Molyneux's Question--good fit for an HTM answer

I was not aware of this ‘question’ until I saw this call for abstracts. Looks a bit like the finger knowing the cup handle or the eyelid feeling the shape of the contact lens…


Visions act as one of the ways to collect inputs, if there is a problem in the vision other routes of getting input gets dominant over time. The brain gives preference to the routes which gives most of the inputs.

1 Like

It takes some time to train the visual system. Experiments with kittens show that there is a critical period in visual perception to learn things like vertical lines. Even very basic things like vergance are learned - see amblyopia for example.
“Fixing” vision would have to take into account this training period to establish basic function, which would render the posed question moot.

The question:

would a person with blindness, able to identify cubes and spheres by touch, immediately identify these shapes by sight alone if suddenly able to see?

Maybe that’s not really what the question is asking, since it probably isn’t designed with low-level processing in mind. It seems to be asking whether we understand shapes in a manner independent from the sensory input. Maybe something like:

A person born blind learns to recognize a tree by touch. Then, that person’s vision starts working, but their sense of touch stops working. Also, the tree is moved into a totally different context (location on earth etc.) Could the person eventually realize it’s the same tree by vision alone?

That way, you give time for learning, but you still need to translate shapes between different senses without being able to experience associations between the senses.

Maybe a tree is too complicated for the answer to be yes, but I think that’s fair. If we’re modelling the world in terms of abstract shapes, you should be able to directly convert from a sensory input to shapes. Comparing language-based descriptions doesn’t count, in my opinion. “It’s metal, it’s 1 foot across, it’s a sphere” is far less detailed than what we see when we look at a tree.

Different sensory modalities have different information to work with, e.g. color in vision and precise texture in touch. So the ambiguities in what you’re sensing are different in each sense, so the ambiguities in the precise spatial configuration of an object is different in each sense. So converting to a fully general spatial system wouldn’t be the main way we recognize complex objects. It’s probably best to still have some fully general representation of shape in a parallel processing manner, but to represent the same object through different senses as being the same, it’d need to be pretty imprecise.


The famous old case is the blind person who feels the lower deck of a double decker bus, because they have used it to travel about. Later in life they get some sight back and start to draw. But they only draw the lower deck of the bus they can now see.

Similar to the young cats who are restricted to seeing only horizontal dimensions, and later don’t see vertical ones.

Hope that helps



The answer is no (?)

These children are born blind due to congenital cataracts that are later corrected. After that, they have to learn to see. Even then, they are unable to recognize faces (including their own).