developing stimulus control. peak shift phenomena where the peak of the generalization curve shifts...

Post on 19-Dec-2015

217 Views

Category:

Documents

3 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Developing Stimulus Control

Peak Shift

• Phenomena where the peak of the generalization curve shifts AWAY from the S-– Means that the most responding does not occur

for the S+– But slightly offset from the S+ and away from the

S-– Question is WHY!

Peak Shift Effect – Hanson (1959)

Two theories

• Spence: Peak shift occurs because of the summation of the excitatory and inhibitory curves– Result is that the most responding is slightly in

favor of the S+– Not as strong of suppression for S-– Thus, get shift away from S-– Just a math phenomenon

Spence’s Theory to Account for Peak Shift

-60

-40

-20

0

20

40

60

80

490 510 530 550 570 590 610 630 650 670Wavelength (nanometers)

Inh

ibit

ory

or E

xcit

ator

y S

tren

gth Inhibitory

ExcitatoryDifference

S+

S-

Two Theories

• Kohler’s Transposition or Interdimensional theory– The animal learns a conceptual rule– “Not” the S- but is the S+– Thus, when given generalization stimuli, take

whatever is the most “not” S-– Shift away from S+ occurs

• Learning a concept: always choose larger or greener, etc.

Interdimensional discrimination

0

100

200

300

400

500

600

700

501 530 555 576 606

Wavelength

# R

espo

nses

PseudodiscrimDiscrimination

Discrimination: S+ = 555nm Light; S- = Tone

Test of Theories• Choose between squares:

• S+ S- Situation:100 cm2 60 cm2 original training 250 cm2 150 cm2 Test 1: Spence’s test500 cm2 250 cm2 Test 2” Kohler’s test 1000 cm2 500 cm2 ??????

• Spence's prediction: will choose closest to original S+• Transposition or interdimensional: choose larger

Which is correct?

• tests for transposition yield evidencesupporting transposition

• tests for generalization yield evidencesupporting Spence

• probably: both effects working part of the time

What is a stimulus

• Any sensory event that elicits a response– Not necessarily the response you want– Need to know that that rat perceives the stimulus

• Establishing a cue– Classically condition (cue to cue)– Operant conditioning: response to reinforcer,

then add predictive cue for contingency

Shaping the cue• Need to shape the response to the cue• Introduce the stimulus cue right when the behavior

occurs– Not as it is starting, not after– E.g., for a dog to learn “sit”, not when it is walking, but

when it starts to sit.– BUT: click once the behavior is underway, not when the

behavior is finished– Then, only introduce the cue when the dog is NOT doing

the behavior• Give the cue• Get the behavior• click

Four rules for stimulus control• Behavior occurs immediately upon presentation

of the cue stimulus

• Behavior never occurs in absence of stimulus

• Behavior never occurs in response to another stimulus

• No other behavior occurs in response to the stimulus

Transferring stimuli

• Once establish one cue, can add a second cue for the same behavior– Again, use only for THAT behavior– Keep stimuli separate. – E.g., does down mean stop jumping on me or lie

down on the ground?

Prompting• Prompts used to increase probability that

organism will engage in correct response

• Use during discrimination training– SD or S+: in presence of this stimulus, do the behavior– S or S-: in the presence of this stimulus, do NOT do

the behavior

• Function of prompts is to produce instance of correct behavior so it can be reinforced

Types of Response prompts• Response prompts: behavior of another evokes desired

response in presence of SD

• Verbal prompts: voice command

• Gestural prompts: physical movement or gesture of trainer

• Modeling prompts: observe model and imitate (not really used with rats)

• Physical prompts: physically assisting rat to engage in behavior

Types of Stimulus prompts• Stimulus prompts: change in stimulus or addition/removal of

stimulus increases likelihood of desired response in presence of SD

• Within-Stimulus prompts: change the salience of a prompt– Bigger or smaller– Louder or softer– More or less

• Extra-stimulus prompts:– Adding a stimulus to increase likelihood of correct discrimination– Picture prompts– Putting line of food pellets for rat to follow

Aversive cues

• Use aversive to STOP behavior• E.g., “no”, a sharp noise, a table slap

correction• Good for dangerous behaviors

• Potential downside:– Squelch behaviors– Organism is startled, now won’t move– More likely to bite/nip/freeze

Limited holds and anticipation

• Limited hold:– Reinforcer only available in a narrow window

• Either take it now, or lose it• Often used to avoid stashing or lazy choice

– If do behavior, can get reinforcer within x seconds or less, otherwise lose it.

• Anticipatory behaviors occur before the cue– If occur, cancels the trial– Withdraw target stick and even put rat in “time out”

Behavior chains• One behavior can be a cue for the next response• This is a behavior chain

– E.g., run through the tunnel, over the balance beam and then through the weave poles

• Can teach forward chain:– Start with first, then add second, third, etc.– Reinforce after successful behavior for criterion

• Or backwards chain:– Start with last response, then next to last, and so on– Reinforce after successful behavior for criterion

Generalization

• May want organism to engage in behavior in many settings/situations/cues

• E.g., go over balance beam on desk or in either of the two big boxes

• Must train so that the cues to which the rat attends are the ones relevant to the task, not the setting

Fading prompts

• Important to fade prompts– Want behavior to occur to cue only– Shows mastery of behavior: • Behavior flows well• Behavior occurs at high rate (or low rate if that is goal)• Occurs evenly and consistently

• Several ways to fade prompts:– Transferring stimulus control• Remove response

Fading prompts

• Several ways to fade prompts:– Transferring stimulus control

• Remove one stimulus cue, then another– Fading across different types of prompts or fading

across prompts• Least to most: fade least invasive prompt first• Most-to-least fading: fade most invasive prompt first

– Prompt delay• Delay the prompt• Increase delay until behavior occurs before prompt would

have been given

example: teach S+ = red; S- = green

• Teaching red vs. green discrimination– S- presented for only 5 or so seconds– S+ presented for 3 minutes– S- begins as unlit (dark) key: pigoens much less likely to peck it

• S- gradually fades to green key• human example: teaching MR children sight words

– word + picture on slide– gradually fade the picture, leaving only the word

RAT

RAT

RAT

RAT

RAT

RAT

RAT

Several important characteristics about procedure:

• is very effective in reducing number of responses to the S-

• improves long term discrimination

• Terrace claims not get aggressive behaviors w/fading procedure

• S- not develop inhibitory properties- thus eliminates peak shift effect

• not get behavioral contrast

top related