience. The only specific trait of humans that separates us from other animals is that we make up stories (and I include mathematics in this). And stories are not new information - they are re-hashed old information. When we get away from that storytelling and experience without a storyline, then the amount of information we can percieve increases substantially. otoh, from my own experience, you appeared to be describing a slowing of time due to boredom - rather different.