A critical form of memory organization, and one people frequently use, is retention of events in the temporal order in which they occurred. Consider, for example, your memory for the events that occurred last summer. If someone asked you what you did during your summer vacation, most likely you would discuss the events in the sequence in which they occurred, beginning with those that occurred at the start of the summer and concluding with those that occurred at the summer's end. Alternatively, you could report together all the parties you attended and report as another group all the times you went hiking or swimming. However, retention in terms of temporal sequence, or serial order, is most common.
Definitions and Distinctions
To study the retention of serial order in the laboratory, the information pertaining to temporal sequence must be distinguished and isolated from other types of related information. The relevant distinctions can be made clear by considering the following hypothetical situation: Imagine a waiter in a restaurant who is taking dinner orders from the people sitting around a table. Usually in such a situation the individuals make their requests in a temporal sequence that follows the spatial arrangement of the seats around the table. However, in the present situation this ordinary practice is not observed. Instead, the waiter takes the requests in an order determined by the individuals' ages and genders, starting with the oldest woman and ending with the youngest man. This situation is illustrated in Figure 1. The first order is for ham, the second for liver, the third for steak, and the fourth for chicken. The temporal sequence of the requests is thus ham, liver, steak, and chicken, a sequence that does not correspond to the spatial arrangement around the table. Hence, the temporal and spatial orders are not the same. When the waiter returns to deliver the dinners, he serves the first person liver, the second turkey, the third steak, and the last chicken. The waiter thus makes two mistakes. In the case of the turkey, he brings a dinner requested by nobody, and in the case of the liver, he gives a dinner ordered by one person to another. The first type of mistake is an item error because the identity of the dinner item is incorrect. The second type is an order error because a correct item is brought but is placed in the wrong position in the temporal sequence. For a discussion of laboratory methods used to distinguish between the retention of item, temporal order, and spatial order information, see Alice F. Healy et al. (1991).
Another important distinction that must be made is the difference between the order of the items and their positions in the sequence. (Some researchers refer to this same distinction as involving relative versus absolute positions.) This distinction can be clarified by considering a related hypothetical situation with the same waiter and diners. In this case, the waiter returns to take the requests for dessert. The first person orders ice cream, the second nothing, the third cake, and the last pudding. The waiter correctly brings the first person ice cream and the last pudding, but he gives the cake to the second person instead of the third. The waiter, therefore, correctly remembers the order of desserts—cake between ice cream and pudding—but he confuses the second and third temporal positions.
Although a number of techniques have been used to study retention of serial order (see Cowan, Saults, Elliott, and Moreno, 2002, for a novel set of techniques), two procedures have been most popular. Much of the early work on this topic used serial learning with the method of anticipation, and many subsequent studies used short-term serial recall with the distractor paradigm. In both methods, the results of primary interest have focused on the serial position curve, which reveals the proportion of correct responses as a function of the position of each item on the list.
Serial Learning Method of Anticipation
In serial learning, subjects attempt to learn an ordered list of items (often nonsense syllables or words) across a number of successive trials. On each trial the list is presented and the subject tries to recall it. With the method of anticipation, the subjects are not required to recite the entire list at one time. Rather, each list item is presented in turn, and the subjects are required to anticipate (i.e., recall) each item before it is presented, in response to the item immediately preceding it on the list. A correct response is scored whenever the subject correctly anticipates an item, and the subject receives feedback (i.e., the subject is told the next item in the sequence) regardless of whether a correct response is made. Usually the experimental trials are continued until the subjects are able to anticipate every item with no errors. At that point, the investigator counts the number of correct responses made at each position in the list, and it becomes evident that the items in the different ordinal positions in the list are not learned with the same ease. Rather, items at the beginning and end of the list yield more correct responses than those in the middle. The point of maximum difficulty is somewhat beyond the center of the list.
Although the total number of correct responses on the list may decrease when the items are more difficult, as when nonsense syllables are used instead of words or when the rate of list presentation is faster, the serial position curve remains constant across such changes in the learning situation when it is plotted as the proportion of the total number of correct responses made at each position in the list. The constancy of the serial position curve when plotted in this manner is known as the Hunter-McCrary law (McCrary and Hunter, 1953). A typical serial position curve for an eight-item list is shown in Figure 2, which presents data reported by Bennett B. Murdock (1960) in an important article relating the serial position function to results of experiments in domains of psychology outside of verbal learning. The serial position function is described as bow shaped because it resembles a bow used in archery. The large relative advantage for the items from the beginning of the list is known as the primacy effect, and the smaller relative advantage for the items from the end of the list is known as the recency effect.
Short-Term Serial Recall with the Distractor Paradigm
On a trial in the distractor paradigm used to study serial recall over a short time interval, subjects are given a short list of items to remember (typically three to five letters), then are required to participate in an interpolated distractor task that is meant to prevent them from rehearsing the list (e.g., they may be told to count backward from a random number), and finally they are asked to recall the list of items according to the order of presentation. The duration of the distractor task, or the length of the retention interval, varies from trial to trial but is usually quite short (no longer than twenty seconds). Also, the list of items to be recalled changes from one trial to the next. A correct response is scored whenever the subject recalls an item that was shown and places it in the ordinal position in which it occurred on the list. The time course of forgetting the serial list is revealed by comparing the proportion of correct responses at each retention interval. The resulting retention function is usually very steep; forgetting is very rapid in this paradigm. A plot of the proportion of trials on which correct responses are made at each ordinal position in the list reveals a serial position curve that is usually bow shaped and nearly symmetrical; the primacy effect is approximately equal in magnitude to the recency effect. Typical serial position curves for three different retention intervals are shown in Figure 3, which presents data reported by Healy (1974). In Healy's study, order information was isolated from item information because the same four items were shown on every trial of the experiment; the subjects knew the identity of the items in advance and had only to reconstruct the order in which they were shown on a particular trial.
A number of theoretical models have been proposed to account for serial organization. Classic models have included simple associative mechanisms, which were elaborated by contemporary models.
Although researchers have proposed many models of serial order retention, two simple opposing models dominated the early research on this topic. Both models include associative mechanisms as the basis for retaining serial order information. According to the associative chaining model, item-to-item associations are constructed so that the first item in a serial list is linked to the second item as a stimulus-response pair, the second item is linked in the same way to the third item, and so on to form an associative stimulus-response chain of items (Crowder, 1968). For example, given the list of dinner orders ham, liver, steak, and chicken in the hypothetical restaurant example discussed earlier, ham would be associated with liver, liver with steak, and steak with chicken. The second model to account for serial order retention involves positional associations. By this account, each item on the list is associated with the ordinal number corresponding to its serial position in the list (Young, Hakes, and Hicks, 1967). In the example described earlier, ham would be associated with the number 1, liver with 2, steak with 3, and chicken with 4. Experimental evidence has refuted both of these simple explanations for serial order retention.
Despite the problems with the simple associative models, two more complex contemporary models have been proposed that can be viewed as extensions of the earlier models. Both models have derived support from a wide range of experimental investigations and observations, including the pervasive serial position functions. According to the theory of distributed associative memory (TODAM), serial order information is represented in memory as a series of pairwise associations linking successively presented items. TODAM resembles the traditional chaining approach except that the items in the list and their associations are stored together in memory. Each item is represented as a list of features or a vector of numbers. Rather than simple links connecting the two items in a pair, the items are connected by means of "convolution," which is a mathematical operation that merges the separate vectors for the items into a single composite vector. The vectors for all the items in the list and for all the pairwise associations are added to a common memory vector. The mathematical details of TODAM and its ability to account for data from many serial order tasks are described in the work of Stephan Lewandowsky and Bennett B. Murdock (1989) and Murdock (1995).
In contrast with TODAM, a second influential model of serial order retention emphasizes positional information. According to this perturbation model, the representation of order information derives from the representation of position information, so that subjects can recall items in the temporal sequence exactly to the extent that the positional information they have stored in memory can adequately prescribe the order. Associations are included in the perturbation model. However, instead of associative bonds linking successive items on the list, there are associative bonds between each item and a single control element, which represents some aspect of the current context in which the list is presented. For example, given two successively presented items X and Y, rather than an association of the form X-Y, the perturbation model includes associations of the form X-C-Y, where C represents the control element. This new associative mechanism has allowed for an elegant description of both short-term and long-term serial order recall. For lucid discussions of the perturbation model, see William K. Estes (1972, 1997).
Late-twentieth-century models have either refined the associative mechanisms (Henson, 1998) or introduced new nonassociative mechanisms for explaining serial recall. For example, the primacy model of Michael P. A. Page and Dennis Norris (1998) is centered on the notion of a primacy gradient of activations reflecting the strength with which the start of the list is associated with each successive list item. According to this model, there is a repeating cycle in which the item with the greatest activation is selected for recall and then suppressed.
Clearly much progress has been made in understanding the processes underlying serial organization in memory, but there is still considerable controversy among researchers because the processes appear to be more complex than originally envisioned.
See also:MEMORY SPAN
Cowan, N., Saults, J. S., Elliott, E. M., and Moreno, M. V. (2002). Deconfounding serial recall. Journal of Memory and Language 46, 153-177.
Crowder, R. G. (1968). Evidence for the chaining hypothesis of serial verbal learning. Journal of Experimental Psychology 76, 497-500.
Estes, W. K. (1972). An associative basis for coding and organization in memory. In A. W. Melton and E. Martin, eds., Coding processes in human memory. Washington, DC: Winston.
—— (1997). Processes of memory loss, recovery, and distortion. Psychological Review 104, 148-169.
Healy, A. F. (1974). Separating item from order information in short-term memory. Journal of Verbal Learning and Verbal Behavior 13, 644-655.
Healy, A. F., Cunningham, T. F., Gesi, A. T., Till, R. E., and Bourne, L. E. (1991). Comparing short-term recall of item, temporal, and spatial information in children and adults. In W. E. Hockley and S. Lewandowsky, eds., Relating theory and data: Essays on human memory, in honor of Bennet B. Murdock. Hillsdale, NJ: Erlbaum.
Henson, R. N. A. (1998). Short-term memory for serial order: The start-end model. Cognitive Psychology 36, 73-137.
Lewandowsky, S., and Murdock, B. B. (1989). Memory for serial order. Psychological Review 96, 25-57.
McCrary, J. W., and Hunter, W. S. (1953). Serial position curves in verbal learning. Science 117, 131-134.
Murdock, B. B. (1960). The distinctiveness of stimuli. Psychological Review 67, 16-31.
—— (1995). Developing TODAM: Three models for serial-order information. Memory & Cognition 23, 631-645.
Page, M. P. A., and Norris, D. (1998). The primacy model: A new model of immediate serial recall. Psychological Review 105, 761-781.
Young, R. K., Hakes, D. T., and Hicks, R. Y. (1967). Ordinal position number as a cue in serial learning. Journal of Experimental Psychology, 73, 427-438.