00100		THE RATIONALE FOR COMPUTER-BASED TREATMENT OF
00200	        LANGUAGE DIFFICULTIES IN NONSPEAKING AUTISTIC
00300		             CHILDREN
00400	
00500		        KENNETH MARK COLBY
00600	
00700		No one person  could  conceive,  implement  and  utilize  the
00800	treatment  method I shall describe shortly. Hence let me say right at
00900	the start that this award honors , besides myself, my  co-workers  at
01000	Stanford  University-  Horace  Enea,  David  Smith, Malcolm Newey and
01100	Maxine Colby, each of whom has put years of effort into this project.
01200	As Casey Stengel, manager of the New York Yankees said once when they
01300	won the World Series, `I couldn't have done it without the players'.
01400		We began about seven years ago with a notion and  two  facts.
01500	The notion was that a nonspeaking autistic child's primary difficulty
01600	was an inability to process symbols, language  being  of  course  the
01700	most important symbolic system used in human communication. The first
01800	fact, oft repeated in the literature, indicated  that  prognosis  was
01900	highly  correlated  with  speech,  the  outlook for nonspeakers being
02000	poor. The second fact was the common clinical observation that  these
02100	children  played  for hours with machines while remaining indifferent
02200	to interactions with people.
02300		There  is  now increasing evidence in the research literature
02400	supporting  this notion of  a  primary  difficulty  in
02500	symbol  processing.[Rutter,  Firth, Campbell, Hermelin and O'Connor].
02600	A dysphasic or aphasic child also has difficulty with language but he
02700	can  acquire  other symbolic systems such a gesturing and drawing and
02800	even learn to read.  But the nonspeaking  autisitc  child  has  great
02900	difficulty with all symbolic processes, not just language.
03000		The cause of this difficulty remains unknown.   Nowadays  few
03100	experts  in  the  field  defend  a  psychogenic  etiology  since  the
03200	supporting evidence is weak and the disconfirming evidence is gaining
03300	in  strength [Rutter]. Regardless of the original cause, if we understand the
03400	crux of the child's difficulty, we should try to  devise  a  remedial
03500	treatment  which  takes  advantage  of  the  fact  that  the child is
03600	fascinated by machines. Our beginning working idea was  to  create  a
03700	machine  a  nonspeaking  child  could play with, a machine which also
03800	happens to be a primitive linguistic entity.
03900		The  machine  consists  of  a  television-like  screen  and a
04000	typewriter-like keyboard. Pressing the keys on  the  keyboard  causes
04100	symbols to appear on the screen accompanied by sounds of human voices
04200	and other noises common in a child's life. It  is  much  like  having
04300	your  own Sesame Street to play with. But rather than being a passive
04400	recipient of what the  television  show  provides,  a  child  in  our
04500	situation  is  an  active  agent  controlling the machine. Instead of
04600	having things done to him, things are done by him. The merits of  the
04700	machine  are that it is untiring, predictable, always saying the same
04800	thing the same way, never  angry,  never  bored  and  controllable-
04900	properties   which   are   notoriously   lacking   in   humans.  This
05000	audio-visual- tactile experience is provided by  a  computer  program
05100	running  on  a  remote PDP-10 in the Stanford Artificial Intelligence
05200	Project. The program is divided into games intended to give the child
05300	various  types  of opportunities at playing with and interacting with
05400	symbols. For example, in one game,  when  a  child  presses  the  key
05500	showing  the  letter  H,  an H appears on the screen and a voice says
05600	"H". In another game pressing the key labelled "H" produces a running
05700	horse on the screen accompanied by the sound of horse's hoofs.  There
05800	exist over 1000  such  experiences  on  the  system.  The  games  are
05900	organized  at various levels of complexity and are designed to show a
06000	child how English is put together from sounds and letters into  words
06100	and  expressions.  I shall not go into the details of the games here.
06200	They are throughly described in the literature.[ ]. Instead  I  shall
06300	try to sketch the major principles underlying this approach.
06400		Let  us  start  by  considering  how  normal children acquire
06500	language.  They are not taught formally as an adult learns  a  second
06600	language.  Children  are  simply  exposed  to members of a linguistic
06700	community and given an  opportunity  to  explore  language  usage  in
06800	everyday communication between themselves and other humans. From this
06900	exposure and exploration they come to pair the sounds and meanings of
07000	words.  Between the ages of roughly 8 months and 36 months they spend
07100	thousands of hours practicing and playing  with  language.  For  some
07200	unknown reason a nonspeaking autistic child does not show this course
07300	of development.
07400		It is  not  simply  the  development  of  language  which  is
07500	important  for a child but also the acquisition of concepts necessary
07600	for a stable and comprehensive model of  the  world, especially  the  human  world.  The
07700	conceptual  or  cognitive  deficits  shown  by  nonspeaking  autistic
07800	children involve those concepts which are normally  acquired  through
07900	language or other symbolizations.  Take  the abstract concept of `danger'. To prevent a child
08000	from becoming hurt, a parent must identify certain concrete objects and
08100	situations  as  dangerous until the child grasps the abstract concept
08200	of danger. All this is done linguistically. Through language  objects
08300	and situations can be referred to and warned about even when they are
08400	not present, and referred to without pointing when they are present.
08500	A  social  function  of language is to mark off for a child
08600	what to pay attention to and what is to be done and not done. Many of
08700	the  behaviors  of  autistic  children  can  be  viewed  as secondary
08800	consequences to an inability  to  form  regular  conceptual  patterns
08900	about  the  world  because  the  necessary  concepts acquired through
09000	language are  missing.  As  one  perceptive  mother  said  about  her
09100	autistic child,` a screw is not loose , a screw is missing'.
09200		A poorly developed ability to  process  symbols  has  further
09300	consequences  besides  conceptual  deficits. Without language a human
09400	has no awareness of being aware. He lacks the ability to self-monitor
09500	and to self-control by talking to himself. He cannot use symbols reflexively, that is, to
09600	to give himself orders and to comment on himself to himself. Having language
09700	a normal child comes to realize the self is really two which can talk to one another.
09800	Finally, to become human one must be treated as a human by other people who treat you as if
09900	you had self-awareness and  self-control.  It  is  easy  to  see  how
10000	disastrous are the many consequences of lacking language.
10100	
10200		Thus  far  I  have  spoken  of  autistic  children  as if the
10300	nosological label stood for a single homogeneous group. Thirty  years
10400	ago  it  seemed that might be the case but now it appears as if there
10500	exist several autistic syndromes, none of which  should  be  confused
10600	with  childhood  schizophrenia  since  they  differ in onset, course,
10700	symptoms, family  history  and  prognosis.  There  are  speaking  and
10800	nonspeaking  autistic children. Among the nonspeaking group there are
10900	those whose linguistic development is normal until  sometime  in  the
11000	second  year when they lose their language abilities. The other major
11100	group are those children who,  from  the  start,  have  trouble  with
11200	language,  understanding  little  and  saying  even less, perhaps one
11300	`mama' or `no' a year. In our experience the most  difficult  problem
11400	for differential diagnosis lies in deciding whether a nonspeaking child
11500	suffers from  dysphasia,  autism  or  perhaps  both.  Over  time  the
11600	distinction  can  be  made  when  it becomes clear that the dysphasic
11700	child can mimic, draw pictures and greet  while  the  autistic  child
11800	cannot.
11900	
12000	     Taking a cue from the normal child who treats language as a toy,
12100	our first principle was that the treatment should provide an opportunity
12200	for exploratory play.  The treatment situation is not one of forced
12300	drill, instruction and training but one of play with the keyboard and
12400	display. Operant conditioning methods reward  the child with candy or food for his
12500	actions.  We do not, believing that food rewards inhibit exploratory
12600	curiousity, as has been shown experimentally by Harlow and by Nissen.
12700	Exploratory learning requires a situation which invites exploration, 
12800	time, security and minimal interference by adults.
12900	
13000	     In each of the sessions the child has a `sitter', an adult whose
13100	main task is to sit and not interfere.  This is difficult to do,
13200	especially if you have been trained to DO things.  We want to give
13300	the child an opportunity to freely self-select those symbols which
13400	interest him, rather than have an adult instruct him or quiz him
13500	about those symbols which he `should' learn.  The sitter's behavior
13600	is crucial to this treatment method if the spirit of play rather than
13700	performance is to be provided.  Of course the sitter offers social
13800	approval and encouragement when it is fitting.
13900	
14000	     An ideal treatment session results when the child is in a good
14100	mood, is interested in working the keyboard display, enjoys imitating
14200	the sounds and is successful in getting the machine to do what he wants.
14300	The principle of success is important here because, in our view, many
14400	nonspeakers have given up on language.  They have failed over and over
14500	and hence withdraw from trying.  We do not let them fail.  As one normal
14600	child said about the experience "it`s fun, you can't lose".  Years ago
14700	it was thought that nonspeaking autistic children were innately withdrawn
14800	from people and hence did not acquire language. We feel it is the other
14900	way around; they have so much difficulty with language they withdraw from
15000	people who unwittingly flood and overwhelm them with meaningless noises.
15100	No wonder they do not call or address or ask questions of these giants who immediately spout
15200	gibberish.  Autistic children are not aloof and indifferent to all 
15300	people -- only to those who talk.  It is not eye-to-eye contact they
15400	avoid but eye-to-a-mouth which provudes, as far as the child is
15500	concerned, gibberish.  If you say little or nothing to a nonspeaking
15600	child, you will soon find him in your lap, as affectionate as any
15700	other child.
15800	
15900	     Incidentally, and to digress, there is another myth in the
16000	literature about autistic children.  They are said to show pronominal
16100	reversal, using the third person pronoun `you' for the first person
16200	pronoun `I`, and `I' for 'you'.  This is not so.  It is the normal
16300	child who reverses personal pronouns.  The autistic child LACKS the
16400	rule for correct pronounc assignments.  Pronouns have deistic functions, 
16500	
16600	i.e. the referents of the symbols varies with the situation in contrast
16700	to, for example, proper names which are individual constants.  The
16800	assignment rule for the variables `I' and `you' is that the speaker
16900	calls himself `I' and calls the listener `you'.  Young autistic children
17000	do not possess this rule, which is not actively taught by adults, but
17100	is grasped by the normal child in observing human dialogues.  The
17200	autistic child, unable to process language, echoes back, in an unprocessed
17300	or uninterpreted form, what is said to him.  If you say to him 'Do you
17400	want your jacket' he will echo it.  He will also echo 'Do I want my
17500	jacket' showing he can say 'I' but doesn't grasp what it refers to.
17600	Everybody calls him 'you' so he believes 'you' must be one of his
17700	names.  Due to the limitations of short-term memory, an autistic child
17800	may echo only the last fragment of a long expression.  `I' usually appears
17900	early in an expression as its subject.  If you place the pronoun `I' at
18000	the end of an expression, an autistic child can echo it as easily as "	`you'.
18100	
18200	     Returning from this digression on deixis, let me discuss our
18300	successes and our failures, especially the latter.  Every treatment
18400	method reports its dramatic successes with one or two cases.  But we need
18500	case series of at least 30 before we can compare methods and decide
18600	which is more effective relative to the effort required.  Thus far our
18700	series of nonspeaking autistic children numbers       with      %
18800	successes and      % failures.  By success we mean only that the child
18900	begins voluntarily to use speech for social communication.  We do not
19000	claim the method results in normal language ability with correct
19100	pronunciation, syntax and grammar.  Our idea is to rekindle the
19200	child's interest in attempting speech, to get him ti try again, to get
19300	him off the ground, to catalyze a damaged or slow-developing natural
19400	process of language acquisition.  By failure we mean the child leaves
19500	us as he came, perhaps with some language understanding but no useful
19600	speech.
19700	
19800	     When one fails several times, as we have, one naturaly tries to
19900	find some common denominator.  We have found only two in our case
20000	series.  The first is that all our failure cases are children who
20100	showed no interest in playing with the machine.  Regardless of our
20200	coaxing and persuading they would have nothing to do with it.  Finally,
20300	in time and becoming desperate, we were forced to violate one of our
20400	basic principles of allowing free play.  We would spend weeks and
20500	months holding the child at the keyboard, pushing the keys for him,
20600	trying to overcome his resistance and negativism.  But to no avail.
20700	Somehow we must be more ingenious in capturing these children's
20800	interest.  We have thought of ways to do this and are currently
20900	trying to implement them.
21000	
21100	     A second characteristic of our failures is late autism.  That is,
21200	in cases where the child shows normal linguistic development until 16
21300	or 22 months and then stops talking, we have failed absolutely.  This
21400	history is suggestive of course of some type of cerebral insult such
21500	as a virus infection or auto-immune reaction, but these hypotheses
21600	remain unconfirmed.
21700	
21800	     Based on our experience we believe there is something worthwhile
21900	in the method.  As I described it, he treatment is a complex variable,
22000	a shotgun prescription, the active ingredients of whih are unclear.
22100	We need other workersin the field to adopt the method, improve it and
22200	hopefully help us find stronger catalysts for language acquisition.
22300	
22400	     Why haven't other workers tried this zero-risk and potentially
22500	promising method?  Aside from the inertia which all new methods must
22600	face, there are two reasons.  One is that everyone has his own pet 
22700	magic which he wants to utilize and even improvve.  The other involves
22800	people's beliefs and misconceptions about computers.  I have come
22900	across people with the following fantasy about our situation --
23000	that we have a small, cowering, bewildered child sitting in front
23100	of a towering bank of flashing lights {the New Yorker-cartoon concept of a
23200	computer} while cold scientists in white coats observve him thru
23300	one-way mirrors.  They say 'the child is already too interested in
23400	machines, he needs relations with people -- you are making him into
23500	even more of a robot'.  But I hope from my brief description you can
23600	see this fantasy is not the case.  It may sound paradoxical, but some
23700	nonspeaking children can become more human, i.e. become language users,
23800	by way of a machine which talks and which they find more acceptable
23900	on their own terms than they do human linguistic entities.  
24000	
24100	     A more realistic objection to computers is their cost.  Few people
24200	in the world have access to million dollar computers for this sort of
24300	work.  But one does not need a large computer to carry out this method.
24400	Mini-computers costing only a few thousand dollars are adequate.  It may
24500	be that one does not need a computer at all if someone finds a way of
24600	rapidly random-accessing both sounds and pictures.  We are currently
24700	working on this possibility and perhaps once we get the computer out
24800	of cthe picture, others will join us in developing these efforts.