The World Memory Championships
Provide useful estimates of the maximum rate of learning from images and digits:
The maximum learning rate from Images
One of the events in the world memory championships record holders, is to rapidly memorise one or more packs of randomly shuffled playing cards.
As there are just 52 cards, we can calculate the information required to memorise each card as 5.7 bits per card (the log to base 2 of 52).
We can calculate the information rate from the time taken per pack.
The world record holders can memorise a pack in just over 21 seconds.
This corresponds to 14 bits per second.
You can see just how difficult this is with the following examples:
One Card per Second = 5.7 bits per second – Click to Try
Two Cards per Second = 11.4 bits per second – Click to Try
Two Cards per Second but displayed in pairs each second – Click to Try
also = 11.4 bits per second. Note that UK champion Ben Pridmore found it faster by memorising 26 pairs of cards.
Note: My calculation of how many bits it takes to memorise a pack of cards in sequence, assumes an equal number of bits per card. However, if the subject remembers which cards have been eliminated, successive cards would require fewer bits and the last card would be obvious! This might appear to be a significant advantage, and it is for people playing games like blackjack where they have time to think, but it would require significantly greater mental processing to make it work, so is probably significantly slower overall.
The maximum learning rate from Decimal & Binary digits
You can experiment with the task of memorising numbers on this website Click Here
The highest rates of all are achieved for simple maths tasks, such as adding a sequence of digits.
This task involves very little introspection as only the running total needs to be memorised.
The record for adding 100 decimal digits is 17.7 bits per second.
and the record for adding ten ten-digit numbers, ten times, is 16.7 bits per second
The results are nearly identical despite there being a factor ten between the task duration.
This suggests that we are seeing the learning bottleneck limit itself. Three subjects gave almost identical results too.