Summary Prepared by Melissa Mitchell (additions and comments by Frawley; for additional comments go to Short summaries of presentations by Landau and Intraub)
A. the SENSORY REGISTER -- a primitive, precategorical memory that stores for a fraction of a second information from each of the senses.
B. the SHORT TERM STORE -- an active, conscious, working memory with a capacity of 7+\-2 chunks of information, lasting several seconds.
C. the LONG TERM STORE -- a large capacity, unconscious memory.
Information from the sensory register is lost or sent to the STS, and info from the STS is lost or, if there is enough time for encoding, sent to the LTS. There is a continuous interchange between the STS and LTS.
This model counters the popular view, that memory is continuous. But many empirical studies, both on normal populaitons and amnesics, point to the compartmentalized view. Experiments on normals involving free recall show a robust phenomenon called a serial curve, a graphical representation of the primacy and recency effects. The primacy effect means there is a greater recall of words at the beginning of the list because there is more time for rehearsal, and the recency effect means there is a greater recall of words at the end of the list, since these are drawn directly from the STS.
Studies of amnesics show further effects. A man named HM was rendered a retrograde amnesic (loss of memory prior to trauma) as well as an antereograde amnesic (loss of memory for new material) by massive brain removal (bilateral removal of hippocampus and amygdala). His short term memory appears normal, but he cannot remember new information. Studies involving HM and other amnesics, such as teaching them to read mirror writing and do Tower of Hanoi puzzles, show that though they do not remember the learning task, their performance improves each time. This is the result of priming, which seems to be stored in a procedural (skill) memory within the LTS. On the other hand, their failure to preserve information over time suggests a deficit in their declarative store. All this in turn suggests that LTS has at least two components, becacause they can be dissociated: procedural memory (or memory for operations and skills; declartative memory, or memory for explicitly stored information or data structures.)
In addition hearing to Prof. Intraub's lecture, the class also read Green et al., Chapter 10, Learning and Memory.
LEARNING can be defined as the acqusition of information, knowledge, or skills, though any definition is incomplete and inaccurate.
One learning method is called CLASSICAL CONDITIONING, developed by Ivan Pavlov in 1929. In this, an UNCONDITIONED STIMULUS (motivational event), causes an UNCONDITIONED RESPONSE (natural reaction). The CONDITIONEDSTIMULUS then replaces the US as a predicative event, causing a CONDITIONED RESPONSE (learned reaction). Classical conditioning depends on a sequence of events, the strength of the conditioning depends on the least possible time interval between the CS and US, and the process can undergo EXTINCTION.
INSTRUMENTAL CONDITIONING, proposed by Edward Thorndike, works similarly, but REINFORCERS (rewards) and punishers are used to shape behavior. His LAW OF EFFECT says that a response is more likely to be repeated if the outcome is pleasant, and vice versa.
There are three main theories of memory: A. Memory is a series of stored facts.
B. Memory is a record of mental processes carried out on stimuli. These may be (1) DATA-DRIVEN, guided by perceptual input, or (2) CONCEPTUALLY-DRIVEN, guided by knowledge and deriving meaning.
C. Memory is actually TRANSFER-APPROPRIATE PROCESSING, meaning that memory benefits from an overlap between the mental operations performed at study and recall.
Memory itself is thought to be compartmentalized. There is a short-term and a long-term memory (see Intraub lecture). Within the short-term may be a WORKING MEMORY, storing for a fraction of a second verbal and visuo-spatial information. The long-term memory contains a PROCEDURAL MEMORY, unconsciously storing skills and habits, and a DECLARATIVE MEMORY, a conscious record of facts (in a part called the SEMANTIC MEMORY) and personal episodes (in another part called the EPISODIC MEMORY). From experimentation, the separation of the STS and LTS is almost certain, and there is a good case for the working memory. However, experimentation on normal minds, as opposed to those of amnesics, have difficulty showing separate procedual and declarative memories.
Memory and learning can be studied by RECALL (retrieving episodic information on cue) or RECOGNITION (judging if information has been seen before). Memory and learning can be said to be two sides of the same coin.
FORGETTING (inability to retreive learned information) is caused by RETROACTIVE INTERFERENCE, when memory traces are interfered with by newly learned information. This may be because the memory was "overwritten" or because of a retrieval failure.
Another issue in learning is the character and properties of CONCEPTUAL KNOWLEDGE. Some believe concepts are stored in mental representations called PROTOTYPES, and learning the concept means abstracting the prototype. All members of a concept were not thought to be equal EXEMPLARS of the idea, depending on how many qualities they shared with the prototype (see Landau lecture).
In aritificial intelligence, this process is analogous to PARALLEL DISTRIBUTED PROCESSING, in which complete entities (prototypes) are represented by "micro-features" (qualities within) distributed across a set of units.