Structure of Sensory Memory


The following two are:

  • High capacity (can hold a *lot* of information for a brief moment)
  • Low duration (iconic - <1 second, echoic - a few seconds)

Visual "iconic" Memory

We have the capacity to store things in our visual memory very briefly (not even a few seconds) before it fades - e.g. A sparkler light - the way the point of light appears as a continuous light is because we're briefly storing a complete representation of where the sparkler was and where it *is*.

Auditory "echoic" Memory

The ability to continue hearing a sound after it's been presented - remembering and replaying. Given that we can only hear a sound once at a time and then wait to process it, it makes sense to have a delay and repeat inbuilt.

A: "What time is it?"
B: "What did you say?…Oh, 2:30"
- Not quite hearing it the first time, hearing it the second time, answering. Happens all the time.

Experiments About Sensory Memory

We can test our iconic memory by flashing grids of letters up for 100ms and asking people to write down everything they can remember. This was down by Sperling in 1960 and is known as Full Report Conditioning, in contrast to his (also 1960) Partial Report Conditioning experiment where participants saw a grid, a row number and just wrote down that line.



The results went from an average of 37.5% to 75% - what we can learn from this is that we know at least 75% of it very briefly (we must have learnt the whole array, being given the row after), but it wanes as we're writing it down.

These results hold if row indicated <300ms after array, else it drops to normal full array equivalence (40%).

Serial Position Curves

If we read out a list of words to people and ask them to write all they can remember after a gap, we find the following results (ordered chronologically).


From this we can see that STM and LTM are different - the first items are stored in LTM, and the last items still in STM. If we count down from 801 in threes whilst you're seeing this list - Primacy still applies, but recency doesn't. Counting prevents rehearsal although the first few are still encoded in LTM.

Primacy Effect

  • Lots of attention on first few items
  • Rehearsed a lot, processed thoroughly
  • Pushed into LTM

Recency Effect

  • Items still in STM - relies on what we've just rehearsed (although I think I did worse here because I wasn't repeating them as much =/)

Von Restorff Effect

  • Distinctive items are better recalled
  • More attention paid
  • Easily pushed into LTM
  • Easier to retrieve - less interference from similar items at retrieval time