User:Graeme E. Smith/Phenomenally Explicit Memory: Scope and Limitations

Phenomenally Explicit Memory Scope and Limitations Graeme E. Smith, GreySmith Institute of Advanced Studies  http://en.wikiversity.org/Portal:GreySmith_Institute   http://en.wikiversity.org/User:GreySmith_Institute  grysmith@telus.net

Fodor and Karmiloff-Smith have between them defined the limitations of explicit memory. Fodor suggested that there had not yet been found a method of addressing phenomenal memory elements in such a way as to create addressing of distinct elements within the memory, and Karmiloff-Smith showed that memory happened in phases that depended on redescription. In the article immediately previous to this article, Phenomenally Implicit Memory: Scope and Limitations, I showed that explicit memory, depended on Phenomenal Memory elements, and that therefore the Scope of Phenomenally Implicit Memory overlapped the scope of explicit memory blurring the distinctions, but the main difference between the two memories was the place-code addressing that made explicit memory access possible. In this article I discuss the scope and limitations of explicit memory and why addressing mechanisms are critical to its classification.

I am a proponent of early redescription. There are some that will talk of implicit memory as if you could recall it. But phenomenally implicit memory cannot be recalled, it can only be triggered by content. This means that even in the case of the short term working memory suggested by Allen Baddely & Lawrence Weiskrantz as promoted in Attention: Selection, Awareness and Control, A tribute to Donald Broadbent there needs to be a separation in function between the implicit memories of primary perception, and the explicitly retrievable memories of the Working Memory that can be retrieved if only by rehearsal. I call this second type of memory phenomenally explicit memory, if only because as I mentioned in Phenomenally Implicit Memory: Scope and Limitations, it may be addressable using place code addressing but it's output is a quale, just as the output of implicit memory was. In The Bottleneck: A New Perspective I described the mechanism for conversion between implicit and explicit forms, and the necessity for a redescription stage was covered in The Dual Mode Cortex. What I want to stress here, of course, is that the first step beyond the conversion between implicit Content Addressing, into explicit place-code addressing, is the point where the Quale is first addressable by place-code addressing right after the bottleneck. Thus the bottleneck defines the earliest stage of Phenomenally explicit memory. If others accept this limitation and scope, and the primary difference between implicit and explicit memory is accepted as being the addressing methodology, then potentially Phenomenally explicit memory could have a scope that covers all forms of memory that involve place-code addressing. In fact, in Annette Karmiloff-Smith's Beyond Modularity: A Developmental Perspective on Cognitive Science she introduces the RR Hypothesis that defines three phases of Explicit Memory. While this interpretation is indeed reasonable if you assume that there are really only two types of representation going on in the brain, However, I believe that in an early redescription mechanism, the 2 year gap between phases of redescription, she notes in childhood development is problematic. While we certainly don't complete the redescription of large systems in a shorter period of time, we do perceive elements that could not be present under the scope and limitations of phenomenally implicit memory, suggesting that there is a relatively short turn around in early redescription, and that there might be a later form of redescription that takes more time. If this is so, then why would we need a new form of redescription? The answer lies in the topic of Declarative Memory. Essentially, Declarative memory is memory that we declare to ourselves that we know. I don't want to get tied up in the mechanisms such as the feeling of knowing (FOK) that tells us that we know something, in this article, because this article is about Phenomenally Explicit Memory, but that is the main difference between vanilla Explicit Memory and Decalartive Memory. However I posit the existence of an index of indexes or Meta-Index that we can use to find a specific memory in Explicit Memory. Because this Index is stored in a separate location in memory and uses a different type of addressing mechanism (I suspect the Subiculum takes the place of the Thalamus)mapping memory into the index, is not as simple as just remapping it in the same explicit memory. As a result, a Redescription Phase is needed to add to the index. This suggests a definition of redescription, If a new phase of memory requires a new area of the brain, and cannot be mapped across because there is no direct mapping convention that can be used, the brain has no option but to redescribe the memory. The mapping in the case of the Meta-index is associated with slow waves, and seems to be a mapping mechanism that maps memories to index elements. Part of the remapping involved might be the mapping of CHUNK arrangements to symbolic addresses in the Striate cortex, and translation back from the symbolic address to the CHUNK arrangements via the Nucleus Accumbens and Nucleus Reticularis Thalami. If this is a true reflection of how the translation is done in the striate cortex, then in episodal memory for instance, a single cell in the hippocampus could represent a symbolic address in the Nucleus Accumbens, which would represent in turn a CHUNK arrangement that would trigger a specific quale from the cerebral cortex, linking that cell to the information about a specific memory element. Using this symbolic form of addressing no object size would affect the mapping of location against the place cells in the hippocampus CA3 Area. a single cell in the hippocampus could trigger a massive functional cluster in the cortex. Such size-invarient coding has been noted at the hippocampus level, which supports this possible scenerio. The idea is that Episodal Memory, is just one of a number of different types of indexing going on in the hippocampus area, that the subiculum is able to act as the place-code addressing mechanism to recover specific elements at the cellular level from as many as 4 separate indexing mechanisms that all output in all probability through the Entorhinal Cortex. Making the Entorhinal cortex the index of indexes for the Cerebral Cortex. If we assume that this meta-index is part of declarative memory, then the fact that it involves multiple areas of the brain such as the striate cortex, the hippocampal area etc., suggests that it also involves a complex redescription mechanism, that maps cerebral cortex addressing to indexing maps. Thus remapping from the explicit memory system is not done directly, and Declarative memory must be seen to be a new representation of Explicit Memory and therefore outside it's limitations. However since the declarative memory has as it's object the selection of specific cerebral cortex memories using the explicit memory system, The explicit memory system is not outside the scope of the Declarative Memory system. Essentially anything that has been converted to explicit memory could be, also converted to declarative memory. However, this is unlikely since the redescription mechanism is complex enough to preclude every element of memory from the explicit conversion reaching the declarative memory. For my own version of RR, what I call the Extended RR Hypothesis, I have therefore suggested a slightly different mapping of terms. Instead of limiting myself to implicit and explicit memory, I have used three different terms Implict, Explicit, and Declarative. Also because I have defined redescription as being required when a memory phase requires a new location in memory, a new process of addressing and cannot be directly mapped to the previous representation due to limitations of the representation, I have defined a number of new forms of each term.
 * I1 Short Term Implicit Memory
 * I2 Long Term Implicit Memory (Requires different cellular process to achieve and is stored in physical processes rather than in synapses Redescription is biochemical)
 * E1 Short Term Explicit Memory (Address CHUNKS stored in working memory requires Bottleneck for conversion from implicit representation)
 * E2 Derivative Explicit Memory (Secondary Areas of Cerebral Cortex, requires recruitment of secondary perception and possibly tertiary perception to produce}
 * D1 Declarative Memory (Hippocampus area mapping from slow wave sequences, probably produced in parietal lobe}
 * D2 Declarative Memory Image in Cerebral Cortex, possibly inferior Temporal Lobe area (requires consolidation to complete archive image }

Because the Declarative Memory parallels the Main Memory Loop, and affects the recovery of individual memories from within the cerebral cortex it's scope overlaps the Explicit memory System needed to actually retrieve the cerebral cortex memories. This means that failures of the Declarative Memory are often seen as failures of the explicit memory such as in amnesia. This is why I have included it in the main memory mapping. Another system parallels the main memory loop, and this system is the SKILL Memory System. However rather than taking advantage of the striate cortex ability to symbolize addresses, the Skill Memory System takes advantage of the ability of the inferior olive to symbolize the actions of the attention mechanism I call the complicit attention system. Essentially, each action/function trigger in the cerebral cortex is echoed onto a single cell in the inferior olive, this symbolizes the action/function call, which is then used to activate a cell called the purkinje cell in the cerebellum. Information on context possibly redirected from the basal ganglia, or directly from the cerebral cortex, is then fed into the cerebellum via a different pathway. The standard implicit memory configuration links multiple patterns of activation to specific contexts. Feedback from the purkinje cells, is used to create pseudo-sequences, where activation of one purkinje cell feeds back into the context of another influencing its willingness to fire. An interesting effect, is one of time-compression, where the relationships are formed without reference to time, and so relatively slow movements become much faster under the influence of adrenaline. The general effect of this processing is that the cerebellum feeds back bottom-up memory addresses in sequences to the thalamus where they could address the cerebral cortex, and trigger actions or functions. In parallel, the Top-Down attention system is in direct connection to the SMA via the ventro-lateral PFC, the SMA collects information about the sequences triggered by the cerebellum, and derives information to allow the PFC to select from among the sequences. This information is sent to the Anterior Cingulate Cortex, where selection of bottom-up sequences is controlled by suppression of alternate quales. The result is that in some cases the ACC selects one sequence to complete without involving any other processing. However if it can't resolve the selection by itself it triggers a process called intention, that uses automation to make the decision if automation is available. Failing intentional selection, then the next step is to invoke Volition to make the decision. In the meantime, a wait process holds up the selection from among the sequences. The volition system either selects from among the sequences, or defines a new, and hopefully better sequence, which is entered into the Thalamus via the PFC and the Nucleus Accumbens/Nucleus Reticularis Thalami connection, thus triggering bottom-up addressing of the required actions/functions, and in doing so programming the cerebellum to offer a new sequence if it is selected often enough to lay in a memory trace in the cerebellum. This means that action/function sequences can be laid in in a top-down fashion even though the Skill Memory does not have a direct connection to the PFC. Because the Skill Memory acts on the cerebral cortex but does not have a direct connection to the PFC and therefore only automates what the PFC demands from the Cerebral Cortex, I do not feel that it needs to be incorporated in the Main Memory description of the Extended RR Hypothesis. However it should be incorporated into the ERR in some manner at some future date.