Perception Flashcards
signal detection theory
basic perceptual task is to determine if a stimulus is present or absent
Signal detection theory outcomes
hit
miss
false alarm
correct rejection
Hit
target present
respond ‘present’
miss
target present
respond absent
false alarm
target absent
respond ‘present’
correct rejection
target absent
respond ‘absent’
hit rate
= hits/hit + misses
accuracy when the target was there
false alarm rate
= false alarms / false alarms + correct rejections
sensitivity
ability to discriminate signal from noise
- depends on distinctiveness and variability of internal responses (overlap distributions)
bias
tendency to provided one answer over the other
- say present or absent when you are unsure
- depends on where the response criterion is
- depends on how lopsided the evidence must be to say that the target is ‘present’ or ‘absent’
What you need to determine sensitivity and bias
hit rate and correct rejection
- need one number from the top and one from bottom
signal detection theory graph
when target is present will get a large number
response criterion
- a line at some value and for any measurement above threshold say present for any below you will say absent
- if you change your bias have to move your criterion
medium sensitivity
overlap
high sensitivity
very little overlap
measurement from target trial is much higher than a target absent
low sensitivity
huge amount of overlap
not good at telling if target is there or not
extreme case they would be perfectly overlaying each other
neutral bias
right in the middle where the 2 distributions cross over
postiver bias
criterion is pushed up to a higher number
- closer to target present
negative bias
criterion is pushed down to a lower number
- closer to target absent
- lots of hits, very few misses but will also get lots of false alarms and fewer correct rejections
sensitivity equation
d’ = Z (hit rate) - Z (false alarm rate)
- d’ = distance between the 2 distributions in standardized coordinates
- 0 = no discrimination, can’t tell distribution at all
- larger value = better discrimination
bias equation
C = - Z(hit rate) + Z(false alarm rate)/2
- <0 bias from ‘present’
- 0 no bias
- > 0 bais for ‘absent’
downward trend
hit rate and correct rejection rate get worse as noise increases
direct perception theories
- bottom up processing
- perception comes from stimuli in the environment
- parts are identified and put together, and then recognition
constructive perception theories
- top down processing
- people actively construct perceptions using information based on expectations
bottom-up processing
recognition by components theory
- irving biederman
- perceive objects by perceiving elementary features
- geons: 3 dimensiona volumes
- objects are recognized when enough information is available to identify objects goons
geons
- discriminability (can be distinguished from other geons from all viewpoints)
- resistance to visual noise (can be perceived in ‘noisy’ conditions
- invariance (recognizable no matter the illumination direction, surface markings, and texture)
- distinctiveness (36 different geons)
principal of componential recovery
the key to object recognition is not the amount of information but the ability to identify its components (geons)
- found that objects with recoverable goons error rates were lower
multiple personalities of a blob
perceive the blob differently in different environments, not just the shape
bottom up theory
context matters