[accessibleimage] Re: Molyneaux's question rephrased

Hi,

I agree with Peter's view in respect to object identification. Theories of generic memory, such as Spreading Activation, put forward the idea that generic memory is made up of concepts, attributes of those concepts, and links between those concepts. Objects are concepts and so identifying them becomes a case of matching the attributes of the object we want to identify with the attributes we know about a concept. Attributes can be gathered from a range of sources, including the sensory modalities.

So, the question becomes how much overlap is there between the attributes we can associate with a concept visually, tactually, and kinesthetically. The greater the overlap then the greater the chance that objects whose identity were learnt tactually or kinesthetically can be identified. To add to the attributes derived from direct sensory stimulation by an object people also learn attributes through knowledge transfer, e.g. the sun is in the sky and the sky is above you. These attributes that have been learnt through knowledge transfer can often help in providing the spatial relationship attributes and other attributes that may only be available visually. Therefore, blindness doesn't necessarily exclude someone from learning these attributes by indirect means. I suspect that in some situations it is in fact possible to identify something just through the attributes that have been indirectly learnt.

To illustrate attribute matching and how attributes acquired tactually or kinesthetically can be applied visually I'm going to borrow Peter's example of a square. Thinking about the attributes of a square then there are certain attributes that, in combination, make a square unique. These attributes are the fact it has four corners, that each corner is a right angle joint, and that all sides are of equal length. These attributes can be learnt tactually and kinesthetically through a variety of means. A corner, which is a concept in itself, can be identified through the attribute of a change in the direction of an edge of something. This can be determined kinesthetically by running someone's finger, or maybe a probe of some sorts, around the edge of a square and checking for the existance of corners. Visually the detection of corners is likely a spatial comparison, given the parallelism of vision, between two edges of the shape. Counting the number of corners then becomes the practice of following the edge of a shape around its perimeter and counting how many times there is a change in direction of the edge, something that can be accomplished both kinesthetically and visually. Judging the equality of length for the sides is probably a simple spatial relationship visually whilst it can be accomplished tactually by feeling two sides at the same time, determining whether those two sides are equal, and repeating the process for the other two sides.

I've seen attribute matching being used in haptics related experiments. The subjects are typically sighted undergrads, well, most research is done at universities *smile*, and so their primary means of object identification is visual. However, when they are presented with shapes, even shapes that they haven't touched before, they can identify simple shapes with fairly good accuracy. Identification of more complex shapes tends to have lower accuracy. This lower accuracy is unlikely due to differences in the primary modality used for object identification but due to the serial nature of haptic devices and, to a lesser extent, touch in general. Complex objects require more attributes on which to make an identification. When exploring something serially each time an attribute is encountered it has to be remembered and this leads to a situation in which people fail to remember all of the attributes comprising a complex object and it is this memory problem that is likely the cause for low accuracy in identifying complex objects. Indeed, when experimental participants are asked to draw what they remember from exploring an object kinesthetically they often recall a number of attributes correctly but fail to recall other attributes, even though they explored them and likely identified them. This lack of attributes in memory results in an inability to successfully match the known attributes of a physical object with those that are known about the object from generic memory or it can result in a situation where the known attributes result in multiple matches and disambiguation isn't possible.

Will
----- Original Message ----- From: "Peter Meijer" <blindfold@xxxxxxxxxx>
To: <accessibleimage@xxxxxxxxxxxxx>
Sent: Sunday, September 03, 2006 4:41 PM
Subject: [accessibleimage] Re: Molyneaux's question rephrased


Hi Lisa,

At the URL

   http://www.seeingwithsound.com/extra/molyneux.ppt

I have a Powerpoint file illustrating how The vOICe lets
you *hear* the difference between a cube (square) and a
sphere (filled circle), even if you have never seen.

I had rephrased Molyneaux's question as "Would someone
born blind, using 'seeing-with-sound', be able to tell
a cube from a sphere by 'sight' alone? The answer being
yes.

Best wishes,

Peter Meijer


Seeing with Sound - The vOICe
http://www.seeingwithsound.com





Other related posts: