Lecture 4: Signal detection theory Flashcards
What is signal detection theory
Separate sensitivity - capacity to percieve the stimulus, from decision of saying you perceived it or not
Describe signal
True sensory info coming from external world
Describe noise
Various physiological or psychological processes influencing our perception of that external stimulus in an unpredictable manner
Not perfect, noise in system
Name physiological noises
Caused by spontaneous activity in our sensory nerves = spontaneous baseline firing
Name psychological sources of noise
Spontaneous fluctuations in attention that alter perception of sensory stimulus in unpredictable manner, expectations too
What are signal and noise
Statistical concepts
Noise is whatever is not measured and that contributes variability to participants responses = boost or decrease perception of stimuli
Need to know difference betwene noise and signal when finding detection threshold
Describe conservative ppl - response bias
Tend to withold correct identifications
Until completely clear that they have perceived something
Cause them to have higher thresholds = could be erroneously interepretd as having lower sensitivity
Describe liberal ppl - response bias
Quick to say have percieve something
Cause them to Have lower thresholds
Does not mean they are more sensitive
Liberal criterion may sometimes cause them to wrongfully detect presence of sensory stimulus when there is none
What is sdt designed for
Getting rid of response bias
Describe sensitivity
Individuals ability to distinguish between signal and noise
Reflect how well can detect target stimulus
Higher sensitivity indicates better discrimination
Describe criterion
Decision threshold set by individual in sdt - determining whether they report presence of signal
Reflects their response bias
Influenced by factors like expectations or consequences of false alarms and misses (could be conscious or unconscious, can consciously change it though)
Conservative ppl = higher criterion
Internally set criterion for saying yes
Describe sdt ex
Perceive phone vibrating??
Signal present Yes response
Hit
Signal absent Yes response
False alarm
Signal present NO response
Miss
Signal absent NO response
Correct rejection
What do blue and green curves stand for
Strength of felt sensations as a function of vibrations or not
What is criterion on sdt graph
Internal trheshold = how something feels
What is probability on sdt graph
Of getting sensation at the exact level
Most of time will feel this way
Describe proportion of hits and misses
Inversely proportional
Describe whole of blue curve = left to right
Few trails where do feel anything at all
Correct rejections
Say no since below threshold
Then say yes and get false alarms bc signal not there
Describe whole of green curve = left to right
Faint sensation - but say no = miss
Then get hits = yes when stimulus
Few exs wehre will feel v strong clear vibrations
What marks sensitivity
Differnce bewteen blue and green peak
Describe sdt graph directly at criterion and middle of blue and green curve
Range of sensations with major overlap = hard to say if feel stimulus or not
What would happen to criterion if experimenter threatened to kill you for saying stimulus present when no stimulus
Become more careful = consciously change criterion
What would happen to criterion if experimented threatened to kill you for saying stimulus present when no stimulus = CRITERION
Move criterion to right = less inclined to say yes
Wait to feel v strong sensation
Say no all teh time = less hits and false alarms
Describe sdt graph when always yes
100% hits and 100% false alarms
Describe sdt graph when always no
0% hit 0% false alarm
Describe receiver operating characteristics (ROC) curves
Probability of hit over probability of false alarm
As a fucntion of where you place criterion
Perfect discriminability = no false alarms, 100% hits
Performance suboptimal unless at specific points
Chance level = linear line, always yes yes or no, same # hits and false alarms, cannot go under this
What happens to roc curve as sensitivity increases
Goes more out
Becomes like square
What does sdt allow
Estimate participants sensitivity = d’ and criterion = c
As a fucntion of their specific ratio of hits and false alarms
What happens if someone detects vibration of cell phone with 40% hits and 20% false alarms
Coordinate falls on roc curve of d’=0.5
= belongs to specific roc curve + criterion
Reflects bias towards saying no
Point falls to left of roc curve = know it was generated by a stricter criterion, participant more likely to say didn’t percieve stimulus
What is tinnitus & phantom limb pain
Cns lowering threshold for us to perceive something or not
Describe stimulus intensity during sdt trials - variation
Kept constant
But may vary form trial to trial = green distribution
Random or unexplained variation due to noise
Describe stimulus intensity during sdt trials - when no stimulus
May feel like there is one = due to noise (blue distribution)
Why is green distribution to right of blue on sdt graph
Indicates that strength of subjectively felt sensation is higher when there is really a stimulus present
Why is there overlap between teh 2 distributions
Sometimes the presence or absence of a stimulus feels exactly teh same
Low probability that subjective sensation is medium to low
Green low
Misses
High probability that a given trial when a stimulus is present will be associated with medium to high subjective sensation
Green high
Hits
Low probability that subjective sensation high
Green low but high sensation
Still hit
Low probability that subjective sensation low
Low blue
Correct rejections
High probability that a given trial when stimulus absent will be associated with medium to low subjective sensation
Very high blue
Correct rejections
Low probability that subjective sensation is medium to high
Low
False alarms
What does sdt do
Describes sensitivity of test
Which test is most sensitive - sdt graphs
Ones with blue and green peaks hella far away
Which test is most sensitive - ROC curve
One that looks most square
Where to put criterion for maximal accuracy
Put at point closest to top left corner = perfect discrimination
So closest to that = greatest accuracy
Where to put criterion for maximal hits
When closest to highest probability of hit
= more false alarms tho too, bc higher chance
What does sdt graph do
How it works = assum it’s working this way
Is stimulus always felt the same - sdt curve
NOOO
Bc noise