In information security a shared secret is often used to authenticate users. (so-called sth-you-know authentication factor; examples include passwords, passphrases, numeric pin-s, graphical passwords (http://blogs.msdn.com/b/b8/archive/2011/12/16/signing-in-with-a-picture-password.aspx , http://passtouch.com/).

A common form of secret, 8-character password over all 95 ASCII chars, would contain ~53 bits of entropy ($95^8$ possibilities).
A password in top 1000 most common passwords would contain only ~10 bits of entropy (1000 possibilities) but would be easier to remember.
A 4-word passphrase over 10'000 most common words would also contain ~53 bits of entropy ($10'000^4$ possibilities), but would maybe(?) be easier to remember that 8 random characters.

My question is based on the assumption that same amount of information represented in different formats (sound, image, text, song, movement, etc.) are not equally easy to remember. Maybe passwords (or text based methods in general) are not the best option.

In which form can humans memorize the largest amount of information and present it to a computer device? I'd be glad if you could link some relevant research too.