Perhaps to test this, you could have two sets of identical statues.

Two dolphins need to interact with the matching statues together to get food.

If you followed the 'five monkeys and a ladder with a banana' experiment and found two dolphins who worked it out. Replace one with another. Presumably one will try and teach the other.

With enough dolphins and statues, you might be able to use the similar methods you'd use to break simple cryptography to work out what part of the communication is about the statue.

Or, you'll find out every dolphin talks to every other dolphin differently. And then and then the government bans dolphins out of fear of being unable to intercept dolphin based communications.

> Presumably one will try and teach the other.

IIRC this has already been tried and demonstrated to happen. This is the first thing that google came up with, but I'm sure Scientific American has also covered the teaching behavior.

http://www.earthintransition.org/2011/09/dolphins-teach-each...

True, but it might just be even more interesting to decipher. Imagine you could rebuilt 3D visualisations of dolphin thoughts.

is there enough data in the clicks and whistles to render an image?

There probably isn't enough data to render a full image, but how many lines do we need to draw to give the impression of a person, or a tree? Surprisingly few. Couldn't the same apply to a sonar image?

It's just an idea. I'm not saying its true, just that its an interesting possibility.

> how many lines do we need to draw to give the impression of a person, or a tree?

Context sensitive, but yes, surprisingly few.

大 is three lines.

木 is four.

人 is just two.

I'm aware, but you need special training before you can recognize that as a person. 木 is "wood" in the modern day; does 树 really make you think "that looks like a tree"?

So perhaps any visual representation of meaning would require shared context to interpret, and would then inevitably become symbolic rather than directly representational.

I don't understand what you're trying to say here, but e.g. Michelangelo's David is widely recognized as representing a person, and no shared context is required for that.