Recognition Memory: Signal Detection Theory and Two-Process Theory, Quizzes of Psychology

Download Recognition Memory: Signal Detection Theory and Two-Process Theory and more Quizzes Psychology in PDF only on Docsity!

Recognition memory

PSY 373, Human Memory

October 22, 2008

Housekeeping

• Hand in experiment report
• Next experiment Remember/Know
• Let’s have a quiz!

Important nomenclature

Response Probe Yes No Old hit miss New false alarm correct rejection

Probabilities

• Hit rate: (number of hits)/(number of old items)
• false alarm rate: (number of false alarms)/(number of new items)
• P (miss) = 1 − P (hit)
• P (correct rejection) = 1 − P (fa)

Measures of recognition performance

• Probability correct: (hits + correct rejections)/(number of probes)
• Measures based on signal detection theory: d′, bias.

SDT citation

Figs taken from

http://white.stanford.edu/ heeger/sdt/sdt.html

A goofy analogy

• Let’s take a group of people who vary in height.
• Half of the people, chosen at random, get little stilts.
• Try to guess whether a person’s wearing stilts or not, just given their height.

SDT, underlying assumptions

• Usually assumed that noise is Gaussian.
• Old item distribution simply shifted.

Criterion, hit rate and false alarm rate

Given disributions and a criterion, you can calculate hit and fa rates

Bias terms

• Where the criterion is placed... called bias.
• A conservative bias means you avoid saying yes.
• A liberal bias means you avoid saying no.

Discriminability

• How far apart the distributions are, in units of the standard deviation, called d′.
• d′^ is a measure of discriminability

Discriminability indpendent of bias

z-transform refresher

Given a normal distribution, how many standard deviations from the mean do you have to go to make the area under the curve x? (It sometimes helps if you remember that .68 of the curve is between μ − σ and μ + σ.)

x z(x) 0 - ∞ .159 - 1 0.5 0 .841 1 1 ∞

d’ and yes/no recognition

• d′^ calculated in this way is probably the most commonly used measure of discriminability in yes/no recognition.
• d′^ should be insensitive to bias.
• (Assumes that old noise and new noise are the same)