1. Associative Theories of Long- Term Memory

2. Network Theory The basic notion that we need to explore is that memory consists of a large number of associations. You know who I am… how? There is an association in your LTM between a visual representation of me, and my name. There are probably other associations there as well…

3. Network Theory - Basics How do we search memory? –We start at one memory node, and let activation spread to the associated nodes via connections. You know that a “rose is red” because the node for “red” and “rose” are connected. Why do I can’t I retrieve some info.? –Some connections are weak. You may have a hard time remembering a particular study it is because the associations/connections are weak. How do I strengthen connections? –Pay attention to information, and use elaborative rehearsal whenever possible.

4. Indirect Evidence for Network Theory Hints –Work to add activation to the appropriate connections. Example: who was Richard Nixon’s VP? Mnemonics –Work by creating several strong, yet indirect paths between to-be-associated nodes. State Dependency –Your emotional state when learning is represented as a node, and can act like a retrieval cue.

5. Direct Tests of Network Theory Lexical decision and associative priming –If a words is preceded immediately by an associatively related word, the RT for that second word will be faster than if it had been preceded by a control word.

6. Associative Priming Example Let’s say that we have a very simple network that represents an association between DOCTOR and NURSE that looks like this: DOCTORNURSE 1.When I present DOCTOR as the prime, it becomes activated. 2. That activation begins to spread to NURSE. 3. NURSE becomes partially activated. So, the RT to respond to NURSE is faster when preceded by DOCTOR because NURSE is partially activated.

7. Collins and Qullian (1969) Used a sentence verification technique to test their theory of associative memory. In this task, Ss are presented w/ sentences such as: A canary has feathers Ss are to respond as quickly as possible whether or not the sentence is true or not. Collins and Quillian argued that the RTs should be dependent on how many “links” between nodes had to be traversed before activation intersected and Ss could respond. Let’s take a look at their model

8. Animals Cat Cow Breath airHas heart Says “moo” Has claws C & Q argued for hierarchical organization with cognitive efficiency… common is stored at the highest level. Given the sentences “A cow says “moo” and “A cow has a heart” which should have the longer RT according to C & Q?

9. Problems with Collins and Quillian Animals Birds Have feathers Peacock Canary The model predicts the same RT for “A peacock has feathers” and “A canary has feathers,” but the peacock sentence is faster. Predicts the same RT for “A peacock is a bird” and “A canary is a bird,” but the canary sentence is faster.

10. So, although Collins and Quillian were not terribly accurate with their conceptualization of the structure and organization of LTM, they did lead us in some interesting directions. Collins (Collins & Loftus, 1975) later argued that memory is organized based on the principle of cognitive distance. More closely associated items are simply closer together in LTM. RT is a function of the distance that needs to be traveled between nodes. So, in the last example, “A peacock has feathers” was responded to faster than “A canary has feathers” because feathers and peacock are closer together in LTM.

11. Direct Evidence cont. Anderson (1974) decided to manipulate the number of associations for each node, by using the “fan” paradigm. In this paradigm, Ss first learn a number of associations between concepts, such as “The Lawyer is in the Park.” Following learning, Ss complete a sentence verification task like we have already discussed.

12. PARK CHURCHLAWYER(FAN SIZE = 3) BANK BOAT SCHOOLPLUMBER(FAN SIZE = 2) HOSPITALARTIST(FAN SIZE = 1) The Fan Effect RT Fan Size 123123

13. Alternative Explanation of Fan Effect Radvansky has offered an alternative explanation of the fan effect, based on mental models. In his situation model theory, he argues that the fan effect occurs because subjects must create individual situation models in Anderson’s studies, and searching separate models takes a long time.

14. Situation Model Explanation of Fan Effects Multiple Location (ML)Single Location (SL) DOCTOR LAWYER PLUMBER PARK STUDENT CHURCHHOSPITAL AIRPORT RT Fan Size 123123 ML SL

15. Evaluating Network Models Computer simulations. –Have shown limited success. Tip of the Tongue (TOT). –Problematic for networks… something more than simple activation must be needed. Homunculus problem. –Activation needs to be guided by something… what is that something?

16. PDP Models Representations are distributed throughout the entire system. –A representation is simply a pattern of activation. PDP models are more biologically plausible than network models. –Nodes can be seen as neuron-like. –Connections can be excitatory or inhibitory. We don’t need a homunculus. –Is there a philosophical dilemma here? What about free will?

17. PDP cont. Learning occurs by adjusting strength of connections. Some problems: –Learning rules don’t always make biological sense. –Models learn well when the deck is stacked by programmers. –Models are very specific in their focus. There are no “grand” models of cognition.