Sternberg Search

Memory set size

(1, 3, 5) (x-axis)

The response time of the participant should reflect the time spent searching STM to determine whether the probe number is part of the list.

By varying the number of items in the list, Sternberg hypothesized that he could test several theories of STM search.

Sternberg's data were consistent with the successive or serial search. Specifically, he found that response times grew linearly with increases in memory set size.

For each additional item in the memory set, participants took (on average) an additional 38 ms to make their responses. Thus, it seems the probe item is compared one-by-one with each item in STM, and each comparison takes approximately 38 milliseconds.

In contrast to a serial search, it is possible that there is a parallel search.

For example, some neural network theories of STM suggest that every item in memory can be accessed simultaneously due to the parallel nature of search in these networks.

Two general types of retrieval theories make quite different predictions.

If memory search requires consideration of each item in succession - a serial search - the response times should increase with memory set size because the participant will, on average, have to search through more items for larger set sizes.

Many researchers of memory believe that there exists a short-term memory (STM) system that holds information for a few seconds.

If the information in STM is not transferred to long term memory (LTM) for more permanent storage, it vanishes.

Your results are in the graph below which plots response time as a function of memory set size, with separate curves for "Present" and "Absent" trials. You should find that both curves increase with set size, and that the curves are nearly straight lines.

Sternberg found that the "Present" and "Absent" curves were nearly superimposed, but you may not find that to be true. Your data would agree with Sternberg's general findings if your curves were parallel.

A fixation point will appear in the middle of the screen. After a second, a memory set consisting of 1, 3, or 5 digits will appear on the screen for 1.2, 3.6, or 6 seconds, respectively.

Study the numbers and commit them to memory. The memory set will disappear and then a probe item will appear 1 to 3 seconds later. Your task is to determine whether the probe item was in the list of just presented numbers (respond present) or was not in the list (respond absent).

At first glance, it might seem that a "Present" trial could terminate as soon as the probe item is matched with the appropriate item in STM. With a self-terminating search, one would expect "Present" trials to be faster, but the data contradict this hypothesis.

The counterintuitive finding from Sternberg's study is that search of STM is always exhaustive. That is, the cognitive processes responsible for searching STM for a particular item search through all items in STM before reporting whether the probe item is in memory or not.

On each trial, you saw 1, 3, or 5 numbers for 1.2, 3.6, or 6 seconds (respectively). Then, 1 to 3 seconds later, a single number was shown. Your task was to indicate as quickly and as accurately as possible whether this probe number was in the list just presented.

The independent variable is the number of digits in the memory set (1, 3, or 5). The dependent variable is the speed with which you made a correct decision.

Participants were shown a short (one to six items) list of numbers and asked to memorize them. After putting them to memory, a probe number was shown. The probe number was either one of the numbers in the list or a new number.

The participant was to respond as quickly as possible, indicating whether the probe number was in the list or not.

When he compared response times for probe "Present" and "Absent" trials (probe item was in the memory set or not, respectively), Sternberg found no differences in response times.

This finding is notable because an "Absent" response can be made only after all items in STM have been searched and found not to match the probe item.

You should find that response time increases with set size, and that when plotted, the lines for present and absent are

nearly straight, parallel lines.

If such a search took place in STM, the prediction is

that response times would not vary as the memory set size increased.