First clear evidence that dogs do not naturally distinguish objects by shape | Hacker News Reader

First clear evidence that dogs do not naturally distinguish objects by shape

lincoln.ac.uk

61 pointsandrewcooke19 comments

Maybe it's just that particular breed of dog used in the experiment (border collie), or even that specific dog. Even the paper title cites a singular dog.

The paper 'Word Generalization by a Dog (Canis Familiaris); is Shape Important?' by Emile van der Zee, Helen Zulch and Daniel Mills is published in PLOS ONE on 21st November 2012 and is available at: http://dx.plos.org/10.1371/journal.pone.0049382*

> Maybe it's just that particular breed of dog used in the experiment (border collie), or even that specific dog.
Excellent point! Now, I would love to read a rebuttal/confirmation study over a larger population and/or varying dog breeds and I think you just stumbled upon an great idea for a research paper.

I think it said that they are planning on testing more breeds as well.

Dogs definitely recognize their acquired family, human and canine (and feline, etc), with no vocalizations. My Shetland Sheepdog had near 100% accuracy going and finding a huge selection of latex toys (caterpillars, alligators, and so on) and knew them all by name, many by name and color. And, yes, I know the obvious retort of "but dogs don't see color", which is false, as it is for cats; the correct statement is that they are both _likely_ color blind in the sense that there exist colors distinguishable by humans which to dogs (and cats) look indistinguishable. So, I could ask her where she put her green alligator and she'd repeatably, for years, with such games return from god knows where later with well over 90% accuracy (even if it was a grey alligator to her).

Seems to me that your evidence is entirely consistent with the OP. Does your dog understand shape, though?

Did you have two identical alligator toys that differed only in color?

Oooh. That would be a good test. Similar to the above comment, our guy has a number of toys and can distinguish by name. Totally going to the pet store this weekend to see if we can try that out.

I thought dogs see in shades of blue and yellow. Basically red and green show up as "yellow" to their cones.
At least thats how it was explained to me when you throw a red toy on the grass that the dog can't "see" it.

Would that suggest our language is connected to our vision?

Or tool use.

Isn't it fairly well accepted that language affects perception and even explicitly vision? Cultures that have words or expressions that other cultures don't have. Groups of people that have more granular colors and then recognize more distinct shades of those colors than those of us that have always just known "that's red or pink" rather than more distinct points between there.

Yep: http://en.wikipedia.org/wiki/Distinguishing_blue_from_green_...

I hate reading stuff like this. It makes me question my grip on reality.

This will blow your mind: http://www.youtube.com/watch?v=G-lN8vWm3m0
http://en.wikipedia.org/wiki/McGurk_effect

> This will blow your mind
Truth in advertising!

So, in effect, they're saying that we use lip-reading to distinguish between very similar phonemes? That's... yes? I'm not seeing what's so remarkable about it, I've been aware of it since forever - it's how I can understand what people next to me are saying at loud concerts and the like. Even further, sometimes I have trouble recognising what someone's telling me if I can't see their lips and no, my hearing is perfectly fine.
Guess it's one of those things where you think everybody senses the world the same way.

Mind blown.

It almost certainly is. You can consider it from a logical perspective without even going into brain science.
When a computer looks at a scene it sees in terms of learned rules or operations on high dimensional elements of a vector space intended to cluster pixels.
Scenes have semantics. When we see a scene we see not just things regardless of color or orientation but That thing BY a thing ON TOP of that other thing. We use not just co-occurence patterns of blobs of pixels but context. No doubt context rules can be brute forced by looking at statistics of large amount of images. But it can be more readily learned by a system with a good language model and rules expressed in terms of it.
Language is more important than vision for making sense of the world.
You can also look at it in terms of the brain. That the parts related to sensory integration and processing are also key to language. And that those sections activate as a coprocessor not a single processor.
There is a part of the brain called the Temporal lobe. The Temporal lobe processes audio, speech and semantics. A section of that, intersecting the parietal lobe (linked to visuospatial integration, symbols, math ability), is the Superior temporal sulcus. It is identified as key in theory of mind, processing motion, social cues, faces as well as integrating audio and visual information. What's interesting is that it shows task related activation. It doesn't do all these things alone but activates differently in conjunction with what networks are currently active.
Recently a paper also made a strong argument that the STS is a part of brain that interprets stimuli specifically, regardless of modality (sound, visual) obviously key to recognizing objects.
http://www.jneurosci.org/content/32/47/16629.abstract
http://en.wikipedia.org/wiki/Temporoparietal_junction
http://www.indiana.edu/~panlab/papers/SraJtw_inverse.pdf

RadioLab did an episode about this: http://www.radiolab.org/2010/aug/09/ - A couple of the bits refer to people with communicative issues who didn't have cognition of certain objects or concepts until they learnt words for them.

Reply on news.ycombinator.com