Memory is an inherently distributed process. The role of the hippocampus, which traditionally was viewed as a centralized store, is more that of an integrator/synthesizer that brings together disparate fragments of experience. Memories are reconstructions rather than recordings. This means that false memories can be formed in the presence of suggestion. Memories can be strongly encoded in one modality (auditory, tactile, etc) but more weakly encoded in others. Our ability to recall events is heavily dependent upon contextual cues. The following notes may be inconsistent with what is stated in the literature, as they are an interpretation / recollection.
Implicit vs Explicit Memories
- Implicit : Subconsciously-recalled memories that can still have an effect on how we behave. These are memories in their 'raw' state, before being integrated by the hippocampus. They are at least partially traceable to cortical regions. Implicit memories can be seen in word completion studies of amnesiacs.
- Procedural : A sub-class of implicit memory. Memories for how to perform a task. Procedural memories are quite often encoded at the level of the basal ganglia / cerebellum.
- Explicit : Consciously-recalled memories. Our experience of explicit memories is a result of the hippocampus synthesizing together our implicit memories.
- A similar taxonomy of memory types includes declarative (explicit) and non-declarative (implicit) memory. Declarative memory is further broken down into semantic (factual) or episodic (event-based) memory. Nondeclarative memory encompasses procedural, priming, conditioned, and nonassociative memories.
- In the programming/computer science world, the functional programming paradigm is analogous to declarative memory, while the procedural paradigm is analogous to procedural memory. There seems to have been another case of interdisciplinary conspiracy at some point.
- Perceiving something in the environment can cause low-level activation of an implicit memory. This can increase the probability that that memory will later be integrated by the hippocampus into conscious experience. It can also influence reasoning/problem-solving on a subconscious level.
- Trying to recall something without any cues (free recall) is much much harder than cued recall. Simply recognizing something is the least cognitively expensive form of recall (and ergo the easiest/fastest).
- Proactive Interference: When old memories prevent the retrieval of new memories (i.e., persistence).
- Retroactive Interference: When new memories prevent the retrieval of old memories (i.e., overwriting).
- Example: Learning key-value pairings with one set of keys/values, and then learning another set with the same keys but different values afterwards. When tested for the recall of the initial key-value pairs, participants have harder time recalling values when presented with a key.
- Fan Effect: The more complex a semantic network / network of relations between concepts, the more time it will take to identify a correct fact. This is because more complex networks take more time to traverse/search. This can partially explain interference effects (via Anderson).
- Interference effects occur when chunks of information have no obvious relationships to each other. Interference does not occur when redundant information is learned.
Working vs Short-term vs Long-term Memory
- "What fires together, wires together."
- The general idea behind cortical reinstatement is that the act of remembering an event directly engages the original modalities that encoded that event's memory.
- By this explanation of memory, during the act of remembering, the hippocampus serves as a nexus that integrates disparate circuits from multiple sensory centers. In this way, every time we remember something we are, to a degree, literally reliving it.
- By this framework, an encoded memory might be more like a set of aggregated C pointers, symlinks, metadata or a book index. A memory would be a structured set of references to circuits in other modalities where the memory is encoded.
- "The cortical reinstatement hypothesis of memory retrieval posits that content-specific cortical activity at encoding is reinstated at retrieval. Evidence for cortical reinstatement was found in higher-order sensory regions, reflecting reactivation of complex object-based information." -- From Bosch et al (2014)
Depth of Processing
- Think of memory as a graph, with encoded aspects of the environment as nodes. The higher the centrality (number of connections to other nodes) that a given encoding has, the more likely it is to be recalled. Since there are more edges to that encoding, the likelihood of reaching that encoding is increased.
- In practical terms, the more that you can tie a fact to your other memories / experience, the more likely you are to remember that fact.
- We are more likely to remember something if in the original context that the memory was encoded. Context can include inner emotional states (mood congruence) and physical locations as well as drug-induced states of mind. For example, studying on the subway will increase your odds of remembering what you studied on the subway. A memory that was encoded while we were angry will be better recalled when we are angry again. Also, a memory that was formed while intoxicated will be better recalled in subsequent intoxicated states.
- In practical terms, the more you vary your learning context (i.e., switch locations), the more likely you are to recall information in new contexts, since that information will not be tied to a particular place/time/state.
Primacy and Recency Effects
- When items are presented in a sequence you are most likely to remember the first and last items on the list.
- Recency effects are slightly stronger than primacy effects.
- One of the purported functions of sleep is to aid in paring down memories and consolidating our knowledge. Studying before sleep might be a good idea.
- Furthermore, learning seems to happen better at night?(Hockey, Davies, Grey, 1972)? Because of a high arousal state? I'm a bit skeptical on this point.