"The average person can hold roughly four unrelated facts in working memory." What is the source for this?

[lynellf]

Great read, and I agree with much of the opinions held.

However, this quote becomes the anchor for many of the following opinions/assumptions:

The average person can hold roughly four facts in working memory

Is this based on Miller's Law? If so, the number is seven instead of four.

[zakirullin]

Hi there, thanks for your warm words!

Miller has one of the most cited papers in psychology titled "The Magical Number Seven, Plus or Minus Two" (1956). Which suggests that the actual capacity falls between 5 and 9.

There are more recent papers like this: "The magical number 4 in short-term memory: a reconsideration of mental storage capacity" (2001).

But that's not the point - we can find papers on the internet to prove just about anything. The point is that the brain is the most complex object in the universe, and we are nowhere close to fully understanding how it works. The only thing we can do is create models.

Number of chunks we can hold in our working memory:

5-9 chunks (suggested by Miller)
3.5 chunks (suggested by a friend of mine)
4 chunks (suggested by Cowan)

These are all simplified models (approximations).

All models are wrong, but some are useful

None of the models above are quite right, but they can be useful as long as they inherit some properties.

For the purpose of this handbook we can choose any model we like, as long as it inherits one important property - "Working memory is limited to a finite not-so-big number of objects."

So I chose four.

[colerss]

Miller's paper is often over-applied; it was by no means a comprehensive test of this capability and strictly measured "information units" rather than "information chunks". It was tested extensively with digits, and focused more on instant recall than it did persistent working memory usage.

With letters, it drops to 6. With words and multi-digit numbers, it drops to 5 or 4 depending on the complexity. Instructions and concepts are typically between 1 and 4, depending on their complexity.

For programming 3-4 would definitely be a more reasonable baseline when all we have to think about is variables, functions and their intentions/connections.

[aquacalc]

Barb Oakley's book, _Learning How to Learn_, and her eponymous Coursera course, which can be audited at no charge, offer good intro-level discussions of this topic, and more.

[jverce]

A handful then?