+ All Categories
Home > Documents > Temporal learning in animals: Empirical tests of two mathematical models of timing * Armando Machado...

Temporal learning in animals: Empirical tests of two mathematical models of timing * Armando Machado...

Date post: 14-Dec-2015
Category:
Upload: adrianna-clowney
View: 213 times
Download: 0 times
Share this document with a friend
42
Temporal learning in animals: Empirical tests of two mathematical models of timing * Armando Machado University of Minho, Portugal Université Charles-de-Gaulle, Lille 3
Transcript

Temporal learning in animals:

Empirical tests of two mathematical models of timing

*Armando Machado

University of Minho, Portugal

Université Charles-de-Gaulle, Lille 3

The Skinner boxThe microscope of learning psychologist

• Two models of timing

• Scalar Expectancy Theory (SET)

• Learning to Time (LeT)

• Critical tests: SET vs LeT

• Temporal Bisection

• Double Bisection

• Successes and Failures: what have we learned?

• The role of context

• Errorless learning and mediating behavior

The pigeon’s sense of time

T

Extinction Reinforc. Time in Interval (s)0 100 200 300

0

10

20

30

40

50

0 10 20 300

20

40

60

80

0 25 50 75 1000

20

40

60

80

Res

pons

es p

er M

inut

eKilleen et al. (1978)

FI-30 s

FI-100 s

FI-300 s

The scalar property

30 s

3,000 s

0.0 0.2 0.4 0.6 0.8

Proportion in Interval

1.0

0.0

0.2

0.4

0.6

0.8

1.0

Res

pons

e R

ate

/ Ter

min

al R

ate

300 s

Dews (1970)

Retrospective TimingThe Bisection Experiment

Red

Retrospective TimingThe Bisection Experiment

Green

Retrospective TimingThe Bisection Experiment

TRAINING

Red

Green

TESTING

4-16

2-8

1-4

Church & Deluty (1977)

3-12

0 1 2 3 4 5

Relative Stimulus Duration

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)Fetterman &

Killeen (1991)

2 4 6 8 10 12 14

Stimulus Duration (s)

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)

Catania (1970)

0 1 2 3 4 5

Stimulus Duration (s)

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)

Pacemaker

Accumulator

Memory S

Choose

Comparator

Memory LChoose

Scalar Expectancy Theory (SET)

n*

T

T0

n*

T

ST0

n*

T

S LT0

SET and temporal bisection

n*

T

S Lt

XS

XL

Xt

SET and temporal bisection

Similarity , SS t

t

XX X

X

SET and temporal bisection

Similarity , SS t

t

XX X

X

( | ) S t

t L

X XP t P

XR

X

Similarity , tt L

L

XX X

X

Stimulus duration (s)1 2 3 4

Pro

babi

lity

("S

hort

")

0.0

0.5

1.0

Learning to Time (LeT)

BehavioralStatesX(t,n)

Response StrengthsR(t), G(t)

Signal

AssociativeConnections

Wn

Behavioral States

0 1 2

Deterministic

interpretation

Stochastic

interpretation

Behavioral States

0 1 2

( ,0) ( ,0)

( , ) ( , 1) ( , )

1 if 0(0, )

0 if 0

dX t X t

dt

dX t n X t n X t n

dt

nX n

n

Machado (1997)

n=3n=2

n=1

...

0 5 10 15 20 25

Time (s)30

0.0

0.1

0.2

0.3

0.4

Act

ivat

ion

X(t

,n)

exp( )( )( , )

!

nt tX n t

n

( , ) 1 , if reinforcement

( , ) , if extinction

n n

n n

WR X S n WR

WR X L n WR

Associative Connections

nWR

( , ) 1 , if reinforcement

( , ) , if extinction

n n

n n

WG X L n WG

WG X S n WG

Associative Connections

nWG

( ) ( , ) nn

G t X t n WG

Response Strength

( ) ( , ) nn

R t X t n WR

Learning to Time (LeT)

1

( | )1 exp ( ) ( )

P tR t G t

R

4-16

2-8

1-4

Church & Deluty (1977)

3-12

0 1 2 3 4 5

Relative Stimulus Duration

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)Fetterman &

Killeen (1991)

2 4 6 8 10 12 14

Stimulus Duration (s)

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)

Catania (1970)

0 1 2 3 4 5

Stimulus Duration (s)

0.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("Sh

ort"

)

LeT and temporal bisection

For the bisection task, SET and LeT make similar predictions.

We need a new procedure to contrast the two models

The Double Bisection task

Double BisectionExperiment 1

TRAINING

Green

Red

Trial Type 1

Violet

Blue

Trial Type 2

TESTING

Green

Signal Durations Choice keys

Blue

vs

TRAINING

Green

Red

Trial Type 1

Violet

Blue

Trial Type 2

TESTING

Green

Signal Durations Choice keys

Blue

vs

Pacemaker

Accumulator

m

Response 1

Comparator

M

Response 2

M M

SET

LeTBehavioral

States

Responses

Associativeconnections

1 s Red

4 s Green

4 s Blue

16 s Violet

PredictionsA context effect?

1 2 84 160.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

("G

reen

")

LeT

SET

Duration of test stimulus (s)

Machado & Keen (1999)

1 s Red

4 s Green

4 s Blue

16 s Violet

DataYes, a context effect!

Machado & Keen (1999)

1 2 84 160.0

0.2

0.4

0.6

0.8

1.0P

roba

bilit

y ("

Gre

en")

Duration of test stimulus (s)

1 s Red

4 s Green

4 s Blue

16 s Violet

Double BisectionExperiment 2

Group 8

1 s Red

4 s Green

4 s Blue

8 s Violet

1 s Red

4 s Green

4 s Blue

16 s Violet

Group 16

0 4 8 12

Stimulus duration (s)

160.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

(G

reen

)

Group 16

Group 8

Double BisectionExperiment 2

SET LeT

Group 8, Group 16

0 4 8 12

Stimulus duration (s)

160.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

(G

reen

)

Group 8: 1-4 and 4-8

Group 16: 1-4 and 4-16

SET vs LeT vs Data

Group 8 Group 16

Duration of test stimulus (s)

0 4 8 12 16

Pro

babi

lity

( G

reen

)

0.0

0.2

0.4

0.6

0.8

1.0

Duration of test stimulus (s)

0 4 8 12 16P

roba

bilit

y ( G

reen

)

0.0

0.2

0.4

0.6

0.8

1.0

LeT: Predictions vs Data

0 4 8 12

Stimulus duration (s)

160.0

0.2

0.4

0.6

0.8

1.0

Pro

babi

lity

(G

reen

)

4 vs 164 vs 8

Stimulus duration (s)

0 4 8 12 16

Pro

babi

lity

( G

reen

)

0.0

0.2

0.4

0.6

0.8

1.0

Double BisectionExperiment 3

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Green

16 s Blue

1 s Blue

16 s Green

Group “Consistent”

Group “Inconsistent”

How fast will the new discrimination be learned?

LeT: Predictions vs dataGroup “Consistent”

Stimulus duration (s)0 4 8 12 16

Pro

babi

lity

Gre

en

0.0

0.2

0.4

0.6

0.8

1.0

Stimulus duration (s)0 4 8 12 16

Pro

babi

lity

Gre

en

0.0

0.2

0.4

0.6

0.8

1.0

Day 1Day 2Day 10

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Blue

16 s Green

LeT: Predictions vs dataGroup “Inconsistent”

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Green

16 s Blue

Day 2

Stimulus duration (s)0 4 8 12 16

Pro

babi

lity

Gre

en

0.0

0.2

0.4

0.6

0.8

1.0

Stimulus duration (s)0 4 8 12 16

Pro

babi

lity

Gre

en0.0

0.2

0.4

0.6

0.8

1.0

Double BisectionExperiment 3

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Green

16 s Blue

1 s Blue

16 s Green

1 s Red 4 s Green

4 s Blue 16 s Violet

Group “Consistent”

Group “Inconsistent”

Which discrimination(s) will be disrupted?

Sessions0 2 4 6 8 10

Pro

babi

lity

cor

rect

0.0

0.2

0.4

0.6

0.8

1.0

Sessions0 2 4 6 8 10

Pro

babi

lity

cor

rect

0.0

0.2

0.4

0.6

0.8

1.0

LeT: Predictions vs dataGroup “Consistent”

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Blue

16 s Green

LeT: Predictions vs dataGroup “Inconsistent”

Sessions0 2 4 6 8 10

Pro

babi

lity

cor

rect

0.0

0.2

0.4

0.6

0.8

1.0

Sessions0 2 4 6 8 10

Pro

babi

lity

cor

rect

0.0

0.2

0.4

0.6

0.8

1.0

1 s Red 4 s Green

4 s Blue 16 s Violet

1 s Green

16 s Blue

Conclusions• SET and its memories

– Failure of independence from alternatives

• LeT and the role of context– Which key to peck after a signal– Which key not to peck after a signal

Current work• LeT and the role of context

– Errorless (temporal) learning?

• Behavioral patterns– Structuring and restructuring of patterns

during the time signal.– Some videos of mediational(?), collateral(?)

behavior.

Behavioral patterns

• P689– 1s: approaching keys– 4s R-G and 4s B-Y: pecking– 16 s: approaching keys, pecking, pausing, pecking

• P65– 16sf: pecking and wing flapping;– 16s: pecking and wing flapping ERROR;– 38: pecking and wing flapping;

• P68– 34 & 54: wall pecking and head-in-the-feeder

Collaborators

• Paulo Rodrigues• Marco Vasconcelos• Paulo Pata• Joana Arantes• Antonio Fidalgo• Ana Paula Leite• Paula Magalhães

• Richard Keen• Ozlem Cevik• Paulo Guilhardi

The full picture

SET LeT

Machado & Keen (1999)

Duration of test stimulus (s) Duration of test stimulus (s)

1 2 8 164

0.0

0.2

0.4

0.6

0.8

1.0

Prob

abili

ty (R

or G

)

1 2 8 164

0.0

0.2

0.4

0.6

0.8

1.0

1 s Red

4 s Green

4 s Blue

16 s Violet

Machado & Keen (1999)

0.0

0.2

0.4

0.6

0.8

1.0

Prob

abili

ty (G

)

1 2 4 8 16

0.0

0.2

0.4

0.6

0.8

1.0

Prob

abili

ty (R

)

1 2 4 8 16

Duration of test stimulus (s)

0.0

0.2

0.4

0.6

0.8

1.0

Prob

abili

ty (G

)

1 2 4 8 16

Duration of test stimulus (s)

1 2 4 8 160.0

0.2

0.4

0.6

0.8

1.0

Prob

abili

ty (R

)


Recommended