This seems like an exercise in looking for a perceptual correlate of the word ‘event’.
Bill uses the monosyllabic word “juice” as his example of an event as a short, well-learned sequence. Other examples include a glissando or arpeggio in skilled musical performance. I think that he was going after the distinction of being well integrated in the learned hierarchy, and that calling it a ‘level’ or ‘order’ of perception is misleading.
Frans claimed a correlation of Event perception with a ‘regression period’ in infants beginning at about 15 weeks of age and consolidating at about 19 weeks. He described it (in the book for parents) as " a short, familiar sequence of smooth transitions from one pattern to the next." A dropped ball bounces and probably will bounce several times successively less high. The description of children’s behavior suggests to me that amid the infinite possibilities of sequences (‘life is just one thing after another’), some recurrent short sequences are becoming familiar, and that unfamiliar “transitions from one pattern to the next” evoke anxiety (the regression).
We are saying that these are after all just sequences. That they are integrated into the learned hierarchy (that is, ‘familiar’) does not make them a distinct level or order in the perceptual hierarchy.
That frees the term ‘event’ for redefinition or reuse. But should we be going from the now undefined term to some phenomena taken as its definition? By what criteria do we identify the phenomena? “Change” is a poor criterion by itself, too general across the hierarchy.
In your broadened sense, Martin, my obtaining the Ph.D. is an event at at least two time scales, the entire 29-year process (an ‘event’ in your sense), and the award at a public event with a stage and audience. Other recognized events punctuated the whole in a partially ordered sequence, such as completing required courses (in absentia), responding to PhD exam questions (a 3-day event, plus additional time for a miffed professor), writing the dissertation, submitting it, receiving word of its approval, etc.
Here is PPC Figure II.8.1, for reference. Just a screen shot of the PDF display, saved to a file, which then is easy to drag and drop into the editing window.
I’ve been postponing replying to this thread, while it has been on my mind often. A few conflicts lurking in the dark. So I’ve gathered some courage and will try to get what I want out of this conversation. Step by step.
First: @MartinT, I am afraid it will happen so many times in the future that someone thinks of something and it has been there all along, in your book.
In I.7.3 The Imagination Loop, you describe exactly what I need, with the reference to predictive coding.
See the excerpts from I.7.3 below:
Taylor (2022, I.7.3):
“There is a hierarchic organization of controlled perceptions functionally equivalent to that proposed by Powers, but with some conceptual advantages. The idea for the alternative circuit was inspired by the following passage from Seth and Friston (2016), taking “prediction” to be equivalent to “reference value”: … descending predictions are compared with lower-level representations to form a
prediction error (usually associated with the activity of superficial pyramidal cells).
This mismatch or difference signal is passed back up the hierarchy, to update higher representations (usually associated with the activity of deep pyramidal cells).
In the Powers hierarchy, the perceptual signal value from one level is sent as an input to the perceptual
functions of control units at the next level above. But since the error value is the reference value minus the perceptual value, the same result could be achieved by sending to the next higher level not the perceptual value but the reference value and the error value (Figure I.7.3).”
In the alternative circuit, the Powers “Imagination connection” is never switched out, but is always active. Noting that the error signal is ordinarily taken to be Reference minus Perception (e = r - p), it is evident that p = r - e, and the higher level function could treat the combination of the two separate signals as though they were simply the perceptual signal of the Powers hierarchy.
Circuits that ignored the reference value but that used the error value would be good candidates as components of a reorganizing system, or as control units whose outputs influenced parameters such as the gains of lower-level units rather than their reference values. They might also serve as alerting units that monitored when an uncontrolled lower-level unit’s perception moved out of its tolerance zone, so that the unit might be brought into control once more, in a short-term form of reorganization. We will not pursue these possibilities here, but should keep them in mind as possibilities.
This alternative connection, inspired by Seth and Friston (2016), resolves a few possible issues with the Powers circuitry. Firstly, consciously we are able to perceive what we want to achieve, namely the reference value, and independently the margin by which the current state differs from it, the error value. Powers assumed that any perception of which we can become conscious must exist somewhere in the control hierarchy of perceptions, but in the standard HPCT hierarchy neither a reference signal nor an error signal does. With the alternative connection, higher levels can perceive both. Furthermore, with this interconnection circuit the higher-level perception can take more nuanced values than simply a choice between the actual sensory perception and the reference value; instead, it can attenuate or augment the perceived error signal from the lower level, so that the higher level perceives the lower level to have exaggerated error (as when one plays the piano for one’s teacher) or diminished error (as when playing the same piece for fun)
With this in mind, and skipping for now the other replies that are about different aspects of the question posed or the replies present: @MartinT, you decide to leave this architectural possibility as it is, and revert to Powers’ structure. But the case you make for more control possibilities is strong. Why not adopt this model?
Rick, @rsmarken, thank you for trying out the error-signal hierarchy!
What about Martin’s proposal, of combining reference and error signal? That would leave the hierarchy of perceptions of Powers’ intact, while opening possibilities mentioned in my first post and in Martins’ book.
You ask why not adopt this model, given that it allows for more possibilities? I have no evidence for or against either model, so I see no reason to adopt either. In fact now that I tend to believe that there are two separate but interlinked “conscious thought” and “non-conscious hierarchy” tracks of control, might both not be simultaneously correct (or neither).
