<<Back to Cognition Class Home Page
---------------------------------------------------------------------------------------
OUTLINE---------------------------------------------------
1. TYPES OF MEMORY
1.1 The Atkinson-Shiffrin Model (Modal model)
The Atkinson-Shiffrin Model is an "information processing" model. It assumes three different systems or memory stores and a series of control processes that allow information to move from one store to another. The three systems are: sensory stores (we have already talked about them), short term memory and long term memory. Each store has a given duration and capacity. The presence of control processes in this model is very important because it states that memory is not a passive recording of events and information, but rather an ACTIVE PROCESS.
For example, we know that sensory stores have a very short duration (e.g., 250 ms for the iconic memory) and a fairly large capacity. Atkinson and Shiffrin proposed that each sensory modality has its own sensory store. Information from sensory stores may be transferred to the second store, short term memory, or decay.
The second store, short-term memory, has a very limited capacity but a longer duration (about 30 s). Information is maintained active in short term memory through rehearsal and is encoded phonologically.
Part of the information in short-term memory is transferred to the last store, long-term memory. This store is of almost unlimited capacity and can store information for a very long time. Memory in this store is encoded semantically. Information from long-term memory can be retrieved and transferred to short-term memory.
Control processes include attention, strategic, and search processes. The most studied process, however, is rehearsal.
Summary:
Structures Control Processes Codes
Sensory stores Rehearsal Physical features
Short-term memory Attention Phonological
Long-term memory Strategic proc. Semantic
Serial position effects:
If we study a series of words and then we are asked to recall them, we will typically recall better the items that are at the beginning and at the end of the list (Figure 3.2 p 72). The fact that we recall better items at the beginning of the list is called primacy effect, whereas the fact that we recall better items at the end of the list is called recency effect. There are different hypotheses that try to account for these two serial position effects. One hypothesis is that primacy and recency effects are associated to different memory stores. In particular, Atkinson and Shiffrin proposed that, because items at the beginning of the list have been rehearsed longer, the probability that they will be encoded in long-term memory is higher than for other items. Therefore, the primacy effect may be explained in term of probability that an item to is encoded in long term memory. According to the Atkinson-Shiffrin model, the recency effect, on the other hand, depends on the fact that the items at the end of the list are still in short-term memory and can therefore be reported with higher accuracy.
Not all researchers agree with this interpretation, though, as we will see below.
Evidence that favors of the modal model:
Kintch and Bushke's study --> In the study phase, participants were asked to learn a list of 16 word in order. In the test phase, the experimenter presented a word, and participants had to report the following word in the list. Two lists were used. In list 1 many of the words were semantically similar (carpet, rug) and in list 2 many of the words were acoustically similar (so, sew).
What are the predictions of the Atkinson-Shiffrin model? If a word is encoded phonologically, as this model proposes for short-term memory, the errors should occur especially for acoustically similar pairs. If a word is encoded semantically, as this model proposes for long-term memory, the errors should occur especially for semantically similar pairs. Therefore, the model predicts a larger amount of error in the semantically similar list for the first part of the list, and a larger amount of errors in the acoustically similar list in the last part of the list. This is exactly the pattern of results observed.
Rundu's Study --> This study showed that the amount of rehearsal affects the recall of the first part of the list, but not the recall of the second part of the list. This result supports the Atkinson-Shiffrin model because agrees with the hypothesis that items in the first part of the list are remembered because they are encoded in long-term memory, and the probability of encoding in long-term memory increases as a function of the number of rehearsals.
Patients' studies --> Lesions in different parts of the brain produce very different memory deficits. H.M. had bilateral removal of the hippocampus and part of the temporal lobes. This patient had normal short term memory and can carry out a normal conversation. However, after just few minutes he will not remember neither that he had that conversation nor the person he was talking with. H.M. seems to have lost the ability to transfer information in long term memory.
Another patient, K.F., that had a left brain lesion after a motorcycle accident, seems to have the opposite problem. He can still learn new information, but he seems very impaired in tasks that require short-term memory.
Evidence against the modal model:
Recency effects can be found also for information stored in long term memory. For example, if I ask you to name the presidents of the United States, you will be better in remembering the last presidents. This effect does not seem therefore just a characteristic of the short term store.
Data from patients, especially K.F, pose a problem to the Atkinson-Shiffrin model. The Atkinson-Shiffrin model assumes that information can be stored in long-term memory only after being in short-term memory for a certain period and through rehearsal. If K.F. has impaired short-term memory, he should not be able to store information in long term memory, according to the modal model, but he does.
Even if the Atkinson-Shiffrin model if very criticized and Shiffrin himself has declared that the model is too simplistic, the distinction between short- and long-term memory is still useful, especially in the study of memory, because it refers to different results and different experimental paradigms.
1.2 The Levels-of-processing approach
Craik and Lockart looked at memory from a different point of view. They were interested in how the type of processing used influences memory. As we saw in the experiment in class that we did at the beginning of the class, different tasks require the activation of different codes. For example, if I ask you to decide whether a word contains a D or a K, you will probably activate only visual or phonological codes, whereas if I ask you to judge the meaning of a word, you will activate semantic codes. As we saw in the experiment, the activation of the semantic code is usually associated with a much better memory for the material than the activation of the phonological code.
Craik and Lockart were the first to report this result and proposed that the type of processing of the information affects how well we will remember this information. Several experimental data support this hypothesis. For example, Sporer (1991) showed that a similar effect can be found for faces: the probability to recognize a face was higher when participants were asked to judge whether that person was honest than when they were asked to judge the gender of the person of the width of the nose.
Another important result of their research is that simple rehearsal does not seem to improve recall. In an experiment they show that the number of rehearsals is not related to the probability of recall of a word. They distinguished between "maintenance rehearsal" that does not improve recall, and "elaborative rehearsal", which requires a deeper and more meaningful analysis of the stimulus and does improve recall. For example, Craik and Watkins (1973) presented a list of words such as "daughter, oil, rifle, garden, grain, table, football, anchor, giraffe". Participants were asked to repeat each word beginning with "g" until the next word beginning with "g" was presented. As you can see from the list, "garden" was repeated for a short time until "grain" was presented, whereas the interval between "grain" and the next word to be rehearsed, "giraffe" is four words. However, the results showed that words that were rehearsed for a long time were not remembered better than word rehearsed for a shorter time.
On the same line of research are the data on the generation effect that show that we are better in remember things when we "generate" them rather than just trying to memorize them. For example, Slamecka and Graf (1978) showed that people were much better in remembering words if the instructions were "find a word that is the synonym of "sea" and starts with "o"" rather than when the two words "sea-ocean" were presented and had to be read aloud.
SHORT TERM MEMORY
According to Atkinson and Shiffrin short term memory has a short duration (about 30 seconds), and has a limited capacity.
Some authors do not even agree in the existence of a separate memory storage with these characteristics. Those who do share different opinions on what short-term memory is.
A more modern approach talks of "working memory", emphasizing the aspect of dynamic and active process more that the one of passive "memory store". Neuroscience evidence supports the idea the different brain regions are involved in retrieval from long-term memory and in the use of working memory. However, neuroscientists do not think that working memory is a "storage" but rather they propose that working memory has a role in maintaining active certain types of information that we need and in producing mental transformation of this information, for example in understanding a sentence or in doing mental arithmetic.
Brown-Peterson task (demonstration p. 117)
Size of short-term memory:
Memory span: it is usually tested with digits and it measures how many digits we can store in short-term memory. This measure is not highly correlated with intelligence but most measures of intelligence include a memory span test.
What is the capacity of short-term memory? Originally, researchers thought that it was possible to specify a precise number of items that could be stored in short-term memory. Take in account that in these studies only verbalized material (words, letters, numbers) was used to measure memory span. However, more recent studies show that one factor is the time required to pronounce the item in short term memory.
A very important study was "The magical number seven, plus or minus two" that George Miller published in 1956. In this study Miller showed that the working memory capacity seemed to be between 5 and 9. The unit that Miller considered was a "chunk". A chunk is a "well learned cognitive unit made up of a small number of components representing a frequently occurring and consistent perceptual pattern." So, if we have to remember a telephone number, 541-346-4940, it is possible that the area code 541 will be a chunk (we know that it is the area code of this part of Oregon) and 346 may be another chunk (all the telephone number on campus start with 346) whereas each of the following digits 4-9-4-0 will be remembered as units. The important thing in the concept of chunk is that it emphasize the active nature of our information processing.
Pronunciation time (Demonstration p.120)
More recent studies in the '80s demonstrated that another important factor affecting memory span is the time required to pronounce the items (again, we are talking about verbalizable items!). In particular, some researchers found that we can only remember items that can be pronounced in 1.5 s. The hypothesis is that there is an auditory trace of the verbal material that rapidly decays and that requires to be reactivated within 1.5 s. (What is a possible way to reactivate the trace?).
This result strongly suggests that acoustic or phonological codes are very important for short term memory, at least when we are trying to remember verbal material.
Other researches suggest that also visual and semantic codes can be used in short-term memory.
Proactive inhibition
Release from proactive inhibition
This effect can be affected by the semantic categories used in the task (Wickens and al. 1976). Demonstration p. 125
Baddeley and Hitch: WORKING MEMORY
There are several issues that suggest that the concept of short-term memory as proposed by Atkinson and Shiffrin is too simplistic. First, there are experiments that suggest that some short term memories that rely heavily on visual-spatial input are disrupted by visual tasks as visual search, but not by phonological rehearsal.
The second problem is that there a confounding between "remembering" things (such as a telephone number read in the phone directory) and more complex processing that involves problem solving, comprehension, etc.
Baddeley's concept of working memory is more complex and more dynamic than the concept of short-term memory proposed by Atkinson and Shiffrin. First, this model includes different "stores" depending on the format in which the information is encoded. Second, it includes a "central executive", a mechanism similar to Shallice and Norman's SAS, which emphasize that information in working memory is manipulated. According to Baddeley, working memory is the essential mechanism that allows us to manipulate and process information that is not physically present. The emphasis in the working memory model is not so much on remembering, but on using information. For example, in order for us to understand a spoken sentence, we need to have active for a certain time part of the information (Demonstration p 116). This is very different from the function of remembering digits of a telephone number.
Phonological loop
It has two components, the phonological store and the subvocal rehearsal process. The phonological store is where phonological information is maintained, and is of limited capacity. The subvocal rehearsal process is the process that maintains information in the phonological store activated, so that this information does not decay after few seconds.
Visuo-spatial working memory
This component is associated to the storage and maintenance of information coded visually and is also called "visuo-spatial scratchpad" Demonstration p.129
Central executive
According to Baddeley, the central executive works like a supervisor. It decides which issue deserves attention and which should be ignored. The executive also selects strategies. It has limited capacity and has the function to integrate information from the two stores and from long-term memory and to perform mental operations on them.
---------------------------------------------------------------------------------------