Reaction time How I learned to stop worrying and love the ... · Reaction time Roozbeh’Kiani ......

Post on 22-Jul-2020

3 views 0 download

Transcript of Reaction time How I learned to stop worrying and love the ... · Reaction time Roozbeh’Kiani ......

Reaction time

Roozbeh  Kiani    

Center  for  Neural  Science  and    Department  of  Psychology,  NYU  

How I learned to stop worrying and love the reaction time

What  is  a  decision?  

•  A commitment to a proposition or plan of action based on evidence, prior knowledge, payoff, and urgency.

•  Decisions often involve interpretation and contingency.

•  The study of decision-making offers insight into how higher brain areas read out or interpret information acquired through the senses.

Roitman & Shadlen, 2002 Kiani et al, 2011

Motion direction discrimination task

Reward for correct choice

Roitman & Shadlen, 2002 Kiani et al, 2011

Motion direction discrimination task

Psychometric  funcBon:  Accuracy  

Chronometric  funcBon:  Speed  

Rea

ctio

n tim

e [m

s]

Hubel, 1988

Sensory  “Evidence”  

From: Hubel & Wiesel

From: Celebrini & Newsome

Responses of MT neurons

Britten et al., 1993

Rightward Leftward

Trivial

Easy

Difficult

100 ms

50 s

pike

s/s

– 0 +

Momentary evidence (Δr)

µ= kC

6.4%

99.9%

25.6%

Coherence

Adapted from Caret Van Essen, 2002 URL: http://www.nitrc.org/projects/caret/

Lateral intra-parietal cortex (LIP)

V1 V4

7a 7b

V2d

V2v

Posterior Anterior

MT/MST

Frontal eye field

Roitman & Shadlen, 2002

saccade

motion onset

select Tin

select Topp

Roitman & Shadlen, 2002

High motion strength!

Low motion strength!

Time"Stimulus

on!Stimulus

off!

Spikes/s"

Time"Stimulus

on!Stimulus

off!

Low motion strength!

accumulaBon-­‐to-­‐bound  model  

MT:  Sensory  Evidence    

MoBon  energy  

“step”  

LIP:  Decision  FormaBon    

AccumulaBon  of  evidence    “ramp”  

 

Bound

Choose  right  

Accumulated  evidence  for  rightward  target  

–                      +  

Momentary    evidence  (Δr)  

Accumulated  evidence  for  leSward  target  

Choose  leS  

Time  

μ=  kC  

accumulaBon-­‐to-­‐bound  model  

Momentary evidence e.g.,

∆Spike rate: MTRight– MTLeft

Decision variable (v)

Link 1992 Ratcliff & Smith, 2004 Palmer et al, 2005 Laming,  1968  Luce,  1986  

Choose left

Time  

–                      +  

Momentary    evidence  (Δr)  

μ=  kC  

Choose right

3 degrees of freedom: B, k, t0

accumulaBon-­‐to-­‐bound  model  

Best  fi]ng  chronometric  funcBon    

t(C) = BkCtanh(BkC)+ tnd

Predicted  psychometric  funcBon  

P =1

1+ e−2k C B

tnd = 200, 300, 400msB= 0.55, 0.71, 0.83 1/k = 0.03, 0.06, 0.12ms

Gaining  intuiBon  about  the  model  

Palmer et al, 2005

Gaining  intuiBon  about  the  model  

t(C) = Bk C

2Pcor C( )−1⎡⎣ ⎤⎦ + tnd Luce 1986, Link 1992

Coupling  response  Bme  and  accuracy  

(halfway  Bme  threshold)  /  (halfway  accuracy  threshold)  ~  3.5  

Why  should  you  care  about  decision  Bme?    

Why  it  isn’t  enough  to  say  “I  always  have  fixed  sBmulus  duraBons  in  my  experiments”?      Because  the  same  decision-­‐making  process  applies,  curtailing  the  effect  of  the  sBmulus.    

Viewing duration varies from 80ms to 1500ms

Tin Tout

Kiani et al, 2008

Correct response probability improves with stimulus duration

Probability correct

Stimulus duration (ms)

coherence

Kiani et al, 2008

Discrimination threshold (%coh)

Stimulus duration (ms)

data Perfect accumulation

Failure to integrate all the information for longer stimuli

Kiani et al, 2008

Effect of motion pulses on the choice

•  Do early motion pulses have stronger effect on the behavior? •  Does the monkey ignore late motion pulses?

Fixation Targets on

Motion Go

Eye movement Motion pulse: 3.2%, 200 ms

Kiani et al, 2008

Motion pulses bias the monkey’s behavior

-50 -25 0 25 50

Motion strength (%coh)

1

0.75

0.5

0.25

0

Proportion of rightward choices

Kiani et al, 2008

The effect of the motion pulse diminishes by pulse time

Pulse time (ms) 0 400 800

1

0

Shi

ft (%

coh)

Motion strength (%coh)

Pro

porti

on o

f rig

htw

ard

choi

ce

-20 0 20 -20 0 20

1

0.5

0

1

0.5

0

Pulse time < 175 ms 362-563 ms

175-362 ms > 563 ms

Kiani et al, 2008

Time

Bound

The pulse effect is expected to depend on both the pulse time and the strength of primary motion

0 200 400 600 800

0.3 0.2 0.1 0 -0.1

Pulse time (ms)

Change of rightward choice proportion

The effect of the motion pulse depends on the primary motion strength

0%

12.8%

Kiani et al, 2008

What  are  the  implicaBons?  

Visualstimulus

Neuronalresponse

Behavioraljudgement

Britten, Shadlen, Newsome & Movshon, 1993

MT responses depend on motion coherence

0 25 50 75 1000

50

100

150

200

0 10 20 30 40 500

50

100

0 25 50 75 1000

10

20

30

40

50

0 10 20 30 400

25

50

75

100

Firi

ng r

ate

(impu

lses

/tria

l)

Coherence (%)

Preferred directionNull direction

Barlow, Levick and Yoon, 1971

Britten, Shadlen, Newsome & Movshon, 1992

Britten, Shadlen, Newsome & Movshon, 1992

Coherence (%)

Pro

port

ion

corr

ect

Britten, Shadlen, Newsome & Movshon, 1992

Coherence (%)

Pro

port

ion

corr

ect

0

20

40

60

Num

ber

of c

ells

0.1 1 10Threshold ratio (neuron/behavior)

MT cells are as sensitive as monkeys to visual motion

When measured correctly!

Cohen & Newsome, 2009

Saccade

Motion

Targets

Fixation

Time

]

ReactionTime

Hanks, Kiani & Shadlen, 2014

Speed-Accuracy Tradeoff

a b

cBound for Tin

0

A Bound for Tout

0

A

Tin accumulator Tout accumulator

0 20 40 600.5

0.6

0.7

0.8

0.9

1

Motion strength (% coh)

Prop

ortio

n co

rrect

0 20 40 60

0.4

0.5

0.6

0.7

Motion strength (% coh)R

eact

ion

time

(s)

Wednesday, November 6, 2013

Speed-Accuracy Tradeoff

Hanks, Kiani & Shadlen, 2014

a b

cBound for Tin

0

A Bound for Tout

0

A

Tin accumulator Tout accumulator

0 20 40 600.5

0.6

0.7

0.8

0.9

1

Motion strength (% coh)

Prop

ortio

n co

rrect

0 20 40 60

0.4

0.5

0.6

0.7

Motion strength (% coh)

Rea

ctio

n tim

e (s

)

Wednesday, November 6, 2013

Speed-Accuracy Tradeoff

0 200 400

0.40.60.8

Time from dots onset (ms)

Nor

mal

ized

firin

g ra

te Accuracy condition

0 200 400

0.40.60.8

Time from dots onset (ms)

Speed condition

−50 0 500123

Motion strength (% coh

Build

up ra

te(N

orm

aliz

ed F

R /

s)

StrongTin

StrongTout

Weak

a b cStrong Tin

Strong Tout

Weak

Wednesday, November 6, 2013

Speed-Accuracy Tradeoff

Hanks, Kiani & Shadlen, 2014

−400 −200 0

0.5

1

Time from saccade (ms)Nor

mal

ized

firin

g ra

te Accuracy condition

−400 −200 0

0.5

1

Time from saccade (ms)

Speed condition

0 20 40 60

0.8

1

1.2

Motion strength (% coh)

Nor

mal

ized

FR

befo

re s

acca

de

Faster

Slower

Faster

Slower

a b c

Wednesday, November 6, 2013

Speed-Accuracy Tradeoff

Hanks, Kiani & Shadlen, 2014

−100 0 100 200 300

0.4

0.6

0.8

1

1.2

Time from motion onset (ms)

Nor

mal

ized

firin

g ra

te

−300 −200 −100 0 100Time from saccade (ms)

0 20 40 600.50.60.70.80.9

Motion strength (% coh)

Prop

ortio

n co

rrect

0 20 40 60

0.4

0.5

0.6

0.7

Motion strength (% coh)

Rea

ctio

n tim

e (s

)

Speed vs. Accuracya b

c

0 0.05 0.1 0.150

0.2

0.4

0.6

0.8

Decision time (s)

Urg

ency

sig

nal (

Frac

tion

boun

d)

Wednesday, November 6, 2013

Speed-Accuracy Tradeoff

Hanks, Kiani & Shadlen, 2014

Saccade

Motion

Targets

Fixation

Time

]

ReactionTime

Hanks, Kiani & Shadlen, 2011

Integration of priors in the decision-making process

0

0.2

0.4

0.6

0.8

1

Proportion positive choice

−50 −25 0 25 500

500

1000

1500

2000

Motion strength (%coh)

Reaction time (ms)

0

0.2

0.4

0.6

0.8

1

−50 −25 0 25 500

500

1000

1500

2000

Motion strength (%coh)

0

0.2

0.4

0.6

0.8

1

−50 −25 0 25 50200

400

600

800

Motion strength (%coh)

Subject SK Subject LH Monkeys B, I, N, and S

50:50 prior80:20 prior

Integration of priors in the decision-making process

Hanks, Kiani & Shadlen, 2011

dt

Accumulatedevidence

A

-A

time

a

b

time

Decision variable

dt

Accumulatedevidence Bias signal

+

(Decision variable)

Momentaryevidence

Momentaryevidence

A

-A

Integration of priors in the decision-making process

Hanks, Kiani & Shadlen, 2011

−30 −20 −10 0 10 20 30−100

0

100

200

300

Motion Strength (% coh)−30 −20 −10 0 10 20 30

−100

0

100

200

300

Motion Strength (% coh)

Time from motion onset (ms) Time from motion onset (ms)

250 300 350 400 450 500 550−10

01020304050

Time from motion onset (ms)250 300 350 400 450 500 550

−60−40−20

020

Time from motion onset (ms)

Buildup rate (sp/s )

Time−dependent bias signal (sp/s)

Difference in buildup rate (sp/s )

a b

dc

2

2

300 400 500 600 7000

10

20

30

300 400 500 600 700−20

−15

−10

−5

0e f

50:50 prior80:20 prior

50:50 prior20:80 prior

Tim

e-de

pend

ent

Bia

s sig

nal (

sp/s

) Integration of priors in the decision-making process

Hanks, Kiani & Shadlen, 2011

−50 −25 0 25 500

0.2

0.4

0.6

0.8

1

−50 −25 0 25 50200

400

600

800

Motion strength (%coh)

Proportion positive choice

Reaction time (ms)

a

b

50:50 prior80:20 prior

Integration of priors in the decision-making process

Hanks, Kiani & Shadlen, 2011