Coding, capacity, duration of memory

The words used in the study had no personal meaning to the participants.

When processing more meaningful information, people may use semantic coding even for STM tasks. 

This means the results of this study have limited application. We should be cautious about generalising the findings to different kinds of memory task. 

Early research in psychology often lacked adequate control of extraneous variables.

For example, some participants may have been distracted while they were being tested so they didn't perform as well as they might.

This would mean that the results may not be valid because there were confounding variables that were not controlled. However, these results have been confirmed in other research, supporting its validity. 

For example, Cowan reviewed other research.

He concluded that the capacity of STM was only about 4 chunks.

This suggest that the lower end of Miller's estimate (5 items) is more appropriate than 7 items.

Trying to memorise consonant syllables does not reflect most real-life memory activities where what we try to remember is meaningful.

So it could be argued that this study lacked external validity.

On the other hand, we do sometimes try to remember fairly meaningless things, such as phone numbers. So the study is not totally irrelevant. 

Real-life meaningful memories (e.g. of people's faces and names) were studied.

When lab studies were done with meaningless pictures to be remembered, recall rates were lower (e.g. Shepard).

The downside of such real-life research is that confounding variables are not controlled, such as the fact that Bahrick's pariticpants may have looked at their yearbook photos and rehearsed their memories over the years.