Forgetting in Short-Term and Long-Term Memory

Positional Uncertainty in the Brown-Peterson Paradigm

Joshua A. Quinlan, Ian Neath, and Aimée M. Surprenant Memorial University of Newfoundland

Since McGeoch’s (1932) influential article, no accounts of long-term memory have invoked decay as a cause of forgetting. In contrast, multiple accounts of short-term memory (STM) invoke decay, with many appealing to results from the Brown-Peterson paradigm as offering support. Two experiments are reported that used a standard Brown-Peterson task but which scored the data in 2 ways. With traditional scoring (was the entire 3-letter consonant trigram recalled?) performance decreased with increasing delay. With immediate serial recall scoring (e.g., was the first letter recalled first, was the second letter recalled second?), standard position error gradients (Experiment 1), and protrusion gradients (Experiment 2) were observed. That is, when the first letter of the consonant trigram was not recalled first, it was more likely to be recalled second than last. In addition, if a letter from a previous list was mistakenly recalled in a later list, it most likely retained its original position. The presence of such gradients is inconsistent with claims of decay but is predicted by SIMPLE, a local distinctiveness model of memory. Moreover, the presence of such gradients is consistent with the claim that forgetting in the Brown-Peterson paradigm follows the same principles observed in other memory tasks.

Keywords: Brown-Peterson, short-term memory, forgetting, SIMPLE, serial recall

One theoretical approach to the study of memory is to divide it into multiple different memory systems, each of which oper- ates according to different principles (e.g., Schacter, Wagner, & Buckner, 2000). One common distinction that is made is between a system for retaining information over the short term, variously termed short-term or working or primary or immediate memory, and one that retains information for longer durations, generally termed long-term memory. Although there are many different models of memory for the short term, there are sufficient similar- ities that one can talk about a “standard” model (see Nairne, 2002). According to this standard model, forgetting occurs due to time- based decay. One frequently cited line of evidence taken in support of time-based decay comes from studies using the Brown-Peterson paradigm (Brown, 1958; Peterson & Peterson, 1959; see also Daniels, 1895). Although there exists a large literature demonstrat- ing numerous problems for the decay account of Brown-Peterson (for a review, see Neath & Surprenant, 2003), the decay interpre- tation first offered by Brown has been frequently invoked (e.g., Atkinson & Shiffrin, 1968; Baddeley, 1990) and continues to be invoked even as recently as 2014 (e.g., Ricker, Vergauwe, & Cowan, 2014). In this article, we report two experiments that further question the decay interpretation. Instead, the results sup- port a view in which memory follows the same principles regard-

less of whether the task is nominally thought to tap short- or long-term memory (Surprenant & Neath, 2009a).

In the typical Brown-Peterson experiment, the subject sees a single item (usually three consonants presented simultaneously, a so-called consonant trigram) and is asked to recall the trigram after a delay of between 3 to 20 s. During this delay, the subject engages in a distractor task to prevent rehearsal. The distractor task is carefully chosen to avoid interference, so with recall of letters a common distractor task is counting backward. The key finding is that performance decreases systematically with increasing delay. Given that the number of to-be-recalled items is substantially below the supposed capacity of short-term memory (STM), this result has been interpreted as showing that items that are not rehearsed in a short-term store fade away or decay over time. Rather than reviewing the literature that shows the many problems with a decay account (see Neath & Surprenant, 2003 for a review of that literature), we focus instead on one aspect of the original task that has not received much attention.

Despite the large quantity of literature on the Brown-Peterson paradigm, relatively little experimental work has examined the errors made when a consonant trigram is not correctly recalled. Instead, the majority of studies use all-or-none scoring: The re- sponse is scored as correct only if the whole consonant trigram is correctly reproduced.1 If the trigram is not recalled correctly, the response is scored as incorrect, even though the subject may have correctly recalled one or two of the letters. That is, researchers have viewed the consonant trigram as a single item that is either recalled or not recalled.

1 There are, of course, experimenters who did not use all-or-none scor- ing. For example, Kincaid and Wickens (1970) presented trigrams and used a scoring system similar to immediate serial recall: They awarded 1 point for each part of the trigram recalled in the correct position and also awarded a bonus point if all three items were correctly recalled in order.

Joshua A. Quinlan, Ian Neath, and Aimée M. Surprenant, Department of Psychology, Memorial University of Newfoundland.

Joshua A. Quinlan is now at York University. This work is based on Joshua A. Quinlan’s undergraduate honours thesis

and was supported by grants from the Natural Sciences and Engineering Research Council of Canada to Ian Neath and Aimée M. Surprenant.

Correspondence concerning this article should be addressed to Ian Neath, Department of Psychology, Memorial University of Newfoundland, St. John’s, NL A1B 3�9, Canada. E-mail: [email protected]

Canadian Journal of Experimental Psychology / Revue canadienne de psychologie expérimentale © 2015 Canadian Psychological Association 2015, Vol. 69, No. 1, 64 –71 1196-1961/15/$12.00 http://dx.doi.org/10.1037/cep0000038

64

In contrast, many other tests of STM use a more fine-grained scoring method. For example, in standard serial recall tests, the subject may see a list of five items and be asked to recall them in order. In this task, the first item is scored as correct if it was recalled first, regardless of whether the remaining items were correctly recalled. In addition to analyzing correct responses, this method of scoring serial recall tasks also allows for the analysis of error data. Data analyzed in this way yield two common findings.2

First, when an error is made and an item is not recalled in the correct position, the item is most likely to be recalled in an adjacent position. In general, the probability of recalling the item in an incorrect position when an error is made is inversely related to how far that position is from the original (e.g., Estes, 1972; Healy, 1974). This pattern of errors is usually referred to as a positional uncertainty gradient or position error gradient and is readily observed with both immediate recall and delayed recall (Nairne, 1992). Second, when an item is mistakenly recalled in the wrong list, a so-called protrusion error, the item is most likely to be recalled in its original position (Conrad, 1960; Henson, 1998; Melton & von Lackum, 1941; Nairne, 1991).

It is only if the consonant trigram is viewed as a list of three letters rather than as a single item that such error analyses are seen as relevant. This may be one reason that so few Brown-Peterson studies have examined whether these patterns are observable in the Brown-Peterson paradigm. One notable exception is the paper by Fuchs and Melton (1974), who reported the existence of protrusion gradients in a Brown-Peterson task but used words rather than letters as the to-be-remembered stimuli and their method of pre- sentation may have induced the subjects to process them as indi- vidual units. In Fuchs and Melton’s task, the stimuli were pre- sented in a left-to-right, downward stair-step pattern, such that the subject might see dome on one line, time on the next line down, spot on the next line down, and so on. One purpose of the current studies is to examine both position error gradients and protrusion errors in the Brown-Peterson paradigm when a single consonant trigram is presented.

There are a number of reasons why one would predict that standard-looking position error gradients and protrusion gradients will be observed in a Brown-Peterson task. First, the Brown- Peterson task can be thought of as a delayed serial recall task and such tasks are already known to produce these gradients (e.g., Healy, 1974). There are some differences, including that in serial recall tasks the to-be-remembered items are usually presented serially whereas in Brown-Peterson tasks the items are usually presented simultaneously, but it seems reasonable to predict that if the two tasks are scored similarly, the same pattern of results will obtain.

A second reason to predict such gradients is that a model of memory that has accounted for many of the Brown-Peterson results predicts them. SIMPLE (scale independent memory, per- ception, and learning; Brown, Neath, & Chater, 2007; Neath & Brown, 2006) views memory as fundamentally a discrimination task: To-be-remembered items are represented as positions along one or more dimensions in psychological space and in general, those items with fewer close neighbors on the relevant dimensions at the time of test are more likely to be recalled than items with more close neighbors. In the typical episodic memory task, the experimenter carefully chooses a set of to-be-remembered stimuli so that they are equated on as many dimensions as possible; as a

result, one of the few dimensions along which the stimuli vary systematically is presentation time. Another way of thinking of this is that in a standard episodic task, the subject already knows all of the words or letters that will be shown. What the task really requires is remembering that a particular item was presented at a particular time (i.e., on the list just seen) rather than at another time (i.e., on a list seen several minutes ago). SIMPLE posits that the Brown-Peterson task is essentially an immediate serial recall task. Indeed, there are relatively few changes to SIMPLE when it is fit to Brown-Peterson data compared to when it is fit to immediate or delayed serial recall data. The details of fitting SIMPLE to Brown- Peterson data have been provided elsewhere; for the main fits, see Brown et al. (2007, pp. 552 onward) and for supplementary details, see Neath and Brown (2012) and Neath, VanWormer, Bireta, and Surprenant (in press).

SIMPLE accurately predicts position error gradients for serial recall tasks (see Brown et al., 2007, p. 557 onward). Because SIMPLE views Brown-Peterson as a type of immediate serial recall—specifically as a variant of a complex span task (Neath et al., in press)—it follows that SIMPLE predicts that position error gradients will be observed in that task also. Note that this predic- tion holds independent of any particular parameter settings or specific model fits because a central characteristic of the model is that items near to one another in psychological space will be more confusable than items that are more distant.

In addition, SIMPLE accurately predicts appropriate protrusion gradients in serial recall tasks (see Brown et al., 2007, p. 558 onward). That is, SIMPLE can represent items on multiple dimen- sions, one of which indicates position within a list and another of which represents position within a set of trials. Just as position error gradients arise due to increased similarity between close neighbors on the list compared to more distant neighbors, similar gradients arise because of increased similarity with close neigh- bors across lists. That is, Item 3 of List 3 has close neighbors (relatively speaking) at both Positions 2 and 4 of List 3, but also at Positions 2, 3, and 4 of List 2.

There are a number of computational models of short-term/ working memory (for a review, see Nairne & Neath, 2013), but relatively few address predictions about recall errors in the Brown- Peterson paradigm. Of those that posit forgetting is due to decay, only two address results from the Brown-Peterson paradigm. Both the Primacy Model (Page & Norris, 1998) and the Time-Based Resource-Sharing Model (Oberauer & Lewandowsky, 2011) can account for decreasing recall as the duration of the distracting activity increases, and both invoke decay as a causal factor. How- ever, because both invoke decay, neither can account for protru- sion errors.3

The two experiments reported here were designed as a first step in assessing the predictions of SIMPLE that both position error and protrusion gradients will be observed in a standard Brown-

2 One influence that led us to this line of research was a study by Mewhort, Campbell, Marchetti, and Campbell (1981) who did a similar analysis on the Sperling task.

3 Oberauer and Lewandowsky (2011) instantiated the Time-Based Resource-Sharing Model of Barrouillet and colleagues (e.g., Barrouillet, Bernardin, & Camos, 2004), which invokes decay as a cause of forgetting. However, it should be noted that Oberauer and Lewandowsky argued against the model.

65POSITIONAL UNCERTAINTY IN BROWN-PETERSON

Peterson task. Experiment 1 was designed to focus on position error gradients and Experiment 2 was designed to focus on pro- trusion gradients. Neither experiment was designed to produce data suitable for modelling. Rather, should the prediction be con- firmed and gradients be observed, then subsequent experiments can focus on the time- and resource-intensive studies necessary to obtain data suitable for modeling.

Experiment 1

Experiment 1 was designed to be similar to the method used by Peterson and Peterson (1959). Subjects saw a consonant trigram and then counted backward, out loud, by 3 s for between 3 and 15 s. They were then asked to report either the final number they had counted back to or to report the consonant trigram.

Method

Subjects. Sixty-three students from Memorial University of Newfoundland volunteered to participate in exchange for a small honorarium. All reported that English was their first language.

Procedure. Subjects were tested individually in a single ses- sion that lasted about 35 min. On each trial, three consonants were randomly selected and were presented simultaneously for 1 s. Then, a three-digit number between 200 and 999 (inclusive) was randomly selected to be the start number. The subjects were asked to count backward by 3 s out loud at a rate of one answer every 1.5 s. The pace was indicated by a circle displayed on the computer screen that alternated colors once every 1.5 s. The duration of the distractor task was either 3, 6, 9, 12, or 15 s. Following the distractor task, the subjects were asked to recall either the conso- nant trigram or the final number they had said out loud. For both tests, the subject used a mouse to click on appropriately labelled buttons. As in the original Peterson and Peterson (1959) study, the consonant trigram was to be recalled exactly (i.e., the first letter reported first, the second letter reported second, and the final letter reported last). Feedback was given after a response for both tasks: For the letter recall, the subject was informed only that the re- sponse was correct or not, but for the counting backward task, the correct answer was provided if an incorrect number had been reported. The next trial began when the subject clicked on a button; thus, the experiment was self-paced.

There were 40 trials, half of which tested memory for the consonant trigram and half of which tested accuracy in counting backward. There were eight trials at each distractor duration. The order of conditions was randomly determined for each subject.

Results and Discussion

The overall accuracy on the math task was 0.495 (SD � 0.257). The data from all 63 subjects are included in the analyses below, however the results and conclusions do not differ if subjects are included based on level of performance on the math task.4

Three different analyses were performed. First, the recall data were scored in the way typical of Brown-Peterson studies: The response was counted as correct only if the entire consonant trigram was recalled in the correct order. As in numerous other demonstrations, recall decreased as the duration of the distrac- tor task increased from 3 to 15 s (see Figure 1). A one-way

repeated measures analysis of variance (ANOVA) revealed a significant effect of delay, F(4, 248) � 15.890, MSE � 0.046, partial �2 � 0.204, p � .001.

Second, the data were rescored as if the test was immediate serial recall. That is, each letter was scored correct if it was reported in the correct position regardless of the other letters. A 5 (Delay) � 3 (Serial Position) repeated-measures ANOVA

4 Most studies of the Brown-Peterson task do not report performance levels on the distractor task. We did analyze the data using various criteria for inclusion based on performance on the math task. Regardless of the criteria, the results and conclusions remain the same. For example, when the criteria was set at 0.5 accuracy or greater, Figure 3 was largely unchanged (e.g., 0.83 vs. 0.86 and 0.17 vs. 0.14) and the chi-square test becomes �2 (1, N � 32) � 30.96, p � .001). All of the significant F values remained significant, and all of the nonsignificant F values remained nonsignificant.

Figure 1. The proportion of times the consonant trigram was recalled in Experiment 1 and Experiment 2 as a function of the duration of the distractor task when standard scoring was used.

66 QUINLAN, NEATH, AND SURPRENANT

revealed a significant main effect of delay, F(4, 248) � 15.606, MSE � 0.092, partial �2 � 0.201, p � .001; with recall decreasing from 0.690 at 3 s to 0.491 at 15 s. There was also a significant main effect of position, F(2, 124) � 22.617, MSE � 0.034, partial �2 � 0.267, p � .001, with better recall for items in the first position (0.596) than in Positions 2 (0.526) and 3 (0.500). The interaction was not significant, F(8, 496) � 1.259, MSE � 0.025, partial �2 � 0.020, p � .25.

Third, in addition to analyzing correct responses, the incor- rect responses were also examined. Figure 2 shows the position error gradients collapsed over delay (top left panel) and also the gradients for each delay condition (remaining panels). As can be seen, at all delays, the errors are systematic and not random. When a consonant from the trigram is recalled out of order, it is more likely to be recalled in an adjacent position than a more distant position.

Analysis of the data shown in Figure 2 is easier if the data are replotted to show the proportion of errors as a function of the distance between the original position and the reported position. Figure 3 shows the proportion of all movement errors as a function of distance of movement. The plot also shows chance performance. In this study, an item recalled in the incorrect position could be either one or two positions distant. There are four opportunities for movements to adjacent positions (e.g.,

Item 1 could be recalled in Position 2; Item 2 could be recalled in Position 1 or Position 3; and Item 3 could be recalled in Position 2) but only two opportunities for movements of Dis- tance 2 (e.g., Item 1 could be recalled in Position 3 and Item 3 could be recalled in Position 1). A chi-square test revealed that the observed differed significantly from what would be ex- pected by chance, �2(1, N � 63) � 46.12, p � .001. That is, there are more errors at near positions than one would expect by chance and fewer errors at more distant positions than one would expect by chance.

Brown, Preece, and Hulme (2000, Figure 2) plotted move- ment errors from six different experiments, which included studies with no retention interval; studies with a retention interval of up to 24 hr; and studies with list lengths of up to 16 items. In all cases, the patterns plotted by Brown et al. are consistent with that shown in Figure 3. Despite variations in the duration of the retention interval, the method of presentation (whether simultaneous or sequential), and the list length (whether the list has three items or 16 items), the pattern remains consistent: More errors occur at close distances and fewer errors at longer distances. In this regard, then, results from Brown-Peterson are the same as those seen in all those other paradigms and are consistent with predictions of SIMPLE.

Figure 2. The proportion of times each of the three consonants was recalled in each of the three possible positions. The top left panel shows the data collapsed over delay, and the remaining panels show the gradients for each delay in Experiment 1.

67POSITIONAL UNCERTAINTY IN BROWN-PETERSON

Experiment 2

The data in Experiment 1 show that recall of consonant trigrams produces movement gradients consistent with those observed from other paradigms. Experiment 2 was designed to measure protru- sion errors, where an item from an earlier list is produced as a response to a later list. The same basic design and procedure were used, but some changes were made to increase the number of observations per condition. First, only three delays (3 s, 6 s, and 12 s) were used rather than five, and second, 42 trials were presented rather than 40.

Method

Subjects. Thirty-four different undergraduates from Memo- rial University of Newfoundland volunteered to participate in exchange for a small honorarium. All identified themselves as native speakers of English.

Procedure. The procedure for Experiment 2 did not differ from that of Experiment 1 except in the ways already discussed: namely, the completion of more trials and the use of fewer delay conditions.

Results and Discussion

Overall accuracy on the math task was higher than in Experi- ment 1, with a mean of 0.666 (SD � 0.221). This is most likely due to the elimination of the longest delay.

As in Experiment 1, Experiment 2 replicated the oft-reported finding that with traditional scoring, recall of consonant trigrams in a Brown-Peterson task decreases with increasing delay, in this case from 0.601 to 0.361, as can be seen in Figure 1. A one-way repeated-measures ANOVA revealed a significant effect of delay, F(2, 66) � 18.010, MSE � 0.032, partial �2 � 0.353, p � .001.

The data were rescored as if the test was immediate serial recall. A 3 (Delay) � 3 (Serial Position) repeated-measures ANOVA revealed a significant main effect of delay, F(2, 66) � 21.143, MSE � 0.067, partial �2 � 0.391, p � .001, with recall decreasing from 0.745 at 3 s to 0.524 at 12 s. There was also a significant main effect of position, F(2, 66) � 3.955, MSE � 0.066, partial �2 � 0.107, p � .05, with recall decreasing with position from 0.634 to 0.616 to 0.584 for Positions 1, 2, and 3, respectively. Unlike in Experiment 1, the interaction was significant, F(4, 132) � 3.346, MSE � 0.014, partial �2 � 0.092, p � .05. For the 12-s delay, recall of the third letter was much worse than recall of the first two whereas in the other two delays, the difference was smaller.

The main data of interest are the protrusion errors. For this analysis, the only protrusions scored were those from the imme- diately prior list, as protrusions from lists more remote were too few to produce reliable findings. Figure 4 shows the data in two forms. The left panel shows the frequency that a letter in each position on the prior list was recalled in each of the possible positions on the following list. The right panel shows the data replotted as movement gradients along with chance performance. A chi-square test revealed that the observed differed significantly from what would be expected by chance, �2(1, N � 34) � 5.53, p � .02.

As with the within-list errors, the between-list errors shown in Figure 4 resemble those observed in other paradigms that assess memory for order (e.g., Brown et al., 2000) and are also consistent with the predictions of SIMPLE.

General Discussion

The traditional way of scoring data in the Brown-Peterson paradigm is to count a response as correct only if all three letters are reported in the correct order. If one letter is missing, or if two letters swap position, no credit is given for being partially correct.

Figure 3. The proportion of errors as a function of the distance between the original position and the reported position in Experiment 1 (data points) and chance performance (line).

Figure 4. The number of times each item from List N - 1 was incorrectly recalled in each of the three positions in List N (left panel) and the same data replotted as the proportion of errors as a function of the distance between the original position and the reported position in Experiment 2.

68 QUINLAN, NEATH, AND SURPRENANT

Such scoring yields the classic forgetting function that shows accuracy decreasing rapidly as the duration of the distractor task increases. Both Experiment 1 and 2 found this pattern. This method of scoring makes sense if one conceives of the consonant trigram as a single unit.

However, the consonant trigram can also be thought of as a list of three letters and therefore the task can be considered as an example of a serial recall task. With this scoring method, subjects are given partial credit if they report only one or two of the three consonants correctly. With this scoring method, accuracy also decreases with increasing delay, but the errors can also be ana- lyzed. When a letter was not recalled in its original position, it was more likely to be recalled in an adjacent position than a more distant position than would be expected by chance. Similarly, when a letter from an earlier list was recalled in a later list, it was more likely to be recalled in its correct position (albeit in the wrong list) than in a different position than would be expected by chance. These error gradients resemble those observed in other serial order tasks (see Figure 2 of Brown et al., 2000). These tasks include both immediate recall (a task thought to tap STM) and recall delayed by as much as 24 hr (a task that must tap long-term memory). The tasks also include lists with as few as three or four items (within the capacity of short-term and working memory) and list with as many as 16 items (well beyond the capacity of either short-term or working memory). Moreover, these gradients can be observed with incidental learning, when there is no reason to suppose that a person would be rehearsing an item to maintain it in STM or working memory (e.g., Nairne, 1992).

Historically, decay theories have had difficulty in accounting for both position error gradients and protrusion gradients. As noted by Healy (1974), decay theories predict that when an item cannot be recalled accurately, it is because the information stored about the item has decayed too much so that it is no longer useful. Therefore, the subject is forced to make a guess from a pool of likely responses (e.g., if the lists have all contained consonants, then the guess will be a consonant; if the lists have all contained digits, then the guess will be a digit). This account cannot predict that the subject will be more likely to recall a near neighbor than a more distant neighbor. It also cannot predict that when the subject guesses an item that happens to be from the prior list, the item will most likely be placed in its original position. Therefore, observing both position error and protrusion gradients in a typical Brown-Peterson task is consistent with the numerous previous studies showing that decay is not a viable explanation of forgetting in this task. Both the Primacy Model (Page & Norris, 1998) and the Time-Based Resource-Sharing Model (as implemented by Oberauer & Lewandowsky, 2011) view forgetting in the Brown-Peterson paradigm as being primarily due to decay, and although both can account for the major finding of decreased recall with increasing delay, neither can account for the systematic pattern of protrusion errors, for the reasons noted by Healy. Note also that neither can explain why the pattern of within-list and between-list errors in the Brown-Peterson task resembles those seen in long-term memory settings and incidental learning settings.

It may be possible to invoke a multiple store account of the results from Brown-Peterson, in which some aspects of the data are attributable to decay from STM whereas other aspects are attributed to recall from long-term memory. In addition to its lack of parsimony, this account suffers from problems in pre-

dicting, a priori, which store will be responsible for which result (see, e.g., the discussions in Nairne, 2002, and Surprenant & Neath, 2009b). Moreover, it seems to us that such an account would need to predict that the store responsible for performance needs to change as a function of the scoring method. For example, decay of information in STM is responsible for the findings when the task is scored using the standard all-or-none method, but not responsible when the task is scored using the standard serial recall method.

In contrast, SIMPLE posits that the Brown-Peterson task is simply another example of a serial order test. Because SIMPLE is a local relative distinctiveness model (as opposed to a global distinctiveness model; see Neath, Brown, McCormack, Chater, & Freeman, 2006), a central characteristic of the model is that items near to one another in psychological space will be more confusable than items that are more distant. It is this feature that makes SIMPLE predict that both position error gradients and protrusion errors will be observed in the Brown-Peterson task and that both types of gradients will resemble those observed in many different types of tasks.

The results also give additional support to the idea that general principles of memory do exist and do apply widely regardless of the hypothetical underlying memory system (Surprenant & Neath, 2009a). It has previously been suggested that the gradients that are the focus of this article are a general characteristic of human memory whenever the task involves order (Brown & Vousden, 1998), and indeed, these gradients likely qualify as a “principle” according to the definition offered by Surprenant and Neath. One reason this may qualify as a principle is that these gradients are observed in many different memory tasks, not only serial recall tasks such as memory span and Brown-Peterson, but also in other tasks such as speech production (see Brown et al., 2000). Of importance, these tasks tap a broad range of cognitive activities and are thus inconsistent with a fractionated collection of memory systems, each following its own rules and principles. If the differ- ent hypothetical memory systems all follow the same principles, there is no compelling reason to maintain the proposed distinc- tions. Instead, the data support the idea that differences in memory arise, not because the information is processed and recalled using different stores, but because the relative distinctiveness of items in memory varies as a function of task, stimulus materials, and processing.

We have demonstrated that position error gradients and protru- sion gradients are observable in a Brown-Peterson task, a finding consistent with the claim that such gradients may be an example of general principle of memory. Although it is not possible to prove that a principle truly is general (i.e., that a principle does always apply in all possible tasks), it is trivially easy to disprove the generality of a principle. For example, SIMPLE has to predict these gradients, but it could have been the case that the Brown- Peterson task is unique and does not give rise to this pattern of errors. Had this been the case, then the generality of SIMPLE and the generality of the gradient principle would have been severely compromised. Given that it was not, however, it provides yet another demonstration of the similarities that exist over many different kinds of tasks that are thought to tap many different kinds of memory systems (Surprenant & Neath, 2009a).

69POSITIONAL UNCERTAINTY IN BROWN-PETERSON

Résumé

Depuis l’article déterminant de McGeoch (1932), aucun autre document sur la mémoire à long terme n’a évoqué le déclin de la trace en tant que cause de l’oubli. À l’inverse, de multiples articles sur la mémoire à court terme cernent ce même déclin comme cause, offrant les résultats de tâches Brown-Peterson en appui. Voici la description de deux expériences ayant eu recours à une tâche Brown-Peterson standard, mais pour laquelle les résultats ont été évalués de deux façons. Selon la notation traditionnelle (le rappel intégral ou non du trigramme à 3 consonnes), le rendement a diminué à mesure qu’augmentait le délai. Selon la notation du rappel sériel immédiat (le rappel des lettres selon l’ordre qu’elles avaient), on a pu constater les gradients d’erreur de position habituels (Expérience 1) et en saillie (Expérience 2). Plus préci- sément, lorsque la lettre initiale du trigramme à consonnes n’était pas la première dont le sujet se souvenait, il était plus probable qu’il la nomme en deuxième lieu plutôt qu’en dernier. De plus, si une lettre d’une liste antérieure était mentionnée par erreur pour une liste subséquente, il était plus probable qu’elle le soit à sa position originale dans le trigramme. La présence de tels gradients va à l’encontre de la notion de déclin de la trace, mais elle est prévue par le SIMPLE, modèle de la mémoire à distinctivité locale. En outre, leur présence est plus conforme à la notion que l’oubli dans une tâche Brown-Peterson suit les mêmes principes observés dans d’autres tâches de mémorisation.

Mots-clés : Brown-Peterson, mémoire à court terme, oubli, SIM- PLE, rappel sériel.

References

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation (Vol. 2, pp. 89 –195). New York, NY: Academic Press. http://dx.doi.org/10.1016/S0079- 7421(08)60422-3

Baddeley, A. D. (1990). Human memory: Theory and practice. Oxford: New York, NY: Oxford University Press.

Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time constraints and resource sharing in adults’ working memory spans. Journal of Experi- mental Psychology: General, 133, 83–100. http://dx.doi.org/10.1037/ 0096-3445.133.1.83

Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory. Psychological Review, 114, 539 –576. http://dx.doi.org/ 10.1037/0033-295X.114.3.539

Brown, G. D. A., Preece, T., & Hulme, C. (2000). Oscillator-based mem- ory for serial order. Psychological Review, 107, 127–181. http://dx.doi .org/10.1037/0033-295X.107.1.127

Brown, G. D. A., & Vousden, J. (1998). Adaptive sequential behaviour: Oscillators as rational mechanisms. In M. Oaksford & N. Chater (Eds.), Rational models of cognition (pp. 165–193). Oxford, UK: Oxford Uni- versity Press.

Brown, J. (1958). Some tests of the decay theory of immediate memory. The Quarterly Journal of Experimental Psychology, 10, 12–21. http:// dx.doi.org/10.1080/17470215808416249

Conrad, R. (1960). Serial order intrusions in immediate memory. British Journal of Psychology, 51, 45– 48. http://dx.doi.org/10.1111/j.2044- 8295.1960.tb00723.x

Daniels, A. H. (1895). The memory after-image and attention. The Amer- ican Journal of Psychology, 6, 558 –564. http://dx.doi.org/10.2307/ 1411192

Estes, W. K. (1972). An associative basis for coding and organization in memory. In A. W. Melton & E. Martin (Eds.), Coding processes in human memory (pp. 161–190). Washington, DC: Winston.

Fuchs, A. H., & Melton, A. W. (1974). Effects of frequency of presentation and stimulus length on retention in the Brown-Peterson paradigm. Jour- nal of Experimental Psychology, 103, 629 – 637. http://dx.doi.org/ 10.1037/h0037196

Healy, A. F. (1974). Separating item from order information in short-term memory. Journal of Verbal Learning and Verbal Behavior, 13, 644 – 655. http://dx.doi.org/10.1016/S0022-5371(74)80052-6

Henson, R. N. A. (1998). Short-term memory for serial order: The Start– End Model. Cognitive Psychology, 36, 73–137. http://dx.doi.org/ 10.1006/cogp.1998.0685

Kincaid, J. P., & Wickens, D. D. (1970). Temporal gradient of release from proactive inhibition. Journal of Experimental Psychology, 86, 313–316. http://dx.doi.org/10.1037/h0029991

McGeoch, J. A. (1932). Forgetting and the law of disuse. Psychological Review, 39, 352–370. http://dx.doi.org/10.1037/h0069819

Melton, A. W., & von Lackum, W. J. (1941). Retroactive and proactive inhibition in retention: Evidence for a two-factor theory of retroactive inhibition. The American Journal of Psychology, 54, 157–173. http://dx .doi.org/10.2307/1416789

Mewhort, D. J., Campbell, A. J., Marchetti, F. M., & Campbell, J. I. (1981). Identification, localization, and “iconic memory”: An evaluation of the bar-probe task. Memory & Cognition, 9, 50 – 67. http://dx.doi.org/ 10.3758/BF03196951

Nairne, J. S. (1991). Positional uncertainty in long-term memory. Memory & Cognition, 19, 332–340. http://dx.doi.org/10.3758/BF03197136

Nairne, J. S. (1992). The loss of positional certainty in long-term memory. Psychological Science, 3, 199 –202. http://dx.doi.org/10.1111/j.1467- 9280.1992.tb00027.x

Nairne, J. S. (2002). Remembering over the short-term: The case against the standard model. Annual Review of Psychology, 53, 53– 81. http://dx .doi.org/10.1146/annurev.psych.53.100901.135131

Nairne, J. S., & Neath, I. (2013). Sensory and working memory. In A. F. Healy & R. W. Proctor (Eds.), Comprehensive handbook of psychology: Vol. 4: Experimental Psychology (2nd ed., pp. 419 – 445). New York, NY: Wiley.

Neath, I., & Brown, G. D. A. (2006). SIMPLE: Further applications of a local distinctiveness model of memory. In B. H. Ross (Ed.), The psy- chology of learning and motivation (pp. 201–243). San Diego, CA: Academic Press. http://dx.doi.org/10.1016/S0079-7421(06)46006-0

Neath, I., & Brown, G. D. A. (2012). Arguments against memory trace decay: A SIMPLE account of Baddeley & Scott. Frontiers in Cognition, 3, 35. http://dx.doi.org/10.3389/fpsyg.2012.00035

Neath, I., Brown, G. D. A., McCormack, T., Chater, N., & Freeman, R. (2006). Distinctiveness models of memory and absolute identification: Evidence for local, not global, effects. Quarterly Journal of Experimen- tal Psychology: Human Experimental Psychology, 59, 121–135. http:// dx.doi.org/10.1080/17470210500162086

Neath, I., & Surprenant, A. M. (2003). Human memory: An introduction to research, data, and theory (2nd ed.). Belmont, CA: Wadsworth.

Neath, I., VanWormer, L. A., Bireta, T. J., & Surprenant, A. M. (in press). From Brown-Peterson to continual distractor via operation span: A SIMPLE account of complex span. Canadian Journal of Experimental Psychology.

Oberauer, K., & Lewandowsky, S. (2011). Modeling working memory: A computational implementation of the Time-Based Resource-Sharing the- ory. Psychonomic Bulletin & Review, 18, 10 – 45. http://dx.doi.org/ 10.3758/s13423-010-0020-6

Page, M. P. A., & Norris, D. (1998). The primacy model: A new model of immediate serial recall. Psychological Review, 105, 761–781. http://dx .doi.org/10.1037/0033-295X.105.4.761-781

70 QUINLAN, NEATH, AND SURPRENANT

Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individ- ual verbal items. Journal of Experimental Psychology, 58, 193–198. http://dx.doi.org/10.1037/h0049234

Ricker, T. J., Vergauwea, E., & Cowan, N. (in press). Decay theory of immediate memory: From Brown (1958) to today (2014). Quarterly Journal of Experimental Psychology. http://dx.doi.org/10.1080/ 17470218.2014.914546

Schacter, D. L., Wagner, A. D., & Buckner, R. L. (2000). Memory systems of 1999. In E. Tulving & F. I. M. Craik (Eds.), The Oxford handbook of memory (pp. 627– 643). New York, NY: Oxford.

Surprenant, A. M., & Neath, I. (2009a). Principles of memory. New York, NY: Psychology Press.

Surprenant, A. M., & Neath, I. (2009b). The 9 lives of short-term memory. In A. Thorn & M. Page (Eds.), Interactions between short-term and long-term memory in the verbal domain (pp. 16 – 43). Hove, UK: Psy- chology Press.

Received June 28, 2014 Accepted October 14, 2014 �

E-Mail Notification of Your Latest CPA Issue Online!

Would you like to know when the next issue of your favorite Canadian Psychological Association journal will be available online? This service is now available. Sign up at http://notify.apa.org/ and you will be notified by e-mail when issues of interest to you become available!

Avis par courriel de la disponibilité des revues de la SCP en ligne!

Vous voulez savoir quand sera accessible en ligne le prochain numéro de votre revue de la Sociétè canadienne de psychologie préférée? Il est désormais possible de le faire. Inscrivez-vous à http://notify.apa.org/ et vous serez avisé par courriel de la date de parution en ligne des numéros qui vous intéressent!

71POSITIONAL UNCERTAINTY IN BROWN-PETERSON