85-411/711 Cognitive Processes and Problem Solving Homework #5-Solutions Memory for Lists Simulation A. Forward memory strategy From these data, draw the serial position curve. First, tabulate how many words in the each position were remembered. For example, how many times was the first word in each of the lists remembered? Second, plot how often words were remembered (out of 20 experiments) as a function of serial position in the list. I should point out that the specifics of the rehearsal strategy being employed, both in the forward and backward memory experiments, are stochastic. Some of you noticed, for instance, that the number of words in each group and the rate of change from one group to another often varied. That was due to this stochasticity. Because there are some random processes involved, not everyone got the exact same serial position curve. But, of course, this is true when you collect human data as well. Here are the average data from 15 sets of homework assignments that were handed in. > See graph1.gif How does this serial position curve compare to the one found with human subjects (as described in the Memory for Lists experiment from assignment 1)? There are several features to notice. First, the EPAM simulation produces both a primacy effect and a recency effect. Second, the primacy effect is much greater for EPAM than for the human data from the memory for lists experiment. Third, the recency effect is much smaller for EPAM than for the human data. Several people reported a small spike in the curve at position 6. However, the large error bars for position 5 and 6 suggest that spike is mostly just a random effect. It is difficult to draw too many conclusions from a comparison between EPAM and the data from the memory for lists experiment because one used lists of 10 words, and the other used lists of 20 words. Furthermore, subjects in the memory for lists experiment could report the answers in any order, whereas EPAM had to report the answers in a particular order (Forwards or Backwards). Note, it is not the case, as some people suggested, that this curve is most similar to the delayed recall condition, as delayed recall produces no recency effect, whereas the EPAM simulation has a small, but significant recency effect. B. Forward memory strategy in closer detail After the to-be-remembered words are displayed, the contents of the auditory loop are displayed each time that EPAM rehearses or adds a new word to the loop. The auditory loop is a list of short-term memory items. Each STM item is composed of three things: a) a time stamp; b) the auditory name of the chunk; and c) to which group the item belongs. The time stamp indicates how many msec are left before that item will be lost from short-term memory. Items must be rehearsed at least every 2000msec or they will decay from memory. As each item is rehearsed it is given a time stamp, which indicates how much time is left before the item decays from memory. The time stamp is initialized to 2000msec. The group label can have one of 4 different values: 1 (b-group), 2 (m-group), 3 (e-group), or NIL (no group). When EPAM rehearses a group, it rehearses all the items in the loop with that label. Words are assigned to each group until it is full (each group has a fixed capacity). For the Visual Fowards experiment, each group is filled in order. In the Visual Backwards experiment, there is a strong preference to put the most recent item in group 3, and shift the other items down a group. 8. For each auditory loop printout, indicate which group has just been rehearsed. A few students had trouble with this question. Here is a printout of the contents of the auditory loop after a rehearsal: LOOP: 280-AIRPLANE-1 620-WOLF-1 880-MONKEY-1 1260-ROBIN-3 1640-BIRD-2 The oldest item (the one with the most time left before decay) is bird. It is part of the M-group (because it has the group label 2). Therefore, the M-group was just rehearsed. The auditory loop is printed out only after the rehearsal of a whole group, or after the addition of a new item. In this case, there was only one item in the M-group, so only one item was rehearsed: bird. Here is another auditory loop printout: LOOP: 200-ROBIN-3 580-BIRD-2 960-AIRPLANE-1 1300-WOLF-1 1560-MONKEY-1 In this case, the B-group was just rehearsed (label 1). There are three items in the B-group, so three items were just rehearsed: Airplane, Wolf, and Monkey. Sometimes, an item doesn't have a group label. For example, in the following printout: LOOP: 200-AIRPLANE-NIL 580-AIRPLANE-NIL 960-WOLF-NIL 1300-AIRPLANE-1 1640-WOLF-1 In this case, the older occurrences of Airplane and Wolf are not part of any group (designated by the label NIL), because there are more recent occurrences of both words that are part of a group already. That is, if a word is rehearsed, and an older copy of the word is still in the auditory loop, then the older copy no longer belongs to any group. Here is a highlighted example to show you what happens to old items in the auditory loop. Time Contents of the auditory loop Group 0 1560-AIRPLANE-1 B (new) 380 1180-AIRPLANE-NIL 1560-AIRPLANE-1 B 760 800-AIRPLANE-NIL 1180-AIRPLANE-NIL 1560-AIRPLANE-1 B 1060 500-AIRPLANE-NIL 880-AIRPLANE-NIL 1260-AIRPLANE-1 1640-WOLF-1 B (new) 1740 200-AIRPLANE-NIL 580-AIRPLANE-NIL 960-WOLF-NIL 1300-AIRPLANE-1 1640-WOLF-1 B 2120 200-AIRPLANE-NIL 580-WOLF-NIL 920-AIRPLANE-1 1260-WOLF-1 1560-MONKEY-1 B (new) 3180 200-WOLF-NIL 500-MONKEY-NIL 960-AIRPLANE-1 1300-WOLF-1 1560-MONKEY-1 B 3480 200-MONKEY-NIL 660-AIRPLANE-1 1000-WOLF-1 1260-MONKEY-1 1640-BIRD-2 M (new) 4540 200-MONKEY-NIL 580-BIRD-2 960-AIRPLANE-1 1300-WOLF-1 1560-MONKEY-1 B 4920 200-BIRD-2 580-AIRPLANE-1 920-WOLF-1 1180-MONKEY-1 1560-ROBIN-3 E (new) 5220 280-AIRPLANE-1 620-WOLF-1 880-MONKEY-1 1260-ROBIN-3 1640-BIRD-2 M 6280 200-ROBIN-3 580-BIRD-2 960-AIRPLANE-1 1300-WOLF-1 1560-MONKEY-1 B The group that was rehearsed is labeled at the end of each line. (new) means that a new item was added to the list at that point. The items that were just added or rehearse are underlined. Along the left is the time in msec since the beginning of the simulation. So, in the first line, the word Airplane was added to the B-group. In the second line, 380 msec later, the B-group was rehearsed. In the third line, 380 msec later, the B-group was rehearsed again. Essentially the time difference is how long it takes EPAM to retrieve the chunk from LTM and say it. The more words in the group that are being rehearsed, the longer it takes to rehearse that group. In this part of the simulation, the groups that were rehearsed are: B-B-B-B-B-B-B-M-E-M-B C. Backwards memory strategy When subjects are asked to report the list starting from the last item, the serial position curve changes. 4. Draw the serial position curve as you did in part A using this new data. Here is the average "backwards memory strategy" serial position curve from 15 homework assignments: > See graph2.gif 7. For each auditory loop printout, indicate which group has just been rehearsed. Here is a sample printout with the group that has just been rehearsed *highlighted*. The name of the group is given at the end of each line. Contents of the auditory loop Group *1560-TABLE-1* B (new) 1180-TABLE-NIL *1560-TABLE-1* B 800-TABLE-NIL 1180-TABLE-NIL *1560-TABLE-1* B 420-TABLE-NIL 800-TABLE-NIL 1180-TABLE-1 *1560-LIGHTNING-2* M (new) 40-TABLE-NIL 420-TABLE-NIL 800-TABLE-NIL 1180-LIGHTNING-2 *1560-TABLE-1* B 40-TABLE-NIL 420-TABLE-NIL 800-LIGHTNING-NIL 1180-TABLE-1 *1560-LIGHTNING-2* M 120-TABLE-NIL 500-LIGHTNING-NIL 880-TABLE-1 1260-LIGHTNING-2 *1640-TRAIN-2* M (new) 120-LIGHTNING-NIL 500-TABLE-NIL 880-LIGHTNING-2 1260-TRAIN-2 *1560-TABLE-1* B 120-TABLE-NIL 500-LIGHTNING-2 880-TRAIN-2 1180-TABLE-1 *1560-RABBIT-3* E (new) 200-TRAIN-NIL 500-TABLE-1 880-RABBIT-3 1300-LIGHTNING-2 *1640-TRAIN-2* M 200-TABLE-1 580-RABBIT-3 1000-LIGHTNING-2 1340-TRAIN-2 *1640-DOG-3* E (new) 320-LIGHTNING-2 660-TRAIN-2 960-DOG-NIL 1300-RABBIT-3 *1640-DOG-3* E 20-LIGHTNING-1 360-TRAIN-1 660-DOG-NIL 1000-RABBIT-2 1340-DOG-2 *1640-FROG-3* E 320-RABBIT-NIL 660-DOG-NIL 960-FROG-3 1300-RABBIT-2 *1640-DOG-2* M 280-DOG-NIL 580-FROG-3 920-RABBIT-2 1260-DOG-2 *1560-THUNDER-3* E (new) 240-RABBIT-2 580-DOG-2 880-THUNDER-NIL 1300-FROG-3 *1560-THUNDER-3* E 200-DOG-1 500-THUNDER-NIL 920-FROG-2 1180-THUNDER-2 *1560-AIRPLANE-3* E (new) 240-FROG-NIL 500-THUNDER-NIL 880-AIRPLANE-3 1300-FROG-2 *1560-THUNDER-2* M 200-THUNDER-NIL 580-AIRPLANE-3 1000-FROG-2 1260-THUNDER-2 *1640-BOOK-3* E (new) 320-FROG-2 580-THUNDER-2 960-BOOK-NIL 1300-AIRPLANE-3 *1640-BOOK-3* E 150-THUNDER-1 530-BOOK-NIL 870-AIRPLANE-2 1210-BOOK-2 *1550-CEDAR-3* E (new) So, the groups that were rehearsed in this simulation (excluding the lines where a new word was added) were: B-B-B-M-B-M-E-E-M-E-M-E Note, items occasionally change groups because EPAM is trying to remember the order of the items. The new items are added to the E-group because it designates items at the end of the list. However, a group can only have so many chunks. So, when the E-group gets full, some of the new items in that group are transferred to the M-group, and the M-group items are transferred to the B-group. Note, it is not the case that EPAM is actively "letting" items decay from the loop (contrary to what some students suggested); rather, EPAM is busy adding new items or rehearsing other groups, and items decay before EPAM can rehearse them again. How does EPAM choose which group to rehearse? In an effort to hang on to all the items in the loop, EPAM rehearses the group with the oldest group items. Even this simple rule, however, can be over-ridden by the strategy determined by the recall condition. In the trace above from the Backwards Memory condition, for instance, we see that "table" is lost from STM after it wasn't rehearsed in time - the model having put its effort into rehearsing the middle and end groups instead. 8. Is the serial position curve different this time? Explain why this is so. The serial position curve is different from the one derived in the Visual Fowardsexperiment, in that there is no longer a primacy effect, and the recency effect is much larger. Because EPAM is trying to maintain the items in reverse order, it pays special attention to the latest item at any given point in time. This delays EPAM from being able to rehearse the early items often enough to prevent the first items from decaying from the auditory loop. Because items are presented so quickly, and EPAM pays attention to each item as it is presented, there is insufficient time to transfer the items to LTM. 9. What does this say about the cause of the primacy effect in the normal recall condition (when words are to be recalled in forward order)? The fact that the primacy effect can be completely eliminated by the retrieval condition, suggests that the primacy effect in the normal condition is purely an effect of strategy, rather than some deep biological property of memory. This effect of retrieval condition is evidence against theories of memory such as Proactive Inhibition (mentioned in the previous assignment) which posit that the serial position curve is due to properties of the basic memory processes.