W8 - Thresholds Flashcards
What are the 2 types of thresholds?
- Absolute = detecting the presence of a stimulus “can you detect it?”
- Relative = telling two suprathreshold stimuli apart “are they different?”
What are absolute thresholds used to study?
Asks: “Can you see the stimulus” / what CONTRAST level / how bright does a dot of light need to be to see it?
Contrasts = difference in luminance between the stimulus and the background (V1 = spatial componency)
Importance = skills to work out clinical and research applications and reading/critiquing science
What does a psychometric curve (cumulative Gaussian) plot?
- Plot of detectability (performance) vs. stimulus intensity
“how well you can see it vs. how intense/bright it is”
Why is detectability over stimulus brightness a curve shape?
There is a gradual increase in performance of detectability, NOT instantaneous transition
The midpoint of the psychometric curve is where the slope is the steepest
What 2 factors affect the shape of the curve? (VARIATION in noise)
- External noise = variation in the stimulus or the environment that affect the performance
* glare the light source monitor over time
* distracting noise or window light, - Internal Noise = variation in the observer:
* neuronal noise including how much variation is in the activity of the cells involved in doing that task
* individual differences in attention, eg. waning attention
Both types of external and internal noise can be either….
- Random Noise = the direction of the error varies randomly from trial to trial
Eg. light levels might vary randomly - Systematic Noise = the direction of the error is constant
Eg. the light level is always measured too low
What are the effects of random and systematic noise on the SHAPE of the psychometric curve?
In random noise = “the curve will become more SHADOW with more variability in the data, lager ERROR bars”
In systematic noise = “The shape of the curve will stay the SAME, but the curve will be SHIFTED and the measured threshold will be wrong, which is bad if the systematic noise affects one of the conditions and not the other, eg. fatigue
If you have identified random noise, can you still make conclusions of the IV on the DV?
Yes - but bigger variance in the data (random noise) needs a GREATER EXPERIMENTAL EFFECT to be able to make conclusions about the differences between conditions
More error needs to run more trials, but that might increase random noise! eg. participants get tired or bored of the task
Is the presence of systematic noise always bad?
Systematic noise can be fine if you notice it and correct it, but is worse if you don’t detect it and if it only affects ONE condition
If you are getting systematic variation in your results, look at FIRST HALF of trials compared to the SECOND HALF to determine if systematic differences are affecting one part of your trials
Out of the three possibilities, which would be the optimal threshold to use as a measure of threshold (just being able to detect the stimulus?
A = start, 0% detectability
B = middle, 50% detectability
C = end, plateau 100% detectability
B = because you want to get above 50% chance level to avoid floor effects
A = cannot see the stimulus, so it is NOT A VALID MEASURE OF THRESHOLD, Need a measure of absolute threshold that indicates person can actually see it
C = not a valid measure of threshold as you need to show that the reason for different thresholds is due to experimental manipulation, but after effect has plateaued, the increase in stimulus intensity has no real effect on performance and you can’t detect improvements in performance
If you were running an experiment, what point on the psychometric curve would give you the most sensitive measure? (The greatest ability to detect an effect of the experimental manipulation if there is one)
Answer, B = The midpoint gives you the greatest sensitivity as it is the steepest part of the curve (B), which means the smallest change in stimulus intensity will lead to the greatest change in performance
What are the desired properties of the threshold?
RELIABILITY & VALIDITY
Unreliable measures makes it difficult to determine whether any differences between conditions is due to experimental manipulation or unreliable variables producing inconsistent results
What performance level would be ideal for the CONTROL CONDITION?
B (at 50% chance level) = as A and C may be due to floor/ceiling effects, while B is at chance level of performance
What is the drawback of just asking someone if they can see the stimulus?
- Subjective criterion = levels in a participants confidence may influence whether they believe they can see the stimulus or not
There is naturally individual variation in subjective criterion, some may say they see it only when they can clearly see it, while others might be more confident in seeing something more faint
- For the same person, detectability can vary from trial to trial, leading to high variability and lower validity, eg. someone might be less confident to see the dot, but gain confidence or fatigue after many trials, affecting the rate of their responses
What’s the negative implication of the subjective criterion on the psychometric threshold?
The subjective criterion raises issues in validity of the threshold - since differences in thresholds between groups may be due to differences in INDIVIDUAL criterion levels over ACTUAL DIFFERENCES in their ability to see the stimulus
After comparing the ability of 2 groups (high anxious vs. control group) to perceive motion stimuli, its found the anxious group have higher thresholds to perceive the stimulus.
Does this mean that anxious people have worse motion ability than the control group?
NO, it may be due to RESPONSE CRITERIA - individual differences in confidence/temperament/decision making, anxious people may be less confident to make judgements on their ability to detect a stimulus compared to non-anxious people
How do we improve the issue of the subjective criterion?
Use FORCED CHOICE design, whereby you measure what the participant’s saw in the stimulus instead of whether they saw the stimulus
- It’s no longer about presence/absence but a feature of a present stimulus in a fixed number of choices. Now asking: “what was the nature of the stimulus?”
