ABSTRACT
introduction
People notice the situations in which they forget, and they judge them to require memory. Cognitive psychologists see memory wherever information is maintained over time. We are as interested in the processes that underlie remembering as those that cause forgetting.
The Episodic-Semantic Distinction Tulving (1972) distinguished episodic from semantic memory. Episodic memory is autobiographical; it preserves the temporal and spatial features of a person's past experience. Semantic memory preserves people's general knowledge of symbols, concepts, and the rules for manipulating them. This chapter is concerned only with how people remember episodes from their own past.
Structural Features and Control Processes This chapter is organized around Atkinson and Shiffrin's (1968) account of the episodic memory system. It describes the evidence for various structural features of that system. By structural features, Atkinson and Shiffrin mean the permanent aspects of memory over which people exercise no voluntary control, and which they use unvaryingly from one situation to another. By control processes they mean memory routines or strategies that are selected, constructed, and used at the option of the person, depending on the characteristics of the situation facing him.
Three Memory Stores According to Atkinson and Shiffrin, the basic structural features of episodic memory are three memory stores, called the sensory register, the short-term store and the long-term store. The three stores are said to be structurally distinct because they preserve information in different formats, for different durations and for different purposes, and because they lose information in different ways and have different anatomical bases. 211
The Sensory Register Environmental information enters the memory system through the various senses, and it goes first to the sensory register. Information in the sensory register is an unconscious, nearly literal, and complete record of its sensory image, but it is lost from the register very rapidly. It either decays spontaneously, just dissipating with the passage of time, or new information from the senses writes over already registered information, erasing it. The purpose of the sensory register is to hold information long enough for the cognitive system to run off routines that select from its sensory inputs those that will be further processed. When Atkinson and Shiffrin wrote, there was experimental evidence for a visual register only.
The Short-Term Store The short-term store is a working memory where conscious mental processes are performed on information from both the sensory register and the long-term store. Information can be maintained indefinitely in the short-term store, provided it is given constant attention. As soon as attention is shifted from information in the short-term store, it begins to decay, and it will be gone completely in 15 to 30 seconds. The purpose of the short-term store is to briefly hold small amounts of carefully selected information from the sensory registers and the long-term store in order for people to execute the optional conscious processes necessary to tailor their behavior to the changing demands of their environment. When Atkinson and Shiffrin wrote, there was experimental evidence for only auditory-verbal-linguistic information in the short-term store.
The Long-Term Store Information enters the long-term store from the short-term store and perhaps also from the sensory register. From the short-term store, transfer can be effected either consciously, by using control processes to relate its contents to those of the long-term store, or unconsciously, by processes that are not understood. Long-term store contains an unlimited amount of information coded in many forms. It loses information in several apparently different ways: by decay, by disruption from new inputs, and by efforts to retrieve for recall. The purpose of the long-term store is to keep information and the rules for its processing when they are not being used.
Historical Antecedents of the Three-Store Model Atkinson and Shiffrin's view of memory is typical of the views of other information-processing psychologists, like Broadbent and Neisser. It also reflects the thinking of earlier workers like James and Hebb. Waughand Norman proposed a similar theory.
Experimental Reasons for Positing a Sensory Register
Research on sensory memory for visual information is called the study of iconic storage. Experimentation on auditory sensory memory is called the study of echoic storage.
Iconic Storage
The Prototypic Phenomenon Sperling (1960) used a partial report procedure to show that people can see far more in a brief glance than they can report when they are required to tell everything they saw. He found evidence for a brief visual storage system, called the icon, of all the letters. The icon faded rapidly, and Sperling estimated that it was gone completely in 250 msec.
Repeatability of the phenomenon Averbach and Coriell (1961) were among the first to establish the repeatability of Sperling's demonstra212tion that a relatively complete and raw visual image persists after the offset of brief visual displays, and that it fades completely in about .25 seconds.
A converging operation Haber and Standing (1969) used a very different method for estimating the duration of the icon, but they too obtained an estimate of 250 msec. Haber and Standing also showed that the icon exists centrally, in the brain, not peripherally, in the eye.
Visual Masking Visual masking experiments present two brief stimuli in rapid succession to see how the second, masking stimulus affects perception of the first, target stimulus. Cognitive psychologists have distinguished icon formation from icon identification, called its readout.
Disrupting icon formation Sperling (1960), Eriksen (1966), Spencer (1969), and Smith and Schiller (1966) employed a bright flash of light as their mask. In sum, they showed that (1) brightness masks disrupt perception maximally when they coincide with target presentation; (2) the disruption occurs equally for masks that precede and masks that follow the target; but that (3) the effect falls off rapidly as the mask is delayed, disappearing completely for delays of 100 msec; and (4) no disruption occurs when target and mask are presented to separate eyes. These results show that icon formation occurs peripherally and is completed in 100 msec.
Disrupting read-out of the icon Averbach and Coriell (1961), Smith and Schiller (1966), and Spencer and Shuntich (1970) used patterned masks. They found that: (1) pattern masks disrupt perception maximally at 100 msec delay; (2) after 100 msec the brightness of the mask is irrelevant; and (3) disruption is as great when the pattern and mask go to different eyes. These results show that icon read-out occurs centrally and that it can be interrupted by the arrival of a second icon.
Decoupling and Operationism We explain how studies of iconic memory illustrate the decoupling of one cognitive subsystem from another, and how they justify a focus upon underlying processes rather than on the operational features of experiments alone.
Echoic Storage Is there an auditory analogue to the iconic store?
Storage Capacity, Rate of Decay, and Form of Stored Information as Shown by Auditory Partial Report Darwin, Turvey, and Crowder (1972) performed an auditory analogue of Sperling's partial report experiments. Their findings suggest an echoic store with a relatively large capacity for raw auditory information but with a slower decay rate than iconic storage. They estimated a 2-second decay rate.
Attempts to Find Convergence A chief interpretive question is whether the partial-report technique overestimates the echo's duration due to its failure to incompletely decouple the auditory sensory register from the short-term store. Several converging operations have been used in efforts to answer this question.
Estimating the echo's duration Efron (1970, a, b, c) used a procedure like Haber and Standing used for visual stimuli. His estimate of 130 msec is much briefer than the 2-second estimate of Darwin, Turvey, and Crowder.
Recall of unattended input Glucksberg and Cowen (1970) studied recall accuracy for unattended verbal materials. Their estimate of the echo's duration was 5 seconds.
213 Auditory recognition masking Massaro (1970) used tones in a masking procedure to arrive at an estimate of 250 msec for the echo's duration.
Conclusions The research on echoic storage supports the idea of an auditory sensory register that preserves the temporal pattern of sounds long enough to allow higher-level cognitive processing. The failure to find full operational convergence on an estimate of the echo's duration probably resulted from a failure to separate echoic memory from the short-term store.
Characteristics of the Short-Term Store
Looking up a new phone number and trying to dial it under various conditions is used to illustrate the space and time limits of the short-term store.
Space Limit Miller's (1956) analysis of the limits of short-term memory at 7 ± 2 items is described. Informal experiments that a student can perform are suggested to demonstrate that the limit on the number of items applies regardless of their size.
Time Limit Peterson and Peterson (1959) demonstrated that items are lost from the short-term store about 20 seconds after attention is withdrawn from them. Murdock (1961) showed that this rate of loss applies regardless of the size of the items from which attention is withdrawn. Bjork's (1972) findings on intentional forgetting illustrate the adaptive value of a relatively rapid loss of information from the short-term store.
The Modalities of Short-Term Storage
Auditory-Verbal-Linguistic Atkinson and Shiffrin posited only an auditory-verbal linguistic mode of operation for the short-term store. Conrad's (1964) findings support this position. He showed that recall errors for visually presented letters were based on acoustic confusions.
Visual Storage Brooks (1968) and Shepard and Metzler (1971) showed that visual information can also be held in the short-term store.
Differentiating the Long-Term Store From the Short-Term Store
Clinical, Anatomical Findings Clinical reports by Shallice and Warrington (1970) and Milner (1959, 1966) suggest different anatomical bases for short-and long-term memory.
Experimental Evidence Glanzer has shown that the serial position curve for free recall is a discontinuous composite of two underlying curves, one reflecting output from a long-term store and one from a short-term store. Craik's (1970) use of final free recall to separate the short- and long-term stores is cited as an elegant example of psychological experimentation.
A Reminder
This chapter described data that are consistent with the model by Atkinson and Shiffrin. Data collected in the 1970s led psychologists to other views of episodic memory.
