Task 1 Reaction time and cognitive architecture Flashcards
What is reaction time?
a) operational definition: RT = time interval between stimulus onset and overt response to that stimulus
b) theoretical definition: RT = minimum amount of time needed to a correct response
underlying assumption: pp intends to be as fast as possible without making errors
What are errors?
- it is extremely rare that participants make no errors at all
- if pp tries to be faster → errors occur more often
- pp may act faster → still make correct responses
- pp may be slower→ but perfectly accurate
- pp may be slower → due to moments of inattentiveness
What is the speed-accuracy tradeoff?
- average RT on x axis
- performance accuracy on y axis
a) extreme speed emphasis:
- responses are very fast and accuracy at chance level
b) moderate speed emphasis:
- responses are slower and more accurate responses
c) extreme accuracy emphasis:
- really slow responses but high accuracy level
=> theoretical definition is right between moderate speed emphasis and extrem accuracy emphasis
What is the role of anticipation?
- if pp can predict stimuli, RTs get shorter
- anticipation can be perceptual (spatial, temporal, etc.) or motor-related (learning after multiple trials)
- neural basis of anticipation is not yet fully understood
Examples of anticipation:
- auditory signal preceding stimulus (“warning”) reduces RT
- duration of foreperiod (period between cue and actual stimulus) affects RT: choice RT is shorter for 0.5 s foreperiod than for 2.5 s foreperiod
What are outliers and how do we deal with them?
Outliers:
a) extremely fast responses:
- pp acts prematurely without having fully analyzed the stimulus
- guesses made after failure to reach a decision
b) extremely slow responses:
- inattentiveness
How do reduce outliers:
- give pp feedback for their performance
- keep experiments brief
- posing response deadline
Methods for dealing with outliers:
a) measures instead of avg. reaction time:
- medians: relatively insensitive to outliers -> robust measure of central tendency
- trimmed means: deleting small proportion (e.g., fastest and slowest 5%) and then take mean of remaining ones
- C standard deviation: compute mean and standard deviation for each pp and then delete RTs that deviate more than some number “C” of SDs from mean (e.g., C=2.5)
- Fixed criterion: deleting all RTs exceeding some value and compute mean of rest → leads to bias
- Doing nothing: not remove and do an analysis on outlier-treated data → see if important effects depend on extreme values
What are processing stages?
Processing stages:
- encoding
- comparison
- decision-making
- response
–> essential for understanding how information is processed and manipulated by the mind
What is Donder’s 3’ task?
Task A: Simple RT → one possible stimulus and one possible response
sensory time + motor time = 201ms
Task B: 2-choice RT → two stimuli, each requests a different response
sensory time + discrimination time + response selection time + motor time = 284 ms
Task C: go-no go task → two possible stimuli and one possible response
sensory time + discrimination time + motor time = 237ms
What is the subtraction method?
- Donders’s substraction method:
- two tasks that differ in only a single component of processing
- measure RT in both
- subtract RTs ⇒ outcome is duration of that single component
- simple subtraction forms basis of modern functional brain imaging work
⇒ Donders is seen as father of “mental chronometry”: study of organization and timing of mental processes
Assumptions of Subtraction method
- “seriality”: processing stages are carried out in a strictly serial manner/ Total RT is equal to the sum of durations of individual stages
- “pure insertion”: the duration of all other processing stages remains the same when an extra stage is added or removed
- validity of assumption 2 has been topic of investigation
- ERP study showed that motor processes were unlikely to remain the same ⇒ in many cases, the assumption is unlikely to be valid
Sternberg’s additive factor method
Main propositions:
- information is being processed in a series of successive functional stages and each stage performs transformation on its input, producing output that is passed on to the next stage
- in experimental manipulation to increase RT: output of a stage is constant across conditions and does not depend on its duration
- because of seriality of stages: total RT is simply sum of stage durations
- if two manipulations affect two different stages, they will produce independent effects on total RT ⇒ the effects of two experimental factors do not interact
- if two manipulations mutually modify each others’ effect (interaction), they must affect some stage in common
Language of factorial designs
- factor: independent variable in an experiment (RT is dependent variable)
- levels: set of values of a factor (e.g., size of memory set)
- condition: combination of factor levels
-
effect of factor X: difference between mean RTs at its two levels: e.g., mRT(X2) - mRT(X1)
- simple effect: effect of quality at one level of dose
- main effect: mean of all simple effects of X
- interacting factors: if simple effect of one factor varies with level of another factor
- additive effects: if simple effect of one factor is invariant across levels of other factor ⇒ both factors have additive effects
-
higher-order interactions:
- two-way interaction: difference between simple effects of F at two levels of G ⇒ interaction contrast → shows where interaction lies
- three way interaction/second-order interaction: compute interaction contrast of F and G at each level of H and subtract two interaction contrasts
- lower-order interactions: interaction between two factors can supersede the main effect of one factor
Diffusion models
- theory that explains the distribution of RTs of correct responses or errors in two-choice task
- assumption: information is accumulated continuously during time between stimulus onset and response
- thresholds for either A or B –> counter runs until one threshold is reached
drift rate: rate of information accumulation
decision threshold: amount of information needed to make a decision
- lower threshold = more errors
non-decision time: time required for processes unrelated to decision making (e.g., motor response)
Random walk models
the process can be simulated with a computer program to get simulated RT distributions
represents the dynamic and stochastic nature of the accumulation of evidence over time during decision-making
crucial for capturing the inherent uncertainty and variability in decision processes observed in experimental data.
The model successfully predicts the main features of performance in speeded decision tasks: accuracy, the ordering of mean response times for correct responses and errors, the shapes of response time distributions, and the effect of instructions.
evidence for one response is evidence against the other.
Random walk models are the discrete-time counterpart of diffusion process models and historically preceded them
