Autoshaping. II. Applicability of the Autoshaping Principles to Some Natural Learning Phenomena

Tbe

lapanese

lburnal

of

lvchenomic

Scr'ence

1ee6,

Vol.

5,No. 1,27-36

Appl

to

Autoshaping.

II.

icability

of

the

Autoshap

Some

Natural

Learning

ing

Principles

Phenomena

Shinya

S.SUZUKIi)

'

Nburobiology

Research,

V17terans

Administration

Ddedicat

Center,

Sopalveda,

Caldernia

91343

and

Pepartment

of

Rsychiatzy

and

Biobehavioral

Sciences,

School

of

Medicine,

Uhiversity

of

Calijbrnia,

Les

Angeles,

Calijbrnia

90024,

USA

Although

the

principles of autoshaping

derived

irom

laboratory

experiments using mostly the adult pigeon were shown to

have

some cross-species and cross-situational generality,

it

is

yet

to

be

demonstrated

that the

_principles

are of any use

in

ttnderstanding

Iearning

phenomena

outside the

raboratory.

Three

natural

learning

_phenomena

were

discussed

that

might

involve

processes

or mechanisms common to autoshaping.

First,

the ontogenetic

development

of

food-

or water-ingestive

behavior

in

the

_yoting

chick was considered to

parallel

the

formation

of autoshaped responses to the signal of

food

or water

delivery

in

the

adult

_pigeon.

Second,

like

autoshaping,

imprinting

in

_young

_precocial

birds

coulcl

be

regarded as an example of the classical conditioning of signal-directed

behavior.

Third,

search

image

iormation

or

learning

to recognize

food

items

against

background

in

birds

might

involve

autoshaping-like

_processes

in

that

it

requires visual

cliscrimination

based

on

stimulus-rein-forcer

asseciation,

It

is

concluded

that

the

demonstrated

applicability of

_the

autoshaping

principles

to

these

natural

learning

_phenomena

should encourage a

functional-biologicai

(evolutionary)

approach

to

learning.

Key

words: autoshaping, natural

learning,

ingestive-behavioraldevelopment,

imprinting,

search

image

formation.

'

.

I.

Introduction

The

ways

in

which

learning

occurs

may

reflect

the

causal relationships

among

en-vironmental events

in

the

natural situation

where

the

learning

occurs.

Considerations

of

the

function

of

learning

in

its

natural

context may

_predict

what and

how

a

partic-1)

The

author thanks

Professor

H.M.

_Jenkins

of

the

Department

of

Psychology,

McMaster

versity,

Ontario,

Canada

for

reading an earlier

version of

the

manuscript.

The

author's

_present

address

is

the

Departrnent

of

Neuroscience,

Mitsubishi-Kasei

!nstitute

of

Life

Sll

ooya,

Machida-shi,

Tokyo

194,

Japan.

ular

species

will

learn

(Shettleworth,

1972,

p.

59).

The

coping

behavior

_of

_an

_organism

in

_the

artificial

niche

(laboratory)

can

only

be

under-stood

in

_terms

_of

its

behavior

_in

_its

natural

(evolutionary)

niche

(Garcia,

Clarke,

&

Hawkins,

1973,

p.

1).

Faced

with a number of recently

discovered

"anomalies"

(e.g.,

learned

_taste

aversion,

auto-shaping)

which

do

not

fit

into

the

traditional

framework

of

learnlng

but

do

indicate

the

existence

of

"biological

constraints

"

on

learn-ing,

many animal

behaviorists

have

begun

to

realize

the

importance

of statements such as

28

The

Japanese

Journal

of

1973;

Rozin

&

Kalat,

1971;

Seligman,

1970;

Seligman

&

Hager,

1972).

Since

learning

in

the

laboratory

is

considered

to

be

infiuenced

by

not only experimental manipulations

but

also

organismic

_{predispositions,}

it

is

essentiar

to

examine

both

kinds

of

infiuence

simultaneously

in

the

study of animal

learning.

It

should

be

noted

that

the

laboratory

situation cannot

be

regarded as a

_proper

_place

to

study

these

or-ganismic

predispositions,

since

it

is

consid-erably

different

from

the

natural environment

in

which

these

_{predispositions}

have

evolved

and

thus

are adaptive

(i,e,,

functional).

There-fore,

it

seems necessary

to

examine

these

pre-dispositions

in

the

natural setting.

However,

there

are a number of obstacles

in

carrying out such an attempt.

Three

of

them

rnay

be

noted.

First,

the

lack

of

experi-mental control and

the

resulting

inability

of

dissociating

the

effects

of

various

factors

are

the

major reason

for

discouraging

the

field

study

by

the

experimental

_{psychologist.}

Sec-ond,

it

may

be

dificult

to

find

natural

learning

phenomena

which

correspond

well

to

laboratory

learning

_phenomena.

Third,

even

if

such

cor-responding cases are

found,

notions such as

stimulus, response, reinforcer, and

their

con-tingencies

developed

through

laboratory

studies may not

be

translatable

into

natural situations.

In

spite

of

these

dithculties,

an

attempt

will

be

made

here

to

find

and examine natural

phenomena

which may

be

useful

in

under-standing

the

nature of autoshaping

in

the

laboratory

(see

Suzuki,

1985).

Such

an attempt

seems

tenable

_partly

because

autoshaping

re-presents

a relatively

_primitive

form

of

learn-ing

(see

Moore,

1973)

and

is

already

known

to

be

constrained

by

species-specific

predispo-sitions.

The

following

three

phenomena

will

be

con-sidered

that

might

involve

autoshaping-like

processes.

First,

the

ontQgenetic

development

of

ingestive

behaviors

(eating

and

drinking)

in

birds

is

examined

in

the

hope

that

it

may

reveal

the

ontogenetic

origins

of

the

standard

auto$haped

response

'(pecking).

Secend,

im-printing

in

precocial

birds

is

related

to

auto-shaping.

Since

both

involve

stimulus-directed

approach

behavior,

they

may

have

_common

underlying mechanisms.

Third,

it

is

suggested

that

an autoshaping-like

_process

_might

be

in-Psychonomic

Scjence

VoL

5,

No.

1

volved

in

the

formation

of so-called "specific

search

images

" established

through

an animal's

encounter with

different

_types

_of

food

items.

II.

Ontogeny

of

ingestiye

behavior

In

manyspecies ontogenetic experience seems

necessary

for

an

individual

to

be

able

to

rec-ognize an object as a

food

item

and resptind

to

it

in

an appropriate way.

For

example,

young

chicks

initially

displaying

indiscriminate

pecking

at

any

small

objects

apparently

learn

to

peck

at only edible objects

(for

a

discussion

of maturational and experiential

factors

in-volved

in

this

process,

see

Hess,

1973,

_pp.

289-306;

see

･also

Hogan,

1973).

If

an association

between

the

visual characteristics

_of

_an

_edible

object

and

the

reinforcing consequence of

its

ingestion

is

found

to

be

important

in

this

learn-ing,

then

the

situation can

be

conceptualized

as

an autoshaping experiment where

the

CS

and

US

are

located

in

the

same

_place

or rather

two

aspects

of

the

same object.

In

an

experiment

by

Hess

(1964)

chicks

were

presented

with

_two

stimuli, a white

triangle

on

a

_green

background

and

a

white

circle

on

a

blue

background.

Grain

was

available

only

in

back

ot

the

former

stimulus which

had

been

shown

to

be

le$s

_prefered

(innately)

than

the

latter.

The

chicks

were

given

this

discrimi-native

training

at various ages and

then

tested

for

learned

shape and color

_preference

at older

ages,

The

results showed

that

the

discrimi-native

training

was effective

in

_producing

more

pecking

at

the

initially

less

_prefered

triangle

in

later

extinction sessions

only

when

the

chicks

were

about

3-5

days

old.

This

learned

preference

was shown

to

be

very resistant

to

modification.

An

interpretation

of

these

results

in

terms

of

the

autoshaping

principles

alone

is

diMcult

partly

because

it

is

not clear whether

the

triangle

acts as

a

Pavlovian

CS

(which

elicits

the

pecking

response) or a

Skinnerian

discriminative

stimulus

(which

sets an occasion

for

the

emitted respon$e

to

be

reinforced).

However,

if

we assume

that

autoshaping

represents a more

_primitive

and rapid

type

of

learning

than

conditional

discrimination

(see

Hearst

&

Jenkins,

1974,

_p.45),

it

may

be

possible

that

the

above example can

be

char-acterized as

a

learning

process

similar

to

auto

-'

S.S.

Suzuki:

Autoshaping

shaping,

The

_persistence

of

learned

_preference

seems

adaptive since

in

nature

it

is

highly

unlikely

that

signals of

food

objects change

their

significance

(i.e.,

their

food

value) over

(evolutionary)

time.

ln

his

ethological study

on

the

transition

from

dependent

to

independent

feeding

in

the

young

ring

dove,

Wortis

(1969,

_pp.44-45)

made

an

illuminating

observation,

which

seems worth

quoting

at

length.

Hovv

does

the

behavior

of

the

parents

direct

the

squab's

behavior

so

that

pecking

at

_grain

may

become

a

_probable

response

to

_parents'

behavior

when

the

squab

is

approximately

2

weeks

_old?

While

_the

_present

study

has

concentrated

on

experiments

_with

squabs

which

were already

2

weeks old, a

ation

_of

_the

_ontogeny

of

the

behavioral

relationship

of

_parents

and

_young

_suggests

a

hypothesis.

_From

_the

_time

_that

_begging

emerges as a

dominant

behavior

_pattern

in

the

hungry

young

squab,

a characteristic of

the

behavior

is

that

the

squab

_pecks

at an

area of

the

_parents'

bill,

When

tion

feeding

occurs,

the

squab's

bill

must

be

opened

and closed

in

order

for

any crop-milk

to

be

ingested.

Therefore,asaresult

of

the

experience

of regurgitation

feeding,

it

is

suggested

that

bill-opening-and-closing

ments

become

conditioned

te

the

_pecking

movements

that

the

squab

directs

towards

the

parents'

bill.

When

the

_young

are about

2

weeks

old, and

the

parent

doves

begin

to

respond

to

begging

by

_pecking

at

_grain,

as

demonstrated

in

this

study,

it

is

suggested

that

the

squab

will attempt

to

maintain

contact with

the

_parent's

bill

when

the

_parent

begins

to

_peck

at

_grain.

While

the

squab

is

begging,

the

parent

lowers

to

the

level

of

the

floor,

or

the

area where

there

is

_grain,

and

the

squab's

bill

will

therefore

touch

the

grain

in

the

vicinity

where

_the

_parent

is

pecking.

If

biil-opening-and-closing-move-ments, are conditioned-to

the

squab's

_pecking

movements,

then,

when

the

squab

_pecks

at

the

surface near

the

_parent's

bM,

some

_grain

would enter

the

mouth as a consequence

_of

the

occurrence

of

the

sequence of

begging,

pecking,

bill-opening-and-closing

movement$,

The

ingestion

of

_grain

would

be

expected

to

and

Natural

Learning

29

be

tolerated

since

the

texure

of

_grain

is

not

a

new experience

for

squabs of

this

age

which

have

been

fed

crop-milk mixed with

grain

since

the

middle

of

the

lst

week of age.

On

the

basis

of

these

observations,

it

seems

possible

to

draw

the

following

conclusions.

(1)

The

squab

_pecks

at

the

_parent's

bill

which

signals

the

imminent

delivery

of

food

(crop-milk)

into

its

mouth.

(2)

The

bill-opening-and-closing movements appropriate or necessary

for

food

ingestion

develop

due

to

the

associa-tion

between

the

close view of

the

_parent's

bill

and

food

in

the

(squab's)

mouth.

(3)

The

parent's

deliberate

movements make

it

_possible

for

the

squab

to

direct

the

already established

sequence

of

_pecking

and

bill-opening-and-clos-ing

rnovements

at

grain.

(4)

This

experience

enables

the

squab

to

excute

the

same sequence

of

rnovements

toward

_grain

without

the

_parent's

guidance.

The

relevance

of

these

conclusions

to

autoshaping

may

be

self-evident.

Another

good

example of

the

involvement

of

autoshaping-Iike

_processes

in

the

ontogeny of

ingestive

behaviors

was

_provided

by

Hunt

and

Smith

₍₁₉₆₇₎

who examined

the

development

of

drinking

re$ponses

in

_young

domestic

chicks.

They

first

noted

that

expetienced chicks showed

different

movement

_patterns

in

_pecking

･grain

and

drinking

water,

Nthough

both

involve

a

similar sequence

_of

orientation and

downwardf

forward

thrust

of

the

head,

_pecking

involves

bill

opening near

the

end of

the

thrust,

object

seizing,

bill

closing,

and

head

withdrawal,

whereas

drinking

involves

bill

holding

in

the

water, movement of

throat

muscles,

head

lower-ing

with a

forward

scooping movement,

head

raising,

and

swallowing with movement of

the

tongue

and

throat.

It

was

indicated

that

naive

chicks

initially

do

not show

the

drinking

re-sponse,

consisting

of

the

movements

described

above,

directed

to

the

sight

of

water,

but

ap-parently

learn

to

do

so after experiencing water

inside

the

bill

by

pecfeing

at water

clrops.

It

should

be

noted

here

that

this

finding

is

in-compatible with

the

response-reinforcement

account

since

the

_pecking

response was

replac-ecl

by

the

drinking

response

in

spite of

the

former's

reinforcement with water.

This

finding

can

be

compared

to

Woodruff

and'

pi-30

The

_Japanese

_Journal

ofPsychonomicScience

_Vol.

5,

No.

1

geons

learned

to

respond

to

the

signal of water

(lighted

key)

with

the

drinking

movements.

Finally,

Woodruff

and

Starr

(1978)

made

a

direct

observation on

the

involvement

of

auto-shaping

in

the

development

_of

feeding

_and

drinking

responses

in

chicks.

They

employed

a

discriminative

omissien

training

in

which

CS+

_(a

colored

key)

was

followed

by

a

US

(intraorally

injected

food

or

water)

on!y

if

key

contact responses

did

not

occur.

In

spite

of

the

omission contingency,

the

chicks

displayed

species-specific

feeding

or

drjnking

responses

directed

toward

the

CS+

key

when

_paired

with

food

or water, respectively.

In

summary,

the

studies

described

in

this

section

_suggest

_that

_young

birds

_learn

_to

_direct

species-specific appetitive,tconsummatory

re-sponses appropriate

for

the

ingestion

of

dif-ferent

reinforcers

to

the

visual characteristic$

of

the

reinforcers

_on

_the

basis

_of

visual-oral

stimulus associations

(i.e.,

autoshaping),

There-fore,

autoshaping

i$

implicated

in

the

develop-ment of visual recognition and

ingestion

_of

food

and water.

This

suggests

an

interesting

possibility

that

"

autoshaping

in

adult organisms

may

be

characterized

_as

reactivation

of

that

process

which originally

led

to

visual

recogni-tion

of

the

reinforcer

during

ontogeny

(Wood-ruff

&

Starr,

1978,

_p.

271)."

III.

Imprinting

Imprinting

refers

to

the

_phenomenon

in

which

_young

_precocial

birds

learn

_to

follow

moving

objects

(normally

their

_parents)

once

they

are exposed

te

these

objects early

in

their

life.

An

obvioussimilaritybetween

imprinting

and

autoshaping

is

that

in

both

cases

the

in-dividual's

behavior

is

modified

to

be

directed

toward

certain environmental

features

after

some

form

of experience.

Although

there

are

consideral)le

differences

in

situations

in

which

they

occur

and

their

biological

functions,

the

above similarity

invites

a search

for

mecha-nisms

common

to

the

two

_phenomena.

Among

many models of

imprjnting

(for

re-views

of

imprinting

models, see

Hess,

1973,

pp.351-423;

Rajecki,

1973),

a

classical

condi-tioning

model

_proposed

by

Hoffman

and

Ranter

(1973)

seems most useful

in

relating

imprinting

to

autoshaping.

This

model attempts

to

explain

major a$pects of

imprinting

on

the

basis

of

the

following

_premises.

The

first

three

account

for

the

formation

of

the

bird's

attachments

to

imprinting

stimuli.

(1)

Precocial

birds

have

an

innate

disposition

to

respond

fiIially

to

cer-tain

kinds

of stimuli such as moving objects.

(2)

Stimuli

that

innately

elicit

filial

responses

(e.g.,

following)

are also

innately

reinforcing.

(3)

Initially

neutral

features

of an

imprinting

stimulus

_gradually

acquire

the

releasing

(elic-iting)

and reinforcing

_properties

because

of

their

spatial and

temporal

juxtaposition

with

innately

reinforcing stimulation.

The

remain-ing

_premises

de,al

with maturational constraints

on

imprintiRg

and

expression

of

social

attatch-ments,

(4)

There

is

an

increasing

tendency

in

an

immature

_precocial

bird

to

respond

fear-fully

to

unfamiliar

imprinting

_stimuli,

(5)

The

bird's

responses

to

a

_given

imprinting

stimulus

represents a reso!ution of competing

tendencies

generated

by

the

stimulus

to

react

filially

(ap-proach)

or

fearfully

(withdrawal).

The

first

three

premises

can easily

be

trans#

lated

into

the

Pavlovian

terminology.

Some

aspect

(e,g,,

movement) of an

imprinting

stim-ulus acts as a

US

which elicits

filial

behavior

(e,g,,

following)

as a

UR.

Other

features

(e.g.,

size, shape, color) of

the

same stimulus which

are

initially

neutral

(CS)

acquire

the

capacity

to

elicit

filial

behavior

(CR)

similar

to

that

elicited

by

the

US

as a result of

their

_pairings

(or

coexistence)

with

the

US.

The

major

the-oretical reason

for

distinguishing

the

uncon-ditional

and conditional aspects of

the

imprint-ing

stimulus may

be･stated

in

the

following

way.

Although

any of a wide variety of

mev-ing

stimuli can elicit

the

following

response

in

a

newly

hatched

_precocial

bird,

a

_prolonged

exposure

to

one

_particular

stimulu$ makes

the

bird

direct

its

following

response

to

only

that

stimulus; other stimuli

that

would

have

elicited

the

same

response

then

appear

to

elicit

fear

(withdrawal)

reactions.

This

_phenomenon

was

called "emergent

discrimination"

by

Jaynes

(e.g.,

1958),

and

is

one of

the

defining

charac-teristics

of

imprinting.

If

we assume

that

the

US

aspect

of

a

_given

stimulus

does

not

change

its

significance

(i,e,,

US

_properties)

as

a result of

the

subject's exposure

to

the

stim-ulus

(this

must

be

_the

defining

characteristic

of any

US),

the

above

_phenomenon

can

be

.

'

S.S,

_{Suzuki:Autoshaping}

taken

to

indicate

that

the

subject's exposure

to

the

stimulus must

have

changed

the

signif-icance

of

its

non-US

aspect.

In

other

words,

a non-imprinte.d

_stimulus

loses

its

_capacity

to

elicit

filial

behavior

because

fear

reactions

caused

by

the

bird's

unfamiliarity with

its

CS

aspect

suppresses

filial

behavior

covertly

elic-ited

by

its

intact

US

aspect.

This

conclusion remains

speculative

unless

it

is

_possible

to

distinguish

the

two

aspects

of

the

imprinting

stimulus

experimentally.

James

(1959)

suggested earlier

that

so-called retinal

flicker

_produced

by

a

moving obj'ect might act as a

US

for

imprinting.

Retinal

fiicker

refers

to

the

fiuctuation

in

illumination

on

the

retina

of an avian eye as an object moves across

the

visual

field,

James

(1959)

demonstrated

that

chicks could

be

imprinted

on astationary source

of

fiashing

light.

Furthermore,

it

was shown

that

if

a stationary

object

(plastic

ball)

was

placed

near

the

flashing

light

source,

the

chicks

would subsequently

follow

the

object moved

acress

a runway.

These

observations support

Hoffman

and

Ranter's

(1973)

hypothesis

_that

the

initially

neutral

ieatures

of an

imprinting

object

acquire

the

capacity

to

elicit

filial

be-havior

through

their

association with

the

in-nately reinforcing

aspect

of an

imprinting

object..

It

_should

be

noted

here

that

as

in

autoshaping

the

US

acts

as

an

innate

releasing

stimulus as well

as

a

Pavlovian

reinforcer.

It

should

_also

be

emphasized

that

in

natural

set-tings

a

US

and

its

_signals

are normally

jux-taposed

or

integrated

in

a single

_object,

_thus

ensuring

their

rapid and stable association.

In

summary,

it

was suggested

_that

_the

formation

of

a

following

response

in

a

_young

_precocial

bird

directed

toward

a moving object may

be

regarded as an

instance

_ot

_the

Pavlovian

con-ditioning

of

skeletal

behavior

similar

to

auto-shaping.

IV.

Search

image

formation

The

concept of "search

image"

was

origi-nally used

by

ecologist

L.

Tinbergen

(1960)

in

order

to

explain certain nonlinear relations

between

the

density

of

_prey

species

(insects)

and

the

actual number of

_prey

of

these

species

taken

by

predatory

birds

(great

tits).

For

exarnple, when

the

density

of

a

given

prey

and

Natural

Learning

31

species

is

relatively

low,

the

_predator

tends

to

ignore

the

_prey

when encountered.

However,

when

the

_prey

density

increases

and

thus

the

predator's

chance enco"nter with

the

_prey

in-creases,

the

rate of

_predatien

suddenly

in-creases

_presumably

as a result of

the

bird's

learning

to

recognize

the

_prey

against

back-ground.

In

Tinbergen's

terminology,

the

bird

is

said

to

be

"

adopting

a

specific search

image

"

for

the

_prey

species.

Therefore,

in

_general,

the

nonlinear relations

between

the

_prey

den-sity and

_the

number of

_prey

actually

eaten

can

be

taken

to

result

from

the

fact

that

pre-dators

adopted "specific _search

_images"

for

certain

_species

_and

_concentrated

_on

_these

spe-cies

by

neglecting

other

species

for

which

no

search

images

were

formed.

Since

Tinbergen's

seminal study,

the

notion of search

image

has

been

used

by

some

ethologists

and

ecologists

interested

in

foraging

behavior

_(see

Curio,

1976,

pp.

58-84;

Krebs,

1973,

pp.81-93).

The

most

rigorous

experimentar

study on

this

subject so

far

reported was

done

by

M.

Dawkins

_(1971a,

b).

In

her

experiments

(1971a)

young

chicks were

observed

taking

colored rice

_grains

(green

or orange)

_placed

on

fioors

of either matched

(cryptic)

or nonmatched

(conspicuous)

color.

The

basic

finding

was

that

the

chicks

took

cryptic

_grain

at a much slower

rate

than

conspicuous

_grain

but

after several

minutes of

foraging

they

started

taking

cryptic

grain

at an accelerating rate.

She

interpreted

this

finding

as

indicating

"that chicks

did

not

take

cryptie rice at

first

because

they

did

not

see

it

and

that

the

sharp

increase

in

the

rate

at which

the

_grains

were

taken

later

in

the

test

shows

that

the

chicks soon

learnt

to

detect

them

(p.

569)."

How

are

these

observations related

to

auto-shaping?

It

was shown earlier

(Section

II)

that

_young

chicks might

learn

to

discriminate

food

and nonfood

or

food

and

water

on

the

basis

of visual-oral stimulus associations.

It

is

quite

possible

to

reason

that

the

similar

_process

may

be

operative

in

the

_phenomenon

described

above.

That'

is,

as a result of

first

several

encounters with cryptic

_grain,

chicks

learn

its

visual characteristics which

_predict

or

_precede

the

imminent

arrival of reinforcing stimuration

(i.e.,

intraoral

_grain).

Since

background

32

'

The

Japanese

Journal

of

the

discrimination

is

made

between

the

visual

characteristics of

the

background

and

those

of

the

_grain.

This

argument

_naturally

leads

_to

a

general

conclusion

that

the

animal's encounter with a

biologically

significant object

(rein-forcer)

modifies

its

_perception

_of

_that

_object.

V.

Discussion

I

have

examined

the

three

natural

learnin.cr

phenomena

that

might

involve

_processes

or

mechanisms common

to

autoshaping.

Arnong

these,

the

development

of

food-

or

water-inges-tive

behaviors

in

the

_young

chick

directly

parallels

the

formation

of

autoshaped

responses

to

the

signal

of

food

or water

delivery

in

the

adult

_pigeon.

The

major

difference

between

the

twe

cases

(other

than

the

species and age

of

the

_subject)

is

_that

in

_the

former

_the

_signal

consists

of

the

visual

characteristics

of

a

rein-forcer

itself,

whereas

in

the

Iatter

it

is

the

key

light

spatially separated

from

the

rein-forcer.

Or

more

_precisely,

in

the

autoshaping

situation

the

key

light

acts as a secondary

signal added

to

the

already established,

_prircary

signal

(i.e,,

the

visual characteristics) of

the

reinforcer.

The

autoshaping experiment

indicates

that

in

the

absence of

the

_primary

signal

(i.e,,

dur-ing

the

CS

_period)

the

animal

tends

to

direct

itS

reinforcer-appropriate

behavior

_toward

_the

secondary

signal.

However,

when

both

signals

are

_present,

the

bird

would choose

the

_primary

signal

presumably

because

it

is

far

more valid

or

reliable

than

the

secondary signal,

This

is

simply

due

to

the

fact

that

in

comparison with

the

secondary signal,

the

_primary

signal

has

(1)

optimal or near optimal

CS

_properties

for

?avlovian

conditioning

in

terms

of

its

temporal

and spatial contiguity

to

the

US,

and

(2)

a

considerably

long

history

of

_pairings

with

the

US

_(i,e.,

a conditioning

trial

occurs

every

time

the

bird

eats

_grain).

Recently,

Boakes

(1977,

1979)

has

contrasted

autoshaping

(sign-tracking)

with what

he

calls

"goal-tracking"

which

refers

to

behavior

di-rected

toward

the

US

site.

In

his

experiments

with rats, a

Iight

(CS)

mounted

in

a

nonre-tractable

lever

signalled

the

delivery

of

food

into

a

food

tray.

Lever

depression

was

taken

as a sign-tracking

behavior,

while

head-poking

Psychonomic

Science

Vo].

5,

No.

1

into

the

food

tray

was

defined

as a

goal-track-ing

behavior,

He

usually obtained

both

types

of

behavior

during

the

CS

_period,

Holland

(1980),

using a similar experimental sittiation,

found

that

localized

CSs

tended

to

evoke

be-haviors

directed

toward

the

CS

source

(sign-tracking),

while more

diffuse

CSs

tended

to

evoke

behaviors

mostly

directed

toward

the

US

site

(goal-tracking).

Although

further

studies

are needed

to

clarify what species and

situa-tional

factors

determine

the

occurrence

of

the

two

ty･pes

of

tracking

behavier,

it

seems

pos-sible

to

hypothesize

that

stimulus

features

of

the

US

site acquire response-evoking

_properties

in

the

same way as nominal

CSs

do

during

conditioning, and

that

the

relative

(associative)

strength

of

these

two

sets of stimuli

deter-mines which

type

of

behavior

and

hew

tt

will

be

generated.

In

the

natural

setting

sign-tracking

and

_{goal-tracking}

are

nermally

in-separable, or sign-tracking

inevitably

leads

the

subject

to

the

_goal.

From

this

viewpoint, one

can

argue

that

somewhat maladaptive

sign-tracking

behavior

found

in

the

autoshaping

experiment

(i.e.,

key

_pecking

is

unnecessary

and

disadvantageous

to

the

subject)

is

simply

due

to

the

use of an abnormal

laboratory

situ-ation where

the

CS

and

the

US

are spatially

separated

(see

Hearst

&

Jenkins,

1974,

_p.45).

I

hax're

argued

that

imprinting

and

autoshap-ing

can

be

related

in

the

sense

that

both

are

examples of

the

classical conditioning of

skel-etal

behavior.

Although

he

does

not accept

the

classical conditioning model of

imprinting,

Hess

_(1973,

pp.291-306)

regards

the

formation

of

food-signal

_preference

as

``

_food

_imprinting."

The

major reason

for

this

interpretation

seems

to

Iie

in

the

rapidity and relative

irreversibility

of

learned

food-signal

_preference

and

the

ex-istence

of sensitive

(critical)

_periods.

Although

the

latter

aspect

is

not

found

in

the

autoshap-ing

of adult

birds,

the

rapidity of acquisition

and relative resistance

to

extinction were also

characteristics of autoshaping

(e.g.,

see

Boakes,

1979).

So

there

appear

te

be

some similarities

between

imprinting

and autoshaping other

than

the

mere occurrence of signal-directed

be-havior.

Since

there

are only a

few

studies of

auto-shaping using reinforcers ether

than

food

or

water

<e.g.,

heat

reinforcer,

Wasserman,

1973;

`

S.S.

Suzuki:

Autoshaping

sexual reintorcer,

Blackman,

1971,

cited

in

Hearst

&

_Jenkins,

1974,

and also

Moore,

1973),

the

importance

or

_{pervasiveness}

of

autoshap-ing-like

learning

in

the

development

and

modi-fication

of non-ingestive

behaviors

is

_yet

to

be

clarified.

Numerous

imprinting

studies

together

with

these

few

autoshaping studies seem

to

expand

the

scope of autoshaping research

by

indicating

that

a

catalog of

learned

behaviors

in

which

the

autoshaping

principles

are

appli-cable are not

limited

to

ingestive

behavior.

Such

a catalog might

be

expanded

further

by

careful analyses of other

examples

of

the

learned

modifications of

instinctive

behaviors

involved

in

mating,

_parenting,

aggre$sion,

etc,

I

suspect

this

to

be

the

case

because

a releaser

of

a

_given

innate

action

is

likely

to

function

also as a

Pavlovian

US

and consequently stimuli

coexistent with

the

releaser

acquire some

response-evoking

properties.

In

any case, a

combined

use of etl]ological and

experimental-psychological

analyses will

be

required

to

ex-amine such

a

_pessibility.

Search

image

formation

is

essentially an

example of

learning

to

recognize

food

objects.

However,

the

transient

nature of a search

image

distinguishes

itself

from

ether examples

of

food

recognition

learning.

A

search

image

formed

of

a

_particular

prey

species

is

not

per-manent

but

variable

depending

on

the

current

density

of

that

species.

When

the

density

decreases

to

a certain

level,

the

search

image

is

_presumed

to

be

erased or suppressed.

How

is

this

erasure or suppression

is

achieved

is

unknown and

dificult

to

explain

in

terms

of

the

autosEaping

principles.

However,

as

des-cribed earlier,

the

formation

or acquisition

of

search

images

is

certainly

compatible

with

the

autoshaping

interpretation.

One

of

the

virtues

of

the

naturalistie

ap-proach

to

learning

is

that

one

can easily

infer

the

biological

function

of each

instance

of

learn-ing

by

looking

at

the

natural context

in

which

it

occurs.

The

biological

function

or adaptive

significance of search

image

formation

is

to

facilitate

foraging

ediciency.

That

is,

the

presence

of search

images

for

relatively

abun-dant

_prey

species

facilitates

hunting

those

species,

while

the

absence of search

images

for

relatively rare

_prey

species eliminates

costly or

time-consuming

efforts of

hunting

and

Natural

Learning

33

these

rare species.

These

considerations

sug-gest

that

autoshaping

and

resultant

sign-track-ing

behavior

can

be

regarded as refiecting such

an adaptive

learning

mechanism as

involved

in

search

image

formation.

IV.

Concluding

remarks

The

mainstream

_psychology

of animal

learn-ing

has

been

facing

a conceptual crisis

for

the

last

two

decades

or so

in

the

sense

that

the

validity

of

its

fundemental

tenet

is

questioned

on

both

theoretical

and empirical

grounds

(see

Jenkins,

1979,

for

a comprehensive

history

of

animal

learning

_psychology).

According

to

the

tenet,

there

are

general

laws

of

learning

ap-plicable

to

all animals

including

man,

and

such

laws

can conveniently

be

discovered

through

laboratory

experiments

using

a

few

"repre-sentative "

species

and

_paradigms.

Perhaps

the

strongest empirical attack

on

this

so-called

general

process

learning

theory

came

from

the

the

study

of

taste

aversion

learning

which

revealed an

inadecuacy

of

its

(implicit)

assump-tion

of equal associability among events

(e.g.,

Garcia

&

Koelling,

1966).

There

accumulated

many other examples suggesting

that

there

are

no

_general

rules

for

_predicting

how

different

stimuli, responses, and reinforcers are

asso-ciated

in

different

species

in

different

situations

(e.g.,

Shettleworth,

1972).

One

major alternative

to

the

_general

_process

theory

of

learning

has

been

adovocated

by

ethologists and ethologically-oriented

psychol-ogists who consider

learning

as an adaptive

specialization of

the

species

to

which a subject

belongs

(e.g,,

Bolles,

1970

;

Rozin

&

Kalat,

1971).

According

to

this

view,

laws

of

learning

may

be

different

in

different

species which

have

evolved

under

different

evolutionary

_pressures.

And

such

laws

of

learning

can only

be

under-stood

when viewed

in

light

of

the

natural

(evolutionary)

context.

There

are

two

major criteria

for

deciding

whether a

law

of

learning

derived

from

labor-atory experiments on one sepcies

in

one

labor-atory

situation

has

any

_generality.

First,

a

law

should

be

applicable

to

the

species

(used

in

laboratory

experiments)

in

its

natural

en-vironment

(Criterion

1).

This

criterion

is

often

nor-34

The

_Japanese

_Journal

ofPsychonomicScience

Vol.

5.

No.

1

mally

learn

in

their

natural environments are

vastly

different

_from

_what

_they

_{are expected}

or

forced

_to

learn

in

laboratory

settings,

For

example,

laws

of

learning

based

on analyses

of

bar-pressing

in

the

laboratory

rat may

be

meaningless

if

_there

_is

no equivalent of

bar-pressing

in

the

wild

_rat's

life.

Second,

a

law

of

laboratory

learning

should

be

_{generalizable}

to

other species

in

_their

natural

environments

(Criterion

2).

For

example,

laws

of

bar-press

learning

in

the

rat should

be

_applicable

_to

verbal

learning

in

_human

_infants.

_Usually,

_the

fulfi11ment

of

Criterion

1

seems

_prerequisite

_to

that

of

Criterion

2.

That

this

is

not necessarily

the

case was suggested

by

Schwartz

(1974,

_p.

196).

According

to

him,

it

is

_possible

that

"the

study of

lower

organisms

in

_arbitrary

situations may

_yield

_principles

that

do

not

describe

the

behavior

of

that

species

in

nature,

･but

that

do

generalize

to

more complex species."

However,

he

could not

indicate

specifically

how

it

is

possible.

Furthermore,

this

_paradoxical

possibility

seems remote

in

_view

_ot

_the

evolu-tionary

theory

of

behavior.

An

additional,

somewhat weak

(from

_the

evolutionary

view-point),

criterion

that

has

been

_employed

by

some comparative

_{psychologists}

such

as

Bitter-man

(e.g.,

1975)

is

whether

laws

of

learning

established

through

studies on one species

in

one

laboratory

situation can

be

applied

to

other

species

in

comparable or similar situations

(Criterion

3).

For

example,

_this

criterion

is

concerned with whether

laws

of

bar-press

learning

in

the

rat

are

applicable

to

key-peck

learning

in

the

_pigeon,

In

what

follows

I

will

briefly

discuss

sorne

implications

of

the

analyses

of

the

autoshaping

(see

Suzuki,

1985)

and autoshaping-like

phenom-ena

(described

in

the

_preceding

sections of

this

_paper)

for

the

ongoing

_debate

on

the

generality

ef

laws

oi

learning,

We

have

_seen

earlier

that

the

autoshaping

_principles

derived

mainly

from

studies

_on

_the

_pigeon's

key-peck-ing

behavior

have

some

_generality

in

_terms

_of

the

three

criteria

described

above.

First,

al-though

it

is

not

clear whether adult wild

pigeons

show autoshaping-like

learning

in

_their

natural

habitat,

such

learning

is

indeed

im-plieated

in

the

development

of

ingestive

be-haviors

in

young

birds

including

pigeons

(Criterion

1).

If

we ac,cept

Woodruff

and

Starr's

(1978)

view

that

atttoshaping

in

adult

organisms represents reactivation of

the

_process

responsible

for

the

development

of

ingestive

behaviors,

the

_parallel

between

laboratory

and

naturalistic autoshaping

becomes

apparent.

Second,

it

was

_possible

to

apply

the

autoshap-ing

principles

to

the

_phenomena

of

imprinting

and search

image

formation

found

in

eertain

species of

birds.

This

suggests some

degree

of cross-species and cross-situational

_gener･

alities of

the

autoshaping

_princlples

(Criterion

2).

Third,

the

autoshaping

_phenomenon

was

demonstrated

in

several species

in

comparable

settings

(Criterion

3).

In

spite of

these

hopeful

indications

of

the

generality

of

the

autoshaping

_principles,

it

was

already

shown

that

the

_general

_principles

are

not so

helpful

in

understanding some

details

of a

_given

instance

of autoshaping.

We

have

seen

that

any

instance

_of

autoshaping cannot

be

free

from

species-specific

and

situation-specific

influences.

_These

_inHuences

_are

_by

definition

cannot

be

infered

from

_{species-general}

and situation-general

_principles,

The

_problem

becomes

clear:

Even

though

_general

laws

of

autoshaping may

_exist,

such

laws

are not･

enough

to

_provide

a complete

_account

_of

_each

instance

of autoshaping,

Furthermore,

since our

knowledge

of what

_i$

_general

_is

always

based

on

our

knowledge

of what

is

specific,

it

seems

essential

to

first

examine

specifics

of

auteshaping using

_a

variety of species and

situations.

Perhaps

the

principal

strength of

the

evolu-tionary

approach

to

learning

lies

in

its

ability

to

deal

with

both

diversity

and universality of

learning.

That

is,

although

the

evolutinary

approach emphasizes

diversity

of specialized

adaptations,

it

is

capable of explaining

univer-sality of

learning

among a variety of species

by

assuming

the

existence of common

evolu-tionaryi'ecological

_pressures

and

_phylogenetic

continuity

(e,g.,

Alexander,

1975;

Johnston,

1981

_;

Plotkin

&

Odling-Smee,

1979).

However,

the

evolutionary approach should not

_preclude

the

use

of

the

traditional

learning

_paradigms

such as classical and operant conditioning.

Rather,

as

the

_present

essay on autoshaping

has

indicated,

the

traditional

_paradigms

are

indispensable

in

the

analysis of natural

learn-ing

phenomena.