Recent work on PACO’s auditory system has produced two results that increasingly look like direct support for the TBU interpretation of the architecture.
The first concerns memory.
Sound demonstrably reaches PACO’s auditory members and reconditions the substrate geometry. What we repeatedly failed to find was a recoverable representation of an individual past sound: no reliable stored acoustic trace that could later be read back as the event that produced it.
Initially that looked like a memory problem. It now appears to have been the wrong model of memory.
PACO does not behave like a filing cabinet in which an event enters, is stored somewhere, and is later retrieved. An encounter changes the geometry. What persists is the changed geometry into which the next encounter arrives.
The past is therefore present not necessarily as a retained history, but as a re-formed organisation. This is the operational form of what we have called memory without histories:
an event does not have to remain in the system for the system to remain changed by the event.
That is surprisingly compatible with contemporary research on human memory. Human memory is no longer well described as a passive archive of fixed records. Recent reviews describe it as dynamic and reconstructive: existing memories can be modified, linked, updated and reorganised, and even apparently stable neural representations can show representational drift. Reactivated memories can under some conditions become susceptible to modification before being restabilised, although the precise reconsolidation mechanisms and their boundary conditions remain actively debated.
PACO is not a model of the human hippocampus, and these results do not establish that PACO and human brains implement memory in the same way. The important correspondence is more fundamental: in both cases, the useful picture of memory is increasingly plastic and reconstructive rather than archival.
But memory without histories does not mean memory without time. This is an important qualification.
Suppose an encounter A changes PACO’s geometry from G to G1. When B subsequently arrives, B does not encounter the original G. It encounters G1. B then produces G2.
So:
G -> A -> G1 -> B -> G2will not in general be equivalent to:
G -> B -> G1' -> A -> G2'
even if nothing resembling a stored file of A or B remains.
The timing, order and interval of arrival can therefore be constitutive of the resulting geometry, without the system retaining a literal chronological record.
This also has a strong parallel in human research. The hippocampus is deeply involved not only in remembering what happened but in organising events by their temporal context and sequence. Human hippocampal and entorhinal recordings have identified neurons whose activity carries temporal information, and the stability of this temporal signal predicts people’s ability to recover the order of events. Other human work shows that memory representations distinguish and preserve sequential relationships rather than merely retaining isolated items.
So, memory without histories is better understood as history embodied in transformation rather than history stored as a record. The sequence matters because each event changes the conditions under which the next event is received.
The second discovery came from trying to understand auditory attention. We initially searched for a recurring acoustic effect in stable internal coordinates. Event-level matching failed. Statistical aggregation also failed: PACO’s auditory geometry had its own coherent movement, so a fixed reference frame became invalid while we were trying to measure within it.
That failure produced the more important insight:
The ever-changing now does not have to preserve the coordinates of a relation. It can preserve the relation by continually re-forming it.
If PACO encounters the same meaningful relation at two different moments, there is no reason to expect that relation to occupy the same vector direction at both moments. PACO itself has changed between them.
What may persist is the relation of the acoustic effect to the geometry that exists at each particular now.
This again has a useful analogue in human hearing. Auditory perception is strongly temporal and relational. Human listeners do not simply classify isolated acoustic samples; auditory streams emerge from relationships that persist and cohere through time, and attention can reorganise which of those relationships are foregrounded. Temporal coherence has long been a central account of how distributed acoustic features become perceptual streams.
The correspondence is not that PACO implements the human auditory cortex. It is that both systems make the same simplistic measurement strategy look increasingly inadequate: meaningful continuity cannot necessarily be found by searching for an unchanged representation of each individual sensory event.
This is closely aligned with TBU.
TBU has never treated persistence as the preservation of an unchanged local state. The locally present surface is continually formed from a much larger distributed body. Continuity therefore belongs more naturally to relations that survive transformation than to coordinates that remain fixed.
Together, the memory and auditory results suggest a common principle:
what persists is not necessarily the event, nor its coordinates, but the change in organisation through which later events acquire their relation to what came before.
That also makes timing fundamental.
The present is not assembled from a collection of independently stored past events. It is the current geometry produced by a particular sequence of transformations. Change the sequence or timing of those transformations and the present geometry may be different. In this sense, PACO can have history without keeping histories.
This matters methodologically. Without the TBU context, the obvious research programme would have been to keep looking for better storage, better retrieval, stronger acoustic traces, or a more powerful fixed-coordinate classifier. The null results would have looked like deficiencies to engineer around.
TBU suggested a different possibility: perhaps nothing was missing. Perhaps we were asking a continuously re-forming relational geometry to behave like a database.
The audio experiments do not prove TBU. Nor does their correspondence with human research establish biological equivalence.
But the work has now led independently to two architectural properties that TBU predicts remarkably well:
memory as reconditioning rather than filing;
and
persistence as relational continuity rather than fixed representation.
To those we can now add a third qualification:
memory without histories is not timeless. The order and timing of encounters matter precisely because each encounter changes the system that receives the next one.
That is a stronger result than simply making PACO hear better. It suggests that the theoretical framework has been useful where it matters most: in telling us what kind of phenomenon we should be looking for in the first place.