How stimuli are recognized and identified
People are excellent at pattern recognition. People are the best at this. Example: Recognizing an old friend on the street.
Computers in contrast are very bad at pattern recognition. They have a hard time dealing with shapes changing as you get closer etc.
Pattern Recognition Models:
appealing because it is simple. The claim: we compare things we see and hear against some sort of internal template. In this context, template is a specific pattern stored in memory. The assumption: the match must be precise, or else recognition won’t occur. Often thought of like a lock and a key.
It really does work. There are a lot of systems that use this model. Example: Automatic check sorting using the numbers at the bottom. Example: Universal scanner codes also employ this technique (bar codes).
Totally inadequate for explaining pattern recognition in humans. One argument against this model is (a) the cognitive system does not have enough memory to store this huge number of templates. You’d have to store in memory every pattern you’ve ever seen. Vast amount of memory would be needed. (But who knows what the capacity of human memory is?) In the end, the argument against the model based on amount of memory is not too strong.
(b) It’s too slow. People can recognize familiar patterns almost instantly. Searching through a huge number of templates would take a lot of time. But the argument of it being too slow assumes a serial search of memory (making comparisons one at a time). Perhaps a parallel search of memory would be fast enough (making many comparisons at once).
( c ) It’s too inflexible. This argument against this model is what kills the model. People can recognize patterns they’ve never actually seen before. Recognizing an old friend on the street who has changed dramatically. So it’s a different pattern and doesn’t match your image of him from earlier in childhood. People are really flexible compared to computer recognition systems. Apple Newton - had problems recognizing handwriting. It took time to learn and this took time and was a pain.
Example: CAPTCHA used on webpages. The goal is to determine if the page is being requested by a human being.
(2) Prototype Model
We store prototypes in memory. Prototypes are abstract idealized representations. When we see or hear a stimulus, we compare it to a prototype.
not as rigid as templates. Minor variations are okay
Assumes that the cognitive system creates prototypes. Is there any evidence for this?
Pseudo-memory = false memory (fooled by a prototype into believing something false) Even though you haven’t seen picture with 5 different elements in it, having seen pics of all five elements, prototype has put them together.
(3) Distinctive features model
We make distinctions between patterns based on a small number of characteristics. Distinctive features are characteristics that differentiate one stimulus from another. The first letter of the alphabet could be rendered as various colors and shapes. They are all the same thing but are different. What do the four things have in common? What do the four share that differentiates them from other letters in alphabet? Two lines that converge at the top and a crossbar. These could serve as distinctive characteristics. They allow a to be differentiated from all other capital letters. How many distinctive features would be needed to differentiate all of the capital letters? Relatively few
Feature lists are simpler than templates or prototypes. It can deal with variation. Assumption is that cognitive system creates feature lists.
A full-blown model of pattern recognition
Pandemonium (Selfridge, 1959) – It is an example of an early information processing model.
Uses “demons” to process incoming stimuli.
Image demons – loosely hold on to representation to image in outside world (icon)
Feature demons - Recognize specific patterns and begin shouting if they are used within the image.
Cognitive demons – look up and see if any of its feature demons (the features used to make it up) are being used, if they are it begins shouting sort of
Decision Demon – decide who is shouting the loudest
Empirical support :
Kinney, Marsetta, and Showman (1966) – Gave subjects presentations of individual letters in a t scope, and asked them to identify them.
Errors were plotted in a visual confusion matrix. It classifies the errors that were made. People confuse 8 and B or C and G. This supports Pandemonium.
Hubel & Wiesel (1959) - Found that cortical cells act like feature detectors. Led to nobel prize in 1981. Inserted electrodes into the visual cortex of cats. Electrodes are very thin wires that can detect and record electrical activity. Some cells fired strongly in response to lines of a particular orientation (like /) ….But fired weakly for lines in other orientations (like or | or –)…
Is this ability innate or learned in cats? Were those cells hard wired for that process?
They are not hard wired for feature detection. Cortical cells aren’t hard wired for feature detection. They become specialized via exposure to the environment. They raise kittens in environments with only vertical stripes. The cats’ cells do not fire when shown horizontal lines later in life.
Pattern Recognition Deficits
We can study the effects of brain injury on the cognitive system. Brain injuries such as strokes and accidents. People show very selective damage to a particular cognitive function. Typical pattern: damage quite selective. For example, agnosia is the loss of the ability to recognize objects (such as visual agnosia). An example: a book by Sacks (1987) The Man who Mistook his Wife for a Hat. It was a collection of unusual neurological conditions: includes the case of a music professor who had agnosia.