harris, r. carl title the effects of type, identifiability, and … · 2014. 1. 2. · document...

18
DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability, and Variety of Instructional Instances on Subject Matter Concept Attainment. PUB DATE Apr 72 NOTE 17p.; Paper presented at 1.:e meeting of the American Educational Research Association in Chicago, Illinois on April, 1972 EDRS PRICE MF-$0.65 HC-$3.29 DESCRIPTORS *College Students; Concept Formation; *Concept Teaching; *Generalization; alized Instruction; Instruction; *Instrut tonal Improvement; Instructional Staff; Research Projects; Students; *Teaching Methods; Teaching Techniques ABSTRACT Subject matter concept attainment by adult SS was studied by experimentally manipulating instructional instances which were either positive or negative, clear or obscure, or presented in large or small variety. Training and testing instances were composed of case studies communicated via prose descriptidn that had been analyzed for relevant and irrelevant attributes. Conceptualization was defined as generalization within and discrimination between classes and was assessed by having 176 undergraduate volunteers classify new positive and negative case studies after instruction from a definition and selected instances. Two, three-way analyses of variance followed by t-tests revealed that instance type, identifiability, and variety have differential effects on generalization and discrimination. References are included. (Author)

Upload: others

Post on 17-Sep-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

DOCUMENT RESUME

ED 078 315 CG 007 111

AUTHOR Harris, R. CarlTITLE The Effects of Type, Identifiability, and Variety of

Instructional Instances on Subject Matter ConceptAttainment.

PUB DATE Apr 72NOTE 17p.; Paper presented at 1.:e meeting of the American

Educational Research Association in Chicago, Illinoison April, 1972

EDRS PRICE MF-$0.65 HC-$3.29DESCRIPTORS *College Students; Concept Formation; *Concept

Teaching; *Generalization; alizedInstruction; Instruction; *Instrut tonal Improvement;Instructional Staff; Research Projects; Students;*Teaching Methods; Teaching Techniques

ABSTRACTSubject matter concept attainment by adult SS was

studied by experimentally manipulating instructional instances whichwere either positive or negative, clear or obscure, or presented inlarge or small variety. Training and testing instances were composedof case studies communicated via prose descriptidn that had beenanalyzed for relevant and irrelevant attributes. Conceptualizationwas defined as generalization within and discrimination betweenclasses and was assessed by having 176 undergraduate volunteersclassify new positive and negative case studies after instructionfrom a definition and selected instances. Two, three-way analyses ofvariance followed by t-tests revealed that instance type,identifiability, and variety have differential effects ongeneralization and discrimination. References are included.(Author)

Page 2: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

FILMED FROM BEST AVAILABLE COPY

CC 007 111'

THE EFFECTS OF TYPE, IDENTIFIABILITY, AND VARIETY

OF INSTRUCTIONAL INSTANCES ON

SUBJECT MATTFR CO\Cr--97:

R. CARL HARRIS

DEPARTMENT OF EDUCATION

THE CHURCH COLLEGE OF HAWAII

LAIE, HAWAII 96762

U S DEPARTMENT OF HEALTH,EDUCATIONAL WELFARENATIONAL INSTITUTE OF

EDUCATIONTHIS DOCUMENT HAS BEEN REPF0DUCED E %ACT L Y AS RECEIVED FROM

THE PERSON OR ORGANIZATIONORIGIN

ATING IT POIN TS OF VIEW OR OPINIONS

STt TED DO NOT NECESSARILY REPRE

SENT OFFICIAL NATIONAL INSTITUTE OF

EDUCATION POSITION OR Poucv

A PAPER PRESENTED AT THE MEETING

OF THE

AMERICAN EDUCATIONAL, RESEARCH ASSOCIATION

CHICAGO, APRIL 1972

Page 3: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

The Effects of Type, Identifiability, and Variety

of Instructional Instances on Subject

Matter Concept Attainment

by

R. Carl Harris

The Pennsylvania State Univelsity

Subject matter concept attainment by adult Ss was studied by

experimentally manipulating instructional instances which were

eitho;tr

positive or negative, clear or obscure, or presented in large or

small

variety.

Training and testing instances were composed of case studies

communicated via prose descriptions that had been analyzed for

relevant and irrelevant attributes.

Con optualization was defined as

generalization within and discrimination between classes and was

assessed by having 176 undergraduate volunteers classify new positive

and negative case studies after instruction from a definition and

0

selected instances.

Two, three-way AVM* followed by t-tests revealed

that instance type, identifiability, and variety have differential

effects on qeneralization and discrimination.

How at the Church College of Hawaii

The

Effe

cts

of T

ype,

Iden

tifta

bilit

y. a

nd V

arie

ty

of In

stru

ctio

nal I

nsta

nces

on

Sub

ject

Mat

ter

Con

cept

Atta

inm

ent

R, C

arl H

arris

Pen

nsyl

vani

a S

tate

Uni

vers

ity

Fro

m b

asic

res

earc

h on

con

cept

lear

ning

hav

e co

me

tlwee

maj

or v

aria

bles

(DeC

ecco

, 196

8)ro

n w

hich

inst

ruct

iona

l ins

tant

er, m

cy d

iffer

:(I

) In

,lanc

e

may

be

eith

er p

ositi

ve if

all

rele

vant

attr

ibut

es a

re p

rese

nt o

r ne

gatp

.e it

I)IIC

or m

ore

rele

vant

attr

ibut

e is

mis

sing

giv

en c

onju

nctiv

e co

ncep

ts (

ilmla

nd

(2)

Inst

ance

lden

tlfla

bilit

y m

ay r

ange

from

obs

cure

to c

lear

dep

endi

ng o

n ho

w

embe

dded

the

rele

vant

attr

ibut

es a

re a

mon

g th

e irr

elev

ant a

ttrib

utes

(A

rdie

r.

1966

), a

nd (

3) In

stan

ce v

arie

ty w

hich

may

ran

ge fr

om n

kw

to .1

1.tr

gt.

of d

iffer

ent i

nsta

nces

(P

odel

I, 19

58).

Stu

dies

whi

ch h

ave

inve

stig

ated

the

influ

ence

of i

nsta

nce

type

on

cont.cin

lear

ning

hav

e fo

und

sign

ifica

nt s

uppo

rt fo

r th

e su

perio

rity

of p

ositi

ve u

sei

Live

inst

ance

s w

hen

the

conc

epts

wer

e de

fined

as

a co

njun

ctiv

e co

mbi

natio

n of

rele

vant

attr

ibut

es a

nd w

hen

geom

etric

des

igns

wer

e us

ed a

s st

imul

us m

ater

ials

(Hov

land

& W

eiss

, 195

3; B

rune

r, e

t. al

., 19

56; B

ourn

e, l9

6n).

The

find

ings

have

bee

n ao

con

vinc

ing

the

Cla

rk (

1971

) ha

s be

en a

ble

to s

tate

that

of a

ll th

e

poss

ible

pre

sent

atio

n se

quen

ces,

sequ

ence

of a

ll po

sitiv

e in

stan

ces

is m

ost

Now

at t

he C

hurc

h C

olle

ge o

f Haw

aii,

Late

, Haw

aii

Page 4: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

2

effe

ctiv

e fo

rton

junc

tive

conc

epts

whi

le a

seq

uenc

e of

neg

ativ

e in

stan

ces

is le

ast

effe

ctiv

e. T

he p

robl

em e

ncou

nter

ed w

hen

tryi

ng to

gen

eral

ize

the

resu

lts o

f

stud

ies

cond

ucte

d on

the

inst

ance

type

var

iabl

e as

with

the

inst

ance

iden

tifia

bilit

y

and

vari

ety

vari

able

s G

that

the

effe

cts

on g

ener

aliz

atio

n an

d di

scri

min

atio

n

have

not

bee

n st

udie

d se

para

tely

. Als

o, m

ost o

f th

e st

udie

s ha

ve u

sed

artif

icia

l

conc

epts

whi

ch a

re h

ighl

y di

ssim

ilar

to th

ose

taug

ht in

sub

ject

mat

ter

conc

ept

lear

ning

(G

lase

r, 1

968)

.

The

infl

uenc

e of

irre

leva

nt a

ttrib

utes

on

conc

ept l

earn

ing

has

also

bee

n

wid

ely

inve

stig

ated

(e.

g., A

rche

r, .

Bou

rne,

& B

raw

n, 1

955;

Bul

gare

lla &

Arc

her,

1962

) sh

owin

g th

at id

entif

iabi

lity

of in

stan

ces

vari

es d

irec

tly w

ith th

e pr

opor

tion

of ir

rele

vant

attr

ibut

es. I

n th

e tr

aditi

onal

exp

erim

enta

l par

adig

m, o

f w

hich

the

cite

d st

udie

s ar

e ex

ampl

es, t

he f

ocus

is o

n or

igin

al le

arni

ng in

con

tras

t to

subj

ect

mat

ter

acqu

isiti

on w

here

tran

sfer

of

lear

ning

is th

e m

ajor

var

iabl

e of

inte

rest

.

Ove

ring

& T

rave

rs (

1966

. 196

7) h

ave

dem

onst

rate

d th

at th

ough

ori

gina

l lea

rnin

g

may

be

som

ewha

t les

s ra

pid,

the

abili

ty to

tran

sfer

con

cept

s an

d ru

les

is s

igni

fi-

cant

ly b

ette

r if

the

inst

ruct

iona

l ins

tanc

es w

ere

typi

cal o

f re

ality

, i.e

., ha

d m

any

irre

leva

nt a

ttrib

utes

in p

ropo

rtio

n to

rel

evan

t attr

ibut

es.

It r

emai

ns o

pen

to q

ues-

tion

whe

ther

obs

cure

or

clea

r in

stru

ctio

nal i

nsta

nces

are

bes

t for

gen

eral

izat

ion

and

disc

rim

inat

ion.

The

re I

s ev

en le

ss a

gree

men

t con

cern

ing

the

effe

cts

of v

arie

ty o

f in

stru

ctio

nal

inst

ance

s on

con

cept

lear

ning

than

with

the

vari

able

of

inst

ance

iden

tifia

bilit

y. T

he

earl

y in

vest

igat

ions

of

Lub

orsk

y (1

945)

alo

ng w

ith m

ore

rece

nt s

tudi

es c

ondu

cted

3

by P

odel

l (19

58),

May

zner

(19

62)

and

Duk

es &

Bev

an (

1967

) le

nd s

uppo

rt to

the

thes

is th

at a

larg

e va

riet

y of

inst

ruct

iona

l ins

tanc

es p

rodu

ces

supe

rior

con

cept

lear

ning

. In

cont

rast

, sev

eral

oth

er in

vest

igat

ors

(Ada

ms,

195

4: S

hore

&

Sech

rest

, 196

1; A

mst

er &

Mar

ascu

ilo, 1

965,

) ge

nt r

ally

con

clud

e th

at s

uper

ior

conc

ept l

earn

ing

obta

ins

whe

n in

stru

ctio

nal i

nsta

nces

are

pre

scnt

td in

sm

all

vari

ety

with

hig

h fr

eque

ncy.

In

none

of

the

stud

ies

have

the

effe

cts

of v

arie

ty

of in

stru

ctio

nal i

nsta

nces

on

gene

raliz

atio

n an

d di

scri

min

atio

n be

en im

estig

ated

.

The

pre

sent

stu

dy w

as d

esig

ned

to in

vest

igat

e th

e ef

fect

s of

inst

ince

type

,

iden

tifia

bilit

y, a

nd v

arie

ty o

n ge

nera

lizat

ion

and

of d

iscr

imin

atio

n us

ing

conn

ect-

ed d

isco

urse

trai

ning

mat

eria

ls to

teac

h a

com

plex

con

junc

tive

subj

ect m

atte

r

conc

ept.

thes

es:

HIa

H:

lb

Hlc

H:

Ila

H:

lib

Mor

e pr

ecis

ely,

the

rese

arch

was

des

igne

d to

test

the

follo

win

ghv

po-

Lea

rner

s w

ho a

re in

stru

cted

by

posi

tive

inst

ance

s w

ill g

ener

aliz

esi

gnif

ican

tly b

ette

r th

an le

arne

rs w

ho a

re in

stru

cted

with

neg

ativ

ein

stan

ces.

Lea

rner

s w

ho a

re in

stru

cted

by

obsc

ure

inst

ance

s w

ill I

.:inc

r.,'

1/C

sign

ific

antly

bet

ter

than

lear

ners

who

are

inst

ruct

ed w

ith c

lear

inst

ance

s.

Lea

rner

s w

ho a

re in

stru

cted

by

a la

rge

vari

ety

of in

stal

iacs

nIll

gene

raliz

e si

gnif

ican

tly b

ette

r th

an le

arne

rs w

ho a

re in

stru

cted

with

a s

mal

l var

iety

of

inst

ance

s.

Lea

rner

s w

ho a

re in

stru

cted

by

nega

tive

inst

ance

s w

ill d

iscr

imi-

nate

sig

nifi

cant

ly b

ette

r th

an le

arne

rs w

ho a

re I

nstr

ucte

d w

ithpo

sitiv

e in

stan

ces.

Lea

rner

s w

ho a

re in

stru

cted

by

obsc

ure

inst

ance

s w

ill d

iscr

imi-

nate

sig

nifi

cant

ly b

ette

r th

an le

arne

rs w

ho a

re ip

stru

cted

with

clea

r in

stan

ces.

:L

earn

ers

who

are

inst

ruct

ed b

y a

larg

e va

riet

y of

inst

ance

s w

illtI

cdi

scri

min

ate

sign

ific

antly

bet

ter

than

lear

ners

who

are

inst

ruct

edw

ith a

sm

all v

arie

ty o

f in

stan

ces.

Page 5: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

ME

TH

OD

Des

ign

and

Subj

ects

4

The

exp

erim

ent w

as c

ondu

cted

acc

ordi

ng to

the

ratio

nale

for

a 2

x 2

x 2

fixe

d fa

ctor

des

ign.

Fac

tor

A w

as tw

o le

vels

(po

sitiv

e an

d ne

gativ

e) o

f in

stan

ce

type

, fac

tor

B w

as tw

o le

vels

(cl

ear,

and

obs

cure

) of

inst

ance

iden

tifia

bilit

y,

and

fact

or C

was

two

leve

ls (

larg

e an

d sm

all)

of

inst

ance

var

iety

. The

inde

-

pend

ent v

aria

bles

wer

e th

e th

ree

fact

ors

liste

d ab

ove,

i.e.

, typ

e, id

entif

iabi

-

lity,

and

var

iety

of

inst

ruct

iona

l ins

tanc

es. T

he d

epen

dent

var

iabl

e w

as b

road

-

ly d

efin

ed a

s cl

assi

fyin

g be

havi

or b

ut m

ore

spec

ific

ally

was

com

pris

ed o

f ge

n-

eral

izat

ion

and

disc

rim

inat

ion

with

reg

ard

to th

e co

ncep

t Ind

ivid

ualiz

ed I

nstr

uc-

tion.

Gen

eral

izat

ion

was

test

ed f

or b

y ha

ving

Ss

clas

sify

cas

e st

udie

s th

at w

ere

judg

ed b

y S

to b

e ex

ampl

es o

f th

e co

ncep

t. D

iscr

imin

atio

n w

as te

sted

for

by

havi

ng S

s cl

assi

fy c

ase

stud

ies

that

wer

e ju

dged

by

E to

be

non-

exam

ples

of

the

conc

ept.

Mat

eria

ls

The

targ

et c

once

pt w

as I

ndiv

idtia

lized

Ins

truc

tion.

The

inst

ruct

iona

l and

test

ing

mat

eria

ls w

ere

brie

f, tw

o-pa

ge, c

ase

stud

ies

(Edl

ing,

197

0) o

f pr

ogra

ms

arou

nd th

e co

untr

y w

hich

per

port

ed to

be

indi

vidu

aliz

ed in

som

e m

anne

r. E

ach

case

stu

dy w

as c

lass

ifie

d as

eith

er a

pos

itive

or

nega

tive

inst

ance

of

the

conc

ept

Indi

vidu

aliz

ed I

nstr

uctio

n an

d w

as f

urth

er c

lass

ifie

d as

eith

er h

avin

g cl

ear

or

obsc

ure

iden

tifia

bilit

y.

The

bas

is f

or d

eter

min

ing

whe

ther

a g

iven

cas

e st

udy

was

a p

ositi

ve o

r

nega

tive

inst

ance

of

the

conc

ept I

ndiv

idua

lized

Ins

truc

tion

was

a 'I

t tin

ition

whi

ch s

peci

fied

fou

r re

leva

nt a

ttrib

utes

in c

onju

nctiv

e co

mbi

natio

n. F

or a

case

stu

dy to

be

clas

sifi

ed a

s a

posi

tive

inst

ance

all

foi.r

rel

evan

t attr

ibut

es

had

to b

e re

pres

ente

d in

one

or

mor

e of

the

case

stu

dy s

ente

nces

. A c

ase

stud

y w

as d

esig

nate

d as

a n

egat

ive

inst

ance

if o

ne o

r m

ore

of th

e re

leva

nt

attr

ibut

es w

as m

issi

ng. A

ll se

nten

ces

in e

ach

case

stu

dy w

ere

anal

yzed

for

expr

essi

ons

of r

elev

ant a

ttrib

utes

.

The

rel

ativ

e id

entif

iabi

lity

of c

ase

stud

ies

stem

med

fro

m f

urth

er c

lass

i-

fica

tion

base

d on

an

oper

atio

nal d

efin

ition

of

ulen

tilla

lulit

y.Fo

r po

sitii

e

stan

ces

it w

as a

ssum

ed th

at c

lari

ty o

f id

entif

icat

ion

incr

ease

d as

the

prop

or-

tion

of r

elev

ant t

o ir

rele

vant

attr

ibut

es in

crea

sed

whi

le f

or n

egat

ive

it w

as a

ssum

ed th

at c

lari

ty o

f id

entif

icat

ion

incr

ease

d as

the

prop

ortio

n of

rele

vant

to ir

rele

vant

attr

ibut

es d

ecre

ased

.

Thr

ough

task

ana

lysi

s ea

ch c

ase

stud

y w

as a

naly

zed

in te

m:,

of th

e ta

rget

conc

ept d

efin

ition

and

the

defi

nitio

n of

iden

tifia

bilit

y re

sulti

ng in

the

ItIt'l

ltaf

ica-

tion

of f

our

case

stu

dy g

roup

s. T

hese

fou

r gr

oups

, i.e

., t

lc..

'-

nega

tive,

obs

cure

-pos

itive

, and

obs

cure

-neg

ativ

e, o

f in

stru

Ltio

n.d

inst

.inie

s

mad

e up

the

four

stim

ulus

poo

ls f

rom

whi

ch in

stan

ces

wer

e ra

ndom

ly s

thu

ed

to c

onst

ruct

the

inst

ruct

iona

l and

test

ing

mat

eria

ls. T

hree

inst

all,

es f

rom

each

gro

up w

ere

rand

omly

sel

ecte

d to

com

pris

e a

larg

e va

riet

y w

here

as o

ne

inst

ance

fro

m e

ach

grou

p w

as s

elec

ted

to r

epre

sent

a s

mal

l var

iety

.

Page 6: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

67

Proc

edur

e'

For

pret

estin

g, in

stru

ctio

n, a

nd p

ost-

test

ing,

176

vol

unte

ers

from

the

unde

rgra

duat

e ed

ucat

iona

l psy

chol

ogy

cour

se a

t the

Pen

nsyl

vani

a St

at.:

Uni

ver-

sity

rep

orte

d to

a s

elf-

inst

ruct

iona

l lea

rnin

g ce

nter

for

two

one-

hour

ses

sion

s.

The

fir

st s

essi

on in

clud

ed th

e pr

etes

t com

pose

d of

fou

r ca

se s

tudi

es a

nd tw

elve

mul

tiple

cho

ice

ques

tions

and

the

inst

ruct

ion

cons

istin

g of

a d

efin

ition

of

the

conc

ept a

nd il

lust

ratin

g in

stan

ces

as p

resc

ribe

d by

the

eigh

t tre

atm

ent c

ondi

-

tions

of

the

expe

rim

enta

l des

ign.

The

lece

nd s

essi

on in

clud

ed a

rev

iew

of

the

inst

ruct

iona

l mat

eria

ls a

nd a

dmin

istr

atio

n of

the

post

-tes

t com

pose

d of

eig

ht

case

stud

iesa

RE

SUL

TS

Res

ults

Rel

ated

to G

ener

aliz

atio

n

To

test

the

Mal

ik a

nd in

tera

ctiv

e ef

fect

s O

f th

ree

sour

ces

of v

aria

nce

on

gene

raliz

atio

n. a

2 X

2 X

2 a

naly

sis

of v

aria

nce

was

con

duct

ed o

n th

e nu

mbe

r

of c

orre

ct c

ase

stud

y cl

assi

fica

tions

. The

mea

ns a

nd v

aria

nces

for

eac

h ce

ll

in th

e 2

X 2

X 2

dat

a m

atri

x ar

e pr

esen

ted

in T

able

i.

Inse

rt T

able

1 a

bout

her

e

The

ana

lysi

s yi

elde

d a

sign

ific

ant m

ain

effe

ct o

n in

stan

ce ty

pe (

F10.

42,

1/16

8, p

(.00

1) a

nd a

sig

nifi

cant

thre

e-w

ay in

tera

ctio

n be

twee

n in

stan

ce M

e,

iden

tifia

bilit

y, a

nd v

arie

ty (

124.

76, 1

/168

, p4.

02),

Bec

ause

of

the

com

plic

at

Table 1

Cell means, standard deviations, and n's for correct category

responses on the generaliraticn test by tyos, a.:!tIntLflaLlAxty,

and variety of instructional instances

TYPE

IDENTIABILITY

Cle

arObscure

VARIETY

VARIETY

Large

Small

Large

Snell

Positive

a

Negative

a

2.32

2.05

.78

.95

22

22

2.09

.92

22

2.18

1.22

22

1.32

2.23

1.68

1.64

.65

.97

.72

.95

22

22

22

22

Page 7: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

lag

effe

cts

of th

e se

cond

-ord

er, i

nter

actio

n, m

eani

ngfu

l tes

ts o

f hy

poth

esis

coul

d on

ly b

e ac

com

plis

hed

by te

stin

g fo

r si

mpl

e ef

fect

s on

the

eigh

t cel

l mea

ns.

Acc

ordi

ngly

, a s

erie

s of

fou

r i -

test

s w

ere

com

pute

d fo

r ea

ch o

f th

e th

ree

gee

-

eral

izat

ion

hypo

thes

es.

Hyp

othe

sis

I (a

).

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

re in

stru

cted

by

posi

tive

inst

ance

s

will

gen

eral

ize

sign

ific

antly

bet

ter

than

lear

ners

who

are

inst

ruct

ed w

ith n

ega

tive

inst

ance

s, a

naly

sis

of s

impl

e ef

fect

s w

as c

ondu

cted

. The

mea

n di

ffer

ence

betw

een

posi

tive

and

nega

tive

inst

ance

s fo

r cl

ear

iden

tifia

bilit

y an

d la

rge

vari

-

ety

was

exa

min

ed f

or s

tatis

tical

sig

nifi

canc

e us

ing

a t-

test

. As

hypo

thes

ized

,

the

test

rev

eale

d th

at p

ositi

ve in

stan

ces

(X =

2.3

2) p

rodu

ced

sign

ific

antly

bet

ter

(t =

3.6

4, d

f =

168,

p(,

.001

) ge

nera

lizat

ion

than

neg

ativ

e in

stan

ces

(Tr,

a 1

.32)

,

The

hyp

othe

sis

was

als

o su

ppor

ted

(t a

1.9

8. d

f =

168

, p<

.05)

by

a te

st f

or s

ig-

nifi

ciuw

e di

ffer

ence

bet

wee

n m

eans

for

pos

itive

(X

= 2

.18)

ver

sus

nega

tive

(TC

a 1

.69)

inst

ance

s on

obs

cure

iden

tifia

bilit

y an

d sm

all v

arie

ty. H

owev

er, t

he

pred

icte

d ou

tcom

e w

as n

ot o

btai

ned

(t a

.66.

df

168,

p).

05)

whe

n po

sitiv

e

(X a

2.0

5) w

as c

ompa

red

with

neg

ativ

e (X

a 2

.23)

on

clea

r id

entif

iabi

lity

and

smal

l var

iety

nor

was

pos

itive

(X

a 2

.09)

sig

nifi

cant

ly s

uper

ior

to n

egat

ive

(X a

1.6

8) w

hen

obsc

ure

iden

tifia

bilit

y an

d la

rge

vari

ety

was

the

case

(t =

1.4

9,

df a

168

, p<

.10)

. Tha

t pos

itive

inst

ance

s pr

oduc

e si

gnif

ican

tly b

ette

r ge

nera

l-

izat

ion

if th

ey a

re e

ither

cle

ar a

nd p

rese

nted

in la

rge

vari

ety

or o

bscu

re a

nd

pres

ente

d in

sm

all v

arie

ty c

an b

. see

n by

insp

ectin

g Fi

gure

1.

Inse

rt F

igur

e 1

abou

t her

e

2.75

2

2.32

2.25

2.10

2.09

2.00

2.05

4.

1.75

-

81.

50 -

1.25

-

2.23 Clear-small

1.68

Crs

cure

-Iar

vo1.

64 C

zscu

re-.

.^ a

ll

1.32 Clear-Large

Posi

tave

Type

Figure 1

Plot of generalisation category

response cell means

showing three-way, disordinal

treat:rent interactions

emphasising instance type factor

t,:eg

atl.v

e

5

Page 8: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

4

10

The

tren

d in

fav

or o

f po

sitiv

e in

stan

ces

is a

lso

stro

ng u

nder

obs

cure

ider

i

llfia

bilit

y an

d la

rge

vari

ety

but n

ot s

igni

fica

ntly

so.

The

onl

y ra

dica

l dep

artu

re

from

exp

ecta

tion

obta

ined

whe

n in

stru

ctio

nal i

nsta

nces

wer

e pr

esen

ted

in s

mal

l

vari

ety

and

thei

r id

entif

iabi

lity

was

cle

ar. H

ere,

neg

ativ

e in

stan

ces

prod

uced

supe

rior

gen

eral

izat

ion

but w

ithou

t sta

tistic

al s

uppo

rt. R

easo

nabl

y fi

rm s

up-

port

was

fou

nd f

or th

e hy

poth

esis

that

pos

itive

inst

ance

s pr

oduc

e be

tter

gene

ral-

izat

ion

than

neg

ativ

e in

stan

ces

with

not

ed q

ualif

icat

ions

.

Hyp

othe

sis

I (b

).

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

re in

stru

cted

with

obs

cure

in-

stan

ces

will

gen

eral

ize

sign

ific

antly

bet

ter

than

lear

ners

who

are

Ins

truc

ted

with

cle

ar in

stan

ces,

an

anal

ysis

of

sim

ple

effe

cts

on c

ell m

eans

was

con

duct

ed.

The

mea

n di

ffer

ence

bet

wee

n cl

ear

and

obsc

ure

inst

ance

s ov

er ti

t.' f

our

inst

ance

type

and

var

iety

com

bina

tions

wer

e te

sted

by

runn

ing

four

t-te

sts.

The

onl

y

sign

ific

ant d

iffe

renc

e be

twee

n cl

ear

and

obsc

ure

iden

tifia

bilit

y (t

2.15

, df

= 1

68.

p405

) w

as in

fav

or o

f cl

ear

inst

ance

s, c

ontr

ary

to e

xpec

tatio

n (R

c *2

.23,

= 1

.64)

, und

er c

ondi

tions

of

a sm

all v

arie

ty o

f ne

gativ

e in

stan

ces.

Obs

cure

0in

stan

ces

did

prod

uce

supe

rior

gen

eral

izat

ion

(Y1o

= 1

.68.

XC

.1.3

2) w

hen

nega

tive

inst

ance

s w

ere

pres

ente

d in

larg

e va

riet

y bu

t non

sign

ific

antly

(t =

1.3

1,

df a

168

, p41

0). T

houg

h th

e tr

end

was

ver

y w

eak,

obs

cure

inst

ance

s w

ere

also

bet

ter

than

cle

ar in

stan

ces

unde

r th

e co

nditi

ons

of p

ositi

vity

and

sm

all v

ari-

ety

(7=

2.1

8. X

= 2

.05;

t = .4

7. d

f =

168.

p7.

05).

In

agre

emen

t with

the

0fi

rst t

est m

entio

ned,

cle

ar w

as s

uper

ior

over

obs

cure

(3i

c =

2.3

2, 3

to =

2.0

9)

in a

mod

erat

e bu

t stil

l nan

sici

fica

nt tr

end

(t =

.84,

df

= 1

68, p

7.05

) w

hen

the

II

inst

ruct

iona

l ins

tanc

es w

ere

posi

tive

and

pres

ente

d in

larg

e va

riet

y.

Lio

n of

Fig

ure

2 re

veal

s th

at th

e hy

poth

esis

reg

ardi

ng th

e sa

lient

,' 'iv

et o

bscu

re

inst

ance

s Is

sup

port

ed, h

owev

er, n

onsi

gnif

ican

tly, b

y tw

o of

11w

tren

ds b

ut n

ot

supp

orte

d by

two

othe

r tr

ends

one

of

whi

ch w

as s

tatis

tical

ly s

igni

tican

t

Inse

rt I

nsur

e 2

abou

t her

e

Such

res

ults

mak

e an

y po

wer

ful g

ener

aliz

atio

n im

pass

ible

con

cern

ing

the

effe

cts

of id

entif

iabi

lity

on g

ener

aliz

atln

n.It

is n

otab

le th

at id

entif

tabi

lity

seem

s to

hav

e lit

tle d

iffe

rent

ial e

ffec

ts w

hen

posi

tive

inst

ance

s ar

e pr

esen

ted

in e

ither

larg

e or

sm

all v

arie

ty (

see

Figu

re 2

) bu

t a g

ood

deal

of

tithe

ren

t ial

effe

ct w

hen

nega

tive

inst

ance

s ar

e pr

esen

ted

in la

rge

nr s

mal

l val

et%

esp

ecia

lly

unde

r th

e cl

ear

cond

ition

.

Hyp

othe

sis

I (c

).

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

re in

stru

cted

with

a la

rge

tat t

ut)

of in

stan

ces

will

gen

eral

ize

sign

ific

antly

bet

ter

than

lear

ners

who

are

intu

it d

with

a s

mal

l var

iety

of

inst

ance

s, a

n an

alys

is o

f si

mpl

e ef

fect

s on

the

eigh

tte

ll

mea

ns w

as c

ondu

cted

. The

mea

n di

tfer

enee

s be

twee

n la

rge

and

a: it

. 1,

of in

stan

ces

over

the

tour

inst

ance

type

and

iden

tifia

bilit

y in

min

s di

ens

wer

e

test

ed b

y ru

nnin

g fo

ur t

-tes

ts. C

ontr

ary

to e

xpec

tatio

n. s

mal

l %on

.=

2.2

1)

resu

lted

in s

uper

ior

gene

raliz

atio

n (t

a 3

.32,

df

a 16

8,p.

,!...

(00i

) i.h

en e

lr.,r

neg

a-

tive

inst

ance

s w

ere

pres

ente

d th

an la

rge

vari

ety

(R -

1.3

2) lo

t tbk

link

sta

tis-

tical

ly s

igni

fica

nt d

iffe

renc

e of

this

ser

ies

of te

sts.

The

onl

y In

stan

ct w

ir

larg

e va

riet

y di

d pr

oduc

e be

tter

gene

raliz

atio

n to

any

cle

ar d

egre

e th

an s

mal

l

Page 9: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

12

.2.75

8 p 2.50

4 2.25

2.32

2.23

2.18 Positive-Small

E 2.00

2.09

Posi

tive-

Lar

ge0

1.75

1.68 Negative-..arge

1.64 Negative-Small

1.50

0

1.25

1.32

Clear

Obscure

Identifiability

Figure

2Plot of generalization category

response cell means

showing three-way disordinal treatment

interactions

emphasising instance identifiability

factor

I i

vari

ety

(Ra

2.32

,X

S2.

05)

was

whe

n cl

ear

posi

tive

inst

ance

s w

ere

pre-

sent

ed. T

his

diff

eren

ce, h

owev

er, w

as in

suff

icie

nt to

rea

ch s

igni

iican

ce

(t a

.98,

df

a 16

8, p

705)

. Har

dly

any

reco

gniz

able

dif

fere

nce

obta

ined

be-

twee

n la

rge

and

smal

l var

iety

whe

n in

stan

ces

are

posi

tive

and

obsc

ure

(XL

a 2

.09,

g2.

18; t

.33,

df

-168

, p7

05)

or w

hen

inst

ance

s ar

c im

p-

tive

and

obsc

ure

a a

1.68

, X1.

64; t

df16

8, p

,>.0

5).

It a

ppea

rs

that

und

er th

e co

nditi

ons

of th

is e

xper

imen

t lar

ge v

arie

ty d

oes

not n

eces

sari

ly

faci

litat

e gr

eate

r ge

nera

lizat

ion

than

sm

all v

arie

ty c

ontr

ary

to p

redi

ctio

n.

The

onl

y su

ppor

ted

conc

lusi

on th

at c

an b

e dr

awn

is th

at s

mal

l var

iety

is s

upe-

rior

giv

en th

e co

nditi

ons

of c

lear

, neg

ativ

e in

stru

ctio

nal i

nsta

nces

. (se

e I

igur

e

3).

Inse

rt F

igur

e 3

abou

t her

d

Res

ults

Rel

ated

to D

iscr

imin

atio

n

The

dat

a fr

om th

e di

scri

min

atio

n ta

sk c

onsi

sted

of

the

num

bei o

f co

rrec

t

clas

sifi

catio

ns o

f ne

gativ

e in

stan

ces

of th

e co

ncep

t ind

ivid

ualiz

ed la

st r

oc to

m

The

se d

ata

wer

e an

alyz

ed u

sing

the

sam

e ra

tiona

le a

nd p

roce

dure

As

thos

e lo

r

gene

raliz

atio

n.

The

mea

ns, s

tand

ard

devi

atio

ns, a

nd n

umbe

r of

Ss

for

each

c4.

11 in

ita.

2 x

2 x

2 da

ta m

atri

x ar

e pr

esen

ted

in T

able

2. E

xam

inat

ion

of th

e an

alys

ts

reve

aled

two

sign

ific

ant i

nter

actio

ns; o

ne b

etw

een

inst

ance

type

at 4

C1

inst

atax

vari

ety

(F *

6.1

5, 1

/168

, p<

.05)

and

ano

ther

bet

wee

n in

stan

ce td

enta

fiat

,iiity

Page 10: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

Table 2

cell means, standard deviations, and n'S for correct category

responses on the discrimination test by type, identifiability,

and variety of instructional instances

TYPE

IDENTIFIABILITY

Clear

Obscure

VARIETY

VARIETY

Large

Small

Large

Small

Y.

2.46

3.00

2.73

2.59

Positive

s.96

.62

.es

1.01

a22

22

22

22

2.86

2:64

3.41

2.73

negative s

.99

.79

.73

.99

22

22

22

22

O 0

2.75-

2.50-

2.25 -

2.00

1.75 -

1.50

1.25

2.32

A, 2.23 Negative-Clear

2.09

,e-

2.18 Pcsitive-Chscste

2.05 Positive-Clear

1.68

1.32

1.64 Negative-Cbcure

Large

Small

vari.ty

Pigure3

Plot of generalization

category response cell

*tears

showing three-way theorelinal

treatment interatticLe

emphasizing instance

variety factor

Page 11: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

16

and

inst

ance

var

iety

(F

= 4

.57,

1/1

68. p

<.0

5). T

he f

irst

inte

ract

ion

resu

lted

from

the

diff

eren

tial e

ffec

t of

inst

ance

var

iety

on

disc

rim

inat

ion

for

posi

tive

or n

egat

ive

inst

ruct

iona

l ins

tanc

es. T

he s

econ

d in

tera

ctio

n is

a r

esul

t of

the

diff

eren

tial e

ffec

t of

inst

ance

iden

tifia

bilit

y on

dis

crim

inat

ion

for

clea

r or

ob-

scur

e in

stru

ctio

nal i

nsta

nces

.

Hyp

othe

sis

n (a

)

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

re in

stru

cted

by

nega

tive

in

stan

ces

will

exh

ibit

sign

ific

antly

mor

e di

scri

min

atio

n th

an le

arne

rs w

ho a

re

Inst

ruct

ed w

ith p

ositi

ve in

stan

ces,

an

anal

ysis

of

sim

ple

effe

cts

was

con

duct

ed.

Col

umn

mea

ns w

ill b

e co

mpa

red

for

this

ana

lysi

s w

here

as c

ell m

eans

wer

e

com

pare

d to

test

Hyp

othe

sis

I (a

). 7

ne d

iffe

renc

e in

pro

cedu

re r

esul

ts f

rom

sign

ific

ant t

wo-

way

inte

ract

ions

in th

e pr

esen

t cas

e w

hile

a s

igni

fica

nt th

ree-

way

inte

ract

ion

had

to b

e ha

ndle

d in

the

form

er c

ase.

'''he

diff

eren

ce b

etw

een

posi

tive

(k-

a2.

59)

and

nega

tive

(R =

3.1

4) in

stan

ces

for

larg

e va

riet

y is

sig

-

nifi

cant

(t =

2.9

3, d

f a

168,

p<

.05)

and

sup

port

a th

e hy

poth

esis

with

neg

ativ

e

inst

ance

s pr

ovid

ing

for

supe

rior

dis

crim

inat

ion.

How

ever

, for

sm

all v

arie

ty,

posi

tive

(3(

= 2

.80)

was

bet

ter

than

neg

ativ

e ((

= 2

.68)

but

not

sig

nifi

cant

ly s

o

(t =

.64.

df

= 1

68. p

,.05)

. The

hyp

othe

sis

was

not

sup

port

ed in

this

latte

r

case

thou

gh th

ere

is a

str

ong

sugg

estio

n fr

om th

e fi

rst s

igni

fica

nt t-

test

and

Flu

whi

ch c

an b

e gr

aphi

cally

obs

erve

d in

Fig

ure

4 th

at n

egat

ive

inst

ance

s ar

e m

ore

faci

litat

ing

at le

ast f

or th

e la

rge

vari

ety.

Inse

rt F

igur

e 4

abou

t het

e

a a a 6 8 O

2.25

3.50

3.25

3.00

2.75

2.50

2.80

2.59

3.14 Large Variety

1.68 Small Variety

Positive

Negative

,Instance Type

Figure

yPlot of discrimination category response colt:-.m reenr

showing

two-way, disordinal treatment interaction

emphasising instance type factor

400

17

Page 12: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

18

Hyp

othe

si;1

1 (b

)

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

rein

stru

cted

with

obs

cure

In-

stan

ces

will

exh

ibit

sign

ifica

ntly

mor

edi

scrim

inat

ion

than

lear

ners

who

are

inst

ruct

ed w

ith c

lear

inst

ance

s, s

impl

e ef

fect

s w

ere

unal

yzet

.. B

ecau

se o

f the

thre

e-w

ay in

tera

ctio

n, H

ypot

hesi

s I(

b) w

as te

sted

by

com

parin

g ce

ll m

eans

.

In th

e pr

esen

t cas

e on

ly tw

o-w

ayin

tera

ctio

ns a

re s

igni

fican

t so

colu

mn

mea

ns

will

be

com

pare

d. T

he h

ypot

hesi

s w

assu

ppor

ted

whe

n th

e di

ffere

nce

betw

een

uma.

,,se(

1 =

3.0

6) a

nd c

lear

a 2

.66)

was

test

ed o

ver

larg

e va

riety

(t v

2.13

,

df v

164

. p<

.05)

but

was

not

whe

n th

edi

ffere

nce

betw

een

obsc

ure

a =

2.6

6)

and

clea

r (1

a 2

.62)

was

test

ed o

n sm

all v

arie

ty(t

11

.75,

df =

168

,

Con

trar

y to

exp

ecta

tion,

wah

t tre

nd d

id a

ppea

rin

this

sec

ond

com

paris

on w

as

in fa

vor

of c

lear

inst

ance

s. in

acc

orda

nce

,with

the

hypo

thes

is th

at le

arne

rs

expo

sed

to o

bscu

re in

stan

ces

will

dis

crim

inat

ebe

tter

than

thos

e ex

pose

d to

clea

r in

stan

ces

and

as c

an b

e se

en in

Fig

ure

5, th

is r

esul

t did

obt

ain

whe

n a

larg

e va

riety

of i

nsta

nces

was

pre

sent

ed.

No

stat

istic

ally

sig

nific

ant d

iffer

-

ence

s w

ere

tkle

eled

whe

n n

smal

l var

iety

of i

nsim

icea

was

pre

sent

ed,

Inse

rt F

igur

e 5

abou

t her

o

!1nr

i..t1

1111

To

test

the

hypo

thes

is th

at le

arne

rs w

ho a

rein

stru

cted

by

a la

rge

varie

ty

of in

stan

ces

disc

rimin

ate

sign

ifica

ntly

bette

r th

an le

arne

rs w

ho a

re in

stru

cted

with

a s

mal

l var

iety

of i

nsta

nces

add

ition

alsi

mpl

e ef

fect

s w

ere

anal

yzed

with

mea

ts. A

gain

, the

col

umn

mea

ns w

ere

com

pare

d in

con

tras

t to

com

paris

on

8 :I)2.

253.50

3.25

3.00

2.7S

2.50

Piq

ue 5

2.42

2.66

3.07 Large Varlety

2.66 Small Variety

Char

Inst

ance

Iden

tifxa

bilit

yO

bscu

re

Plo

t of d

iscr

imin

atio

nca

teep

try

ren;

twin

eI .

show

ing

a tw

o-w

.y, c

ilt04

41;..

.1(1..A1161, itot.t.4

emph

asis

ing

inst

ance

idun

tifia

bilit

y fa

ct

Page 13: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

of c

ell m

ein

as w

as d

one

for

Hyp

othe

sis

I (c

). T

he s

econ

d or

der

inte

ract

ion

in th

e ge

nera

lizat

ion

hypo

thes

es n

eces

sita

ted

the

appl

icat

ion

of t

-tes

ta to

cel

l

mea

ns w

here

as th

e fi

rst -

orde

r in

tera

ctio

n ir

. the

pre

sent

dis

crim

inat

ion

ana-

lysi

s re

quir

es c

olla

psin

g ov

er c

olum

ns f

or m

eani

ngfu

l com

pari

sons

. Whe

n

the

mea

ns f

or la

rge

vari

ety

vers

us s

mal

l var

iety

for

obs

cure

iden

tifia

bilit

y

wer

e co

mpa

red

via

a t -

test

, lar

ge v

arie

ty (

31 =

3.0

7) w

as, a

ccor

ding

to e

x-

pect

atio

n, s

igni

fica

ntly

gre

ater

(t =

2.1

8, d

f =

168

, p(1

5) th

an s

mal

l var

iety

(X =

2.6

6). T

he t-

test

com

pari

ng la

rge

(31

= 2

.66)

and

sm

all (

R =

2.8

2) v

ari-

ety

for

clea

r id

entif

iabi

lity

did

not,

how

ever

, rev

eal a

sig

nifi

cant

dif

fere

nce

(t =

.85,

df

= 1

68, p

7.05

). T

his

part

icul

ar tr

end,

thou

gh, n

onsi

gnif

ican

t,

was

con

trar

y to

exp

ecta

tion.

Add

ition

al p

ost h

oc c

ompa

riso

ns o

f si

mpl

e ef

-

fect

s w

ere

com

pute

d w

hich

com

pare

d la

rge

and

smal

l var

iety

on

the

inst

ance

type

fac

tor.

For

neg

ativ

e in

stan

ces

the

adva

ntag

e of

larg

e va

riet

y a

= 3

.14)

over

sm

all v

arie

ty (

5( =

2.6

8) w

as s

uppo

rted

(t =

2.4

5, d

f =

168

, p4.

05)

as

hypo

thes

ized

. For

pos

itive

inst

ance

type

the

tren

d w

as in

the

oppo

site

dir

ec-

tion

(XL

= 2

.59,

51

= 2

.80)

but

the

diff

eren

ce w

as t

It s

igni

fica

nt (

t = 1

.12,

df =

168

, p)

05).

Con

trar

y to

wha

t was

fou

nd f

or g

ener

aliz

atio

n, it

app

ears

from

the

pres

ent d

ata

and

anal

ysis

that

larg

e va

riet

y do

es p

rodu

ce s

uper

ior

disc

rim

inat

ion

whe

n id

entif

iaoi

lity

is o

bscu

re (

see

Figu

re 6

) or

whe

n in

stan

ces

are

nega

tive

(see

Fig

ure

7).

It is

of

inte

rest

to n

ote

that

thou

gh n

eith

er o

f

Inse

rt F

igur

es 6

& 7

abo

ut h

ere

them

rea

ched

sig

nifi

canc

e, tw

o of

the

four

com

pari

sons

did

sho

w a

tren

d fo

r

qu

3.50

K3.25

ca o

3.00

'3.

07

U o2.75

2.66

2.50

92.

25 -

I

Lar

ge

2.82 Clear

2.66 Coscure

Variety

S all

Figu

re6

Plot of discrimination category response colurn rears

showing a two-way, disordinal treatment intervztion

emphasising instance variety factor

21

Page 14: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

3.50

3.25

3.00

a2.

75

82.

50

02.

25

Fig

ure

3.14

2.59 La

rge

Var

iety

7 Plot of discrimination

category response columnmeans

show

ing

a two-way, disordinal

treatment interaction

emphasising instancevariety factorsm

all

2.80

Pos

itive

2.68

Neg

ativ

e

22

4sm

all v

arie

ty o

f ins

tanc

es to

be

supe

rior;

that

is, w

hen

inst

ance

s w

ere

clea

r

or p

ositi

ve s

mal

l var

iety

see

med

to b

e be

st. T

he c

omm

on it

ribut

cs o

f cla

ri-

ty a

nd p

ositi

vity

see

m to

com

bine

with

sm

all v

arie

ty to

pro

duce

bet

ter

disc

ri-

min

atio

n th

an la

rge

varie

ty. H

owev

er, w

ith g

ener

aliz

atio

n th

e at

trib

utes

of

clar

ity a

nd n

egat

ivity

see

n to

com

bine

with

sm

all v

arie

ty to

be

mor

e ia

cili-

tativ

e th

an la

rge

varie

ty. I

n th

e la

m:.

case

sta

tistic

al s

uppo

rt w

as fo

und

whe

reas

In th

e fo

rmer

cas

e on

ly s

tron

g tr

ends

wer

e ob

serv

ed.

DIS

CU

SS

ION

inst

ance

Typ

e

The

find

ings

of t

his

stud

y su

bsta

ntia

te M

arkl

e an

d T

ietn

ann'

s (1

9b9,

197

0)

supp

ositi

on th

at th

e ap

prop

riate

kin

d of

inst

ruct

iona

l ins

tanc

es to

pro

mot

e

gene

raliz

atio

n ar

e po

sitiv

e in

stan

ces

and

to p

rom

ote

disc

rimin

atio

n ar

e m

o-

tive

inst

ance

s. U

nfor

tuna

tely

, the

se a

utho

rs a

void

any

psc

holo

gi, a

l ecp

lana

-

bon

as to

why

thes

e pr

edic

ted

func

tiona

l rel

atio

nshi

ps m

ight

exi

st.

IltrA

c %

Lt.,

it do

es a

ppea

r th

at a

n ex

plan

atio

n w

hich

wou

ld c

oinc

ide

with

thei

r le

ason

ing

coul

d st

em fr

om a

dis

cuss

ion

of th

e ab

stra

ctio

n pr

oces

s. B

riefly

, it g

cner

al-

izat

ion

is b

ased

t the

lear

ner's

abs

trac

tion

of w

hat t

he c

once

pt is

and

is n

ot

and

thes

e ab

stra

ctio

ns a

re fo

rmed

from

the

lear

ner

anal

yzin

g th

e at

trilm

ics

In s

earc

h of

thos

e w

hich

are

rel

evan

t and

irre

leva

nt, i

t wou

ld to

laiu

tiein

-

stan

ces

that

car

ry th

e in

form

atio

n m

ost d

irect

ly w

ould

be

supe

rior

for

loth

proc

esse

s. A

s de

mon

stra

ted

in th

is in

vest

igat

ion,

pos

itive

inst

ance

s di

d ap

-

pear

to c

arry

the

criti

cal i

nfor

mat

ion

best

for

gene

raliz

atio

n w

hile

neg

ativ

e

Page 15: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

24

inst

ance

s ca

rrie

d it

best

for

dis

crim

inat

ion.

Tho

ugh

Bri

nier

, et.

al. (

1956

) m

ade

no d

istin

ctio

n be

twee

n ge

nera

lizat

ion

and

disc

rim

inat

ion

in c

once

pt le

arni

ng, t

he a

bove

rat

iona

le s

eem

s to

sub

stan

-

tially

agr

ee w

ith th

eir

theo

rizi

ng a

bout

pat

tern

s of

dec

isio

n m

akin

g. T

hese

inve

stig

ator

s ob

serv

ed, a

fter

stu

dyin

g Ss

' con

cept

lear

ning

str

ateg

ies,

that

they

muc

h pr

efer

red

mak

ing

"dir

ect"

teat

s of

thei

r hy

poth

eses

c.,

cern

ing

wha

t the

con

cept

was

. Pos

itive

inst

ance

s fo

r ge

nera

lizat

ion

and

nega

tive

in-

stan

ces

for

disc

rim

inat

ion

wou

ld a

ppea

r to

mak

e di

rect

test

ing

poss

ible

.

Inst

ance

Ide

ntif

iabi

lity

In th

e pr

esen

t stu

dy, p

ositi

ve in

stan

ces

whi

ch h

ad m

any

irre

leva

nt a

ttri-

bute

s as

com

pare

d to

rel

evan

t attr

ibut

es o

r ne

gativ

e in

stan

ces

that

had

man

y

rele

vant

attr

ibut

es a

s co

mpa

red

to ir

rele

vant

attr

ibut

es w

ere

thou

ght t

o be

anal

ogou

s to

the

"rea

listic

" or

mor

e co

mpl

ex tr

aini

ng s

ituat

ion

in th

e O

veri

ng

and

Tra

vers

(19

66, 1

967)

stu

dies

. The

se h

ard-

to-i

dent

ify

or o

bscu

re in

stan

ces

wer

e ex

pect

ed to

fac

ilita

te b

oth

supe

rior

gen

eral

izat

ion

and

disc

rim

inat

ion

sinc

e th

ey w

ould

for

ce th

e le

arne

r to

mak

e fi

ne g

ener

aliz

atio

ns a

nd d

iscr

imi-

natio

ns in

the

boun

dary

are

a of

his

con

cept

. The

fin

ding

s of

this

inve

stig

atio

n

did

not c

ompl

etel

y su

ppor

t thi

s lin

e of

rea

soni

ng. W

hen

disc

rim

inat

ion

was

bein

g te

sted

for

, obs

cure

inst

ance

s di

d in

deed

pro

duce

sup

erio

r re

sults

ove

r

clea

r in

stan

ces

if th

ey w

ere

pres

ente

d in

larg

e va

riet

y. B

ut w

hen

gene

raliz

a-

tion

was

bei

ng te

sted

for

, cle

ar in

stan

ces

prov

ed to

fur

nish

mor

e as

sist

ance

than

obs

cure

inst

ance

s if

the

inst

ance

s w

ere

nega

tive

and

pres

ente

d in

sm

all

vari

ety,

i

The

res

ult f

or d

iscr

imin

atio

n is

in h

arm

ony

with

Mar

kle

and

Tie

man

n's

(197

0) p

resc

ript

ion

that

eac

h no

nexa

mpl

e sh

ould

lack

onl

y on

e of

the

rele

vant

attr

ibut

es, 1

,e.,

nega

tive

inst

ance

s sh

ould

be

diff

icul

t to

tell

from

pos

itive

inst

ance

s if

fin

e di

scri

min

atio

ns w

ere

to b

e le

arne

d.It

app

ears

that

bot

h

the

resu

lts f

or d

iscr

imin

atio

n an

d ge

nera

lizat

ion

obse

rved

in th

is in

vest

iga-

tion

may

be

expl

icat

ed in

term

s of

per

form

ance

set

(R

eese

, 196

4), B

ecau

se

gene

raliz

atio

n is

bas

ed p

rim

arily

on

an a

bstr

actio

n of

wha

t the

con

cept

is,

and

sinc

e th

is a

bstr

actio

n ca

n be

com

e qu

ite s

tabl

e as

the

defi

nitio

n sp

celli

es

rele

vant

attr

ibut

es th

at r

emai

n co

nsta

nt o

ver

all p

ositi

ve I

nsta

nt i

s, a

per

tor-

man

ce s

et f

or m

atch

ing

abst

ract

ion

with

test

inst

ance

wou

ld s

eem

mos

t efl

ect-

ive.

On

the

othe

r ha

nd, d

iscr

imin

atio

n is

a p

roce

ss b

ased

pri

mar

ily o

n an

abst

ract

ion

of w

hat t

he c

once

pt is

not

. Bec

ause

the

irre

leva

nt a

ttrib

utes

nia

y

and

do v

ary

over

inst

ance

s, th

e ab

stra

ctio

n of

wha

t the

con

cept

is n

ot c

an n

ever

beco

me

as s

tabl

e an

d w

ell d

efin

ed a

s th

e ab

stra

ctio

n of

wha

t the

con

cept

is.

Acc

ordi

ngly

, agr

eat d

eal m

ore

sear

chin

g an

d an

alyz

ing

wou

ld s

eem

to b

ane

to a

ccom

pany

the

disc

rim

inat

ion

proc

ess.

The

refo

re, i

t app

ears

that

.1 p

er-

form

ance

set

for

ana

lyst

s w

ould

be

mos

t eff

ectiv

e in

the

i ase

of

disc

i 'ul

ulat

ion.

Obs

cure

trai

ning

inst

ance

s w

ith th

eir

adde

d co

mpl

exity

exity

wou

ld 6

rid

to p

ro b

ier

an a

naly

sts

perf

orm

ance

set

whe

reas

cle

ar in

stan

ce's

wou

ld ti

nil

io p

rf,(

1,«

a

mat

chin

g pe

rfor

man

ce s

et. A

s th

e re

sults

of

this

stu

dy r

evea

led

clea

r in

-

stan

ces

duri

ng in

stru

ctio

n sh

ould

fac

ilita

te s

uper

ior

gene

raliz

atio

n bc

i aus

e of

the

perf

orm

ance

set

for

mat

chin

g th

ey w

ould

fos

ter

whi

le o

bscu

re in

stan

ces

duri

ng in

stru

ctio

n sh

ould

fac

ilita

te s

uper

ior

disc

rim

inat

ion

beca

use

of th

e pe

r-

Page 16: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

26

iorm

ance

set

for

ana

lyzi

ng th

ey w

ould

stim

ulat

e.

Inst

ance

Var

iety

The

aut

hor

alon

g w

ith o

ther

s (B

rune

r, e

t. at

, 195

6; C

arro

ll, 1

964;

DeC

ecco

, 196

8; G

agne

, 197

0) r

easo

ned

that

a v

arie

ty o

f in

stru

ctio

nal i

nst-

sces

was

nee

ded

for

adeq

uate

con

cept

for

mat

ion

to ta

ke p

lace

.

Som

ewha

t sur

pris

ingl

y, th

e re

sults

of

this

stu

dy r

evea

led

that

a la

rge

vari

-

ety

of in

stru

ctio

nal i

nsta

nces

was

not

nec

essa

rily

the

best

for

bot

h of

thes

e

lear

ning

out

com

es. W

hen

gene

raliz

atio

n w

as th

e in

tend

ed o

utco

me,

a s

mal

l

vari

ety

of in

stan

ces

prod

uced

mor

e co

rrec

t cla

ssif

icat

ions

if th

e in

stan

ces

wer

e al

so c

lear

ly n

egat

ive.

In

cont

rast

, whe

n di

scri

min

atio

n w

as th

e in

tend

ed

outc

ome,

a la

rge

vari

ety

of in

stru

ctio

nal i

nsta

nces

res

ulte

d ui

a g

reat

er r

m."

-

ber

of c

orre

ct c

ateg

oriz

atio

ns b

oth

whe

n in

stan

ces

wer

e ob

scur

e an

d w

hen

they

wer

e ne

gativ

e.

In c

ontr

ast t

o th

e bu

lk o

f re

late

d st

udie

s, A

mst

er a

nd M

aras

cuilo

(19

65)

foun

d ev

iden

ce w

hich

cor

robo

rate

s th

e pr

esen

t fin

ding

s w

ith r

egar

d to

the

supe

-

rior

ity o

f sm

all v

arie

ty f

or g

ener

aliz

atio

n. T

hese

aut

hors

fou

nd n

o si

gnif

ican

t

diff

eren

ces

in m

ean

num

ber

of c

orre

ct r

espo

nses

to in

ters

ectio

n an

d un

ion

pro-

blem

s du

ring

trai

ning

bet

wee

n th

e la

rge

and

smal

l var

iety

con

ditio

ns. H

owev

er,

whe

n fo

urth

-gra

de c

hild

ren

wer

e as

ked

to g

ener

aliz

e th

eir

lear

ning

to n

ew in

-

ters

ectio

n an

d un

ion

prob

lem

s, th

e m

ean

num

ber

of c

orre

ct r

espo

nses

for

the

larg

e an

d sm

all v

arie

ty c

ondi

tions

was

sig

nifi

cant

ly d

iffe

rent

in f

avor

of

smal

l

vari

ety.

No

expl

anat

ion

was

pro

pose

d by

Am

ster

and

Mar

ascu

ilo f

or th

ese

un-

expe

cted

res

ults

exc

ept t

o in

fer

that

som

e so

rt o

f in

terf

eren

ce m

ight

hav

e be

en

Z7

oper

atin

g in

the

larg

e vp

ety

cond

ition

and

to p

oint

out

that

Pod

ell (

1963

a. b

)

also

fou

nd s

ome

evid

ence

for

the

adva

ntag

e of

a s

mal

l var

iety

mer

a la

rge

vari

ety.

IMPL

ICA

TIO

NS

FOR

IN

STR

UC

TIO

N

Bas

ed o

n th

e co

nclu

sion

s ar

rive

d at

in th

is s

tudy

the

reco

mm

enda

tion

is

mad

e th

at, w

hen

the

educ

atio

nal g

oal s

peci

fies

an

inte

ntio

n fo

r tin

stu

di.n

ts

to a

cqui

re s

ubje

ct m

atte

r co

ncep

ts, t

each

ers

and

inst

ruct

iona

l des

igne

rs

shou

ld s

peci

fy th

eir

inst

ruct

iona

l obj

ectiv

es in

term

s of

gen

eral

izat

ion

and

disc

rim

inat

ion.

Obj

ectiv

es d

ealin

g w

ith g

ener

aliz

atio

n sh

ould

com

mun

iL m

e

to th

e st

uden

t the

nee

d fo

r re

cogn

izin

g ne

w e

xam

ples

of

the

conc

ept u

hen

the%

are

pres

ente

d. I

n th

e ca

se o

f ob

ject

ives

that

dea

l with

dis

crim

inat

ion,

stu

-

dent

s sh

ould

und

erst

and

from

them

that

they

will

be

expe

cted

to1L

eogn

i/e

new

non

exam

ples

of

the

conc

ept.

Aci

ditio

nal.y

, whe

n te

st it

ems

are

Lon

stru

Lt-

ed f

or a

sses

sing

the

degr

ee o

f co

ncep

t acq

uisi

tion

they

mus

t be

cfm

pris

eti

,.1

exam

ples

and

non

exam

ples

of

the

conc

epts

bei

ng te

sted

for

to w

hixi

i the

stu

-

dent

s ca

n m

ake

gene

raliz

atio

n an

d di

scri

min

atio

n re

spon

ses.

11 h

oih

apes

of

item

s ar

e no

t inc

lude

d th

e co

mpl

ete

test

of

conc

ept l

earn

ing

will

not

haN

c;w

en

atta

ined

.

A s

econ

d re

com

men

datio

n is

that

whe

n th

e te

ache

ror

inst

ruct

iona

l tit

sign

-

er c

onst

ruct

s th

e in

stru

ctio

nal m

ater

ials

, dif

fere

nt c

onsi

dtra

tions

he

gin

for

faci

litat

ing

gene

raliz

atio

n th

an th

ose

give

n fo

r fa

cilit

atin

g di

scri

min

atio

n.It

is r

ecom

men

ded,

as

a ge

nera

l rul

e, th

at in

stru

ctio

nal

inst

ance

s th

at c

onta

in

Page 17: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

28

all o

f th

e te

leva

nt a

ttrib

utes

as

spec

ifie

d by

the

defi

nitio

n be

giv

en to

enh

ance

the

likel

ihoo

d of

acq

uiri

ng th

e ca

paci

ty to

gen

eral

ize.

As

a ge

nera

l rul

e, it

is a

lso

reco

mm

ende

d th

at in

stru

ctio

nal i

nsta

nces

whi

ch la

ck o

ne o

r m

ore

of

the

rele

vant

attr

ibut

es b

e gi

ven

to f

acili

tate

dis

crim

inat

ion.

With

som

ewha

t

less

con

fide

nce,

it is

sug

gest

ed th

at e

asy-

to-i

dent

ify

inst

ruct

iona

l ins

tanc

es

Will

be

bette

r fo

r ge

nera

lizat

ion

whi

le h

ard-

to-i

dent

ify

inst

ance

s v

.1 in

flu-

ence

sup

erio

r di

scri

min

atio

n. A

lso

with

cau

tion,

it is

adv

ised

than

a s

mal

l

num

ber

of in

stan

ces

be p

rese

nted

for

gen

eral

izat

ion

and

a la

rge

num

ber

of

diff

eren

t ins

tanc

es b

e gi

ven

to p

rom

ote

disc

rim

inat

ion.

Prob

ably

the

area

of

mos

t dir

ect a

pplic

abili

ty o

f th

ese

conc

lusi

ons

is th

at

of r

emed

iatio

n w

hen

ther

e is

insu

ffic

ient

sub

ject

mat

ter

conc

ept l

earn

ing.

If

the

teac

her

or in

stru

ctio

nal d

esig

ner

know

that

a s

tude

nt is

fai

ling

to a

cqui

re

a co

ncep

t, th

ey a

re u

rged

to a

ttem

pt to

dis

cove

r w

heth

er h

is p

robl

em s

tem

s

from

a f

ailu

re to

gen

eral

ize

or d

iscr

imin

ate.

The

lear

ner's

fai

lure

to c

orre

ct-

ly c

lass

ify

exam

ples

of

a co

ncep

t wou

ld in

dica

te a

lack

of

gene

raliz

atio

n w

hile

failu

re to

cor

rect

ly c

lass

ify

none

xam

ples

of

a co

ncep

t wou

ld in

dica

te a

lack

of

disc

rim

inat

ion.

The

app

ropr

iate

rem

edia

tion

wou

ld f

ollo

w th

e ge

nera

l con

clu-

sion

s of

this

stu

dy; t

hat i

s, if

the

stud

ent's

wea

knes

s w

as g

ener

aliz

atio

n, th

en

posi

tive

Inst

ance

s sh

ould

be

pres

ente

d, a

nd if

the

stud

ent's

wea

knes

s is

dis

cri-

min

atio

n, th

en n

egat

ive

inst

ance

s sh

ould

be

pres

ente

d. I

t may

be

that

ther

e

are

part

icul

ar k

inds

of

over

gene

raliz

atio

n (i

nsuf

fici

ent d

iscr

imin

atio

n) o

r un

der-

gene

raliz

atio

n (

insu

ffic

ient

gen

eral

izat

ion)

whi

ch c

onsi

dera

tion

of id

entil

iabi

lity

and

num

ber

of in

stru

ctio

nal i

nsta

nces

dur

ing

rem

edia

tion

can

help

. How

ever

,

addi

tiona

l res

earc

h is

nee

ded

to d

eter

min

e ex

actly

wha

t the

se r

ecom

men

da-

tions

wou

ld b

e.

An

addi

tiona

l im

plic

atio

n fo

r ef

fect

ive

inst

ruct

iona

l man

agem

ent c

once

rns

the

anal

ysis

of

inst

ruct

iona

l mat

eria

ls to

be

used

in e

ffec

ting

conc

ept l

earn

ing.

With

out r

egar

d to

the

find

ings

of

this

stu

dy a

nd o

ther

s, a

teac

her

or in

$tru

c-

tiona

l des

igne

r m

ay h

ave

his

own,

wel

l-fo

unde

d re

ason

s fu

r be

lievi

ng th

at

inst

ruct

iona

l ins

tanc

es w

hich

are

cle

ar o

r ob

scur

e or

bot

h, in

a p

artic

ular

sequ

ence

, are

bes

t for

teac

hing

a g

iven

con

cept

.Ii

this

is th

e st

ate

of a

ltair

s

then

the

inst

ruct

or n

eeds

som

e cr

iteri

a fo

r de

cidi

ng w

hich

inst

ance

s ar

e si

m-

ple

and

whi

ch a

re c

ompl

ex. O

ne c

rite

ria

that

cou

ld b

e us

ed, a

s el

emoi

,st r

ated

In th

is s

tudy

, is

that

of

rela

tive

prop

ortio

ns o

f re

leva

nt a

nd ir

rele

vant

attl

!-

bate

s. T

o us

e th

ese

crite

ria

wou

ld r

equi

re th

at th

e in

stru

ctio

nal d

esig

ner

know

wha

t the

rel

evan

t and

irre

leva

nt a

ttrib

utes

are

and

then

to m

ake

som

e

sort

of

freq

uenc

y co

unt o

f ho

w m

any

are

pres

ent i

n th

e va

riou

s in

stan

ces,

Onc

e

this

was

don

e, th

e in

stan

ces

coul

d be

ran

k or

dere

d, a

s w

.is d

one

iol t

ins

stud

y,

and

then

seq

uenc

ed o

r pr

esen

ted

as th

e te

ache

r de

emed

mos

t app

ropr

iate

. Tha

i,

sam

e pr

oces

s is

als

o re

corn

raen

ded

for

dist

ingu

ishi

ng b

etw

een

posi

tive

and

ne-

gativ

e in

stan

ces

whe

n th

e co

ncep

t is

conj

unct

ive

and

the

inst

ruct

iona

l ins

tanc

es

are

very

com

plex

.

Page 18: Harris, R. Carl TITLE The Effects of Type, Identifiability, and … · 2014. 1. 2. · DOCUMENT RESUME ED 078 315 CG 007 111 AUTHOR Harris, R. Carl TITLE The Effects of Type, Identifiability,

31 r

LIO

GR

APH

Y

Ada

ms,

J. A

. Mul

tiple

ver

sus

sing

le p

robl

em tr

aini

ng in

hum

an p

robl

emso

lvin

g. J

ourn

al o

f E

xper

imen

tal P

sych

olog

y, 1

954,

48,

15-

19.

Am

ster

, H. &

Mar

ascu

ilo, L

. Eff

ect o

f ty

pe p

retr

aini

ng a

nd v

arie

ty o

fin

stan

ces

on c

hild

ren'

s co

ncep

t lea

rnin

g. J

ourn

al o

f E

xper

imen

tal

Chi

ld P

sych

olog

y, 1

965,

2, 1

92-2

04.

Arc

her,

E. J

. The

psy

chol

ogic

al n

atur

e of

con

cept

s. I

n H

. J. K

laus

mel

er&

C. W

. Har

ris

(eds

.), A

naly

ses

of c

once

pt le

arni

ng. N

ew Y

ork:

Aca

dem

ic P

ress

, 196

6.

Arc

her,

E. J

., B

ourn

e, L

. E.,

& B

row

n, F

. G. C

once

pt id

entif

icat

ion

asa

func

tion

of ir

rele

vant

info

rmat

ion

and

inst

ruct

ions

.ou

rnal

of

Exp

erim

enta

l Psy

chol

ogy,

195

5. 4

9, 1

53-1

64.

Bou

rne.

L. E

. Lea

rnin

g an

d ut

iliza

tion

of c

once

ptua

l rul

es. I

n T

. R.

Dix

on &

D. L

. Hor

ton

(Eds

.), V

erba

l beh

avio

r an

d ge

nera

l beh

avio

rth

eory

. Eng

lew

ood

Clif

fs, N

. J.:

Pren

tice

Hal

l, 19

68, 2

30-2

44.

Bru

ner,

J.,

Goo

dnow

, 3.,

& A

ustin

, G. A

stu

dy o

f th

inki

ng. N

ew Y

ork:

Wile

y, 1

956.

Bul

gare

lla, R

. G. &

Arc

her,

E. J

. Con

cept

iden

tific

atio

n of

aud

itory

stim

uli a

s a

func

tion

of a

mou

nt o

f re

leva

nt a

nd ir

rele

vant

info

rmat

ion.

Jour

nal o

f E

xper

imen

tal P

sych

olog

y, 1

962.

63,

254

-257

.

Car

roll,

J. B

. Wor

ds, m

eani

ngs,

and

con

cept

s. H

arva

rd E

duca

tiona

lR

evie

w, 1

964,

34,

178

-202

.

DeC

ecco

. J. P

. The

psy

chol

ogy

of le

arni

ng a

nd in

stru

ctio

n: E

duca

tiona

lPs

ycho

logy

. Eng

lew

ood

Clif

fs. N

. J.:

Pren

tice

Hal

l, 19

68.

Duk

es, W

. F. &

Bev

an, W

. Stim

ulus

var

iatio

n an

d re

petit

ion

in th

eac

quis

ition

of

nam

ing

resp

onse

s. J

ourn

al o

f E

xper

imen

tal P

sych

olog

y.19

67, 7

4, 1

78-1

81.

Edl

ing,

J. V

. Ind

ivid

ualiz

ed in

stru

ctio

n: A

man

ual f

or a

dmin

stra

tors

.C

orva

llis,

Ore

gon:

DC

E P

ublic

atio

ns, 1

970.

Gag

ne, R

. M. T

he c

ondi

tions

of

lear

ning

. (Se

cond

Edi

tion)

New

Yor

k:H

olt,

Rin

ehar

t, &

Win

ston

; 197

0.

Gla

ser,

R. C

once

pt le

arni

ng a

nd c

once

pt te

achi

ng.

In R

. M G

egne

W. J

. Gep

hart

(E

ds.)

, Lea

rnin

g re

sear

ch a

nd s

choo

l sub

ject

s.It

asca

, Ill.

:F.

E. P

eaco

ck P

ublis

her,

196

8. p

p. 1

-38.

Hov

land

, C. I

. A "

com

mun

icat

ion

anal

ysis

" of

con

cept

lear

ning

. Psy

cho-

logi

cal R

evie

w, 1

952,

59,

461

-472

.

Hov

land

, C. I

. & W

eiss

, W. T

rans

mis

sion

of

info

rmat

ion

conc

erni

ngco

ncep

ts th

roug

h po

sitiv

e an

d ne

gativ

e in

stan

ces.

ourn

al o

fE

xper

imen

tal P

sych

olog

y, 1

953,

45,

165

-182

.

Lub

orsk

y, L

. F. A

ircr

aft r

ecog

nitio

n: 1

. The

rel

ativ

e ef

fici

ency

of

teac

hing

pro

cedu

res.

Jou

rnal

of

App

lied

Psyc

holo

gy, 1

945,

29,

385-

398.

Mar

kle,

S. M

. & T

iem

ann.

P. W

. Rea

lly u

nder

stan

ding

con

cept

s.C

ham

paig

n, I

ll.: S

tipes

, 196

9.

M-

kle,

S. M

. & T

iem

ann,

P. W

. "B

ehal

aora

l" a

naly

sts

of "

cogn

itive

"co

nten

t. E

duca

tiona

l Tec

hnol

ogy.

Jan

uary

, 197

0, p

p. 4

1-45

.

May

zner

, M. S

. & T

ress

elt,

M. E

. Ver

bal c

once

pt a

ttain

men

t: A

func

tion

of th

e nu

mbe

r an

d st

reng

th o

f po

sitiv

e in

stan

ces

pres

ente

d.T

he J

ourn

al o

f Ps

ycho

logy

. 196

2, 5

3, 4

69-4

74.

Ove

ring

, R. L

. R. &

Tra

vers

, R. W

. M. E

ffec

t upo

n tr

ansf

er o

fva

riat

ions

in tr

aini

ng c

ondi

tions

. Jou

rnal

of

Edu

catio

nal P

sy...

,,olu

gy,

1966

, 57,

170

-188

.

Ove

ring

, R. L

. R. &

Tra

vers

, R. M

. W. V

aria

tion

in th

e am

ount

of

irre

leva

nt c

ues

in tr

aini

ng a

nd te

st c

ondi

tions

and

the

effe

ct u

pon

tran

sfer

. Jou

rnal

of

Edu

catio

narP

scho

logy

, 196

7, 5

8, 6

2-68

.

Pode

ll, H

. A. T

wo

proc

esse

s of

con

cept

for

mat

ion.

Ps)

chol

ogic

r.1

Mon

ogra

phs,

195

8, 7

2. W

hole

No.

468

.

Ree

se, H

. W. D

iscr

imin

atio

n le

arni

ng S

et in

chi

ldre

n.In

L. 1

1: ,L

)vitt

& C

. C. S

pike

r, (

Eds

.), A

dver

ces

in c

hild

dev

elop

men

t and

beh

avio

r.V

ol. 1

. New

Yor

k: A

cade

mic

Pre

ss, 1

964,

pp.

115

-145

.

Shor

e, E

. & S

echr

est,

L. C

once

pt a

ttain

men

t as

a fu

nctio

n of

num

ber

ofpo

sitiv

e in

stan

ces

pres

ente

d. J

ourn

al o

f E

duca

tiona

l Ps}

chol

on,

1961

, 52.

303

-307

.