Attention - Control

marblefreedomAI and Robotics

Nov 14, 2013 (3 years and 11 months ago)

114 views

Washington University Medical School

Attention
-

Control

Control has been studied, in part, by focusing
on processes or tasks that demand “attention”
or “control”, in comparison with those that
don’t, i.e. those that are more “automatic”.


Automaticity is often realized by extensive
experience: skilled vs. unskilled typist


Washington University Medical School

Attention
-

Control

Two studies for today



Posner and Snyder



Neely


Washington University Medical School

Posner and Snyder

Reviewed studies, and performed
experiments of their own and postulated:




Controlled

Automatic

Speed


Slow



Fast

Expec.


Effect



No effect

Fac/Inh


Both



Fac. only





Washington University Medical School

Neely

Decided to test Posner/Snyder in a lexical
decision task with semantic priming


Lexical decision
-

see a word (word),
respond yes, see a nonword (vorl), respond
no.


Semantic priming
-

word is processed faster
if it is preceded by a related word (cat
-
>
dog) than unrelated or neutral stimulus
(xxxx
-
> dog)

Washington University Medical School

Two explanations for semantic priming

Grandma
-

when you see “cat”, you start thinking
about related things like “dog”, so you respond
more quickly to it (control/attentional explanation).


Word nerd
-

when you see “cat”, there is automatic
“spreading activation” to related lexical items like
“dog”, and you respond more quickly (automatic
explanation)


Washington University Medical School

What is spreading activation?

TIGER

CAT

IRRITATING

DOG

FURRY

LOVING

EXAMS

SLOBBER

Washington University Medical School

What is spreading activation?

TIGER

CAT

IRRITATING

DOG

FURRY

LOVING

EXAMS

SLOBBER

Washington University Medical School

What is spreading activation?

TIGER

CAT

IRRITATING

DOG

FURRY

LOVING

EXAMS

SLOBBER

Washington University Medical School

What is spreading activation?

TIGER

CAT

IRRITATING

DOG

FURRY

LOVING

EXAMS

SLOBBER

Activated nodes
are responded to
more quickly

Washington University Medical School

Posner and Snyder

Reviewed studies, and performed
experiments of their own and postulated:




Controlled

Automatic

Speed


Slow



Fast

Expec.


Effect



No effect

Fac/Inh


Both



Fac. only





Washington University Medical School

Neely
-

Experimental conditions

To distinguish automatic from controlled in time
-



Manipulate SOA (stimulus onset asynchrony), time
between prime onset and target onset


Short (emphasize automatic)
-

200 ms

Long (emphasize control)
-

2000 ms

Washington University Medical School

Neely
-

Experimental conditions

To manipulate expectation
-



Non
-
shift
-

When you see BIRD, think of birds

Shift
-

When you see BODY, think of building parts

Washington University Medical School

Neely
-

Experimental conditions

To allow for both facilitation and inhibition
-



Related
-

prime
-
target relation (BIRD
-
ROBIN)

Neutral
-


prime stimulus is XXXX (

XXXX
-

SOFA)

Unrelated
-

No prime
-
target relation (BIRD
-
SOFA)


Washington University Medical School

Neely predictions

250

2000

Inhibit

Facilitate

Neutral

Washington University Medical School

Neely predictions
-

Rel./Non
-
shift/Exp.

BIRD
-
ROBIN

250

2000

Inhibit

Facilitate

Neutral

Washington University Medical School

Neely predictions
-

Rel. /Non
-
shift /Exp

BIRD
-
ROBIN

250

2000

Inhibit

Facilitate

Neutral

X

X

Washington University Medical School

Neely predictions
-

Unrel. ./Shift/Exp

BODY
-
DOOR

250

2000

Inhibit

Facilitate

Neutral

Washington University Medical School

Neely predictions
-

Unrel. ./Shift/Exp

BODY
-
DOOR

250

2000

Inhibit

Facilitate

Neutral

X

X

Washington University Medical School

Neely predictions
-

Unrel. /Shift/Unexp.

BODY
-
MAPLE

250

2000

Inhibit

Facilitate

Neutral

Washington University Medical School

Neely predictions
-

Unrel. /Shift /Unexp.

BODY
-
MAPLE

250

2000

Inhibit

Facilitate

Neutral

X

X

Washington University Medical School

Neely predictions
-

Rel. /Shift/Unexp.

BODY
-
ARM

250

2000

Inhibit

Facilitate

Neutral

Washington University Medical School

Neely predictions
-

Rel. /Shift /Unexp.

BODY
-
ARM

250

2000

Inhibit

Facilitate

Neutral

X

X

Washington University Medical School

Who’s right? Grandma or Nerd
-
boy?

Grandma
-

when you see “cat”, you start thinking
about related things like “dog”, so you respond
more quickly to it (control/attentional explanation).


Word nerd
-

when you see “cat”, there is automatic
“spreading activation” to related lexical items like
“dog”, and you respond more quickly (automatic
explanation)

Washington University Medical School

Who’s more right? Grandma or Nerd
-
boy?

Do the two types of processes seem to interact, or do
they appear to act independently?


They appear to act independently
-

facilitations and
inhibitions are the same whether they would help
each other, or cancel each other out.

Washington University Medical School

Any neuroscientific evidence?

Studies of practice on functional anatomy

A lot of similarities: vision, motor output

Some important differences: Meaning, grammar, attention

Some regions (e.g. motor cortex) are

similar in the two tasks

Other regions are more active

in one task than the other

Original verb generation study

It was surprising that there were
regions that emphasized reading
over verb generation.

Practicing verb generation

A musing
--

reading is skilled, and verb
generation is effortful.

Practicing verb generation

A musing
--

reading is skilled, and verb
generation is effortful.

Practicing verb generation

What if you practiced generation
over and over?

Practicing verb generation

What if you practiced generation
over and over?

Use less

Practicing verb generation

What if you practiced generation
over and over?

Switch

areas

Practicing verb generation

Marc Raichle

Julie Fiez

Tom Videen

Ann
-
Mary McLeod

Jose Pardo

Peter Fox

Paradigm

Behavior

Behavior

Behavior

Brain

“Generate” regions

Same prediction for the
“generate” regions:


They will decrease with
practice and come back to
original level with novel
words.

Left Frontal

Reading baseline

Left Frontal

Reading baseline

Reading regions

Different predictions for the regions
more active in the reading task:


In the “use less” scenario, these regions
would remain unchanged across
conditions.


In the switching pathway scenario, after
practice, the generate task would look
like the reading task.

Left Insula

Reading baseline

Left Insula

Reading baseline

Practicing verb generation

What if you practiced generation
over and over?

Switch

areas

Use less

Any reason to buy this idea?

Switch

areas

Use less

Tono Tono Tono