Letters are crucial features that the visual system

Letters are crucial features that the visual system attempts to detect during the
process of word recognition – became too hard to identify so many letters in just
200msec if the contrast energy was too low
o Letters with low contrast energy are weak signals that the visual system will
‘squelch’
 Squelching = tendency of the nervous system to inhibit the processing of
unclear features
 Prevents us from just guessing, otherwise we would see many false
features
Recognition by Components
- What are the fundamental elements / size of these features we put together, i.e. for
complex 3D objects such as people/cars
- Biederman : Recognition by components (RBC) theory = theory that we recognize
objects by breaking them down into their fundamental geometric shapes
- Geons = the set of 36 basic 3D shapes from which all real-world objects can be
constructed
o Objects broken into geons, which are then compared with existing geon
configurations stored in memory – match input and memory to recognize the
object
- Tf if objects are parsed into geons, recognition should be a function of the number of
geons available to perceive
- Experiment – varied number of geons used to depict a given object
o More complex objects (penguin) needed more geons than less complex
(flashlight)
o Objects were flashed for 100msec then covered to interrupt perceptual processing
- FINDINGS:
o People were good at recognizing objects represented by only 2-3 geons, and
recognition improved when more geons were added to add detail
o Tf accuracy increased for more detailed penguins than less
o Also – more complex objects (penguin) were recognized more efficiently than
less complex (flashlight)
o ***would have thought more complex require more processing time
- Conclusion: complexity and detail lead to more geons and more geons lead to better
recognition
- Deconstructing objects into geons is a critical component of object recognition

Context and Knowledge
- Context in which features appear is important
- Context effects = influence of proximate stimuli and the situation on the perceptual
experience of the stimulus
o I.e. you see a hand Tf likely on wrist is a bracelet / watch not an elephant
- Interesting example of context: Moon illusion = tendency for the moon to appear
different in size depending on where it is in the sky
Letters in Context
- Profound example of the effect of context on letter/word perception:
- Jumbled word effect = ability to read words in sentences despite having mixed-up
letters in the middle of some of the words.
o Expectations regarding what the words will be help you determine what the words
actually are
- Word superiority effect = its easier to identify a letter if it appears in the word than if it
appears alone (e.g. P versus WARP)
o Can more efficiently map strings of tentatively identified letters to real words
- Context effects can be modelled using connectionist approaches
- Parallel distributed processing (PDP) = model of perception according to which
different features are processed at the same time by different ‘units’ (simple processing