Prof. Greg Francis 1/2/12 PSY 200: Intro. to Cognitive Psychology 1 Purdue University Working memory PSY 200 Greg Francis Lecture 16 Why there is a gate at the first floor stairway in the Psych building. Purdue University Modal Model of Memory  Atkinson & Shiffrin (1968)  Today we focus on the Short-term store (Short term memory) Purdue University Search of memory  How is memory searched?  Sternberg hypothesized three types of searches  Explore by varying the number of items in memory set (similar to visual search experiments)  measure reaction time  Sternberg (1969) 5 3 2 9 5 3 2 9 8 5 3 2 9 NO Purdue University 5 3 2 9 Types of searches  (1) parallel: target item is compared to all the items in memory at the same time  the answer (yes or no) is returned after all items have been checked 8 8 8 8 8 8 8 8 8 8 8 NO Purdue University 5 3 2 9 Types of searches  (1) parallel: target item is compared to all the items in memory at the same time  the answer (yes or no) is returned after all items have been checked 3 3 3 3 3 3 3 3 3 3 3 YES Reaction time is the same for a yes response Purdue University Memory search  If parallel search  number of items does not matter  Yes and No responses are both flat Set size Reaction timeProf. Greg Francis 1/2/12 PSY 200: Intro. to Cognitive Psychology 2 Purdue University 5 3 2 9 Types of searches  (2) serial terminating: target item is compared to each item one after the other  the answer (yes or no) is returned after the target is found or all items are searched 8 8 8 8 8 NO 8 8 8 Purdue University 5 3 2 9 Types of searches  (2) serial terminating: target item is compared to each item one after the other  the answer (yes or no) is returned after the target is found or all items are searched 3 3 3 3 3 YES 3 Reaction time is faster for a yes response Purdue University If self-terminating search  Go through items one-by-one until find target  RT increases with set size  YES RT’s shorter than NO RT’s  Lines have different slopes Set size Reaction time 01002003004005006007001 2 3 4 5 6YESNOPurdue University 5 3 2 9 Types of searches  (3) serial exhaustive: target item is compared to each item one after the other  the answer (yes or no) is returned after all items are searched (regardless of whether target is found or not) 8 8 8 8 8 NO 8 8 8 Purdue University 5 3 2 9 Types of searches  (3) serial exhaustive: target item is compared to each item one after the other  the answer (yes or no) is returned after all items are searched (regardless of whether target is found or not) 3 3 3 3 3 YES 3 Reaction time is the same for a yes response as for a no response 3 3 Purdue University If exhaustive search  Go through every item and then report answer  RT’s increases with set size  YES RT increases the same as NO RT’s  Lines are parallel Set size Reaction time 01002003004005006007001 2 3 4 5 6YESNOProf. Greg Francis 1/2/12 PSY 200: Intro. to Cognitive Psychology 3 Purdue University Hypothetical searches  So, we have three hypothetical ways of searching STM  They predict very different patterns of reaction time as a function of memory set size  Sternberg runs the experiment to see how the data comes out  You ran a version of the experiment in CogLab Purdue University Search of memory  Sternberg’s data support exhaustive search  Sternberg’s data support exhaustive search  Here’s the CogLab data (191 participants) =“Yes” =“No” Purdue University Search of memory  Implications: Search of STM  1) is serial, one item at a time » and checking each item takes approximately the same length of time » Approximately 40 milliseconds (CogLab data is a bit slower, 50 milliseconds)  2) is exhaustive » search always goes through all items Purdue University Search of memory  These results were a bombshell in 1969  finer analysis of cognition than anyone expected was possible  used a thought experiment about different types of searches to generate precise testable predictions about cognition » subsequent research found that there were other types of searches that complicate the conclusions  counter-intuitive finding » why should search be exhaustive? » seems inefficient! Purdue University Interpretation  Exhaustive search makes sense if search of STM is done by some process that is  very efficient (can search very quickly)  dumb (doesn’t bother to stop itself)  initiated by some other system (a controller) Controller Search memory process STM Purdue University Controller  Controlling attentional system  supervises  coordinates  starts and stops relatively independent processes  e.g.  Search short term memory  Search long term memory  walking down stairs  gate in psychological sciences building  DoorsProf. Greg Francis 1/2/12 PSY 200: Intro. to Cognitive Psychology 4 Purdue University Other aspects of STM  At about the same time, another study indicated important characteristics of phonological and visuo-spatial systems  Brooks (1968)  two types of tasks (visuo-spatial and phonological)  two types of responses (visuo-spatial and phonological)  Identifies two types of systems that are relatively separate Purdue University Separate systems  A complicated experiment  Part 1: spatial mental task (diagrams)  visual imagery  classify corners (top or bottom corner?)  “yes” if top or bottom  “no” if not top or bottom F Purdue University Separate systems  Part 2: verbal mental task  read sentence  categorize words (noun or not? A bird in the hand is not in the bush. Purdue University Two response types  Either  verbally  spatially no, yes, no, no, yes, no, ... YES NO YES NO NO YES YES NO NO YES Purdue University Results  Measure time to finish mental task for each response type  diagrams -- pointing  sentence -- pointing  diagrams -- verbal  sentence -- verbal Purdue University Results  Results  when you have to respond by pointing, it is easier to work with sentence information than diagram information  when