Search

This paper introduces two schemes to correct bit errors caused by defective memory cells in high-speed random-access memory systems. The schemes are addressed to word-organized memories produced by the integrated technologies. One of the two schemes calls for encoding of input information and the other does not. The schemes are simple, economical for the technologies concerned, and exhibit a regularity which makes it possible to fabricate the necessary additional hardware within the same technology.

Ss decided whether or not a test digit had occurred in a previously presented memory set. Decision time increased with the size of the memory set at equal linear rates for positive and negative responses. Positive decision times were sensitive to the serial position of the test digit. Together, these results are incompatible with serial-exhaustive models; some modifications are necessary in order to accommodate the obtained results.

A Sternberg high-speed memory scan task was used as a retention interval activity in a standard Brown-Peterson paradigm. Two levels of Sternberg search set, one or five items, were employed to see if the differential memory load produced by the different size search sets resulted in differential recall in the Brown-Peterson task. A comparison was made to a control condition in which S was required to shadow digits during the retention interval. No differential effect on recall of memory load was found when the Sternberg tasks were compared across three different retention intervals. Very little forgetting was produced by the Sternberg tasks as compared to the digit shadowing condition. Results are discussed in terms of the need for verbal activity to prevent rehearsal in short-term memory.

Four groups of 13-14-year-olds, classified on the basis of Raven's Progressive Matrices scores as intellectually above average, average, below average and retarded, were subjects in a Sternberg (1966) type memory scanning task in which memory sets of two, three and four letters were used. Slope values of the memory search function did not differ between groups, whereas intercept values decreased with higher intelligence up to the average intelligence level. There was an overall inverse relation between intercept values and intelligence (r = -0.77) and similar relations obtained within the three non-retarded groups (rs greater than or equal to -0.57), but not within the retarded group (r = -0.06). Possible connections between these results and findings from inspection time studies are discussed.

Third-grade, high-school, and adult normal subjects and two diagnostic categories of educable mental retardates (cultural familials and encephalopathies) were tested in a Sternberg-type “memory scanning” recognition task. All five groups showed the characteristic linear increase in correct reaction time as the number of items in the memorized set increased. The slopes of the linear functions, however, were steeper for the two retardate samples than for the normal samples, suggesting a central processing deficit which could not be attributed to a lag in development.

The additive factors model of memory retrieval (Sternberg, 1969, 1975) proposes that the slope of the choice reaction time (RT)/set-size function represents the time taken to execute serial, exhaustive comparisons in response to the probe stimulus on the item recognition task. An alternative model of memory retrieval (Ratcliff, 1978) proposes that the decision as to whether the probe is a member of the positive set is achieved through a parallel search that is self-terminating if a match is detected or exhaustive across nonmatching trials. An empirical test, which relies on the additive factors methodology of Sternberg (1969, 1975), is to ascertain whether a factor, likely to affect comparison time, produces an interactive or an additive effect with set size in its influence on choice RT. Such a factor, stimulus familiarity, was examined in the present study, employing the item recognition task (fixed-set procedure) and a completely within subjects design. The findings supported Sternberg's (1969, 1975) serial, exhaustive comparisons view of the memory scanning process. Stimulus familiarity did interact with set size in its influence on choice RT.