Lecture 10 - Perceptual Enhancement & Feature Binding Flashcards
“My experience is what I agree to attend to,” sounds like which kind of attention?
Endogenous
internal!
what you’re trying to focus your attention on!
consistent with some meaningful goal
Perceptual enhancement
– Increasing the processing of some specific channel information.
When attending to a specific channel (e.g. location - spatial attention), we often try to increase the amount of information we can extract.
- increase processing power
“Enhancement” is often measured by
the speed at which you can make an identification (i.e. reaction time), the accuracy of the identification, or using signal detection measures (d’).
Each channel is thought to have
signal + noise (the stuff you don’t want).
right out of information processing
Noise is both ____ an _____
external (e.g. fog, distance, occlusion, etc.)
internal (neural variability and spontaneous activity).
- spontaneous activity of neurons is background noise that could be interfering with detection
When attending to a specific channel (e.g.
location), do you ‘turn up’ the entire channel
(so you turn up the signal AND the noise) or reduce the internal noise?
Lu & Dosher (1998) - attempt to answer this question
• had subjects attend to visual stimuli under lower or higher external noise conditions. The task was to judge orientation of the filters.
• Subjects were better at making judgments at
attended locations, but only under low noise
conditions.
• This suggested that attending enhanced all
elements in the channel (the signal & the noise), rather than reducing internal noise. [Like
turning up the volume on a radio station with
static.]
• This is a bit oversimplified…
Based on the result from Lu & Dosher (1998), does perceptual enhancement seem to operate early or late in the perceptual process?
most consistent with early
you’re doing the whole thing (SIGNAL AND NOISE): turing that one volume knob: you’re getting it all in
if it was later we would think that we could turn down the internal noise or use top-down information
bottleneck models limits
suggest that ‘selective filters’ prevent
too much information from being processed at any one time. Input is limited (constrained) and sometimes lost (e.g. decays in sensory memory).
these filters block out some of the info from coming in so you only process a few things at a time (so you don’t get overloaded)
stuff you don’t need decays (goes away)
early selection theory limits
the limits are on attention, when processing low level properties.
analyzing basic physical properties (how loud it is)
late selection theory limits
the limits are in SHORT-TERM MEMORY (or later), after stimuli have been recognized.
when you’re trying to dredge something out of memory that matches what you’re perceiving
Broadbent gave us an ________, in which only attended messages made it past analysis of physical properties.
early selection filter model
Because some information (names, “fire”, and other salient information) can get through, some (Treisman) have suggested a….
filter model with priorities.
you’re filtering out things but you have selective filters that let unattended things through (e.g. your name)
these things get passed right through and then you start analyzing meaning
tricky: priorities can change so it’s hard to test universally
Early example
analysis of basic physical properties –> low level perceptual processing
talking: frequencies, locations (bottleneck here) - (or here)-> phonemes
task: people hearing two stories and shadowing them
Using a filter model with priorities, subjects can shadow meaningful messages that switch from ear to ear (across channels).
if he tracks the meaning he won’t mix up the two different streams
priority allows meaning to be accesses directly = attention puts it together
Late selection models suggest that the limitations come from
short-term memory. Only so many ‘meanings’ can be activated and processed at once.
you can only follow so many words at a time
