tervention targets is largely a matter of establishing prio-rities. The target behaviors under investigation rangedfrom single words or phrases in isolated contexts toconversational interactions in varying contexts.

This review focuses on peer-reviewed research ar-ticles published in the past 20 years that evaluated com-munication treatments with children with autism. Thefollowing minimal selection criteria were used to jus-tify reviewing an article:

1. Program descriptions or case studies with noexperimental design were excluded.

2. Only empirical studies that reported resultswith measures of some aspect of languageform, content, or use in individuals with autismwere included.

3. Only studies reporting reliability estimates for thedependent variables under investigation or stud-ies using standardized instruments were included.

Tables I to VI summarize approximately 60 stud-ies that evaluated the effectiveness of child languagetreatment approaches. The columns provide (a) ratingsof internal validity, external validity, and generalization

This review focuses on experimental research ontreatment effectiveness applied to speech and languageskills of individuals with autism. Studies reviewedsought to evaluate whether changes in communicationskills were associated with treatment protocols ratherthan uncontrolled factors, such as maturation or every-day learning opportunities. The positive effects of treat-ment may cover a wide range of communication skillsincluding comprehension, production, and social use oflanguage form (phonology, syntax, morphology), con-tent (semantics), and use (pragmatics).

Appropriate communication depends on the use ofa vast array of linguistic skills, made all the more com-plicated when they must be applied collectively andadapted to interactions with others. Particular inter-ventions may target one of these domains, but their in-tent is usually to influence children’s ability to usecommunication to control, understand, and participatein their social world. It is important to keep in mindthat most children with autism have deficiencies acrossa variety of linguistic domains, and the selection of in-

Communication Intervention for Children with Autism:A Review of Treatment Efficacy

Howard Goldstein1

Empirical studies evaluating speech and language intervention procedures applied to childrenwith autism are reviewed, and the documented benefits are summarized. In particular, inter-ventions incorporating sign language, discrete-trial training, and milieu teaching procedureshave been used successfully to expand the communication repertoires of children with autism.Other important developments in the field stem from interventions designed to replace chal-lenging behaviors and to promote social and scripted interactions. The few studies of the par-ent and classroom training studies that included language measures also are analyzed. Thisarticle seeks to outline the extent to which previous research has helped identify a compendiumof effective instructional practices that can guide clinical practice. It also seeks to highlightneeds for further research to refine and extend current treatment approaches and to investi-gate more comprehensive treatment packages.

KEY WORDS: Autism; communication intervention; language development; social interaction.

Journal of Autism and Developmental Disorders, Vol. 32, No. 5, October 2002 (© 2002)

3730162-3257/02/1000-0373/0 © 2002 Plenum Publishing Corporation

1 The Florida State University, Tallahassee, Florida, 32306-1200.

Tab

le I

.C

omm

unic

atio

n In

terv

entio

ns In

corp

orat

ing

Sign

Lan

guag

e

Parti

cipa

nts

Expe

rimen

tal

Inde

pend

ent

Dep

ende

nt

Gen

eral

izat

ion

Ref

eren

ces

(n)

desi

gnva

riabl

esva

riabl

esm

easu

res

Res

ults

Dur

atio

n

Bra

dy &

Sm

ouse

,6

yr o

ld (1

)Si

mul

tane

ous

Sign

onl

y, s

peec

h R

ec (s

peec

h, s

ign,

and

!ex

perim

ente

rsTC

sig

nific

antly

bet

ter t

han

21 s

essi

ons

1978

treat

men

t on

ly, &

TC

TC) R

s (th

ree

actio

ns !

spee

ch a

lone

I, IV

, II

desi

gnth

ree

colo

rs !

3 ob

ject

s)C

arr &

Dor

es,

mut

e; 6

–11

MB

L !

Ss &

Rec

(TC

) lab

elin

g C

ompr

ehen

sion

of

Not

ass

esse

d2

chn

acq

rec

sign

s M

edia

n tri

als

to19

81yr

s (6

)tra

inin

g se

tsw

ith g

ood

(4)

spee

ch v

ersu

s si

gns

and

4 ch

n ac

quire

d cr

iterio

n "

97–2

86; 6

vers

us p

oor (

2)

rec

sign

s an

d sp

eech

, se

ts o

f 4 o

bjec

tsim

itato

rspr

edic

ted

by v

erba

l II

, III

, IV

imita

tion

scor

esC

arr,

Prid

al, &

6–

16 y

r C

ompa

rison

of

Rec

(sig

n an

d R

ec (s

ign

and

spee

ch)

Not

ass

esse

dN

o di

ffer

ence

in s

ign

Sign

s→!

20–3

00D

ores

, 198

4(1

0)4

good

ver

sus

6 sp

eech

) lab

elin

g in

la

belin

gdi

scrim

inat

ion,

but

big

trial

s to

crit

erio

n; in

poor

imita

tors

two

grou

psdi

ffer

ence

in s

peec

h10

00 tr

ials

mos

tdi

scrim

(not

cor

rela

ted

spee

ch d

iscr

ims

not

with

MA

or l

angu

age

lear

ned;

3 p

airs

II, I

II, I

Vag

eof

obj

ects

Whe

rry

& E

dwar

ds,

5 yr

(1)

Sim

ulta

neou

s Pr

ompt

ing

and

Rec

Rs

to s

peec

h,N

ot a

sses

sed

Poor

acq

uisi

tion

data

3 da

ily s

essi

ons

18

1983

treat

men

t, di

ffer

entia

lsi

gns,

& T

C) (

9an

d no

diff

eren

ces

!da

ys in

terv

entio

nco

unte

r bal

ance

d re

info

rcem

ent

verb

s !

9 co

lors

!9

cond

ition

stx

s an

d te

ache

rsno

uns)

; als

oin

itiat

ions

and

eye

I, IV

, IV

cont

act

Bar

rera

, Lob

ato-

mut

e; 4

yr

Alte

rnat

ing

Sign

onl

y, s

peec

hEx

pres

sive

labe

ling

!pe

rson

s &

TC

→su

perio

r lea

rnin

g 26

–104

tria

ls to

Bar

rera

, Sul

zer-

old

(1)

treat

men

ts

only

, and

TC

(tot

alan

d le

vel o

fst

imul

ire

plic

ated

!gr

oups

of

crite

rion

for 1

0 TC

Aza

roff

, 198

0de

sign

com

mun

icat

ion)

prom

ptin

gw

ords

wor

ds, 2

1 w

ords

I, IV

, II

lear

ned

in 1

2 se

ssio

nsB

arre

ra &

Sul

zer-

echo

lalic

;A

ltern

atin

gSp

eech

onl

y ve

rsus

Ex

pres

sive

labe

ling

Not

ass

esse

dTC

→fa

ster

and

mor

e 52

–88

trial

s to

Aza

roff

, 198

36–

9 yr

(3)

treat

men

tsTC

and

leve

l of

com

plet

e le

arni

ng

crite

rion

in T

Cde

sign

prom

ptin

gre

plic

ated

!gr

oups

of

cond

ition

and

65–

82w

ords

& c

hild

ren

in s

peec

h co

nditi

on;

6–10

wor

ds le

arne

dI,

II, I

Vin

16–

25 s

essi

ons

Layt

on 1

988

6–9

yr,

Ran

dom

Spee

ch a

lone

ver

sus

Exp

and

rec

!se

tting

s &

All

txs

effe

ctiv

e fo

r 90

40-

min

. dai

lym

od-s

evas

sign

men

t to

sign

alo

ne v

ersu

s(s

igns

/wor

ds)

pers

ons;

3 m

osgo

od im

itato

rs, s

peec

h se

ssio

ns; R

ec M

"au

tism

(60)

four

mat

ched

alte

rnat

ing

vers

usfu

nctio

nal v

ocab

ular

ym

aint

enan

ceal

one,

less

eff

ectiv

e fo

r 11

–31

wor

ds/s

igns

;gr

oups

(goo

dsi

mul

tane

ous

txs

poor

imita

tors

Exp

M "

3–30

and

poor

w

ords

/sig

nsI,

I, II

imita

tors

)

Yod

er &

Lay

ton,

M

in v

erba

l;M

ultig

roup

Spee

ch a

lone

ver

sus

Tota

l num

ber o

f chi

ld-

!se

tting

sSi

gn a

lone

→le

ss c

hild

-90

40-

min

. dai

ly19

883–

7 yr

old

sra

ndom

ass

ign;

sign

alo

ne v

ersu

sin

itiat

ed s

poke

nin

itiat

ed s

peec

h; v

erba

lse

ssio

ns(6

0)ap

titud

e !

tx

alte

rnat

ing

vers

usw

ords

abili

ty p

redi

cted

pos

t-tx

inte

ract

ion

sim

ulta

neou

s tx

svo

cabu

lary

I, I,

III

anal

ysis

Cas

ey, 1

978

6 an

d 7

yr,

MB

L !

SsPa

rent

s ta

ught

to%

inte

rval

s of

com

mu-

!se

tting

sC

lear

incr

ease

s in

20 s

essi

ons

(4)

use

sign

sni

catio

n at

tem

pts

and

com

mun

icat

ion

&in

appr

opria

te b

ehav

ior

decr

ease

s in

inap

prop

riate

beha

vior

for 2

of

II, I

II, I

I4

child

ren

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n, re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p or

sin

gle-

subj

ect d

esig

n ex

perim

ent c

ompa

ring

alte

rnat

ive

treat

men

ts; I

I, pr

e-po

st g

roup

des

ign

or s

ingl

e-su

bjec

t des

ign

with

relia

ble,

obj

ectiv

e ou

tcom

es m

easu

res;

III,

pre-

post

, ex

post

fact

o, o

r sin

gle-

subj

ect d

esig

ns w

ith q

uest

iona

ble

inte

rnal

val

idity

or m

easu

rem

ent.

Exte

rnal

val

idity

: I, R

ando

m a

ssig

nmen

t of w

ell-d

efin

ed c

ohor

ts a

nd a

dequ

ate

sam

ple

size

; II,

nonr

ando

m a

ssig

nmen

t, bu

t wel

l-def

ined

coh

orts

and

ade

quat

e sa

mpl

e si

ze o

r rep

licat

ion

acro

ss 3

or m

ore

sub-

ject

s in

a s

ingl

e-su

bjec

t des

ign;

III

, wel

l-def

ined

pop

ulat

ion

of th

ree

or m

ore

subj

ects

in s

ingl

e-su

bjec

t des

igns

or

sam

ple

of a

dequ

ate

size

in g

roup

des

igns

; IV

, sm

all s

ampl

e in

gro

up d

esig

n or

one

or

two

subj

ects

in s

ingl

e-su

bjec

t des

ign.

Gen

eral

izat

ion:

I, C

onsi

sten

t gen

eral

izat

ion

to f

unct

iona

l set

ting

beyo

nd tr

eatm

ent s

ettin

g w

ithou

t fur

ther

inte

rven

tion

or r

obus

t soc

ial v

alid

atio

n re

sults

; II,

long

-term

mai

nten

ance

or

gene

raliz

atio

n to

one

setti

ng b

eyon

d tre

atm

ent s

ettin

g; II

I, in

terv

entio

n oc

curr

ed in

nat

ural

set

ting

or o

utco

me

mea

sure

with

doc

umen

t rel

atio

n to

func

tiona

l out

com

es; I

V, n

ot a

ddre

ssed

or o

ther

.M

BL

= M

ultip

le B

asel

ine

Des

ign;

TC

= T

otal

Com

mun

icat

ion;

Rec

= R

ecep

tive;

Exp

= E

xpre

ssiv

e; R

= R

espo

nse;

chn

= c

hild

ren;

Tx

= tre

atm

ent.

Tab

le I

I.D

iscr

ete

Tria

l Tra

inin

g

Parti

cipa

nts

Expe

rimen

tal

Inde

pend

ent

Dep

ende

nt

Gen

eral

izat

ion

Ref

eren

ces

(n)

desi

gnva

riabl

esva

riabl

esm

easu

res

Res

ults

Dur

atio

n

Koe

gel,

O’D

ell,

&

Dun

lap,

198

8

II, I

I, II

McI

lvan

e et

al.,

1984

II, I

V, I

V

Han

dlem

an, 1

979

II, I

I, I

Han

dlem

an, 1

981

II, I

I, II

I

Seca

n, E

gel,

&

Tille

y, 1

989

II, I

I, I

Egel

, Sha

fer,

&

Nee

f, 19

84

II, I

I, II

Buf

fingt

on, K

rant

z,

McC

lann

ahan

, &

Poul

son,

199

8

II, I

I, I

seve

re c

omm

u-ni

catio

n de

lay;

3–

11 y

r (4)

18 y

r (1)

6–7

yr (4

)

5–12

yr (

6)

5–9

yr (4

)

6–8

yr (4

)

4–6

yr (4

)

Rev

ersa

l des

ign

Mul

tiple

pro

be

desi

gn !

Rs

MB

L !

Q s

ets,

BC

BC

BC

-re

stric

ted

vers

us

mul

tiple

nat

ural

se

tting

s

MB

L !

Q s

ets,

BC

BC

BC

-re

stric

ted

vers

us

mul

tiple

nat

ural

se

tting

s

MB

L !

Q fo

rms

and

subf

orm

s (e

.g.,

how

Qs

rela

ting

to

actio

n, m

eans

, af

fect

) and

Ss

ATD

(obj

ect v

ersu

s se

lf po

sitio

ning

) an

d M

BL

!Ss

MB

L !

Rs

Rei

nfor

ce v

erba

l atte

mpt

sve

rsus

sha

ping

mot

orsp

eech

pro

duct

ion

Excl

usio

n pr

oced

ure

to

teac

h fo

od n

ames

Diff

eren

tial r

einf

orce

men

tan

d co

rrec

tion

proc

e-du

re

Diff

eren

tial r

einf

orce

men

tan

d co

rrec

tion

proc

e-du

re

Pict

ure

train

ing

usin

g di

f-fe

rent

ial r

einf

orce

men

tan

d m

odel

ing

corr

ectio

npr

oced

ure

Dis

cret

e-tri

al tr

aini

ng w

ithdi

ffer

entia

l rei

nfor

ce-

men

t pro

mpt

ing,

and

m

odel

ing

Mod

elin

g, p

rom

ptin

g, to

ken

rein

forc

emen

t, an

d pr

aise

Aff

ect r

atin

gs a

nd

chan

ges

in

phon

emic

pr

oduc

tion

scor

es

Exp

spee

ch a

nd s

ign

voca

bula

ry

1 w

ord

Rs

to w

h-Q

s

1 w

ord

Rs

to w

h-Q

s

Rs

to w

hat,

how

, and

w

hy-Q

s to

pic

ture

s

Com

preh

ensi

on o

f pr

epos

ition

s—be

hind

, in

fron

t of,

next

to,

Ges

tura

l and

ver

bal

Rs

(atte

ntio

n, a

ffec

t, an

d re

fere

nce)

4 ye

ar fo

llow

-up

for 3

chn

!no

vel w

ords

!se

tting

s an

d ad

ults

(i.e

., ho

me)

!se

tting

s (i.

e.,

corn

er o

f cl

assr

oom

)

!st

oryb

ook

and

natu

ral

cont

exts

and

to

Q-a

skin

g

!no

vel s

timul

ian

d R

topo

gra-

phie

s

!pr

obe

stim

ulus

an

d to

cla

ssro

oman

d so

cial

val

id-

ity ra

tings

of

vide

os

Slig

htly

bet

ter a

ffec

t rat

ings

in v

erba

l atte

mpt

con

di-

tion;

eff

ects

on

spee

ch

prod

uctio

n di

fficu

lt to

in

terp

ret

2–10

rec

mat

chin

g tri

als

befo

re b

egan

nam

ing

food

s

Trai

ning

in m

ultip

le

setti

ngs →

grea

ter

gene

raliz

atio

n to

hom

e fo

r 3 o

f 4 c

hn a

nd e

qual

lyhi

gh g

en fo

r 4th

chi

ld

No

cons

iste

nt d

iffer

ence

inle

arni

ng a

nd g

ener

ally

as

soci

ated

with

2 tr

aini

ngco

nditi

ons

Rap

id le

arni

ng w

ith p

ix

refe

rent

s, n

o ge

n !

Q

subf

orm

s an

d fo

rms;

fair

to g

ood

gen

!co

ntex

ts,

2–5

sess

ions

inte

rspe

rsed

train

ing

stim

uli r

emed

i-at

ed p

oor g

en; a

nd n

o Q

aski

ng

Posi

tioni

ng o

bjec

ts d

id n

otge

n to

pos

ition

ing

self

and

vice

ver

sa w

ithou

t mul

ti-pl

e ex

empl

ar o

r int

ersp

er-

sal t

rain

ing

Zero

bas

elin

e →

robu

st

effe

cts

repl

icat

ed a

cros

s R

cat

egor

ies

and

chn,

ge

n pr

obe

data

par

alle

led

train

ing

data

. Cla

ssro

omda

ta n

ot p

rese

nted

. Goo

dso

cial

val

idity

resu

lts.

14–2

4 se

ssio

ns

10 fo

od n

ames

3 se

ssio

ns p

er

day,

20

train

ing

trial

s→10

0–17

40

trial

s to

crit

erio

n pe

r set

of 5

Rs

3 se

ssio

ns p

er

day,

20

train

ing

trial

s→40

–156

0 tri

als

to c

riter

ion

per s

et o

f 5 R

s

3–9

train

ing

sess

ions

pe

r Q fo

rm

3–12

trai

ning

se

ssio

ns p

er S

fo

r 2 p

repo

sitio

ns

57–6

9 se

ssio

ns

Koe

gel,

Cam

arat

a,

Val

dez-

Men

chac

a,

& K

oege

l, 19

98II

, II,

I

Kra

ntz,

Zal

ensk

i, H

all,

Fens

ke, &

M

cCla

nnah

an,

1981

—Ex

p 1

II, I

I, I

Kra

ntz

etal

.,19

81—

Exp

2II

, II,

I

Kra

ntz

etal

.,19

81—

Exp

3

II, I

V, I

I

Gol

dste

in &

Bro

wn,

19

89

II, I

V, I

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p or

sin

gle-

subj

ect d

esig

n ex

perim

ent c

ompa

ring

alte

rnat

ive

treat

men

ts; I

I, pr

e-po

st g

roup

des

ign

or s

ingl

e-su

bjec

t des

ign

with

relia

ble,

obj

ectiv

e ou

tcom

es m

easu

res;

III,

pre-

post

, ex

post

fact

o, o

r sin

gle-

subj

ect d

esig

ns w

ith q

uest

iona

ble

inte

rnal

val

idity

or m

easu

rem

ent.

Exte

rnal

val

idity

: I, R

ando

m a

ssig

nmen

t of w

ell-d

efin

ed c

ohor

ts a

nd a

dequ

ate

sam

ple

size

; II,

nonr

ando

m a

ssig

nmen

t, bu

t wel

l-def

ined

coh

orts

and

ade

quat

e sa

mpl

e si

ze o

r rep

licat

ion

acro

ss 3

or m

ore

subj

ects

in a

sin

gle-

subj

ect d

esig

n; II

I, w

ell-d

efin

ed p

opul

atio

n of

thre

e or

mor

e su

bjec

ts in

sin

gle-

subj

ect d

esig

ns o

r sam

ple

of a

dequ

ate

size

in g

roup

des

igns

; IV

, sm

all s

ampl

e in

gro

up d

esig

n or

1to

2 s

ubje

cts

in s

ingl

e-su

bjec

t des

ign.

Gen

eral

izat

ion:

I, C

onsi

sten

t gen

eral

izat

ion

to f

unct

iona

l set

ting

beyo

nd tr

eatm

ent s

ettin

g w

ithou

t fur

ther

inte

rven

tion

or r

obus

t soc

ial v

alid

atio

n re

sults

; II,

long

-term

mai

nten

ance

or

gene

raliz

atio

n to

one

setti

ng b

eyon

d tre

atm

ent s

ettin

g; II

I, in

terv

entio

n oc

curr

ed in

nat

ural

set

ting

or o

utco

me

mea

sure

with

doc

umen

t rel

atio

n to

func

tiona

l out

com

es; I

V, n

ot a

ddre

ssed

or o

ther

.M

BL

= M

ultip

le B

asel

ine

Des

ign;

TC

= T

otal

Com

mun

icat

ion;

Rec

= R

ecep

tive;

Exp

= E

xpre

ssiv

e; R

= R

espo

nse;

chn

= c

hild

ren;

Tx

= tre

atm

ent.

!st

imul

i, se

tting

s,an

d pe

ople

!pe

rson

s,

setti

ngs,

nov

el

mat

eria

ls, a

ndm

odal

ity

(writ

ing)

!un

train

ed s

timul

i

!se

tting

s

!un

train

ed s

timu-

lus

com

bina

tions

and

setti

ngs

Qs

lear

ned,

use

d w

ith le

sspr

efer

red

item

s an

d ge

ner-

aliz

ed. A

vera

ge 6

new

exp

noun

labe

ls le

arne

d/w

k

Chn

qui

ckly

lear

ned

and

gen

whe

n ta

ught

incr

easi

ngly

long

er u

ttera

nces

Chi

ldre

n qu

ickl

y le

arne

d an

dge

nera

lized

Zero

bas

elin

e→ro

bust

ef-

fect

s fo

r sen

tenc

es a

ndcl

ear e

ffec

ts fo

r par

a-gr

aphs

64 re

c an

d 64

exp

3-w

ord

Rs

lear

ned

thro

ugh

peer

mod

-el

ing

of 3

rec

#5

exp

Rs

for o

ne c

hild

and

4 re

c #

2 ex

p R

s fo

r sec

ond

child

and

gen

to h

ome

16–3

4 se

ssio

ns

Tota

l !10

0 da

ys

2–3

mo,

36–

54

sess

ions

5–8

mo

13 s

essi

ons

3–5

yr (3

)

7–13

yr (

4)

5–10

yr (

3)

5 yr

(2)

2–3

yr (2

)

MB

L !

Ss

MB

L !

lang

uage

ca

tego

ries

MB

L !

wh-

conc

epts

MB

L !

setti

ngs

(hom

e an

d sc

hool

) #ta

sks

(sen

tenc

es a

nd

para

grap

hs)

MB

L !

Rs

(sub

mat

rices

) an

d m

odal

ities

Mod

elin

g, p

rom

ptin

g im

ita

tion,

rein

forc

ed w

ith

desi

red

item

s in

a b

ag

Toke

n re

info

rcem

ent a

ndm

odel

ing

in d

iscr

ete-

trial

form

at

Ver

bal p

rom

ptin

g/fa

ding

an

d di

ffer

entia

l rei

n-fo

rcem

ent

Mod

elin

g, re

hear

sal,

prom

ptin

g, a

nd fa

ding

Mod

elin

g of

item

s se

lect

edfr

om o

bj-p

rep-

loca

tion

subm

atric

es

Num

ber o

f “w

hat’s

that

” Q

s as

ked,

new

nou

ns

prod

uced

Des

crip

tions

of o

bjec

ts

and

pict

ures

Ver

b #

Col

or #

Shap

e/si

ze #

Labe

l

Rs

to w

h-Q

s (4

–5

conc

epts

eac

h fo

r wha

t,ho

w, a

nd w

hy Q

s

Rs

to w

h-Q

s ab

out p

ast

even

ts

Rec

Rs

to a

nd

prod

uctio

n of

ob

j-pre

p-lo

catio

n ut

tere

nces

Tab

le I

II.

Mili

eu T

each

ing

Parti

cipa

nts

Expe

rimen

tal

Inde

pend

ent

Dep

ende

nt

Gen

eral

izat

ion

Ref

eren

ces

(n)

desi

gnva

riabl

esva

riabl

esm

easu

res

Res

ults

Dur

atio

n

Cha

rlop,

Sc

hrei

bman

, &

Thib

odea

u, 1

985

II, I

I, II

Cha

rlop

& W

alsh

, 19

86

II, I

I, II

Cha

rlop

&

Tras

owec

h, 1

991

II, I

I, I

Inge

nmey

& V

an

Hou

ten,

199

1

II, I

V, I

I

Mat

son,

Sev

in, B

ox,

Fran

cis,

& S

evin

, 19

93

II, I

I, II

McG

ee, A

lmei

da,

Sulz

er-A

zaro

ff, &

Fe

ldm

an, 1

992

II, I

I, II

Nee

f, W

alte

rs, &

Eg

el, 1

984

II, I

I, II

McG

ee, K

rant

z, &

M

cCla

nnah

an, 1

985

II, I

I, II

MB

L !

Ss

MB

L !

Ss

and

setti

ngs

MB

L !

Ss a

nd s

et-

tings

MB

L !

beha

v-io

rs

MB

L !

beha

v-io

rs

MB

L !

Ss

MB

L !

Ss

MB

L !

pairs

of

prep

s w

ith

ATD

and

MB

L!

Ss

Tim

e-de

lay

proc

edur

e

Tim

e de

lay

proc

edur

e an

d pe

er

mod

elin

g in

2nd

se

tting

Pare

nts

taug

ht ti

me-

dela

y pr

oced

ure

Tim

e-de

lay

proc

edur

e (m

odel

de

scrip

tions

of c

ar

play

and

dra

win

g)

Two

Rs

each

taug

ht

usin

g tim

e de

lay

vers

us m

odel

ing

and

visu

al c

ue fa

ding

pr

oced

ures

Peer

inci

dent

al

teac

hing

(wai

t for

R

Q in

itiat

ion,

RQ

im

it, p

rovi

de

rein

forc

er a

nd p

rais

e)

Dis

cret

e-tri

al

tuto

ring

sess

ions

(“

is th

is a

___

?”)

vers

us a

cros

s-th

e-da

ych

ild R

Q-in

itiat

ed

embe

dded

tx

(“do

you

wan

t ___

?”)

Trad

ition

trai

ning

(W

here

is th

e __

_?,

prom

pt, a

rbitr

ary

rein

-fo

rcer

) ver

sus

Inci

den-

tal t

rain

ing

(Whe

re is

___?

, Tel

l me

whe

re,

prom

pt e

labo

ratio

n,

acce

ss to

RQ

ed it

em

Obj

ect R

Qs

(fou

r per

child

); al

l but

one

ch

ild c

ould

labe

l ob

ject

s

Spon

tane

ous

exp

of

“I li

ke/lo

ve y

ou”;

ey

e co

ntac

t, sm

iling

, ap

proa

ch/to

uch

Soci

al p

hras

es to

ad

ults

% o

f tria

ls w

ith s

pont

a-ne

ous

desc

riptio

ns o

fpl

ay

Soci

al p

hras

es—

hello

, pl

ay w

ith m

e, e

xcus

em

e, th

ank

you,

hel

p m

e

% in

terv

als

of

posi

tive

reci

proc

al

inte

ract

ions

, %

inte

rval

s of

in

itiat

ions

% c

orre

ct y

es/n

o R

s

Prep

pro

duct

ion

to

desc

ribe

loca

tion

of

edib

les

and

toys

!se

tting

s,

unfa

mili

ar

pers

ons,

unt

rain

edst

imul

i

!se

tting

s an

d re

spon

se g

en

!lo

catio

ns a

nd

pers

ons

!se

tting

s, 5

wk

and

4 m

o fo

llow

up

!pe

ople

and

se

tting

s

!se

tting

s; te

ache

r ra

tings

of s

ocia

l co

mpe

tenc

e; p

eer

soci

omet

ric

ratin

gs

!ne

w s

timul

us

item

s an

d Q

type

s (a

ctio

ns,

poss

essi

ons

and

spat

ial r

elat

ions

)

!te

ache

rs s

ettin

gs, a

ndst

imul

us a

rran

gem

ents

Chi

ldre

n (w

ho c

ould

labe

l)qu

ickl

y le

arne

d an

d ge

nera

lized

RQ

ing

once

time

dela

y im

plem

ente

d

Tim

e de

lay

and

not p

eer

mod

elin

g ef

fect

ive,

va

riabl

e ge

n !

setti

ngs

(unt

il tim

e de

lay

impl

emen

ted)

and

littl

e ge

n to

anc

illar

y be

havi

ors

Tim

e de

lay →

spon

phr

ase

use

in 4

of 6

, 4 o

f 7, a

nd 2

of 6

con

text

s, re

spec

tivel

y;ge

n or

loss

of c

ontro

l for

othe

r con

text

s; g

ood

mai

nt.

Tim

e de

lay →

imm

edia

te

incr

ease

in s

pont

aneo

ussp

eech

, gen

to u

ntra

ined

ac

tions

w/in

5 s

essi

ons

Bot

h tre

atm

ents

eff

ectiv

e fo

r inc

reas

ing

self-

initi

ated

Rs

to n

onve

rbal

cue

s (e

.g.,

ther

apis

t ent

erin

g ro

om,

pres

enta

tion

of a

gam

e)

Incr

ease

in re

cipr

ocal

in

tera

ctio

ns a

nd le

ss

teac

her i

nvol

vem

ent;

spec

ific

effe

cts

on s

pon

verb

al R

Qs

not c

lear

Embe

dded

inst

ruct

ion

asso

ci-

ated

with

rapi

d im

prov

e-m

ent i

n R

ing;

gen

to “

is

this

a _

__?”

Qs

only

afte

rin

trodu

ced

in e

mbe

dded

inst

ruct

ion,

goo

d ge

n !

Q ty

pes

Bot

h tx

s eq

ually

eff

ectiv

e fo

rac

q an

d re

tent

ion,

inci

den-

tal t

each

ing →

bette

r, bu

tin

com

plet

e ge

n !

setti

ngs

(less

gen

!ar

rang

emen

ts)

and

mor

e sp

onta

neou

spr

oduc

tions

3–6

bloc

ks o

f 10

train

ing

trial

s

!45

–60

days

25–3

4 m

o of

obs

er-

vatio

n an

d 1–

4 m

os o

f tra

inin

g

22 s

essi

ons

33–4

7 se

ssio

ns, 2

–16

sess

ions

to le

arn

each

resp

onse

17–3

8 se

ssio

ns

38–6

2 se

ssio

ns, t

x ef

fect

est

ablis

hed

in 1

–7 d

ays

90 s

essi

ons,

!30

se

ssio

ns o

f tra

in-

ing

to a

ppro

xim

ate

mas

tery

5–11

yr (

7)

6–8

(4)

Echo

lalic

; 7–

8 yr

(3)

10 y

r (1)

4–5

yr, e

chol

alic

an

d lim

ited

Rs

(3)

3–5

yr (3

)

4–6

yr (4

)

6, 8

, 11

yr (3

)

Koe

gel,

O’D

ell,

&

Koe

gel,

1987

III,

IV, I

II

Hw

ang

& H

ughe

s,20

00

II, I

I, I

Koe

gel,

Cam

arat

a,

Koe

gel,

Ben

-Tal

l, &

Sm

ith, 1

998

I, II

, I

Lask

i, C

harlo

p, &

Sc

hrei

bman

, 198

8

II, I

I, I

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p or

sin

gle-

subj

ect d

esig

n ex

perim

ent c

ompa

ring

alte

rnat

ive

treat

men

ts; I

I, pr

e-po

st g

roup

des

ign

or s

ingl

e-su

bjec

t des

ign

with

rel

iabl

e, o

bjec

tive

outc

omes

mea

-su

res;

III,

pre-

post

, ex

post

fact

o, o

r sin

gle-

subj

ect d

esig

ns w

ith q

uest

iona

ble

inte

rnal

val

idity

or m

easu

rem

ent.

Exte

rnal

val

idity

: I, R

ando

m a

ssig

nmen

t of w

ell-d

efin

e co

horts

and

ade

quat

e sa

mpl

e si

ze; I

I, no

nran

dom

ass

ignm

ent,

but w

ell-d

efin

ed c

ohor

ts a

nd a

dequ

ate

sam

ple

size

or r

eplic

atio

n ac

ross

3 o

r mor

esu

bjec

ts in

a s

ingl

e-su

bjec

t des

ign;

III,

wel

l def

ined

pop

ulat

ion

of th

ree

or m

ore

subj

ects

in s

ingl

e-su

bjec

t des

igns

or s

ampl

e of

ade

quat

e si

ze in

gro

up d

esig

ns; I

V, s

mal

l sam

ple

in g

roup

des

ign

or 1

or

2 su

bjec

ts in

sin

gle-

subj

ect d

esig

n.G

ener

aliz

atio

n: I

, Con

sist

ent g

ener

aliz

atio

n to

fun

ctio

nal s

ettin

g be

yond

trea

tmen

t set

ting

with

out f

urth

er in

terv

entio

n or

rob

ust s

ocia

l val

idat

ion

resu

lts; I

I, lo

ng-te

rm m

aint

enan

ce o

r ge

nera

lizat

ion

toon

e se

tting

bey

ond

treat

men

t set

ting;

III,

inte

rven

tion

occu

rred

in n

atur

al s

ettin

g or

out

com

e m

easu

re w

ith d

ocum

ent r

elat

ion

to fu

nctio

nal o

utco

mes

; IV

, not

add

ress

ed o

r oth

er.

MB

L =

Mul

tiple

Bas

elin

e D

esig

n; A

TD =

Alte

rnat

ing

Trea

tmen

ts D

esig

n; R

Q =

Req

uest

; Q =

Que

stio

n; R

= R

espo

nse;

Tx

= tre

atm

ent;

gen

= ge

nera

lizat

ion;

Rec

= R

ecep

tive;

Exp

= E

xpre

ssiv

e; c

hn =

child

ren.

MB

L !

Ss

MB

L !

Ss

Rev

ersa

l des

ign

AB

AC

AB

with

coun

terb

al-

ance

d or

der

MB

L !

Ss

Chi

ld-s

elec

ted

stim

uli,

mod

elin

g (w

/o im

itatio

n R

Q),

com

mun

icat

ive

atte

mpt

s re

info

rced

, with

acce

ss t

o to

ys a

nd

soci

ally

Soci

al in

tera

ctiv

e tra

inin

gw

ith a

dult

partn

er

(con

tinge

nt im

itatio

n,

envi

ronm

ent a

rran

ge-

men

t, ex

pect

ant l

ook,

natu

ral r

einf

orce

rs)

Ana

log

(dis

cret

e-tri

al)

vers

us n

atur

alis

tic

teac

hing

pro

cedu

res

Pare

nts

taug

ht N

atur

al

Lang

uage

Par

adig

m

(rei

nfor

ce a

ttem

pts,

tu

rn ta

king

, var

y ta

sk,

shar

e co

ntro

l)

Imita

tive,

def

erre

d im

itativ

e, a

nd s

pon-

tane

ous

wor

d pr

oduc

tion

Freq

of c

hild

eye

co

ntac

t, jo

int a

tten-

tion,

and

mot

or

imita

tion

Acc

urac

y of

targ

etph

onem

e pr

oduc

tion

and

inte

lligi

bilit

y ra

t-in

gs b

y na

ïve

judg

es

Freq

uenc

y of

par

ent

and

child

ver

baliz

a-tio

ns

!se

tting

s

!se

tting

(nov

el c

lass

-ro

om w

ith te

ache

r) a

ndso

cial

val

idity

ratin

gsof

vid

eota

pes

!se

tting

s (h

ome

with

pare

nts,

sch

ool w

ithpe

ers)

!3

setti

ngs

Follo

win

g 2

and

19 m

o of

ana

-lo

g te

achi

ng, N

atur

al L

ang

Para

digm

impl

emen

ted →

incr

ease

in im

med

imita

tion

and

even

tual

ly s

pont

aneo

usut

tera

nces

(but

afte

r 4 o

r 5

mo)

Incr

ease

s in

thre

e ta

rget

be-

havi

ors

repl

icat

ed a

s tra

in-

ing

was

initi

ated

acr

oss

Ss.

Goo

d ge

nera

lizat

ion,

esp

e-ci

ally

for e

ye c

onta

ct, b

utm

inim

al fo

r joi

nt a

ttent

ion.

Rob

ust e

ffec

ts o

n so

cial

va-

lidity

msr

es

Cor

rect

pro

duct

ion

in c

onve

r-sa

tion

lang

uage

sam

ples

muc

h gr

eate

r for

pho

nem

esin

nat

ural

istic

tx c

ondi

tion

repl

icat

ed w

ithin

and

ac

ross

chi

ldre

n pl

us g

en !

setti

ngs

Cle

ar e

xper

imen

tal e

ffec

ts fo

r7

of 8

fam

ilies

2 h/

2 !

wee

k/ov

er

30 m

o

58 o

bser

vatio

n se

ssio

ns

54–9

4 se

ssio

ns

13–2

0 se

ssio

ns; 5

–9

pare

nt tr

aini

ngse

ssio

ns

4 an

d 5

yr (2

)

Prev

erba

l 32

–43

mo

(3)

3–7

yr (5

)

4 m

ute

and

4ec

hola

lic

5–9

yr (8

)

Tab

le I

V.

Com

mun

icat

ion

Inte

rven

tions

to R

epla

ce C

halle

ngin

g B

ehav

ior

Parti

cipa

nts

Expe

rimen

tal

Inde

pend

ent

Dep

ende

nt

Gen

eral

izat

ion

Ref

eren

ces

(n)

desi

gnva

riabl

esva

riabl

esm

easu

res

Res

ults

Dur

atio

n

Car

r & D

uran

d,

1985

—Ex

perim

ent 1

II, I

I, IV

Car

r & D

uran

d,

1985

—Ex

perim

ent 2

II, I

I, IV

Dur

and

& C

arr,

1987

—Ex

perim

ent 1

II, I

I, IV

Dur

and

& C

arr,

1987

—Ex

perim

ent 2

II, I

I, IV

Dur

and

& C

arr,

1987

—Ex

perim

ent 3

II, I

I, IV

Dur

and

& C

arr,

1992

—Ex

perim

ent 1

II, I

I, IV

Dur

and

& C

arr,

1992

—Ex

perim

ent 2

&

3

II, I

I, II

7–14

yr (

4)

7–14

yr (

4)

7–13

yr (

4)

7–13

yr (

4)

7–13

yr (

4)

3–5

yr (1

2)

3–5

yr (1

2)

AB

AC

AC

AB

A

AB

AC

AC

AB

AC

AB

AC

AC

AB

A

AB

AB

AB

AB

A

MB

L !

Ss

AB

AC

AC

AB

A

(bas

elin

e at

tent

ion

and

dem

and

cond

ition

s)

AB

AB

AC

; 2

mat

ched

gro

ups

(Fun

ctio

nal

Com

mun

icat

ion

Trai

ning

ve

rsus

tim

e-ou

t)

Easy

task

s w

ith 1

00%

at

tent

ion;

Eas

y w

ith

33%

atte

ntio

n;

Diffi

cult

task

s w

ith

100%

atte

ntio

n

Rel

evan

t and

irr

elev

ant R

s (I

don

’t un

ders

tand

and

Am I

doin

g go

od w

ork?

)

Soci

al a

ttent

ion

and

task

dem

ands

Tim

e-ou

t (re

mov

ing

task

dem

ands

) co

ntin

gent

on

ster

eoty

py

Imita

tion,

tim

e de

lay,

cor

rect

ion

proc

edur

e to

teac

h “h

elp

me”

B-A

ttent

ion

redu

ced

from

100

% to

33%

of

mat

chin

g tri

als;

C

-incr

ease

diffi

culty

of

task

s to

33%

cor

rect

le

vel &

CR

F

FCT-

imita

tion,

tim

e de

lay

train

ing

of “

Am

Ido

ing

good

wor

k?”

(in M

"18

min

) TO

-sim

ilar t

rain

ing

of

“My

nam

e is

___

” (1

8 m

in)

% in

terv

als

of

prob

lem

beh

avio

r (a

ggre

ssio

n,

tant

rum

s, s

elf-

inju

ry)

% in

terv

als

of

prob

lem

beh

avio

r an

d re

leva

nt a

nd

irrel

evan

t Rs

% in

terv

als

of

ster

eoty

pic

beha

vior

s

% in

terv

als

of

ster

eoty

pic

beha

vior

s(h

and

flapp

ing,

bod

yro

ckin

g)

% in

terv

als

of “

Hel

p m

e” R

Qs

and

ster

eoty

pic

beha

vior

s

% in

terv

als

of

chal

leng

ing

beha

vior

s,co

rrec

t mat

chin

g R

s,Tr

aine

r beh

avio

rs

% in

terv

als

of

chal

leng

ing

beha

vior

an

d un

prom

pted

at

tent

ion

RQ

s (A

m I

doin

g go

od w

ork?

)

!tra

iner

s (A

BA

B in

Ex

perim

ent 3

)

Littl

e ad

ult a

ttent

ion

and

high

task

diffi

culty

→in

crea

sed

prob

lem

beh

avio

rs fo

r di

ffer

ent S

s

Com

mun

icat

ive

repl

acem

ent

beha

vior

s fo

r sam

e fu

nc-

tion

only

(sol

iciti

ng

atte

ntio

n or

hel

p) →

less

prob

lem

beh

avio

r

Diffi

cult

task

→in

crea

sed

ster

eoty

py

Con

tinge

nt re

mov

al o

f tas

k de

man

ds →

incr

ease

dst

ereo

typy

Teac

hing

chi

ldre

n to

RQ

as

sist

ance

→re

duce

d st

ereo

typy

Red

uced

adu

lt at

tent

ion

was

asso

ciat

ed w

ith in

crea

sed

prob

lem

beh

avio

r in

all

12 c

hild

ren

Bot

h tre

atm

ents

suc

cess

ful,

but o

nly

grou

p th

at

rece

ived

FC

T ac

ross

se

tting

s m

aint

aine

d lo

w

rate

s of

pro

blem

beh

avio

rsw

naï

ve te

ache

rs

31 s

essi

ons

20–4

0 se

ssio

ns

28 s

essi

ons

23–3

6 se

ssio

ns

25 s

essi

ons

27 s

essi

ons

Expe

rimen

t 2—

27se

ssio

ns

Expe

rimen

t 3—

18

sess

ions

Hor

ner &

Bud

d,

1985

II, I

V, I

I

Bird

, Dor

es, M

oniz

, &

Rob

inso

n, 1

989

II, I

V, I

Wac

ker,

Stee

ge,

Nor

thup

, Sas

so,

Ber

g, R

eim

ers,

C

oope

r, C

igra

nd,

& D

onn,

199

0

II, I

I, II

I

Har

ing

& K

enne

dy,

1990

II, I

V, I

II

Schr

eibm

an &

Car

r, 19

78

II, I

V, I

I

McM

orro

w &

Fox

x,

1986

—Ex

perim

ent 1

II, I

V, I

I

McM

orro

w &

Fox

x,

11 y

r (1)

27 a

nd 3

6 yr

(2

)

7 (w

aut

.), 9

, 30

yr (

3)

15 y

r-au

tism

; 19

yr D

own

synd

rom

e (2

)

7 an

d 15

MR

(2

)

21 y

r (1)

MB

L !

beha

vior

s

!M

BL

!Ss

and

MB

L !

Trai

ners

for S

2

ATD

and

Rev

ersa

l de

sign

s

AB

CB

CB

MB

L !

Ss

MB

L !

Q ty

pes

(iden

ti-fic

atio

n, in

tera

ctio

n,fa

cts/

figur

es)

Prom

pts,

rein

forc

e-m

ent,

corr

ectio

n us

ed to

teac

h si

gn

RQ

s in

sim

ulat

ion

setti

ng a

nd th

en

natu

ral s

ettin

g

FCT

(toke

n ex

chan

ge fo

r S1

and

sig

ns fo

r S2)

af

ter s

erie

s of

DR

O a

ndD

RI c

ondi

tions

RQ

s (e

.g.,

Plea

se) t

augh

t w

ith le

ast-m

ost p

rom

pt-

ing;

TO

or g

radu

ated

gu

idan

ce c

onse

quen

ces

for p

robl

em b

ehav

iors

DR

O a

nd ti

me-

out i

n ta

sk

and

leis

ure

cont

exts

Taug

ht “

I don

’t kn

ow”

in

R to

pre

viou

sly

echo

ed

Qs,

pro

mpt

-fad

ing

Pres

ent a

nsw

er c

ard,

sile

nce

cue,

ask

Q, p

oint

to c

ard,

repe

at Q

with

out c

ard

% c

orre

ct s

ign

prod

uc-

tion

Gra

bbin

g, y

ellin

g,an

d si

gn R

Qs

durin

gsc

hool

day

Mea

n se

lf-in

jury

Rs/

RQ

an

d Fr

eq s

elf-

inju

ry #

aggr

essi

on

% o

f int

erva

ls w

ith p

rob-

lem

beh

avio

r; in

de-

pend

ent s

igni

ng o

r sw

itch

pres

ses

(S2)

,pr

ompt

s

3 st

ereo

typi

c be

havi

ors

per p

artic

ipan

t (%

in-

terv

als)

# of

“I d

on’t

know

,” a

p-pr

opria

te, &

ech

olal

ic

Rs

Cor

rect

Rs

to Q

s in

3

cont

ent a

reas

&

echo

lalia

!se

tting

s

!tra

iner

s, s

ettin

gs,

and

task

dem

ands

!th

erap

ists

!be

havi

ors

(rat

e of

cor

rect

Rs

in

task

con

text

)

!un

train

ed Q

s,

wh-

Q ty

pes,

and

pe

rson

s

!pe

rson

s

Alth

ough

sig

ns le

arne

d in

si

mul

atio

n se

tting

, sig

n us

e an

d re

duct

ion

in m

alad

ap-

tive

beha

vior

s oc

curr

ed

only

afte

r tra

inin

g in

nat

-ur

al s

ettin

g

S1—

imm

edia

te re

duct

ion

in

self-

inju

ry a

nd s

pont

aneo

usR

Qs →

!0

afte

r 12

wk

of

FCT

S2—

robu

st e

ffec

ts o

n pr

ob-

lem

beh

avio

r (!

0 af

ter

8 w

k) a

nd in

crea

ses

in

spon

tane

ous

com

mun

icat

ion

S1—

com

pone

nt a

naly

sis

show

edth

at F

CT

and

TO (a

nd n

otD

RO

) →in

crea

se s

igni

ng a

ndde

crea

sed

prob

lem

beh

avio

r;FC

T an

d gr

adua

ted

guid

ance

mos

t eff

ectiv

e fo

r S3;

FC

Tal

one

effe

ctiv

e fo

r S2

DR

O a

nd n

ot T

O re

duce

d pr

oble

m b

ehav

ior (

and

in-

crea

sed

corr

ect R

s) in

task

cont

ext a

nd o

nly

TO re

-du

ced

prob

lem

beh

avio

r in

leis

ure

cont

ext

Gen

eral

ized

redu

ctio

n in

echo

lalic

resp

ondi

ng a

fter

taug

ht “

I don

’t kn

ow”

R,

mai

ntai

ned

at 1

mo

follo

w-u

p

70–1

00%

ech

oes

in b

asel

ine

repl

aced

with

cor

rect

re-

spon

ses

whe

n pa

use-

poin

tpr

ompt

s pr

ovid

ed, p

oint

ing

fade

d

35 s

choo

l da

ys

S1 !

120

wee

ks; S

2!

200

wee

ks

29, 1

4, a

nd 4

2 se

s-si

ons

27 a

nd 3

0 se

ssio

ns

37 s

essi

ons

daily

30–

40

min

, 95

sess

ions co

ntin

ued

1986

—Ex

perim

ents

2

and

3II

, IV

, II

Foxx

, Faw

, McM

orro

w,

Kyl

e, &

Bitt

le, 1

988

II, I

I, II

Dav

is, B

rady

, Will

iam

s,

& H

amilt

on, 1

992

II, I

V, I

I

Koe

gel,

Koe

gel,

&

Surr

att.

1992

I, II

, II

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n, re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p or

sin

gle-

subj

ect d

esig

n ex

perim

ent c

ompa

ring

alte

rnat

ive

treat

men

ts; I

I pre

-pos

t gro

up d

esig

n or

sin

gle-

subj

ect d

esig

n w

ith re

liabl

e, o

bjec

tive

outc

omes

mea

sure

s;II

I, pr

e-po

st, e

x po

st fa

cto,

or s

ingl

e-su

bjec

t des

igns

with

que

stio

nabl

e in

tern

al v

alid

ity o

r mea

sure

men

t.Ex

tern

al v

alid

ity: I

, Ran

dom

ass

ignm

ent o

f wel

l-def

ine

coho

rts a

nd a

dequ

ate

sam

ple

size

; II,

nonr

ando

m a

ssig

nmen

t, bu

t wel

l-def

ined

coh

orts

and

ade

quat

e sa

mpl

e si

ze o

r rep

licat

ion

acro

ss 3

or m

ore

subj

ects

in a

sin

gle-

subj

ect d

esig

n; I

II, w

ell-d

efin

ed p

opul

atio

n of

3 o

r m

ore

subj

ects

in s

ingl

e-su

bjec

t des

igns

or

sam

ple

of a

dequ

ate

size

in g

roup

des

igns

; IV

, sm

all s

ampl

e in

gro

up d

esig

n or

1 o

r 2

subj

ects

in s

ingl

e-su

bjec

t des

ign.

Gen

eral

izat

ion:

I, C

onsi

sten

t gen

eral

izat

ion

to f

unct

iona

l set

ting

beyo

nd tr

eatm

ent s

ettin

g w

ithou

t fur

ther

inte

rven

tion

or r

obus

t soc

ial v

alid

atio

n re

sults

; II,

long

-term

mai

nten

ance

or

gene

raliz

atio

n to

one

setti

ng b

eyon

d tre

atm

ent s

ettin

g; II

I, in

terv

entio

n oc

curr

ed in

nat

ural

set

ting

or o

utco

me

mea

sure

with

doc

umen

t rel

atio

n to

func

tiona

l out

com

es; I

V, n

ot a

ddre

ssed

or o

ther

.M

BL

= M

ultip

le B

asel

ine

Des

ign;

FC

T =

Func

tiona

l Com

mun

icat

ion

Trai

ning

; TO

= T

ime-

Out

; NLP

= N

atur

al L

angu

age

Para

digm

; RQ

= R

eque

st; Q

= Q

uest

ion;

R =

Res

pons

e; T

x =

Trea

tmen

t; ge

n =

gene

raliz

atio

n.

21 y

r (1)

13, 3

6, 4

0 yr

(3

)

7 w

ith D

own

synd

rom

e an

d 5

yr (2

)

3–4

yr o

ld (3

)

MB

L !

Q s

ets

MB

L !

Ss

MB

L !

train

ers

Rev

ersa

l des

ign

(BC

BC

)w

ith c

ount

erba

lanc

ing

Imm

edia

te (E

xper

imen

t 2)

and

dela

yed

(Exp

eri-

men

t 3) m

odel

ing

of a

n-sw

ers

by a

noth

er a

dult

Pres

ent a

nsw

er c

ard,

si-

lenc

e cu

e, a

sk Q

, poi

nt

to c

ard,

repe

at Q

with

-ou

t car

d

Serie

s of

hig

h-pr

obab

ility

R

Q a

dmin

iste

red

prio

r to

eac

h lo

w-p

roba

bilit

y R

Q

Ana

log

(dis

cret

e-tri

al)

vers

us N

LP te

achi

ng

proc

edur

es

Cor

rect

Rs

to Q

s in

2

cont

ent a

reas

&

echo

lalia

Cor

rect

ver

bal l

abel

s in

R

to Q

s, e

chol

alic

Rs

% c

orre

ct R

s to

RQ

s

Dis

rupt

ive

beha

vior

an

d la

ngua

ge ta

rget

be

havi

ors

!no

vel s

timul

i in

Expe

rimen

t 3

!un

train

ed Q

s an

d se

tting

s

!ad

ults

exp

effe

cts

for

lang

uage

targ

ets

not p

rese

nted

gr

aphi

cally

, but

sm

all-m

oder

ate

effe

cts

impl

ied

by

cond

ition

ave

rage

s

Mod

elin

g →

imm

edia

te in

-cr

ease

s in

cor

rect

Rs

to Q

s an

d go

od m

aint

enan

ce

Cue

s-pa

use-

poin

t tra

inin

g →

corr

ect l

abel

ing

Rs

and

de-

crea

sed

echo

lalia

, gen

eral

-iz

ed !

setti

ngs

Imm

edia

te in

crea

se in

Rs

to

low

-pro

babi

lity

RQ

s w

hen

tx a

dmin

iste

red;

gen

afte

r im

plem

ente

d by

2–3

adu

lts

Con

sist

ently

less

than

20%

di

srup

tions

in N

LP c

ondi

-tio

ns a

nd 2

0–90

% in

ana

-lo

g co

nditi

ons;

51 s

essi

ons

and

48

sess

ions

!87

tria

ls

!25

day

s,

max

of 4

0 tri

als/

day

11–1

6 se

ssio

ns

Tab

le V

.In

terv

entio

ns to

Pro

mot

e So

cial

and

Scr

ipte

d In

tera

ctio

ns

Parti

cipa

nts

Expe

rimen

tal

Inde

pend

ent

Dep

ende

nt

Gen

eral

izat

ion

Ref

eren

ces

(n)

desi

gnva

riabl

esva

riabl

esm

easu

res

Res

ults

Dur

atio

n

McG

ee, K

rant

z,

McC

lann

ahan

, 198

4

II, I

I, II

Shaf

er, E

gel,

&

Nee

f, 19

84

II, I

I, II

Gol

dste

in,

Kac

zmar

ek,

Penn

ingt

on, &

Sh

afer

, 199

2II

, II,

III

Cha

rlop

& M

ilste

in,

1989

II, I

I, I

Gol

dste

in &

Cis

ar,

1992

II, I

I, II

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n, re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p, o

r si

ngle

-sub

ject

des

ign

expe

rimen

t com

parin

g al

tern

ativ

e tre

atm

ents

; II,

pre-

post

gro

up d

esig

n or

sin

gle-

subj

ect d

esig

n w

ith r

elia

ble,

obj

ectiv

e ou

tcom

es m

ea-

sure

s; II

I, pr

e-po

st, e

x po

st fa

cto,

or s

ingl

e-su

bjec

t des

igns

with

que

stio

nabl

e in

tern

al v

alid

ity o

r mea

sure

men

t.Ex

tern

al v

alid

ity: I

, Ran

dom

ass

ignm

ent o

f w

ell-d

efin

e co

horts

and

ade

quat

e sa

mpl

e si

ze; I

I, no

nran

dom

ass

ignm

ent,

but w

ell-d

efin

ed c

ohor

ts a

nd a

dequ

ate

sam

ple

size

or

repl

icat

ion

acro

ss th

ree

orm

ore

subj

ects

in a

sin

gle-

subj

ect d

esig

n; II

I, w

ell-d

efin

ed p

opul

atio

n of

thre

e or

mor

e su

bjec

ts in

sin

gle-

subj

ect d

esig

ns o

r sam

ple

of a

dequ

ate

size

in g

roup

des

igns

; IV

, sm

all s

ampl

e in

gro

up d

esig

n or

1 or

2 s

ubje

cts

in s

ingl

e-su

bjec

t des

ign.

Gen

eral

izat

ion:

I, C

onsi

sten

t gen

eral

izat

ion

to f

unct

iona

l set

ting

beyo

nd tr

eatm

ent s

ettin

g w

ithou

t fur

ther

inte

rven

tion

or r

obus

t soc

ial v

alid

atio

n re

sults

; II,

long

-term

mai

nten

ance

or

gene

raliz

atio

n to

one

setti

ng b

eyon

d tre

atm

ent s

ettin

g; II

I, In

terv

entio

n oc

curr

ed in

nat

ural

set

ting

or o

utco

me

mea

sure

with

doc

umen

t rel

atio

n to

func

tiona

l out

com

es; I

V, n

ot a

ddre

ssed

or o

ther

.M

BL

= M

ultip

le B

asel

ine

Des

ign;

RQ

= R

eque

st; Q

= Q

uest

ion;

R =

Res

pons

e; T

x =

treat

men

t; ge

n =

gene

raliz

atio

n.

13–1

5 yr

(3)

5–6

yr (4

)

3–6

(5)

6–7

yr (3

)

Pres

choo

l (3)

intri

ads

with

two

typi

cal p

eers

MB

L !

asse

rtive

R c

lass

es

MB

L !

Ss

AB

CB

reve

rsal

MB

L !

Ss a

ndm

ultip

le p

robe

!co

nver

satio

ns

MB

L !

scrip

ts

Mod

elin

g, re

hear

sal,

and

toke

n re

info

rcem

ent

Peer

s ta

ught

to s

hare

and

give

pla

y di

rect

ions

usi

ng

mod

els,

pro

mpt

s,pr

actic

e, a

nd fe

edba

ck

Tria

ds w

ith tw

o ty

pica

lpe

ers

prom

pted

to

atte

nd to

, com

men

t,an

d re

spon

d to

Ss

durin

g pl

ayPe

er v

ideo

mod

elin

g of

scr

ipte

d co

nver

satio

ns a

ndtim

e de

lay

Thre

e so

ciod

ram

atic

play

scr

ipts

with

10

beha

vior

s fo

r eac

h of

thre

e ro

les

taug

ht

% in

terv

als

with

pos

i-tiv

e an

d ne

gativ

e as

serti

ons

to p

eers

durin

g ca

rd g

ame

and

ball

gam

eM

otor

-ges

tura

l and

voca

l-ver

bal i

nitia

-tio

ns a

nd re

spon

ses

Ver

bal a

nd n

onve

rbal

com

mun

icat

ion

acts

and

adul

t mon

itorin

gbe

havi

ors

Rs

varia

tions

, Q a

skin

g,sc

ripte

d co

nver

sa-

tions

Trai

ned

and

untra

ined

them

e-re

late

d an

dun

rela

ted

soci

albe

havi

ors

Gen

to

untra

ined

pos

itive

asse

rtion

s an

d 4.

5m

o fo

llow

-up

!se

tting

s

No

!un

train

ed

topi

cs,

pers

ons,

se

tting

s, a

ndst

imul

i, so

cial

valid

atio

n!

peer

gr

oupi

ngs

Tx →

clea

r inc

reas

es re

plic

ated

acro

ss fo

ur R

cla

sses

& S

s,st

rong

er e

ffec

ts fo

r pos

itive

as

serti

ons

Rap

id im

prov

emen

t in

soci

al

beha

vior

s (m

ainl

y R

s) d

urin

gpe

er tr

aini

ng; f

ollo

wed

by

gen

to tr

nd p

eer o

nly

in p

lay

grou

p; tr

aini

ng n

eede

d fo

r gen

!se

tting

sPe

er s

trate

gy u

se a

ssoc

iate

d w

ith in

crea

sed

soci

al b

ehav

ior,

but m

ainl

y no

nver

bal f

or a

ll bu

t 1 c

hild

Incr

easi

ng q

uick

and

robu

st g

enpr

oduc

tion

of 3

-beh

avio

r co

nver

satio

ns a

fter m

ax o

f 3 o

f5

or 6

con

vers

atio

ns m

odel

ed

Each

scr

ipt l

earn

ed in

crea

sing

lyqu

ickl

y an

d →

them

e-re

late

din

tera

ctio

n, p

lus

mai

nten

ance

and

gen

to u

ntra

ined

, rel

ated

ut-

tera

nces

18 s

essi

ons

and

4.5

mo

follo

w-u

p

3–6

peer

trai

n-in

g se

ssio

ns

39–5

1 se

ssio

ns

3–20

vid

eovi

ewin

gs to

reac

h cr

ite-

rion

6–15

sess

ions

/scr

ipts

and

30–3

9 pl

ayse

ssio

ns

Tab

le V

I.Pa

rent

and

Cla

ssro

om T

rain

ing

Ref

eren

ces

Parti

cipa

nts

(n)

Expe

rimen

tal d

esig

nIn

depe

nden

t var

iabl

esD

epen

dent

var

iabl

esR

esul

tsD

urat

ion

Har

ris, W

olch

ilk,

& W

eltz

, 198

1

III,

IV, I

VH

owlin

& R

utte

r, 19

89

II, I

I, II

IH

arris

, Han

dlem

an,

Kris

toff

, Bas

s, &

G

ordo

n, 1

990

III,

IV, I

VO

zono

ff &

Cat

hcar

t, 19

98

II, I

I, IV

Joce

lyn,

Cas

iro,

Bea

ttie,

Bow

, &

Kne

isz,

199

8

I, II

, IV

Not

e: R

oman

num

eral

s in

the

first

col

umn

deno

te ra

tings

of i

nter

nal v

alid

ity, e

xter

nal v

alid

ity, a

nd g

ener

aliz

atio

n, re

spec

tivel

y, b

ased

on

the

follo

win

g cr

iteria

:In

tern

al v

alid

ity: I

, Pro

spec

tive

grou

p or

sin

gle-

subj

ect d

esig

n ex

perim

ent c

ompa

ring

alte

rnat

ive

treat

men

ts; I

I, pr

e-po

st g

roup

des

ign

or s

ingl

e-su

bjec

t des

ign

with

relia

ble,

obj

ectiv

e ou

tcom

esm

easu

res;

III,

pre-

post

, ex

post

fact

o, o

r sin

gle-

subj

ect d

esig

ns w

ith q

uest

iona

ble

inte

rnal

val

idity

or m

easu

rem

ent.

Exte

rnal

val

idity

: I, R

ando

m a

ssig

nmen

t of w

ell-d

efin

ed c

ohor

ts a

nd a

dequ

ate

sam

ple

size

; II,

nonr

ando

m a

ssig

nmen

t, bu

t wel

l-def

ined

coh

orts

and

ade

quat

e sa

mpl

e si

ze o

r rep

licat

ion

acro

ss3

or m

ore

subj

ects

in a

sin

gle-

subj

ect d

esig

n; II

I, w

ell-d

efin

ed p

opul

atio

n of

thre

e or

mor

e su

bjec

ts in

sin

gle-

subj

ect d

esig

ns o

r sam

ple

of a

dequ

ate

size

in g

roup

des

igns

; IV

, sm

all s

ampl

e in

grou

p de

sign

or 1

–2 s

ubje

cts

in s

ingl

e-su

bjec

t des

ign.

Gen

eral

izat

ion:

I, C

onsi

sten

t gen

eral

izat

ion

to fu

nctio

nal s

ettin

g be

yond

trea

tmen

t set

ting

with

out f

urth

er in

terv

entio

n or

robu

st s

ocia

l val

idat

ion

resu

lts; I

I, lo

ng-te

rm m

aint

enan

ce o

r gen

eral

-iz

atio

n to

one

set

ting

beyo

nd tr

eatm

ent s

ettin

g; II

I, in

terv

entio

n oc

curr

ed in

nat

ural

set

ting

or o

utco

me

mea

sure

with

doc

umen

t rel

atio

n to

func

tiona

l out

com

es; I

V, n

ot a

ddre

ssed

or o

ther

.

30–7

4 m

o (1

1)

3–11

yr (

30)

4–5

yr (1

0)

2–5

yr (2

2)

2–6

yr a

nd p

aren

ts(3

5)

Two

grou

ps tr

nd (h

ome

vers

us la

b tre

atm

ent?

),re

peat

ed m

easu

res

desi

gn

Mat

ched

gro

up

com

paris

on

Gro

up a

ssig

nmen

t,m

atch

ed b

ut n

ot ra

n-do

mly

ass

igne

d

Trea

tmen

t ver

sus

no-

treat

men

t com

paris

on,

mat

ched

gro

ups

Ran

dom

ized

con

trolle

dtri

al, 1

6 in

exp

erim

en-

tal g

roup

and

19

in d

ayca

re o

nly

grou

p

10 w

eekl

y pa

rent

train

ing

grou

pm

eetin

gs a

nd 6

cons

ulta

tive

hom

evi

sits

Pare

nts

taug

ht b

ehav

-io

ral t

echn

ique

s(q

uest

ions

, rep

eti-

tion,

exp

ansi

on,

corr

ectio

n,pr

ompt

s, a

nd re

in-

forc

emen

t)

Inte

grat

ed v

ersu

sse

greg

ated

cla

sses

TEA

CC

H-b

ased

hom

e pr

ogra

m, b

utno

trea

tmen

t fi-

delit

y m

easu

res

12 w

k of

lect

ures

,co

nsul

tatio

ns a

tda

y-ca

re c

ente

rsfo

r sta

ff a

nd p

ar-

ents

, and

fam

ilysu

ppor

t

Num

ber c

orre

ct a

ndhi

ghes

t ste

p on

hi-

erar

chy

of 2

1 co

m-

mun

icat

ion

skill

s

Freq

uenc

y of

com

-m

unic

ativ

e sp

eech

,ec

hola

lia a

nd n

on-

com

mun

icat

ive

ut-

tera

nces

, inc

reas

eco

mpl

exity

Pres

choo

l Lan

guag

eSc

ale

Psyc

hoed

ucat

iona

lPr

ofile

-Rev

ised

—7

deve

lopm

enta

ldo

mai

ns

Aut

ism

Kno

wle

dge

Qui

z, R

atin

gA

utis

m S

ympt

oms,

EI a

nd P

resc

hool

Dev

elop

men

t Pro

-fil

es, F

amily

as-

sess

men

t and

stre

ssm

easu

res,

sat

isfa

c-tio

n m

easu

re

Larg

e ef

fect

s fo

r 2 o

f 5ve

rbal

chi

ldre

n, m

od

effe

cts

for 2

non

verb

alch

ildre

n, o

ther

eff

ects

min

imal

. Sig

nific

ant v

er-

bal v

ersu

s no

nver

bal !

time

inte

ract

ion.

No

chan

ges

at 1

yr f

ollo

w-u

p.C

lear

exp

erim

enta

l eff

ects

for m

othe

rs a

nd C

hn;

Sign

ifica

nt in

crea

se in

child

ren’

s so

cial

utte

r-an

ces,

mor

phem

es, a

ndre

duct

ion

in in

com

pre-

hens

ible

and

non

verb

albe

havi

or fo

r exp

erim

enta

lgr

oup.

Hig

h co

rrel

atio

nbe

twee

n m

othe

rs’ a

ndch

ildre

n’s

lang

uage

chan

ges

No

pret

est-p

ostte

st g

roup

diff

eren

ces

Sign

ifica

nt g

rp !

time

ef-

fect

s on

imita

tion,

fine

and

gros

s m

otor

, and

nonv

erba

l cog

nitiv

e do

-m

ains

Sign

ifica

nt g

rp !

time

in-

tera

ctio

n fo

r one

of s

ixde

velo

pmen

tal d

omai

ns(i.

e., l

angu

age)

; als

o gr

p!

time

diff

eren

ce in

Aut

ism

Qui

z fo

r mot

hers

and

child

car

e w

orke

rs

10 w

k

18 m

o in

terv

entio

n, c

om-

paris

on a

fter 6

mo

repo

rted

5–11

mos

8–12

trai

ning

ses

sion

s,

10 w

k of

trea

tmen

t, 4

mo

betw

een

pre-

and

post

-test

s

12 w

k

results of each study; (b) summaries of the ages andnumber of subjects included in individual studies; (c)the experimental designs; (d) the independent variablesunder investigation or an outline of the treatment com-ponents; (e) the measures used to evaluate primary ef-fects and generalized effects; (f ) a summary of results,and (g) an estimate of the duration of training or thestudy length. The studies were categorized to reflect sixprimary approaches to the development of communica-tion interventions that are commonly applied to chil-dren with autism. Table I summarizes studies thatincorporated sign language into training. Table II sum-marizes studies using a variety of largely operant ap-proaches to language intervention that appear to bepresented in discrete trial training formats. Table III in-cludes studies that sought to teach language using var-ious milieu language teaching paradigms. Table IVsummarizes studies that investigated the relationship be-tween problem behaviors and communication skills.Table V includes studies that sought to promote socialand scripted interactions that focused on language skills.Table VI includes more general parent and classroomtraining approaches that included measures of language.

COMMUNICATION INTERVENTIONSINCORPORATING SIGN LANGUAGE

Much research has explored the effects of intro-ducing gestural and visual modes of communicationthrough augmentative and alternative communicationsystems. Clearly, the use of augmentative and alterna-tive communication systems has spurred the develop-ment of language skills with a great number of childrenwho had extremely limited communication abilities.Presenting language stimulation using the visual modal-ity seems to capitalize on an area of relative strengthfor many children with autism. Table I includes nineexperiments that investigated the benefits of introduc-ing signs with children with autism. Experiments usingcommunication systems using pictures, symbols, print,vocal output devices, and such are reviewed in theMirenda chapter (this volume). Of course, great interestin the latter area was generated given the miraculousclaims offered by proponents of “facilitated commu-nication.” When those claims have been subjected torigorous experimental scrutiny, however, evidence ofmeaningful communication under the control of indi-viduals with autism was found lacking (see Green,1994; Jacobson & Mulick, 1995).

Dating back to the late 1970s some investigatorsfocused on the use of sign language systems. A num-

Communication Intervention 385

ber of studies compared the effects of teaching ex-pressive vocabulary using speech, sign, or total com-munication (speech # sign) input (BarreraLobatos-Barrera, & Sulzer-Azaroff 1980, 1983; Lay-ton, 1988; McIlvane, Bass, O’Brien, Gerovac, & Stod-dard, 1984; Yoder & Layton, 1988). Other studiestaught receptive vocabulary (Brady & Smouse, 1978;Carr & Dores, 1981; Carr, Pridal, & Dores, 1984;Wherry & Edwards, 1983) or both expressive and re-ceptive vocabulary (Layton, 1988). The results of thesestudies are rather consistent. Sign or total communica-tion training resulted in quicker and more completelearning of vocabulary than speech training for manyparticipants. The participants who benefit most seemto be those with more limited communication reper-toires, whereas more sophisticated participants, and es-pecially those with better verbal imitation skills, aremore likely to learn from systematic discriminationtraining regardless of the mode of communication. Par-ticipants with good verbal imitation skills are morelikely to demonstrate speech production in addition toor in lieu of sign production than participants with poorverbal imitation skills (Carr & Dores, 1981; Carr, Pri-dal, & Dores, 1984; Yoder & Layton, 1988). The onestudy (Yoder & Layton, 1988) that sought to explorewhether other factors might predict spoken vocabularyperformance found verbal imitation to account for amajority of the variance (R2 " .63), and that 78% ofthe variance could be accounted for by adding age andIQ to the regression model. In some studies the extentof the repertoire taught was quite limited; for example,Hinerman, Jensen, Walker, and Peterson (1982) taughta participant to request “milk” and “cookies.” Very fewstudies progressed beyond single words; Keogh, Whit-man, Beeman, Halligan, and Starzynski (1987), in astudy with older students, taught brief dialogues. Mostof the studies employed single-subject experimental de-signs with 1 to 10 participants. Two studies enrolled 60participants each and randomly assigned participants tofour treatment conditions (Layton, 1988; Yoder & Lay-ton, 1988).

Although interest in this area has waned, the gen-eral findings are fairly clear. Total communication ap-pears to be a viable treatment strategy for teachingreceptive and expressive vocabulary (i.e., languagecontent) to individuals with autism. The presentationof speech alone is less effective for individuals whohave poor verbal imitation skills in particular. Althoughfew studies explored the use of sign alone, it is con-traindicated because the likelihood of children pro-ducing speech is diminished. In contrast, totalcommunication often results in comprehension and pro-

duction of both signs and speech. Because the focus ofthese studies has been on teaching vocabulary for themost part, one can assume that investigators have notpursued this approach to the point of teaching sign lan-guage as an alternative communication system to chil-dren with autism. Thus it would appear that presentingsigns as well as speech is mainly an effective adjunctstrategy for jump-starting early vocabulary learning.Signs represent another symbolic system for represent-ing objects and actions in the world that children withautism hopefully will be motivated to request, label, andcomment upon. Signs are less transient than words, andgestures and signs are easier to prompt than verbal pro-ductions. Whether these or other reasons underlie theirbenefits is largely a matter of speculation at this point.

INTERVENTIONS INCORPORATINGDISCRETE TRIAL TRAINING FORMATS

Table II summarizes 12 experiments that investi-gated methods of teaching comprehension and/or pro-duction of various language content (e.g., vocabularyand phrases) and forms (e.g., verb # adjectives # nounutterances). One study focused on speech production(i.e., phonology) (Koegel, O’Dell, & Dunlap, 1988),and one study focused on gestural communication(Buffington, Krantz, McClannahan, & Poulson, 1998).The majority of the experiments taught children to pro-duce single-word responses or single phrases (e.g.,“What’s that?”). However, children’s responses re-quired discrimination of language forms (e.g., differ-ent types of wh-questions). The last four experimentsin Table II taught sentences.

These studies investigated teaching proceduresthat were employed in a variety of discrete-trial train-ing formats. Thus various differential reinforcementand correction procedures with modeling of responsesand prompting-fading procedures are commonalities.Not surprisingly, these procedures are largely effectivein teaching the behaviors of interest. Perhaps the mostimpressive of these studies are those experiments inwhich the training was organized to teach increasinglycomplex productions. Krantz, Zalewski, Hall, Fenski,& McClannahan (1981) taught children to describe ob-jects and pictures using four-term sentences (verbs #colors # shape/size # labels) and to answer a varietyof what, how, and why questions about immediate andthen past events with single or multiple sentences.

One should note that the questions addressed bythese studies typically went beyond straightforward ef-fects of instruction to explore procedures that might

386 Goldstein

produce more efficient learning or generalized treat-ment effects. For example, McIlvane et al. (1984)demonstrated that exclusion procedures in which un-known food names are presented in the context ofknown foods resulted in accurate receptive respondingand that expressive naming emerged without specifictraining after relatively few trials. Although the exclu-sion procedure (akin to process of elimination) is ef-fective for teaching basic concepts or novel words in amanner that permits generative language learning, itsuse with children with autism is not evident in the lite-rature. This work has been replicated and extended withother individuals with severe developmental disabili-ties, however.

Buffington et al. (1998) demonstrated that chil-dren with autism could be taught to include gestures toaccompany simple verbal responses to improve theircommunication. It is not clear in the article whether thetraditional training provided affected gestural and ver-bal communication skills equally, but social validitymeasures indicated that their communicative attemptswere virtually indistinguishable from age-matched typi-cal peers. Although the children generalized to probesusing novel materials, data on carryover to natural en-vironments were not included.

Research on autism also has contributed greatly tothe literature on generalization-promoting strategies. Awell-established technology for producing generaliza-tion is still under development, based largely on theconceptualization of several strategies forwarded byStokes and Baer (1977) in their landmark article. Forexample, when a multiple exemplar approach did notproduce consistent results (Egel, Shafer, & Neef,1984; Handleman, 1979, 1981), investigators devel-oped additional strategies to overcome deficiencies ingeneralization exhibited by children with autism. In-vestigators found that minimal additional training wasrequired. Also, embedding instruction across the dayor interspersing different types of trials proved rathereffective (Egel, et al. 1984; McMorrow & Foxx, 1986;Neef, Walters, & Egel, 1984; Secan, Egel, & Tilley,1989). These findings seem to indicate that general-ized effects are not likely when generalization probesare easily discriminated from the training situation.However, if training exemplars are selected carefully,are sampled broadly from natural contexts relevant tothe child, and are embedded within those contexts,then widespread generalization across settings islikely to result.

Perhaps the experiment in which generalizationwas most clearly elucidated was an application ofmatrix-training procedures to object-preposition-location

utterances (e.g., penny under couch) with two pre-schoolers with autism (Goldstein & Brown, 1989). Inthat study, recombinative generalization to new utter-ances occurred predictably as new preposition and lo-cation words were learned via peer modeling in thecontext of 6 to 8 utterances systematically selected from64 possible combinations of 4 known objects, 4 un-known prepositions, and 4 unknown locations. Gener-alization accounted for 95% of the 128 responsesgenerated. Generalization from production to compre-hension did not occur until modeling in that modalitywas demonstrated, however. The conditions that pro-duce cross-modal transfer remain elusive.

In summary, experiments involving discrete-trialtraining formats have provided a foundation for the de-velopment of procedures needed to teach discrimina-tive performances that children with autism had notlearned on their own. In contrast to the investigationsin the late 60s and the 70s, there has been increasingattention on teaching somewhat more sophisticated lan-guage and on improving the generalized use of new lan-guage skills. These advances have been in response tothe often-disappointing effects of discrete-trial teach-ing on the everyday functioning of children withautism. The need to assess and program generalizationfrom training situations into everyday living contextshas inspired application of many of the same instruc-tional strategies within the natural milieu of home,school, and community contexts.

INTERVENTIONS DESIGNED FORIMPLEMENTATION IN THE NATURALMILIEU

Table III presents summaries of 12 studies that in-vestigated time delay, milieu language teaching, or nat-ural language paradigm interventions. Time delay is aprompting procedure in which a delay is imposed be-tween a stimulus and a prompt or a hierarchy ofprompts. The delay can be held constant (e.g., 5 seconds)or can be lengthened progressively until the child an-ticipates the prompt and produces the desired verbal ormotor response. Time delay procedures have estab-lished rapid and often generalized language produc-tions, especially requesting, when the communicationsskills are in children’s repertoire but rarely used (e.g.,Charlop, Schreibman, & Thibodeau, 1985). In manystudies it is not clear whether children are acquiringnew skills, however. Correction procedures usually ap-pear to accompany the use of time-delay procedures;so modeling with or without prompts to imitate tends

Communication Intervention 387

to be an additional intervention component (Charlop &Trasowech, 1991; Charlop & Walsh, 1986). Charlopand Walsh (1986) found that implementing a time delayprocedure was associated with spontaneous productionsof “I like/love you,” whereas peer modeling of this ut-terance was not effective. Matson, Sevin, Box, Fran-cis, and Sevin (1993) found modeling and the fadingof visual cues as effective as the time delay procedurein teaching social phrases. McGee, Krantz, and Mc-Clannahan (1985) found traditional training with arbi-trary reinforcers as effective as incidental teaching withrequested items serving as reinforcers when teachingchildren to produce prepositions; but better general-ization across settings and more spontaneous produc-tions resulted from incidental teaching. Likewise, Neefet al. (1984) found greater generalization of yes/no re-sponses to various question types when teaching trialswere embedded across the day in comparison to tutor-ing sessions.

Milieu language teaching is a family of proceduresthat are designed to capitalize on children’s desires andinterests in their natural environments to embed teach-ing opportunities. Incidental teaching (cf. Hart & Risley,1975), perhaps the hallmark of milieu language teach-ing, requires that the child initiate an attempt to com-municate, usually to make a request. Mand-modelprocedures were developed to teach fundamental com-munication skills that could be used to make requeststo children who are not initiating. Then time-delay pro-cedures could be used to teach independent use of thoseskills. Other procedures that are designed for use in nat-ural environments with a focus on following the child’slead might be considered forms of milieu languageteaching (e.g., the Natural Language Paradigm).

Milieu language teaching is usually used to teachrequesting, because high motivation is inherent whenindividuals are requesting desired items that presum-ably function as reinforcers. However, an examinationof Table III reveals that a variety of communicativefunctions are being taught to children with autism usingmilieu teaching procedures: preverbal communication(eye contact, joint attention, and motor imitation,Hwang & Hughes, 2000); spontaneous productions of“I like/love you” (Charlop & Walsh, 1986); descrip-tions of drawings and car play (Ingenmey & VanHouten, 1991); social amenities such as “please, thankyou, excuse me, you’re welcome, hello” (Matson et al.,1993); positive interactions with peers (McGee,Almeida, Sulzer-Azaroff, & Feldman, 1992), answersto “Where is ___?” questions (McGee et al. 1985);phoneme production (Koegel, Camarata, Koegel, Ben-Tall, & Smith, 1998); and simply increased talking

(Laski, Charlop, & Schreibman, 1988). It is notable thatparents have been taught to implement time delay pro-cedures (Charlop & Trasowech, 1991) and the NaturalLanguage Paradigm (Laski et al., 1988). The NaturalLanguage Paradigm (NLP) is a multiple component in-tervention that continues to evolve because it has notalways been operationalized the same way (e.g.,Koegel, O’Dell, & Koegel, 1987; Laski et al., 1988).Characteristics of this approach have begun to be clar-ified, especially when contrasted with discrete-trialtraining. For example, Koegel et al. (1998) clearly con-trasted the procedural differences in analog (discrete-trial) versus naturalistic (NLP) intervention approachesbased on selection of stimulus items, steps in training,interaction patterns, response-reinforcer contingencies,and the types of consequences provided.

Research on milieu language teaching procedureshas been extensive and seems to be applicable to teach-ing early language skills to a broad population of chil-dren (see Kaiser, Yoder, & Keetz, 1992). Most of theapplications to children with autism have focused onMand-model and time-delay procedures to accommo-date their deficiencies in communicative initiations. In-vestigators have had success when applying fairlywell-established procedures that have been used withother populations.

At this point, there is no compelling evidence thatmilieu teaching procedures are clearly more effectivethan the procedures that have developed out of discrete-trial procedures. Indeed, one might argue that there isa great deal of commonality in the procedures em-ployed. An analysis of naturalistic language interven-tion procedures offered by Hepting and Goldstein(1996) includes many of the instructional techniquesfound in the discrete-trial teaching literature. One canargue that milieu language teaching procedures can bemore easily incorporated into everyday activities andreduce the need to program for generalization. Suchcomparisons of treatment approaches are not necessar-ily straightforward. An examination of Table II indi-cates that many interventions using discrete trialtraining have produced impressive generalization re-sults. Also, a comparison of Tables II and III shows lit-tle overlap in the target behaviors that were taught usingdiscrete-trial and milieu teaching procedures.

Comparison studies have yielded mixed results. Inone such comparison study, Elliott, Hall, and Soper(1991) taught receptive and expressive vocabulary to23 adults with autism and severe to profound mentalretardation using discrete-trial and natural languageteaching techniques using a crossover design withmatched groups; they found no significant differences

388 Goldstein

in acquisition or generalization. Koegel et al. (1998)contrasted traditional (analog) articulation training witha naturalistic approach (Natural Language Paradigm)in a single-subject experiment (counterbalanced rever-sal design) and found much greater generalization toconversational language samples for phonemes taughtusing the naturalistic procedures.

Yoder et al. (1995) argue that teaching approachesmay differ depending on the language skills beingtaught or the linguistic sophistication of the children.Only in the case of sign language interventions (Yoder& Layton, 1988) have these correlational analyses beendone with children with autism. Nevertheless, a care-ful inspection of the correlational findings used to de-tect aptitude-by-treatment interactions indicates thatfew children consistently fall into the regions of sig-nificance that are differentially predictive of outcomes(e.g., Cole, Mills, Dale, & Jenkins, 1996; Yoder et al.,1995; Yoder & Warren, 1999). Defining profiles thatare predictive of differential treatment effects is thuslikely to be difficult. Experimental investigations ofaptitude-by-treatment interaction hypotheses have yetto be done with any children with developmental dis-abilities, let alone children with autism.

COMMUNICATION INTERVENTIONS TOREPLACE CHALLENGING BEHAVIOR

Table IV presents a summary of 17 experimentsin which communication skills have been taught as al-ternative to a variety of maladaptive or challenging be-haviors (e.g., stereotypy, aggression, tantrums, propertydestruction, self-injury, echolalia). This aspect of com-munication intervention emphasizes a view of languageas a means of control over one’s environment. Inter-ventions represented in this segment of the literaturehave been applied to especially challenging cases, mostoften children and adults with severe developmentaldisabilities and severe behavior problems, many ofwhom have been diagnosed with autism. Investigatorshave developed procedures for identifying antecedentand consequent variables that appear to motivate chal-lenging behavior in individuals with severe disabilities.Carr and Durand (1985) hypothesized that challengingbehaviors could serve a communicative function. Thisobservation has inspired numerous investigations inwhich individuals with autism and other disabilities aretaught specific language skills to serve the same com-municative function as their challenging behavior. Be-cause of the potential to prevent the development ofchallenging behavior by focusing on communication

training early, Table IV and this section of the reviewincludes data on older individuals with autism, manyof whom we can assume were not treated successfullywhen they were younger.

When individuals with autism are taught commu-nication skills that serve efficiently and effectively asalternative behaviors, reductions in challenging be-haviors result (Bird, Dores, Moniz, & Robinson, 1989;Carr & Durand, 1985; Durand & Carr, 1987, 1992;Horner & Budd, 1985). Determining the motivation ofchallenging behavior is important, because a typical re-sponse of allowing individuals to escape demands whenthis is the motivation for stereotypic behavior, for ex-ample, may serve to reinforce the undesired behavior(Durand & Carr, 1987). Although manipulation of con-sequences or teaching communication in other contextsmay be associated with initial reductions in challeng-ing behaviors, functional communication training con-ducted in a variety of natural contexts is associated withbroader generalization and greater maintenance of ef-fects (Bird et al., 1989; Durand & Carr, 1992; Horner& Budd, 1985). Identifying the function of challengingbehaviors and developing effective interventions isoften complicated. Wacker et al. (1990) demonstratedthat multiple-component interventions are sometimesmore effective than functional communication trainingalone. Haring and Kennedy (1990) showed that the mo-tivation of challenging behavior could vary based oncontext (task versus leisure contexts) as well.

Schreibman and Carr (1978) found that they couldreplace echoing of questions with a general verbal re-sponse (“I don’t know”). McMorrow and Foxx (1986)demonstrated that inhibiting the echoing of questionsaccompanied with teaching of correct responses (eitherthrough written cues or peer modeling) resulted in ap-propriate responding. They suggest the benefits ofteaching individuals to pause before responding. In sub-sequent extensions to this work (McMorrow, Foxx,Faw, & Bittle 1987; Foxx, Faw, McMorrow, Kyle, &Bittle 1988), they found that subsequent to cues-pause-point training, participants were able to apply their la-beling repertoire to untrained questions in a variety ofgeneralization settings. Davis, Brady, Williams, andHamilton (1992) demonstrated that embedding requeststhat are unlikely to obtain a response (low-probabilityrequests) within a series of requests that are respondedto readily (high-probability requests) can improve in-struction following. Koegel, Koegel, and Surratt (1992)found that teaching language targets to 3- to 4 year-olds using NLP procedures resulted in less disruptivebehavior than when using analog procedures in a dis-crete-trial format. However, data were not displayed

Communication Intervention 389

adequately to determine whether consistent changes inlanguage responding were under such clear experi-mental control. In general, these studies demonstrateseveral effective strategies for remediating echolalia,noncompliance, and disruptive behavior in childrenwith autism.

This is an especially important area because of theintense interest in managing challenging behavior ofchildren with autism in school, as well as in home andcommunity settings. The replacement of challenging be-haviors with appropriate and increasingly sophisti-cated communication skills has the potential to havefar-reaching implications for academic achievement, so-cial relationship development, and vocational outcomes.

INTERVENTIONS TO PROMOTE SOCIAL ANDSCRIPTED INTERACTIONS

Table V summarizes studies from a larger body ofliterature on facilitating social skill development (seeMcConnell article, this volume). These five studieswere selected because of their inclusion of communi-cation measures. Two studies taught facilitative strate-gies to peers. When peers were taught to share and giveplay directions (Shafer, Egel, & Neef, 1984) or to at-tend to, comment, and respond to peers with autism(Goldstein, Kaczmarek, Pennington, & Shafer, 1992),children with autism demonstrated increased social-communicative interaction (mainly responding) duringplay. Maintenance and generalization across settings islargely dependent upon prompting peers to continueusing the facilitative strategies.

A common concern found in the literature is thedifficulty in getting children with autism to initiate in-teractions. A number of modeling and prompting ap-proaches have been used successfully to increaseinitiations, however. McGee et al. (1984) used model-ing, rehearsal, and token reinforcement to teach olderchildren with autism assertive statements to use withpeers during a card game and a ball game.

Investigators have taught scripted interactions ofvarying length and complexity. Charlop and Milstein(1989) showed videotapes of peers engaging in three-turn conversations. Goldstein and Cisar (1992) devel-oped sociodramatic play scripts involving three rolesthat were taught to triads with two typical peers and achild with autism; 10 behaviors per role were delin-eated with nonverbal, single word, or sentence-longvariations of each of the 30 steps in the sequence. Ineach of these studies children with autism demonstratedimproved peer interaction and the children produced

untrained as well as trained variations or elaborationson the scripts.

Given that a problem relating to others is a coresocial deficit associated with autism, the effectivenessof these interventions in increasing social interactionwith peers in particular is quite striking. Increased so-cial interactions can set the stage for other develop-ments as well, e.g., generalized use of newly acquiredlanguage skills, modeling of new language skills, in-clusion in more normalized educational settings, and,hopefully, the development of positive, long-lasting re-lationships with peers and others. This aspect of theliterature helps provide insights into methods for in-corporating communicative interventions into naturalenvironments of children with autism.

CLASSROOM AND PARENT INTERVENTIONSAPPLIED TO GROUPS

Table VI includes five group design studies thatseemed to focus on teaching communication skills orfocused their evaluations on language measures. SeeRogers (1998) for a review of evaluations of compre-hensive treatment programs, most of which were ap-plied for 2 or more years. Harris, Wolchik, and Weltz(1981) taught parents facilitative interaction skills andexamined effects on a hierarchy of communicationskills. Results were difficult to interpret (e.g., the na-ture of the groups); nevertheless, after 10 weeks, analy-ses revealed a significant verbal ability by timeinteraction, because the children with better imitationskills were the children who benefited most from treat-ment. No further improvements were evidenced afterparent training was terminated based on a 1-year fol-low-up. Howlin and Rutter (1989) conducted a matchedgroup comparison in which parents were taught behav-ioral techniques thought to facilitate language develop-ment. After 6 months of intervention, parents in theexperimental group talked more to their children andused more language facilitative forms. In turn, their chil-dren showed a significantly superior increase in socialutterances and morphemes, as well as a reduction intheir incomprehensible or nonverbal behaviors. Harris,Handleman, Kristoff, Bass, and Gordon (1990) admin-istered the Preschool Language Scale after 5 to 11months of intervention to 10 “matched” children withautism who were enrolled in integrated or segregatedpreschool classes and found no differences. Ozonoff andCathcart (1998) compared 22 “matched” children whoenrolled in the TEACCH program or a wait-list control.After 10 weeks of treatment, results revealed significant

390 Goldstein

group by time interactions on three of seven domainson the PEP-R, but not on the language domain.

The single randomized, controlled trial (n " 16 #19) compared typical day care with a day care that in-cluded 12 weeks of lectures and consultations for staffand families (Jocelyn, Casiro, Beattie, Bow, & Kneisz,1998). A significant group by time interaction effectwas shown for one of six developmental domains as-sessed using the Early Intervention or Preschool De-velopmental Profiles—the language domain.

Our confidence in the results of these studiesshould be tempered for a number of reasons. They suf-fer from a lack of true experimental designs (except forthe Jocelyn et al., 1998 study), small samples, short in-tervention periods, the absence of treatment fidelitymeasures, and sketchy descriptions of the interventions.In general, these group design studies reviewed reveallittle about what specific treatments or components ofintervention programs result in clinically significantimprovements in communication development. Thelack of randomized controlled clinical trials in theautism literature may seem disappointing. It can be ar-gued that sufficient knowledge has been accumulatedin the literature reviewed in previous sections to put to-gether treatment packages that are likely to be highlysuccessful in improving educational outcomes and dailyfunctioning of children with autism. However, the dif-ficulty in doing this should not be underestimated.

DISCUSSION OF THE STATE OF THELITERATURE

Participant Selection and Description

One will notice the lack of specificity in the col-umn used to summarize the participants in Table I toVI. It is difficult to provide brief summaries of devel-opmental levels and severity of impairments of partic-ipants with any confidence based on informationpresented in most journal articles. The level of detailand the nature of participant descriptions vary. Someinvestigators give priority to narrative descriptions, andothers rely largely on standardized assessments.Clearly, the field would benefit from a set of conven-tions that would help standardize the sharing of de-scriptive information about participants.

One of the limitations of this review is that it ig-nores a broader treatment literature that is pertinent tochildren with autism. Developmental disabilities re-searchers do not always report or highlight diagnosesof autism for a number of possible reasons: they maythink that the diagnosis is not relevant given their in-

terest in developing instructional procedures with broadapplicability; they may not have access to diagnosticrecords or may choose not to pursue diagnosticworkups. Nevertheless, many investigations describechildren with severe developmental disabilities who ex-hibit autistic characteristics. For example, Stoddard,McIlvane, and their colleagues (e.g., McIlvane, Dube,Green, & Serna, 1993; McIlvane et al., 1984) at theShriver Center have conducted extensive research withexceedingly difficult-to-teach individuals with devel-opmental disabilities. Their research is largely absentfrom this review. Yet, they have developed effectiveprocedures for individuals who have had difficulty inlearning simple discriminations and have used thosebehaviors as the basis for establishing conditional dis-criminations. Most of their participants have progressedto the point of learning equivalent relationships amongconcepts and various symbolic representations (e.g.,spoken words, pictures, written words, signs, graphicsymbols, and verbal productions). Their advances ininstructional technology and errorless learning proce-dures, as well as the work of others, may have greatrelevance to children with autism. A focus solely onstudies that enrolled children with autism may avert ourattention from other important findings in the educa-tional treatment literature. An alternative approachmight be to develop a database around target behaviorsto a greater extent than was accomplished in this paper.

Treatment Components

The descriptions of the actual training proceduresused by investigators are often quite sketchy. A reviewof “naturalistic” language intervention studies by Hep-ting and Goldstein (1996a) cautions us from assumingthat treatment procedures are equivalent even when in-terventions are called the same thing (e.g., “incidentalteaching” or “interactive”). Further specification isneeded to delineate what treatment components areplanned as part of a treatment program, to determinethe extent to which the treatment components are infact implemented as part of a treatment program (treat-ment fidelity), and to begin to determine what compo-nents are more and less responsible for treatment effects(component analyses). Perhaps the taxonomy used byHepting and Goldstein (1996a) provides a starting pointfor specifying treatment components. They were ableto classify 34 language-treatment studies based on theinclusion of eight procedures that manipulated an-tecedents (prompting imitation, manding verbalization,requesting elaboration or clarification, waiting for ini-tiations or responses, arranging the environment, mod-

Communication Intervention 391

eling, repeating/expanding/recasting, and descriptivetalking) and three procedures that manipulated conse-quences (delivering desired consequences, praising,minimal encouragers). The under-specification of whatinstructional procedures are active in interventions rep-resents a large obstacle to those interested in conduct-ing treatment comparisons.

Treatment Intensity

No literature was found that directly sought toevaluate the effects of delivering treatments with vary-ing frequency or intensity. Some relevant informationcould potentially be extracted from the literature, how-ever. Most single-subject experiments present data thatallow one to assess the rapidity of learning. It is diffi-cult to evaluate treatment intensity data in a straight-forward manner. For example, some investigators mayseek to increase the frequency of communication skillsalready in the repertoire of participants. One might ex-pect quicker effects in those cases than when teachingtotally unfamiliar behaviors. Other investigators indi-vidualize their selection of target behaviors by identi-fying new skills that are developmentally appropriateand functional for their participants, but some investi-gators are better at this than others. The greater the dis-crepancy between targeted behaviors and the presentrepertoire of participants, the longer one might expectthe acquisition process. On the other hand, errorlesstraining procedures may be especially effective withchildren with autism, but they do not necessarily max-imize the efficiency of learning. This can be remediedwith probes of more sophisticated performances, butone necessarily expects errors to occur when task de-mands exceed present capabilities.

Placebo Effects in Single-Subject Experiments

One of the concerns that should be considered inquasi-experimental research is the possibility of placeboeffects. In particular, one would want to be confidentthat treatment effects are not attributable to the effectsof the added attention given to participants enrolled in atreatment study, to the effects of repeated testing or mea-surement, or other threats to internal validity. Repeatedmeasurements over time have the potential to minimizeconcerns about placebo effects if treatment effects areclearly and consistently associated with implementationof treatment conditions and if two other conditions aremet. First, investigators need to have reliable and validmeasures. Second, data collection within conditions, andin baseline or control conditions in particular, must be

sufficient to establish stability in performance. Interpre-tation of results is severely compromised when investi-gators are lax about the latter requirement (cf. Johnston& Pennypacker, 1993; Sidman, 1960).

These requirements do not discount the possibilitythat peers and teachers in educational settings, for ex-ample, are not displaying behaviors other than thoseunder examination that have an influence on childrenwith autism. Consequently, it is advantageous to keepindividuals and data collectors blind to the schedule oftreatment implementation. For example, children whoare taught communication skills through video model-ing might view brief videos that are unrelated to the skillsbeing taught until they are scheduled to begin the lan-guage intervention condition. In that way the classroomobservers, teachers, and others are unaware of when thetreatment is initiated (e.g., Hepting & Goldstein, 1996b).

Also, social validity assessments are gaining wideracceptance in the literature, whereby naïve raters makejudgments about the behavior of participants usuallybased on pre- and post-treatment videotapes presentedin a random order. These social validity assessmentsallow one to judge whether the treatment effects arereadily discernable. They also can help to validate themeasures taken by observers and researchers and alle-viate some of the concerns that changes are attributableto biased recording on their part.

CONCLUSIONS

This body of research has emanated from a num-ber of disciplines, primarily psychology and specialeducation, as well as communication sciences and dis-orders. It is notable that children with autism are over-represented in the language intervention literaturegenerally; approximately 26% of language interventionstudies include this comparatively small population interms of prevalence (Goldstein & Hockenberger, 1991).The bulk of the child language intervention researchhas sought to develop effective treatments for our mostchallenging cases and thus have targeted more rudi-mentary language skills. Two reasons for this empha-sis come to mind. First, teaching language to childrenwith severe communication deficits has profound im-pacts on the lives of both the children and the peoplewho interact with them everyday. Rudimentary lan-guage skills can make a tremendous difference in theability of children to control their environment. In ad-dition, language has been found to be an adaptive al-ternative to highly disruptive forms of communication(such as tantrums, aggression toward others, or self-

392 Goldstein

injurious behavior). Second, more information is avail-able on early language development than later languagedevelopment. Having a more complete knowledge ofdevelopmental milestones and their sequencing makesit easier to identify treatment goals that are develop-mentally appropriate. As the complexity of the lan-guage system grows, one begins to address the subtletiesand nuances of effective communication that are harderto define. The full complement of language skills thatcome into play as the demands of development and newsituations expand has yet to be defined adequately, letalone addressed in treatment studies. Indeed, knowl-edge of the effects of language treatment has developedincrementally. Experimental studies have focused(largely to their credit) on restricted treatment goalsand relatively short-term outcomes.

The accumulated knowledge of the effects of spe-cific treatments on particular target behaviors hasresulted in a compendium of fairly sophisticated in-structional procedures, but the development of aneffective technology is in its infancy. Because mostexperiments are designed to assess the effects of a re-stricted set of treatment components on certain aspectsof language development, it is conceivable that treat-ment programs that draw upon a number of empiricallysupported approaches and are individualized to meetthe specific needs of children with autism have the po-tential to have robust clinical effects. The process ofcompiling this technology into long-term, comprehen-sive treatment programs remains an ambitious, com-plex challenge. Speech-language pathologists andservice delivery teams are aware of the multifacetednature of communication and realize that the com-plexities of the developing language system that chil-dren must acquire to be fully competent communicatorsis not well understood. Nevertheless, they need to drawupon the existing knowledge base to develop treatmentprograms to address the multiple needs of individualchildren and their families. What behaviors should beincluded? How should skills from different communi-cation domains be sequenced? How should the programbe adapted based on the child’s age, educational envi-ronment, families’ wishes, cultural milieu, etc.? Few in-dividuals, be they researchers, teachers, speech-languagepathologists, or parents, would be satisfied with thescope and sequence reflected in available training pro-grams, many of which were developed decades ago.

Investigators interested in evaluating larger treat-ment packages face an even stiffer challenge. First, theyneed to identify appropriate outcome measures. Thereis considerable danger in depending upon global mea-sures of change in language development that may not

be sensitive to changes associated with their interven-tion efforts. Because of the relationship between lan-guage skills and other academic and social domains,one can argue that comprehensive language treatmentprograms are likely to have significant impacts on long-term outcomes, such as academic achievement, the ex-tent of special education services required, schooldropout rates, and vocational placements. How differentaspects of language relate to these important outcomeshas yet to be delineated. To help us understand theserelations and other moderating influences, it is impor-tant to measure primary and short-term outcomes. Also,it is essential that direct intervention effects are docu-mented to help us interpret results, especially whenlong-term outcomes are modest or absent.

Second, researchers must specify, measure, andcontrol the active variables in their interventions. Asmentioned earlier, it is difficult to extract informationon treatment procedures in the literature. This problemtends to be compounded in the case of comprehensivetreatment programs. Furthermore, one must specify theextent to which interventionists should be expected orallowed to individualize treatments to meet the needsof children with a broad range of developmental lev-els. It is critical that when comprehensive interventionsare evaluated that treatment components are specifiedin an operational manner. Measures need to be devel-oped to judge the fidelity of the treatment that is ad-ministered over time.

We need to encourage behavioral scientists tocontinue pursuing the promising results that have beensummarized in this review. When we have evidenceof treatments that are effective at teaching functionalcommunication skills, this is a beginning point andnot an end point. We have not discovered a silver bul-let and we are unlikely to do so. Teaching childrenwith autism tends to be a difficult, painstaking en-deavor. We have made much progress, but we mustrecognize that progress will continue to be slow if thescientific community continues to publish only aboutthree experimental studies per year that address sucha complex problem: teaching communication to chil-dren with autism. More investigators should be en-couraged to pursue these promising findings. A singlestudy may help to identify a promising treatment ap-proach, but a number of systematic replications areneeded to begin to:

● refine treatments by adding useful componentsor by eliminating superfluous ones.

● maximize the breadth of generalization and thedurability of treatment effects.

Communication Intervention 393

● delimit the contextual conditions under whichthe treatment protocol is effective.

● delineate with whom the treatment is more andless effective.

● identify other behaviors to which the treatmentcan be applied effectively.

● elucidate psychological mechanisms or learningprocesses that help us understand why the treat-ment is effective.

Based on this review, one can conclude that sub-stantial evidence exists to claim that effective interven-tions exist to teach communication skills to children withautism. Even though the vast majority of the studiesreviewed were conducted with few subjects, with fewexceptions investigators were able to demonstrateclearly and repeatedly that the initiation of their treat-ment was associated with improved communicativeperformance. In many cases, the effects are dramaticbecause control conditions show conventional com-munication to be so minimal for many children withautism. For children with at least rudimentary speechand language skills, who are verbally imitative, for ex-ample, there are an array of approaches that may yieldfairly rapid learning. For children who are nonverbal,more precise programming and slower progress islikely. Are there interventions that have been shown tobe efficacious in teaching relatively complete commu-nication systems to the most challenging children withautism? I do not think so. But a knowledge base isavailable to help guide efforts to expand the commu-nication repertoires of those challenging cases.

The literature provides little direction in terms ofservice delivery models or the intensity of services thatare more likely to maximize communication interven-tion efforts. Logical arguments can be made related toopportunities to learn. It is hard to imagine a variablethat is more predictive of learning than opportunitiesto respond (Greenwood, 1991). But that begs the ques-tion of the quality of instruction. The literature re-viewed is firmly grounded in principles of learningderived from operant and social learning perspectivesand their descendants. Effective communication inter-ventions will depend on good decision making on a va-riety of fronts. For example, careful programming isneeded: to select useful objectives, to provide envi-ronments that set the occasion for meaningful commu-nication, to provide functional reinforcers that eventuallyare available with regularity in the natural environment,and to provide scaffolds (models, prompts, corrections,and encouragement) that are faded to promote inde-pendent and spontaneous communication. Developing

comprehensive communication intervention programsfor children with autism represents a major challenge,perhaps involving a healthy dose of art. Yet, the avail-able literature provides a scientific foundation thatshould allow us to discriminate interventions that arepurely art and imagination from those that are builtupon treatment procedures with empirical support.

REFERENCES

Barrera, R. D., Lobatos-Barrera, D., & Sulzer-Azaroff, B. (1980). Asimultaneous treatment comparison of three expressive languagetraining programs with a mute autistic child. Journal of Autism& Developmental Disorders, 10, 21–37.

Barrera, R. D., & Sulzer-Azaroff, B. (1983). An alternating treatmentcomparison of oral and total communication training programswith echolalic autistic children. Journal of Applied BehaviorAnalysis, 16, 379–394.

Bird, F., Dores, P. A., Moniz, D., & Robinson, J. (1989). Reducingsevere aggressive and self-injurious behaviors with functionalcommunication training. American Journal on Mental Retarda-tion, 94, 37–48.

Brady, D., & Smouse, A. D. (1978). A simultaneous comparison ofthree methods for language training with an autistic child: Anexperimental single case analysis. Journal of Autism and ChildSchizophrenia, 8(3), 271–279.

Buffington, D. M., Krantz, P. J., McClannahan, L. E., & Poulson,C. L. (1998). Procedures for teaching appropriate gestural com-munication skills to children with autism. Journal of Autism andDevelopmental Disorders, 28, 535–545.

Carr, E. G., & Dores, P. A. (1981). Patterns of language acquisitionfollowing simultaneous communication with autistic children.Analysis & Intervention in Developmental Disabilities, 1, 347–361.

Carr, E. G., & Durand, V. M. (1985). Reducing behavior problemsthrough functional communication training. Journal of AppliedBehavior Analysis, 18, 111–126.

Carr, E. G., Pridal, C., & Dores, P. A. (1984). Speech versus signcomprehension in autistic children: Analysis and prediction.Journal of Experimental Child Psychology, 37, 587–597.

Casey, L. O. (1978). Development of communicative behavior inautistic children: A parent program using manual signs. Jour-nal of Autism and Child Schizophrenia, 8, 45–60.

Charlop, M. H., & Milstein, J. P. (1989). Teaching autistic childrenconversational speech using video modeling. Journal of AppliedBehavior Analysis, 22, 275–285.

Charlop, M. H., Schreibman, L., & Thibodeau, M. G. (1985). In-creasing spontaneous verbal responding in autistic children usinga time delay procedure. Journal of Applied Behavior Analysis,18, 155–166.

Charlop, M. H., & Trasowech, J. E. (1991). Increasing children’sdaily spontaneous speech. Journal of Applied Behavioral Analy-sis, 24, 747–761.

Charlop, M. H., & Walsh, M. E. (1986). Increasing autistic children’sspontaneous verbalizations of affection: An assessment of timedelay and peer modeling procedures. Journal of Applied Be-havior Analysis, 19, 307–314.

Cole, K., Mills, P. E., Dale, P. S., & Jenkins, J. R. (1996). Preschoollanguage facilitation methods and child characteristics. Journalof Early Intervention, 20, 113–131.

Davis, C. A., Brady, M. P., Hamilton, R., McEvoy, M. A., &Williams, R. E. (1994). Effects of high-probability requests onthe social interactions of young children with severe disabili-ties. Journal of Applied Behavior Analysis, 27, 619–637.

394 Goldstein

Davis, C. A., Brady, M. P., Williams, R. E., & Hamilton, R. (1992).Effects of high-probability requests on the acquisition andgeneralization of responses to requests in young children withbehavior disorders. Journal of Applied Behavior Analysis, 25,905–916.

Durand, V. M., & Carr, E. G. (1987). Social influences on “self-stimulatory” behavior: Analysis and treatment applications.Journal of Applied Behavior Analysis, 20, 119–132.

Durand, V. M., & Carr, E. G. (1992). An analysis of maintenancefollowing functional communication training. Journal of Ap-plied Behavior Analysis, 25, 777–794.

Egel, A. L., Shafer, M. S., & Neef, N. A. (1984). Receptive acquisi-tion and generalization of prepositional responding in autisticchildren: A comparison of two procedures. Analysis and Inter-vention in Developmental Disabilities, 4, 285–298.

Elliott, R. O., Hall, K., & Soper, H. V. (1991). Analog languageteaching versus natural language teaching: Generalization andretention of language learning for adults with autism and men-tal retardation. Journal of Autism & Developmental Disorders,21, 433–447.

Foxx, R., Faw, G., McMorrow, M., Kyle, M., & Bittle, R. (1988).Replacing maladaptive speech with verbal labeling responses:An analysis of generalized responding. Journal of Applied Be-havior Analysis, 21, 411–417.

Goldstein, H., & Angelo, D., & Mousetis, L. (1987). Acquisition andextension of syntactic respertoires by severely mentally retardedyouth. Research in Developmental Disabilities, 8, 549–574.

Goldstein, H., & Brown, W. (1989). Observational learning of re-ceptive and expressive language by preschool children. Educa-tion and Treatment of Children, 12, 5–37.

Goldstein, H., & Cisar, C. L. (1992). Promoting interaction duringsociodramatic play: Teaching scripts to typical preschoolers andclassmates with handicaps. Journal of Applied Behavior Analy-sis, 25, 265–280.

Goldstein, H., & Hockenberger, E. H. (1991). Significant progressin child language intervention: An 11-year retrospective. Re-search in Developmental Disabilities, 12, 401–424.

Goldstein, H., Kaczmarek, L., Pennington, R., & Shafer, K. (1992).Peer-mediated intervention: Attending to, commenting on, andacknowledging the behavior of preschoolers with autism. Jour-nal of Applied Behavior Analysis, 25, 289–305.

Goldstein, H., & Wickstrom, S. (1986). Peer intervention effects oncommunication interaction among handicapped and nonhandi-capped preschoolers. Journal of Applied Behavior Analysis, 19,209–214.

Goldstein, H., Wickstrom, S., Hoyson, M., Jamieson, B., & Odom,S. L. (1988). Effects of sociodramatic script training on socialand communicative interaction. Education and Treatment ofChildren, 11, 97–117.

Green, G. (1994). The quality of the evidence. In H. C. Shane (Ed.),Facilitated communication: The clinical and social phenome-non (pp. 157–225). San Diego: Singular.

Greenwood, C. R. (1991). A longitudinal analysis of time, engage-ment, and academic achievement in at-risk vs. non-risk students.Exceptional Children, 57, 521–535.

Greenwood, C. R., & Delquadri, J. (1995). ClassWide Peer Tutoringand the prevention of school failure. Preventing School Failure,39(4), 21–25.

Handleman, J. S. (1979). Generalization by autistic-type children ofverbal responses cross-settings. Journal of Applied BehaviorAnalysis, 12, 283–294.

Handleman, J. S. (1981). Transfer of verbal responses across in-structional settings by autistic-type children. Journal of Speechand Hearing Disorders, 46, 69–76.

Haring, T. G., & Kennedy, C. H. (1990). Contextual control of prob-lem behavior in students with severe disabilities. Journal ofApplied Behavior Analysis, 23, 235–243.

Harris, S. L., Handleman, J. S., Kristoff, B., Bass, L., & Gordon, R.(1990). Changes in language development among autistic and

peer children in segregated and integrated preschool settings.Journal of Autism and Developmental Disorders, 20, 23–31.

Harris, S. L., Wolchik, S. A., & Milch, R. E. (1982). Changing thespeech of autistic children and their parents. Child & FamilyBehavior Therapy, 4, 151–173.

Harris, S. L., Wolchik, S. A., & Weitz, S. (1981). The acquisition oflanguage skills by autistic children: Can parents do the job?Journal of Autism & Developmental Disorders, 11, 373–384.

Hart, B., & Risley, T. R. (1975). Incidental teaching of languagein the preschool. Journal of Applied Behavior Analysis, 8,411– 420.

Hepting, N. H., & Goldstein, H. (1996a). What’s natural about nat-uralistic language intervention? Journal of Early Intervention,20, 249–265.

Hepting, N. H., & Goldstein, H. (1996b). Requesting by preschool-ers with developmental disabilities: Videotape self-modelingand learning of new linguistic structures. Topics in Early Child-hood Special Education, 16, 407–427.

Hinerman, P. S., Jenson, W. R., Walker, G. R., & Peterson, P. B.(1982). Positive practice overcorrection combined with addi-tional procedures to teach signed words to an autistic child. Jour-nal of Autism & Developmental Disorders, 12, 253–263.

Horner, R. H., & Budd, C. M. (1985). Acquisition of manual signuse: Collateral reduction of maladaptive behavior and factorslimiting generalization. Education and Training of the MentallyRetarded, 20, 39–47.

Howlin, P., & Rutter, M. (1989). Mothers’ speech to autistic chil-dren: A preliminary causal analysis. Journal of Child Psychol-ogy and Psychiatry, 30, 819–843.

Hwang, B., & Hughes, C. (2000). Increasing early social-commu-nicative skills of preverbal preschool children with autismthrough social interactive training. Journal of the Associationfor Persons with Severe Handicaps, 25, 18–28.

Ihrig, K., & Wolchik, S. A. (1988). Peer versus adult models andautistic children’s learning acquisition, generalization, andmaintenance. Journal of Autism & Developmental Disorders,18, 67–79.

Ingenmey, R., & Van Houten, R. (1991). Using time delay to pro-mote spontaneous speech in an autistic child. Journal of AppliedBehavior Analysis, 24, 591–596.

Jacobson, J. W., & Mulick, J. A. (1995). A history of facilitated com-munication: Science, pseudoscience, and antiscience. AmericanPsychologist, 50, 750–765.

James, S. D., & Egel, A. L. (1986). A direct prompting strategy forincreasing reciprocal interactions between handicapped andnonhandicapped siblings. Journal of Applied Behavior Analy-sis, 19, 173–186.

Jocelyn, L. J., Casiro, O. G., Beattie, D., Bow, J., & Kneisz, J. (1998).Treatment of children with autism: A randomized controlled trialto evaluate a caregiver-based intervention program in commu-nity day-care centers. Developmental & Behavioral Pediatrics,19, 326–334.

Johnston, J. M., & Pennypacker, H. S. (1993). Strategies and tacticsof behavioral research. Hillsdale, NJ: Lawrence Erlbaum.

Kaiser, A. P., Yoder, P. J., & Keetz, A. (1992). Evaluating milieuteaching. In S. F. Warren & J. Reichle (Eds.), Causes and ef-fects in communication and language intervention (pp. 9–47).Baltimore, MD: Paul H. Brookes.

Keogh, D., Whitman, T. L., Beeman, D., & Halligan, K. & Starzyn-ski, T. (1987). Teaching interactive signing in a dialogue situ-ation to mentally retarded individuals. Research in DevelopmentalDisabilities, 8, 39–53.

Koegel, R. L., Camarata, S., Koegel, L. K., Ben-Tall, A., & Smith,A. E. (1998). Increasing speech intelligibility in children withautism. Journal of Autism and Developmental Disorders, 28,241–251.

Koegel, R. L., Camarata, S., Valdez-Menchaca, & Koegel, L.(1998). Setting generalization of question-asking by childrenwith autism. American Journal on Mental Retardation, 102,346–357.

Communication Intervention 395

Koegel, R. L., Koegel, L. K., & Surratt, A. (1992). Language inter-vention and disruptive behavior in preschool children withautism. Journal of Autism and Developmental Disorders, 22,141–153.

Koegel, R. L., O’Dell, M. C., & Dunlap, G. (1988). Producing speechuse in nonverbal autistic children by reinforcing attempts. Jour-nal of Autism and Developmental Disorders, 18, 525–538.

Koegel, R. L., O’Dell, M. C., & Koegel, L. K. (1987). A natural lan-guage teaching paradigm for nonverbal autistic children. Journalof Autism & Developmental Disorders, 17, 187–200.

Krantz, P. J., & McClannahan, L. E. (1993). Teaching children withautism to initiate to peers: Effects of a script-fading procedure.Journal of Applied Behavior Analysis, 26, 121–132.

Krantz, P. J., Zalewski, S., Hall, L., Fenski, E., & McClannahan, L.(1981). Teaching complex language to autistic children. Analy-sis & Intervention in Developmental Disabilities, 1, 259–297.

Laski, K., Charlop, M., & Schreibman, L. (1988). Training parentsto use the natural language paradigm to increase their autisticchildren’s speech. Journal of Applied Behavior Analysis, 21,391–400.

Layton, T. (1988). Language training with autistic children usingfour different modes of presentation. Journal of CommunicationDisorders, 21, 333–350.

Matson, J. L., Sevin, J. A., Box, M. L., Francis, K. L., & Sevin, B. M.(1993). An evaluation of two methods for increasing self-initiatedverbalizations in autistic children. Journal of Applied BehaviorAnalysis, 26, 389–398.

McGee, G. C., Almeida, M. C., Sulzer-Azaroff, B., & Feldman, R. S.(1992). Promoting reciprocal interactions via peer incidentalteaching. Journal of Applied Behavior Analysis, 25, 117–126.

McGee, G. G., Krantz, P. J., & McClannahan, L. E. (1984). Con-versational skills for autistic adolescents: Teaching assertive-ness in naturalistic game settings. Journal of Autism &Developmental Disorders, 14, 319–330.

McGee, G. G., Krantz, P. J., & McClannahan, L. E. (1985). Thefacilitative effects of incidental teaching on preposition useby autistic children. Journal of Applied Behavior Analysis, 18,17–31.

McIlvane, W. J., Bass, R. W., O’Brien, J. M., Gerovac, B. J., & Stod-dard, L. T. (1984). Spoken and signed naming of foods afterreceptive exclusion training in severe retardation. Applied Re-search in Mentally Retarded, 5, 1–28.

McIlvane, W. J., Dube, W. V., Green, G., & Serna, R. W. (1993).Programming conceptual and communication skill development:A methodological stimulus class analysis. In A. P. Kaiser &D. B. Gray (Eds.), Understanding childrens’ language (pp. 243–285). Baltimore, MD: Paul Brookes.

McMorrow, M. J., & Foxx, R. M. (1986). Some direct and general-ized effects of replacing an autistic man’s echolalia with cor-rect responses to questions. Journal of Applied BehaviorAnalysis, 19, 289–297.

McMorrow, M. J., Foxx, R. M., Faw, G. D., & Bittle, R. G. (1987).Cues-pause-point language training: Teaching echolalics func-tional use of their verbal labeling repertoires. Journal of AppliedBehavior Analysis, 20, 11–22.

Neef, N. A., Walters, J., & Egel, A. L. (1984). Establishing genera-tive yes/no responses in developmentally disabled children.Journal of Applied Behavior Analysis, 17, 453–460.

Ozonoff, S., & Cathcart, K. (1998). Effectiveness of a home programintervention for young children with autism. Journal of Autismand Developmental Disorders, 28, 25–32.

Rogers, S. J. (1998). Empirically supported comprehensive treat-ments for young children with autism. Journal of Clinical ChildPsychology, 27, 168–179.

Schreibman, L., & Carr, E. G. (1978). Elimination of echolalicresponding to questions through the training of a generalizedverbal response. Journal of Applied Behavior Analysis, 11, 453–463.

Secan, K. E., Egel, A. L., & Tilley, C. S. (1989). Acquisition, gen-eralization, and maintenance of question-answering skills in

autistic children. Journal of Applied Behavior Analysis, 22,181–196.

Shafer, M. S., Egel, A. L., & Neef, N. A. (1984). Training mildlyhandicapped peers to facilitate changes in the social interactionskills of autistic children. Journal of Applied Behavior Analy-sis, 17, 461–476.

Sidman, M. (1960). Tactics of scientific research. New York, NY:Basic Books.

Stokes, T. F., & Baer, D. M. (1977). An implicit technology of gen-eralization. Journal of Applied Behavior Analysis, 10, 349–367.

Wacker, D. P., Steege, M. W., Northup, J., Sasso, G., Berg, W.,Reimers, T., Cooper, L., Cigrand, K., & Donn, L. (1990). Acomponent analysis of functional communication training acrossthree topographies of severe behavior problems. Journal ofApplied Behavior Analysis, 23, 417–429.

Wacker, D., Wiggins, B. Fowler, M, & Berg, W. (1988). Trainingstudents with profound or multiple handicaps to make requests

396 Goldstein

via microswitches. Journal of Applied Behavior Analysis, 21,331–343.

Wherry, J. N., & Edwards, R. P. (1983). A comparison of verbal,sign, and simultaneous systems for the acquisition of receptivelanguage by an autistic boy. Journal of Communication Dis-orders, 16, 201–216.

Yoder, P. J., Kaiser, A. P., Goldstein, H., Alpert, C., Mousetis, L.,Kaczmarek, L., & Fischer, R. (1995). An exploratory compari-son of milieu teaching and responsive interaction in classroomapplications. Journal of Early Intervention, 19, 218–242.

Yoder, P. J., & Layton, T. L. (1988). Speech following sign languagetraining in autistic children with minimal verbal language. Jour-nal of Autism & Developmental Disorders, 18, 217–230.

Yoder, P. J., & Warren, S. F. (1999). Facilitating self-initiated proto-declaratives and proto-imperatives in prelinguistic children withdevelopmental disabilities. Journal of Early Intervention, 22,337–354.