R. Conrad

Order error in immediate recall of sequences

The only attempts to account for order errors in serial recall, have been informational analyses leading to models involving separate storage for item and order information. The present study begins by showing that there is a high degree of association between order and item errors. Items which acoustically confuse with each other are likely to transpose in recall. This result suggests that apparent order error could arise from two or more independent item errors substituting for each other. Taking into account the demonstrated association between acoustic similarity of items and liability to confuse in recall, the chance that two independent item errors will form a transposition is much increased beyond the pure chance level. This chance is even further increased because (a) acoustic confusions in recall reciprocate, and (b) when sequences are such that repetition of items does not occur, when Ss have made one mistake, they are more willing to make a wrong report on a later item than to be led into a repetition. It is concluded that memory models do not necessarily need a mechanism which could transpose the order of items in storage. A simpler model is suggested in which items are fixed in the input order, encoded only according to properties of individual items. There is in addition a response availability store from which item substitutions are drawn. The size of this store is independent of vocabulary size, is relatively small, and consists primarily of recent responses only. Thus even when sequence items are drawn from a very large vocabulary, this modified mutual substitution model would still be adequate to account for differences between the order in which items enter and leave a memory store.

Interference or decay over short retention intervals

It is well established that if a formally noninterfering task which minimizes rehearsal is interposed between presentation and recall of a short consonant sequence, recall lessens as retention interval is increased. Superficially, this appears to be a classical example of decay. Keppel and Underwood have proposed an alternative explanation in interference-theory terms suggesting that forgetting in this paradigm is due to proactive interference from extinguished associations from prior trials which recover during the retention interval. An experiment was carried out which varied retention interval after four consonants had been presented. More forgetting occurred as a result of the longer interval, but the main dependent variable was the nature of “intrusions.” For neither the short nor long interval were wrong letters equiprobable, errors tending to be acoustically similar to the correct letter. But the distribution of long-interval errors tended more towards random than did that for the short interval. It is argued that the difference in error distribution between short and long interval is incompatible with the Keppel and Underwood explanation. A modification of decay theory is proposed which regards decay as a loss of discriminative characteristics (in the present case acoustic), and recall as a process involving discrimination of available traces. This model would be supported by the error-distribution data.

The developmental role of vocalizing in short-term memory

When young children perform a short-term memory (STM) task with familiar objects, they may be instructed to identify items aloud, or they may be observed to do so spontaneously. When recall is compared with that from a “silent” condition, conflicting data have been reported. A STM experiment was carried out using 5-year-old children who were sometimes instructed to vocalize and sometimes instructed not to. For both conditions there was an independent assessment of whether naming (verbal mediation) occurred. It was found that changing the vocalizing instruction changed naming behavior for many children. However, with vocalizing held constant, naming always aided recall. The role of vocalizing itself is considered to be confounded with attention. A developmental model is discussed which relates vocalizing, attention and naming in STM.

Practice, familiarity and reading rate for words and nonsense syllables

Two experiments were carried out to test a number of hypotheses regarding the effects of practice and S-R familiarity on choice reaction times. In Experiment 1 subjects read lists of nonsense syllables of varying levels of association value and differing degrees of choice. It was found that the slope of reading time per syllable was linear with respect to the log of the number of choices, but that the slope was steeper for syllables of lower association value. In Experiment 2, 4-and 32-choice lists of the same average association value nonsense syllables were read on three successive days, as were lists of common three-letter words. The words showed zero slope against number of choices, and the slope for the nonsense syllables decreased with practice. These results are considered to support recent modifications to views on the role of information theory in psychology.

Predictability and Immediate Memory for Consonant Sequences

Two measures which have been shown to predict the ease of learning trigrams, namely log letter frequency and sequential predictability, were applied to data from an experiment on short term memory. This involved the immediate recall of 120 six-letter consonant sequences which were presented visually one letter at a time. A significant correlation was found between the probability that a given sequence would be recalled correctly and both its mean log letter frequency (r = 0.308, p < 0.001), and its mean predictability (r = 0.393, p < 0.001). Partial correlation showed only a marginally significant effect of log letter frequency when predictability was partialled out (r = 0.161, 0.05 < p < 0.1). With log letter frequency partialled out, however, a reliable correlation between predictability and recall score remained (r = 0.300, p < 0.001).

Rate of presentation and the acoustic similarity effect in short-term memory

Ss attempted to recall sequences of six consonants drawn from either an acoustically similar set (B C D G P Q T V), or from a relatively dissimilar set (H K M P R S W Y). Letters were presented visually at a rate of 60 or 120 letters per min. Performance was impaired by acoustic similarity (p<.001) but there was no effect of rate of presentation and no interaction between rate and similarity. This does not support a limited channel capacity interpretation of the acoustic similarity effect.

Acoustic factors versus language factors in short-term memory

Forty five Ss recalled 6-consonant sequences immediately after letter by letter visual presentation. The main factor contributing to ease of recall was within-sequence acoustic confusability. Language habits were relatively unimportant. Single-letter language frequency was unrelated to recall; second order effects made a small but significant contribution.

The role of the interpolated task in short-term retention.

It has been proposed that a single set of operations based on classical interference theory is adequate to describe the phenomena of both short- and long-term memory. An article by Keppel and Underwood (1962) argues that short-term forgetting is due to proactive interference and, by implication, not a result of trace decay. An experiment which varied retention interval and the nature of the interpolated task, gave results which indicate that when the amount forgotten and the nature of errors are considered, a decay model is supported, the proactive interference suggestion being untenable.

An apparatus providing signals at “random” intervals

\YHET designing apparatus for the study of skill, it is often difficult to decide how complex the task should be. On the one hand, equipment like the Cambridge Cockpit (Davis, 1946) sets a task which is complex and integrated, overall measures of performance are possible, but may be difficult to interpret in detail. On the other hand, attempts to break the task and its performance into simpler elements, such as are found in the NcDougall Dotter or the Triple ’Tester, restrict the parameters of behaviour that can be studied with confidence. They produce tasks so artificial that it becomes difficult to relate the performance measured to everyday behaviour. There are, however, some daily-life skills where an integrated behaviour is not possible, but in which discrete signals for action occur at irregular-apparently random-intervals of time. The occurrence of these signals can often be predicted with more or less accuracy, but sometimes even this is not possible. Such a task occurs in cotton spinning and winding, the signals then being breakages of the cotton thread. The apparatus which forms the subject of this note was built to enable this type of behaviour to be studied. The apparatus presents a task in which signals occur with an nppyoxiunately random distribution in time. ’There is virtually no repetition of the pattern of signals, the average rate of their occurrence being under the experimenter’s control. The phlsical nature of the sigrials is not fixed, a number of different types being possible, and with some arrangements the subject can predict when they will occur. A recent mathematical discussion by Cox and Smith (1953) has provided the basis for the design. ’The point made in this paper is that, if a number of sources each emit a periodic signal, the combined effect is a series of signals with an approximatel], exponential distribution of the time intervals between them. The limiting factors are that the periods must be slightly different and prime with respect to each other, not too different, and there must be a sufficient number of them. ’tl’hatever the number and length of the periods, the distribution of the time intervals between signals can be calculated, and the divergence from the exponential measured. An apparatus using this principle consists of 16 small dials mounted in a 4 x 4 fashion. Each dial carries a pointer driven through a gear link by a single variable speed motor. This gear link is in each case such that every pointer revolves a t a slightlgr different speed. Behind each dial face and mounted on the pointer spindle, is a cam which actuates a relay once every revolution. Subsequent events, electrical or mechanical, will depend on the type of signal required. For example, the closing of the relay might stop the pointer at an indicated mark until the subject makes a correct response, the time of both events being recorded. Or by responding correctly before the pointer reached the indicated mark, the subject might prevent the pointer from stopping. Again the relevant time relations can be recorded easily. In both cases the beliaviour of the subject would actia2ely affect the nature of the display. Alternatively, closing the cam-actuated relay might leave the pointer unaffected, so that it continued its revolutions, but would