User:Graeme E. Smith/Declarative Memory: Architecture Scope and Limitations

Declarative Memory Architecture, Scope and Limitations Graeme E. Smith, GreySmith Institute of Advanced Studies  http://en.wikiversity.org/wiki/Portal:GreySmith_Institute   http//en.wikiversity.org/wiki/User:GreySmith_Insitute  grysmith@telus.net The Declarative Memory is that portion of the memory that we tell ourselves that we have. It is not meant to be a separate memory, however because there is a requirement that we tell ourselves that we know something, only those elements that get into the Meta-index are truly retrievable. It is for this reason, that H.M. a classic case, developed amnesia. The cerebral cortex was more or less intact, however he had had the medial temporal lobe excised and the excision had removed the hippocampus area where the Meta-index is thought to exist. This article explains the nature of H.M.'s amnesia, and why it argues for a two stage declarative memory. As well it covers a plausible but not yet proven explanation for a meta-cognitive "Feeling of Knowing".

You can know something, and not have access to it. At first this statement seems ridiculous. how can you know something and not be able to retrieve it? The answer is simple, you can't demand a memory that you don't know that you have. The reason we call declarative memory, Declarative, is that at some level you have told yourself that you have that memory, and when you try to recover the memory later, you can often tell yourself again. There are therefore two ways that we know that we know something, 1. we tell ourselves that we know it, and 2 we store information away about what we know so that we can recall the information on what we know, later. This storing away of information about what information we know, requires some form of index, that will allow us to find the information later, and telling ourselves that we know it, means we need some mechanism to search the index, and find the index entry for that element before we try to find the actual element. The first phase in declarative memory is the declaration of knowledge. Somehow we have to map our knowledge onto an index somewhere. Allen Baddeley discussed the requirements for a working memory a number of times over the last 20 years. Part of what he discussed, most recently is the need to have an episodal memory buffer. Episodal memory is memory that has been located in time and space. With the discovery of the place cells in the hipposcampus we have a location for the episodal memory, however there is evidence that the hippocampus is one of about 4 separate memory systems all serviced by the Subiculum. The list includes the hippocampus area CA3, the hippocampus area CA2, the parahippocampus, the entorhinal cortex. Evidence seems to suggest that sequences from the parietal lobe are passed through these areas, as evidenced by slow waves that are detectable in this area both during the waking period and during some sleep states. What some scientists think is happening is that during these slow wave episodes, patterns are detected in the parietal sequences, and segregated out into different mapping mechanisms, which can then be accessed from the central subiculum. Evidence is that the parietal sequences are oriented around the experience of the organism as it goes through its day, while the episodal memories are more oriented towards mapping the invarient aspects of the environment. Thus a rat that does not have the use of it's hippocampus can't orient itself as easily when the maze is turned a quarter turn, as the rat that can orient itself to the cues on the walls of the lab, and has full use of its hippocampus. To add to this oppinion as to the location of the Place cells in the hippocampus, other researchers have found evidence of a mapping grid generator that is part of the Entorhinal cortex. Different layers of the cortex control differently oriented grids, so, by adjusting the source of the grids it is possible to adjust the orientation of the animal with regard to place. This means that the organisms experience of movement and direction can be mapped simply by adjusting the grids to accomodate the changes in direction of the path the organism follows to receive its inputs. If the rat turns left, it traverses the mapping grids in a direction that turns the grid to the right, and visa versa. The result is a semi-stable map of the environment around the organism that includes elements experienced by the organism. We call this map episodal memory. Once we have defined the map, the next stage is setting up access to it, via an index or legend of some type. In other words we had to tell ourselves where we were, now we have to tell ourselves that we told ourselves. This is similar to the effect of mapping the implicit memory onto the explicit memory, and may involve the exact same type of interface. The main difference is that because the map defines a standard location for a particular piece of information we do not need the bottleneck to map across between implicit and explicit forms of the map. There is still probably a learning curve associated with learning to use the map, since the map is not the reality it maps, and an organisms experience of where something should be in implicit memory often needs to be audited against later experience of trying to find it in the real world. However the explicit version of the map, can be easily indexed via the subiculum which acts as a bottom-up attention mechanism for the meta-index, or index of indexes that is implied by the multiple organs making up the index itself. Once we have the memory mapped, and can index it to find it, the next job is to tell ourselves whether a particular memory is in the index or not. We still have no proof how this is done, however, it is possible that we do an index search, and then use that to trigger a signal called a feeling, that indicates success or failure to find the item in the index. This Feeling of Knowing or FOK, is obviously a meta-cognitive signal, in that it tells us whether or not we can remember knowing something. In the model I use, the concept is that a search is done by the subiculum of the meta-index in the hippocampal areas described above, and from this search a signal is generated on a success / failure basis. The index value and the success signal then indicate readiness to attempt to recover the quale from the Cerebral cortex. This does not guarantee success however because the index might be out of sync with the Cerebral Cortex. The Cerebral cortex retrieval operation then might succeed or fail. If it fails a failure detector, triggers an attempt to recover the memory by searching the immediate area around the index entry, if it succeeds no further care is needed, however sometimes the index is so far out of sync with the actual implicit memory that a protracted search is needed. During this protracted search, a signal called Tip of the tongue or TOT is detected. Many people report this Tip of the tongue feeling as lasting for protracted periods as various techniques are tried and discarded until the memory either finds the entry or discards the attempt to find it. A similar meta-cognitive feeling is often experienced with memories that have been accessed recently, in that those memories are often tagged with a feeling of familiarity. Usually both a feeling of familiarity and a feeling of knowing that happen together assures success at retrieval, however, in other circumstances they just add to the frustration of the TOT signal. The reason that we need the TOT signal is simple, phenomenally implicit memory rearranges memory locations over time, the actual location of a memory is only temporarily knowable, and so when the implicit memory moves it, the explicit memory often takes a while to catch up to the move, and the declarative memory must wait for the next time that location is declared. TOT gives us an opportunity to try a few different techniques to access it, without waiting for it to be rewritten to declarative memory. Now lets introduce the idea of the BSO or Beta (frequency) Synchronized Oscillation. Like the implicit memory, the meta-index may produce multiple index entries that all point to a particular memory. What is needed is a method of selecting from among the multiple entries, one that will be used to address the Cerebral Cortex. To do this, we take a leaf from the thalamus and impose a frequency on the individual entries so that the PFC, probably a portion of the ACC, can select from among them. However in order not to confuse the memory system we impose a different frequency range on the signals. In this case, evidence suggests beta waves. If this is correct, then the beta waves signal the ACC that the signal being chosen is an index rather than a functional cluster. There is probably a priority pattern that selects beta frequencies before gamma frequencies. Evidence suggests that this beta frequency gets transferred over to the thalamus, thus double tagging the memories that are referenced by the declarative memory so that they resonate both on the beta and gamma range. This might be a good reason why the memory does not use a carrier wave like technique it is difficult to impose two carrier waves simultaneously. The additive technique used allows the signal to resonate on both bands simultaneously, essentially striking a chord. Theoretically this single stage declarative memory is all that is needed, or at least that was the theory until H.M. had his most unfortunate operation, and showed by his amnesia that the theory was incomplete. H.M. was a brain surgery patient that had his Medial Temporal Lobe taken out as part of a medical procedure that has since been deprecated as a result of his experience. When H.M. recovered, it was found that he had a profound amnesia, that included both a anterograde aspect, where he couldn't form long-term memories, and a retrograde aspect where some memories that where thought to be already formed were found to be missing. The anterograde aspect was explainable by the loss of the episodal memory, but the retrograde aspect puzzled scientists for a while. Eventually they decided that there was a step called consolidation that happened in the hippocampus area, that was not completed for about 2.5 years after a memory was first formed. Within the model that I am proposing here, what was removed was his Declarative Memory System. But what seems obvious now, is that there was evidence of a fall back memory system, and archive if you will of the Declarative Memory that ran about 2.5 years after the actual Declarative Memory System. This archive had to be consolidated, probably into the cerebral cortex, in order for it to be available for H.M. to use. Consolidation stopped at 2.5 years before the operation. Annette Karmiloff-Smith in her book Beyond Modularity: A Developmental Perspective on Cognitive Science noted in her notes on childhood development that there was about a 2 year gap between learning one phase of language and learning the next, that indicated that while children could learn language at about 2 years of age, and could learn to manipulate words at about 4 years of age, that it took until 6 years of age for them to be able to discuss their manipulation of words. This suggests, at least to me, that in fact, what she is describing is the need to complete consolidation, before you can advance your learning based on the results of that consolidation. I have written a book called That explains intuition in respect of consolidation and some techniques about how you can recover the memory that has not yet been released for use because it has not completed consolidation and inclusion in declarative memory as a result. Essentially what we have tripped across here, is a role of consolidation in updating the implicit memory. Since the memories in the declarative memory are based in the hippocampus areas, it makes sense that the consolidation process involves the update of the cerebral cortex from information derived during Declarative processing. This must involve the rehearsal of the involved functional clusters in order to retrieve the implicit contents that the new index entries point to, which is needed to process them and thus to have a parietal sequence in order to load into the Meta-index. In other words, the reason it takes so long to consolidate, is that consolidation happens during sleep, and requires the use of the bottleneck. We think that consolidation might be related to REM sleep states. The heavy use of sleep states for Declarative learning, indicates that the brain puts this at a slightly lower level of priority than quick reactions during daylight hours.