Journal of Autism and Developmental Disorders, Vol. 21, No. 2, 1991
Neologisms and Idiosyncratic Language in Autistic Speakers 1 Joanne Volden Red Deer College, Red Deer, Alberta
Catherine Lord 2 Department of Psychiatry, University of North Carolina, Chapel Hill
Language samples from matched groups of 80 autistic, mentally handicapped, and normally developing children were coded for the presence of neologisms and~or idiosyncratic language use. Cognitive, socia~ or linguistic factors that might account for these errors were identified and assessed. More autistic subjects used neologisms and idiosyncratic language than age- and language skill-matched control groups. No single factor or combination of factors was responsible for this difference. Across diagnostic groups, similar patterns of error were noted, except that the autistic subjects were more likely to use words inappropriately that had no phonological or semantic similarity to the intended English word. For the autistic groups, the frequency of idiosyncratic language increased with language complexity. In contrast, such errors decreased with language skill in the mentally handicapped group. Theoretical implications for determining the nature and source of the language disorder associated with autism are discussed
INTRODUCTION T h e n a t u r e and source of the language disorder associated with autism has generated considerable debate in recent years. Controversy has IThe authors acknowledge the assistance in transcription and coding of Cathy Mulloy, Lynn Anderson-Cook, Heather Jordan, and Sharon Storoschuk, and the financial support of NSERC grant 5547740, and AHFMR grant 04764, held by the second author. 2Address all correspondence to Catherine Lord, Division TEACCH, Medical School Wing E, CB 7180, Chapel Hill, North Carolina 27599-7180. 109
0162-3257/91/0600-0109506.50/0 9 1991PlenumPublishingCorporation
110
Volden and Lord
focused on what sort of model best describes the language deficit of autistic persons. Proposals for such a model have ranged from a linguistic-specific disorder (Rutter, 1978) to accounts of the unusual language in terms of deviant social and/or cognitive development (Fein, Skoff, & Mirsky, 1981; Rutter, 1983). The purpose of the present investigation was to identify two aspects of language seen in autism, neologisms and idiosyncratic use of words and phrases and, at the level of single words and sentences, assess the extent to which these features could be accounted for by simple social, cognitive, or linguistic-specific deficits. Most, if not all, persons with autism show delays in the acquisition of receptive and expressive language compared to development of nonverbal skills. Given overall delays, sequences within the specific areas of phonology (Bartolucci, Pierce, Streiner, & Tolkin-Eppel, 1976), syntax, and semantics (Tager-Flusberg, 1981, 1985; Waterhouse & Fein, 1982) have been reported to follow relatively normal patterns of development. However, seldom has a single study examined more than one of these language domains. Further, most studies have used very simple analyses focusing on aspects of syntax and semantics that are normally acquired very early in development (Bartak, Rutter, & Cox, 1975; Cantwell, Baker, & Rutter, 1978; Howlin, 1984). Yet, in contrast to findings of normal developmental sequences; specific, deviant aspects of autistic speakers' communication that are not readily attributable to delay have also been noted (Baltaxe & Simmons, 1983; Cohen et al., 1978). Descriptions and case reports repeatedly provide vivid examples of markedly unusual utterances. It is these utterances that are the focus of this paper. Neologisms are defined as "non-words" or words that are "obviously peculiar" (LeCouteur et al., 1989). More formally, these are words that are not included in a standard lexicon of adult native language speakers. Kanner, in 1946, first commented on their presence in the speech of persons with autism. The notion emerged again in Rutter's (1965) investigation of the language characteristics of 63 autistic children. Of the 32 children who had some functional language, 3 had, at one time, "invented their own words for things" (p. 41). However, as far as we know, the factors contributing to neologisms in the language of autistic persons have not been addressed specifically. Idiosyncratic use of words and phrases is the second feature examined in this paper. This phenomenon has defied consistent description. The language of autistic persons has been characterized as "literal," "peculiar and out of place in ordinary conversation," "irrelevant and metaphorical" (Kanner, 1946); "stereotyped," "metaphorical," and "inappropriate" (Bartak et al., 1975, p. 345); "formal, [demonstrating] a lack of ease in the use of words" (Rutter, 1965, p. 41). The central characteristic of this feature seems
Neologisms and Idiosyncratic Language in Autism
111
to be the use of standard, familiar words or phrases in idiosyncratic, but meaningful ways. The broadest characterization of the phenomenon was the one coined by Cantwell et al. (1978, p. 356) as abnormal functional use. This term was intended to include all of the above abnormalities as well as immediate and delayed echolalia. For the purposes of this study, idiosyncratic language is defined as the use of conventional words or phrases in unusual ways to convey specific meanings. Our focus is on language abnormalities within words or sentence boundaries. Differences that distinguish this phenomenon from figurative language in the normal population are subtle, but have nonetheless been reliably identified (LeCouteur et al., 1989; Lord et al., 1989). This broad definition differs from neologisms only in that idiosyncratic language specifies the use of standard English words. Thus, while in this study we coded true neologisms separately, it was our assumption that neologisms and metaphorical and peculiar speech fall onto two continua of idiosyncratic language use: (a) conventionality of form ranging from unconventional or nonstandard English forms, through modified or partially conventional words, to fully conventional or standard English usage; and (b) complexity from single words to complex sentences. We expected that neologisms (at the extreme end of both continua) and idiosyncratic utterances (including pedantic speech, which would be the unusual use of conventional and overly complex words and phrases at the other end of the continua) to be accounted for by the same set of linguistic, cognitive and social factors. Like neologisms, idiosyncratic language has received little systematic exploration. Bartak et al. (1975) studied groups of autistic and developmentally receptive language-delayed children similar in terms of mean length of utterance (MLU) and grammatical complexity of speech. When their parents were interviewed, the autistic children were described as using "stereotypic utterances," "metaphorical language," and "inappropriate utterances" significantly more often than the dysphasic group. In a follow-up study of the same subjects, Cantwell et al. (1978) sought to test these parental impressions by analyzing the spontaneous language of 12 autistic and 12 dysphasic boys matched on age and nonverbal IQ. Language samples were analyzed for syntactic complexity by evaluating spontaneous, appropriate use of several elementary morphemes, phrasestructure rules, and transformational rules. Samples were also coded according to several socialized (i.e., normal) functional speech categories and a few "abnormal" categories, of which metaphorical language was one. Diagnostic groups were found to be similar on all measures of syntactic complexity. Differences were, however, found in "functional usage." On the whole, autistic children demonstrated significantly greater use of abnormal
112
Volden and Lord
categories, including metaphorical speech. The authors acknowledged that the tests of syntactic complexity were far from exhaustive and that unmeasured subtle, but significant, differences might account for the language performance of the autistic speakers. Nevertheless, they concluded that the language of autistic speakers was differentiated from that of the languagedelayed adolescents by its deviant pattern of use. Recent efforts to account for deviant language in autism have often focused on the domain of pragmatics, particularly as related to cognitive and social development. In part, this emphasis may come from the recognition that it is severe social impairments that ultimately differentiate autistic adults from those with mental handicaps or other types of language dysfunction (Volkmar, 1987). For example, Fein et al. (1981; Fein, Humes, Kaplan, Lucci, & Waterhouse, 1984) posited that language disturbances considered critical to autism are not truly language disorders, but merely by-products of more fundamental disturbances in social and/or cognitive development. Still, given the relative paucity of empirical data, assigning pragmatics as the sole fundamental deficit appears to have been mostly a function of the failure to find clear, stable differences in other areas of language. Pragmatics has to do with the appropriate social use of language within a communicative context. A pragmatic disorder involves a mismatch between language and its context. One type of social contribution to a neologism might be a speaker who cannot remember the appropriate word and so supplies a semantically or phonologically similar word (such as "curlets" for "ringlets"), but does not inform the listener of his/her uncertainty by saying something like "Does that sound right?" Another social contributing factor might be a failure to achieve joint reference, so that the speaker introduces a new topic out of context without warning. Another possibility to account for an autistic speaker's unusual choice of words or phrases arises from a difference in the cognitive factor of "social meanings" assigned to the concepts to which they are referring (Rutter, 1983). For example, an autistic child used the word "plin" to refer to objects that remain suspended in the air when tossed; referring to a concept unlikely to be shared by many other children. A linguistic approach to idiosyncratic language would be warranted if autistic children created their own words or syntactic systems that followed the same social rules and conceptual boundaries as standard language. It may be that no single factor can account for the noted differences in the language of autistic speakers. Abnormal language in autism could be due to a combination of social, cognitive, and linguistic factors. A similar, but somewhat different, suggestion is that what appears as deviant development in autism may be accounted for by developmental asynchrony among various domains (Tager-Flusberg, 1987). Just as a mentally handi-
Neologisms and Idiosyncratic Language in Autism
113
capped 8-year-old boy may appear abnormal if he jumps up and down and flaps his hands in excitement like a normally developing 18-month-old, so an autistic adolescent who uses complex grammar and sophisticated phrasing but invents words like a 3-year-old may be judged to sound odd. From this perspective, it is important to know the relationship between odd features and measures of language skill. The current paper addresses issues of both the nature and source of the language disorder in autism by examining the abnormal use of conventional and unconventional words, including neologisms, and of more complex combinations of conventional and unconventional words and phrases within individual sentence boundaries. Errors that were solely due to problems with discourse were not addressed. In summary, three questions are asked: (a) To what extent do abnormalities in the use of single words and phrases, including neologisms, occur in the expressive language of autistic children and adolescents compared with other groups of children or adolescents matched on a general index of language skill? (b) Can individual instances of neologisms and idiosyncratic language be accounted for by the presence of specific syntactic, semantic, or nondiscourse pragmatic characteristics? (c) How do idiosyncratic language and neologisms relate to measures of language complexity in autistic and nonautistic populations?
METHOD Subjects
As shown in Table I, four groups of subjects were identified through an autism clinic and schools for normally developing and mentally handicapped children in which the autistic subjects were placed. Twenty higher f u n c t i o n i n g autistic (verbal IQs > 80 on the W I S C - R or W A I S - R ; Wechsler, 1974, 1981), 20 normally developing, 20 lower-functioning autistic (verbal IQs 45-79), 3 and 20 mentally handicapped children and adolescents served as subjects. All groups were matched for chronological age (CA) and sex (12 male, 8 female per group), with slightly younger normally developing children recruited to compensate for the slightly lower than average verbal IQ/PPVT-R scores of the higher autistic group. The first two groups and the last two groups were matched to each other on "lan3It is important to note that our lower functioning autistic subjects were not really "low functioning" (defined often as nonverbal IQs below 30 or 40) but were simply lower than our subjects of near-normal intelligence.
114
Volden and Lord Table I. Description of Subjects a
GroupQ Higher autistic M SD
Normal M SD
Lower autistic M SD
Mentally handicapped M SD
Chronological age
Verbal IQ
PPVT-R
Raven IQ
13.24 5.87
95.10 17.31
92.80 22.50
100.15 15.58
13.07 4.80
60.50 11.17
51.44 21.46
80.05 17.12
12.97 3.10
59.70 10.39
64.00 14.66
66.40 7.92
12.03 4.64
aAll groups contained 20 subjects, except only 18 subjects in the lower functioningautistic group were given the PPVT.
guage age" with the assumption that chronological age equaled language age for the normally developing children (derived by multiplying WISC-R o r WAIS-R verbal IQ x CA; Weschler, 1974, 1981). Verbal IQ was used as a general index of expressive and receptive communication skills. The autistic groups ranged in chronological age from 6 to 18 years. Each child was judged independently by a child psychologist and a child psychiatrist to have met DSM-III-R (American Psychiatric Association, 1987) and ICD-10 (World Health Organization, 1987) diagnostic criteria. All autistic subjects also met research diagnostic criteria for autism on the Childhood Autism Rating Scale (Schopler, Reichler, & Renner, 1986) and as outlined in the Autism Diagnostic Interview (Le Couteur et al., 1989). Nonautistic mentally handicapped subjects were selected from special classrooms and schools for the mentally handicapped on the basis of chronological age, sex, and IQ. Children with sensory or physical handicaps were excluded from the school registers prior to subject selection. Normal subjects were recruited from summer day camps and schools that the autistic children attended. Children from homes where languages other than English were spoken regularly were excluded from all groups. Procedures
Language Measures
As shown in Table II, two lO-minute language samples were taken from videotaped conversations with an unfamiliar adult during administra-
Neologisms and Idiosyncratic Language in Autism
115
tion of the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 1989). The first sample was generated during the subjects' description of a poster depicting a variety of familiar activities (e.g., a busy street with people shopping) and the subjects' telling a story from a sequenced picture book with no text, either showing a child going to bed at night or a bee creating a ruckus in a farmyard. The second sample was obtained from general conversation about topics that had arisen in the course of the interview and from the subjects' responses to specific questions about social and emotional relationships.
Coding Procedures Three sets of procedures were employed: (a) transcription and initial identification of "odd" or unusual utterances, (b) utterance by utterance ratings to determine the dimension(s) of "oddness," and (c) qualitative analysis of unusual words. Specific procedures for each step are described separately. Step A. Transcripts of videotapes of the ADOS administration were made by four trained researchers unaware of the children's diagnoses or
Table 1I. Description of Language Samples Poster/story
Higher autistic M SD Range Lower autistic M SD Range Mentally handicapped M SD Range Normal M SD Range
Conversation
No. of utterances
MLU a
No. of utterances
MLU
54.90 22.10 14-102
6.62 a 2.23 3-11.45
139.58 65.06 17-281
6.29 a 1.99 3.08-9.93
57.80 30.75 21-143
4.11 b 1.57 1.83-6.9
87.20 63.84 17-209
3.91 b 1.25 2.09-7.55
65.85 32.39 31-135
5.36 b 1.29 2.54-7.75
115.90 56.98 36-286
4.72 b 1.49 2.24-7.7
65.85 32.39 31-135
7.77 a 2.07 4.41-12.14
130.05 60.56 50-276
7.19 a 2.21 3.08-13.19
a M L U = M e a n length of utterance. There were 20 subjects per group. Different subscript letters indicate significantly different groups, p < .0001, based on univariate A N O V A . T h e r e were no significant group differences in n u m b e r of utterances.
Voiden and Lord
116
Table IIL Examples of Utterances Flagged in Step A by Diagnostic Group a Higher autistic It makes me want to go as deep as economical with it They're having a meal and they're finishing and siding the table And so he's seriously wounded like cutses and bloosers A n d wave their things on the floor in the bathroom If they even take it true enough . . . . But in the car, it's some In Thames, with about 8 years ago Lower autistic It bells the school turken (N) I had a racket when I'm a little baby It's ready to come and ready to go She's bawcet (N) That woman is wiping her hair Sittbzg bz both chairs A n d driving to all the buses 14qzat does cows do? Mentally handicapped He's gone running to there the horse is I'm going to stump the guy (N) Although if the bus is off sequence, I don't know if I'm on time or not Like it just felt me scaJed b~side The duck run Them go first Normal She was very baM The bee's going to cut him a(N) indicates codings of non-English words.
the purpose of this study. Transcripts of language samples were analyzed using the Systematic Analysis of Language Transcripts (SALT) program (Miller & Chapman, 1983). SALT analyses yielded a variety of standard psycholinguistic measures including number of utterances per speaker and MLU. They also provided the opportunity for inserting "flags" for features of particular interest. Transcribers flagged any utterance that they judged to include either a non-English word or an "idiosyncratic or unusual use of a conventional word or phrase" (i.e., abnormal functional use). For all pairs of transcribers, word by word and code by code transcription reliability exceeded 90% agreement for all measures for the first five consecutive videotapes coded after training was completed. Spot checks for transcription reliability were made by having pairs of researchers unknowingly code approximately every fifth tape. Reliability remained equally high. Disagreements were resolved by discussion among all the coders and the authors. Table III provides examples of utterances flagged by
Neologisms and Idiosyncratic Language in Autism
117
transcribers as either containing a non-English word or an unusual use of a word or phrase. Step B. Flagged u t t e r a n c e s were then c o d e d individually by the authors. F o r e a c h instance of neologism (i.e., n o n - E n g l i s h word) or idiosyncratic use of language, codes were made on several dimensions of 4the utterance:
Social Factors: 1. Social risk: Any verbal indication that some portion of the utterance was felt to be in error such as saying "Is that right?" or "I can't think of the word." 2. Joint reference: Whether or not joint reference was established in the interaction and by whom. Were the speakers both talking about the same topic?
Cognitive Factors: 3. Conventional semantic category: Whether or not the referent of the unusual word or phrase had a label in English (e.g., an example of an unconventional category would be "plin" for pieces of p a p e r or fabric that float in the air when tossed). The purpose of these three codes was to investigate the extent to which neologisms or idiosyncratic language were associated with differences in social contexts across diagnostic groups or with the cognitive factor of conventional categorization. So few of any of these codes occurred that they were not analyzable. However, the absence of these simple social and cognitive factors in some ways is as important as their presence. Linguistic Subsystems. Each unusual utterance was analyzed to determine which language c o m p o n e n t or combination of components was in error. (See Table IV.) 1. Syntactic impairment was indicated by two codes: "developmental syntax error," or "nondevelopmental syntax error." Assignment of type of syntactic error followed a comparison of the actual utterance to a paraphrase of what the investigator thought the speaker was trying to say, using standard English as similar as possible to the actual utterance. Developmental syntax codes were assigned to utterances containing well-docum e n t e d c o m m o n omissions or o v e r g e n e r a l i z a t i o n s of early a c q u i r e d syntactic rules such as "Where bees goed?" (Brown, 1973). All other syntactic errors were coded as nondevelopment syntax errors (e.g., "And driving to all the buses"). In fact, such errors may have followed normal developmental patterns which have not been documented to the extent of those we called developmental errors. It is important to note that the intention of Step A procedure was to flag utterances that sounded odd, not to code syntactic errors. However, we were interested in the extent to which utterances flagged because of their neologisms or unusual uses of words
118
Volden and Lord
Table IV. Step B: Interpretation and Codes Assigned to Sample Flagged Utterances Linguistic categories Flagged utterance
Developmental
Higher autistic "It makes me want to go as deep as economical with it" interpreted to mean "withdraw as much as possible" "They're having a meal and then they're finishing and siding the table" interpreted to mean "clearing the table" "And so he's seriously wounded like cutses and bloosers" interpreted to mean "cuts and bruises" "And wave their things on the floor in the bathroom" interpreted to mean "leave their things" "If they even take it true enough" interpreted to mean "If they take it seriously enough" "But hi the car, it's some" interpreted to mean "But in the car, there's something different" "In Thames, with about 8 years ago" interpreted to mean "In the Thames, about 8 years ago" Lower autistic "It bells the school" interpreted to mean "The bell rings in the school" "turken," - to denote turkeys" "I had a racket when I'm a little baby" interpreted to mean "I had a rattle when I was a little baby" "It's ready to come and ready to go" interpreted to mean "Easy come, easy go" "She's bawcet interpreted to mean "She's bossy" "That woman is wiping her hair" interpreted to mean "drying her hair" "Sitting in both chairs" interpreted to mean "Both sitting in chairs"
Nondevelopmental
Semantic
X
X
X
X
X
X
X
X
X
X
X X
X
X X
X
Neologisms and Idiosyncratic Language in Autism
119
Table IV. Continued
Linguistic categories Flagged utterance "And driving to all the buses" interpreted to mean "And all the buses are driving" or "They're driving all the buses" "What does cows do?" interpreted to mean "What do cows do?"
Mentally handicapped "He's gone running to there the the horse is" interpreted to mean "He's running to where the horse is" "I'm going to stump the guy" interpreted to mean "He's going to stick (this fork) into the guy" "Although if the bus is off sequence, I don't know if I'm on time or not" interpreted to mean "off schedule" "Like it just felt me scaled b~side" interpreted to mean "Like I just felt scared inside" "The duck run" interpreted to mean "The duck runs" or "The duck is running" "Them go first" interpreted to mean "They go first" Normal "She was very bald" interpreted to mean "She was naked" "The bee's going to cut him" interpreted to mean "sting him"
Developmental
Nondevelopmental
Semantic
x x
x x
x x x x
x x
c o n t a i n e d s y n t a c t i c e r r o r s , a n d in d i s c r i m i n a t i n g e r r o r s w e l l - d o c u m e n t e d as part of normal development from unusual errors of syntax that may have c o n t r i b u t e d to h o w o d d t h e u t t e r a n c e s s o u n d e d . 2. S e m a n t i c errors w e r e c o d e d if o n e o r m o r e w o r d s in a f l a g g e d utt e r a n c e was j u d g e d to b e o d d o r u n u s u a l a f t e r c o m p a r i s o n o f t h e a c t u a l u t t e r a n c e to a p a r a p h r a s e g e n e r a t e d o n t h e basis o f t h e i m m e d i a t e c o n t e x t . S t e p C: Q u a l i t a t i v e a n a l y s e s o f s e m a n t i c e r r o r s , e i t h e r n e o l o g i s m s o r o t h e r s e m a n t i c e x a m p l e s o f s e m a n t i c i m p a i r m e n t as d e s c r i b e d a b o v e , w e r e
120
Volden and Lord
carried out on samples from the subset of subjects who produced more than five flagged items in the total language sample. Fifteen higher autistic, seven lower autistic, seven mentally handicapped, and one normal subject met this criterion. Because he was the only one from his group, the normal subject was excluded from these analyses. These subjects produced a range of 5 to 41 errors out of a total of 151 to 216 utterances. Possible similarities to standard English that might have prompted the substitution were systematically rated. As shown in Table V, codes included, first, whether or not the word was a standard English word or a true neologism. Second, we asked whether it had any relationship to a real word in English and, if so, whether that relationship was semantic and/or phonological. Interrater reliability between the principal investigators for these codes was established at over 89% for each individual code over five transcripts selected at random. As the study proceeded, questionable words or utterances were discussed by both authors and differences were resolved by discussion.
RESULTS
PreliminaryAnalyses Preliminary analyses of subject characteristics of the mentally handicapped and two autistic groups included a three-way (diagnosis) A N O V A on the following psychometric measures (illustrated in Table I): verbal IQ, Peabody Picture Vocabulary Test standard score (PPVT-R), and Raven IQ. The purpose of these tests was to confirm the accuracy of matching. Significant effects of group were found for verbal IQ: F(2, 57) = 46.05, p < .0001; PPVT: F(2, 55) = 22.11,p < .0001; and Raven IQ: F(2, 57) = 28.89, p < .0001. Student-Neuman-Keuls analyses were used to make pairwise ~:,-~nparisons. As expected, on verbal IQ and the PPVT-R, significant differences emerged between the higher autistic group and both the lower autistic and the mentally handicapped groups. There were no significant differences between the lower autistic and the mentally handicapped groups on language measures. On Raven IQ, significant differences occurred among all pairs, with the lower autistic group falling between the higher autistic and the mentally handicapped groups. Because groups were not matched on the latter measure, this difference is not surprising. Preliminary analyses were also conducted on the total number of utterances and the mean length of utterance (MLU), using a 4 (Diagnosis) • 2 (Condition) repeated measures MANOVA. As shown in Table II, no significant main effects or interactions were found for the M A N O V A or on univariate statistics for total number of utterances. On MLU, a sig-
Neologisms and Idiosyncratic Language in Autism
121
Table V. Examples of Codes Assigned to Semantic Errors in Step C Type relationshipa Un usu al word/phrase and interpretation (from Table IV) "go as deep as economical with it" vs. withdraw as much as possible" "siding" vs. "clearing" "bloosers" vs, "bruises" "wave" vs. "leave" "tr ue" vs. "seriously" "bells" vs. "rings" " t u r k e n " vs. "turkey" "racket" vs. "rattle" "ready" vs. "easy" " m a s h e b a " - - reference unknown "bawcet" vs. "bossy" "th ere" vs. "where" "stump " vs. "stick" " s e q u e n c e " vs. "schedule" "bald" vs. " n a k e d " "cut" vs. "sting"
English
Non-English (Neologism)
Semantic
Phonologic
+ + + + + + + + + +
+ + + + + + + + + -
X X X X X X X X X X X X X X X X
aA plus sign indicates a positive relationship between the child's utterance and the perceived referent on the dimension marked, while a minus sign indicates absence of such a relationship . A + in each c o l u m n indicates that the word was re l a t e d both s e ma nt i c a l l y and phonologically to the paraphrase, while a - in each column indicates that the child's utterance was related neither in meaning nor by sound to the perceived referent.
nificant main effect for diagnosis occurred, F(3, 76) = 14.01, p < .0001. Student-Newman-Keuls pairwise comparisons among diagnostic groups revealed significant differences between normal and both lower autistic and the mentally handicapped groups and between higher autistic and both lower autistic and mentally handicapped groups. Thus, as expected, MLU was higher for the two nonretarded groups: normal (M = 7.48); higher autistic (M = 6.46) than for the two delayed groups: lower autistic (M = 4.01) and mentally handicapped (M = 5.04). No other significant main effects or interactions were found.
Frequency of Errors in Matched Groups Because of group differences in distributions, a 4 (Diagnosis) • 2 (Condition) repeated measures MANOVA was carried out on rankings of the three major variables. This procedure resulted in the equivalent of a Friedman two-way analysis of variance on percentage of developmental er-
Voiden and Lord
122
Table VI. Percentage of Utterances Containing Errors by Diagnostic Groupsa Developmental syntax
Nondevelopmental syntax
Semantic
Higher autistic M SD
Range
1.96a, b 2.56 0-10.3
0.62 a 0.70 0-2.47
1.60 a 1.78 0-5.62
0.49a, b 0.67 0-2.00
1.30 a 1.80 0-7.82
4.31 a 5.49 0-21.14
0.34 b 0.39 0-1.54
0.63a, b 0.65 0-2.96
0.71 b 1.55 0-21.14
0.11 b 0.26 0-1.54
0.22 b 0.39 0-2.96
Lower autistic M SD
Range
2.25a, b 2.17 0--6.45
Mentally handicapped M SD
Range Normal M SD
Range
aTwenty subjects in each group. Different subscript letters (a, b) indicate significantly different groups according to nonparametric A N O V A s by ranks. The same subscripts indicate groups that were not significantly different according to ANOVA analysis.
rors, percentage of nondevelopmental syntax errors, and percentage of semantic errors for each subject in each condition. Combinations of these errors within an utterance were not analyzed because they were too infrequent. Main effects occurred, as indicated by Pillai's criterion, for diagnosis, F(9, 228) = 3.24, p < .001, and for sample (story-picture vs. conversation), F(3, 76) = 3.63, p < .017, with no interaction. Univariate analyses revealed significant effects of diagnosis on developmental syntax, F(3, 76) = 4.06, p < .01, nondevelopmental syntax F(3, 76) -- 4.70, p < .005, and semantic errors F(3, 76) = 4.52, p < .006. Student-Newman-Keuls analyses were used to make specific comparisons among groups. As shown in Table VI, mentally handicapped subjects made the highest proportion of errors in developmental syntax, the normally developing group made the least, with the autistic subjects in between. For nondevelopmental syntax errors, the higher functioning autistic group made the most errors and the mentally handicapped and normally developing groups the least, with the lower autistic group in between. Semantic errors were made most frequently by the higher autistic group, followed by the lower autistic, the mentally handicapped, and the normal groups, respectively. Significant differences (Fisher Exact, p < .003) were also found at the individual level in the number of subjects in the combined autistic group
Neologisms and Idiosyncratic Language in Autism
123
compared to the number in the mentally handicapped group with more than 2% total nondevelopmental syntax and semantic errors (2% was chosen as it most often equaled 4 or 5 errors). O f 40 autistic subjects, 16 had 2% or more utterances assessed in these categories compared to only 1 of 20 mentally handicapped subjects. Univariate analyses revealed a significant effect of context of sample (e.g., describing poster and book vs. conversation) on developmental syntax errors, F(1, 76) = 9.85, p < .002 (mean percentage errors for the posterstory task = 6.18, SD = 9.61; mean percentage errors for the socioemotional conversational task = 2.89, SD = 4.01). There were no other main effects or interactions.
Relationship Between Frequency of Errors and Other Measures Spearman rank order correlations were run for percentages of each individual error type with verbal IQ, Raven IQ, P P V T - R age equivalent, and MLU. In the combined autistic group, significant positive relationships occurred between the percentage of semantic errors and Raven IQ, r = .31, p < .05, between the percentage of syntax errors and MLU, r = .33, p < .03, and between percentage of semantic errors and MLU, r = .31, p < .05. In the mentally handicapped group, a negative correlation occurred between percentage of semantic errors and verbal IQ, r = -.57, p < .009. Other correlations in the mentally handicapped group were also negative but not significant. All correlations between errors and measures of language complexity compared across the two diagnostic groups (autistic, mentally handicapped) were significantly different from each other, t(58)s > 3.21, p < .001. A N O V A s were conducted on data from the subset of subjects who had five or more examples of idiosyncratic language and/or neologisms to determine the impact of diagnostic group on percentage of flagged terms related to English paraphrases semantically, phonologically, neither semantically nor phonologically, or in both ways. The proportion of flagged terms that contained a non-English word, that is, that contained true neologisms, was also assessed. No significant main effects of group for mean proportion of any of these errors were found. All groups used a relatively low frequency of true neologisms, with a range of 1.40-5.29% of all utterances containing nonstandard English words. The largest proportion (55.76%) of e r r o r s w e r e semantically r e l a t e d to an English word, 18.02% were phonologically related to an English paraphrase, 11.08% were related to both, and 28.23% related to neither. For all groups, most examples of
124
Volden and Lord
idiosyncratic or unusual language consisted of conventional English words used in unconventional ways. To assess the extent to which individual subjects varied across diagnostic group in error type, Fisher Exact tests were carried out on the same measures. Comparisons were made of the number of subjects in each of the two autistic groups and the mentally handicapped group who had one or more error of each type (phonological, semantic, both, neither) versus none. More autistic children from both high and low groups used flagged terms related neither semantically nor phonologically to standard English than did mentally handicapped children (p < .03, with 19 of 22 in the combined autistic group exhibiting one or more of this type of flagged term compared with 3 of 7 mentally handicapped children). Children of the lower autistic group more often used true neologisms than children from the high autistic group and mentally handicapped children (p < .05, 3 of 7 lower autistic children used non-English words compared to 2 out of 15 higher autistic children and 1 of 7 mentally handicapped children).
DISCUSSION The current investigation examined the language disorder in autism through two frequently reported phenomena, neologisms and idiosyncratic use of words and phrases. Our interest was whether or not these two features differed in frequency and form in autistic speakers compared to ageand language-matched control groups and to determine if social-, cognitive-, or linguistic-specific factors could be used to account for their presence. More autistic subjects made frequent semantic errors and nondevelopmental syntax errors than IQ-matched controls. Out of the entire sample, nearly one quarter of the autistic subjects (9 of 40) produced a true neologism, that is, a non-English word, in contrast to 2 of 20 in the mentally handicapped group and none of the normal subjects. This proportion in the autistic group is somewhat higher than described in Rutter (1970), but could be due to our detailed analysis of actual language samples or to a higher proportion of children with significant amounts of expressive language. Similarly, in analyses of samples where five or more terms were flagged, almost all (19 out of 22) subjects from the combined autistic group used at least one unusual English word not related semantically or phonologically to the expected word, compared to fewer than half of the mentally handicapped children. However, contrary to expectations, most of the unusual words in the speech of the autistic subjects resulted from com-
Neologisms and Idiosyncratic Language in Autism
125
bining English morphological terms in logical but incorrect ways, rather than from true neologisms. Examples of true neologisms from these samples ranged from, at one extreme, "Kellogg's nahavities" as a label for a particular stretch of road, to "redundiate" used to connote the condition of being redundant, to "commendment" used in place of "commendation." Because they consist of morphologically modified standard word roots, the latter words sound more familiar to an English speaker. Nonetheless, due to the nonstandard morphological modifications, neither is an English word. Thus, in each of these examples, subjects could be seen as "inventing their own words (or phrases) for things" (Rutter, 1965, p. 41). Altogether, the proportion of unusual usages accounted for by truly non-English words was very low. These words seemed to be one end of continua of conventionality and complexity of unusual use, with pedantic use of correct English words and syntax at the other. If our data are representative, why has such a phenomenon drawn specific attention? An explanation may be found in the overall communicative profile of the autistic subject compared to the normally developing child. In longer samples or diaries, examples of neologisms in the speech of young normal children can easily be found. "Glassable" to denote "breakable" and "poma" to describe a particular configuration of Lego pieces (both uttered by 3-yearolds in play with their mothers) are two of these. Some of the same sort of neologisms were found in our language samples. Perhaps the salient difference is that linguistic errors abound in young normal children as they proceed with language acquisition and are, in fact, considered both normal and endearing. In autistic children, however, the playful quality embodied in such linguistic experimentation is missing, both for the child and the caregiver. When neologisms appear in the speech of older autistic children and adolescents, they are also often part of a complicated profile of linguistic disorder, including problems with pragmatics, semantics, and syntax. Perhaps, due to differing listener expectations for older, bigger children in less playful social situations, neologisms and other abnormal uses of words are more readily noted despite their relatively low frequency of occurrence. They may be rare, but they are salient differences to the listener. Patterns of words and utterances identified as odd were also different across diagnostic groups. In contrast to the mentally handicapped children, the two autistic groups were not significantly different from normal subjects in number of developmental errors. Nondevelopmental syntax errors were, however, more common in higher functioning autistic children than any other group. In fact, autistic subjects of normal verbal intelligence made significantly more of these errors than either their language-matched nor-
126
Volden and Lord
mal controls or mentally handicapped children with significantly lower verbal IQs and receptive vocabulary scores. This finding suggests the presence of linguistic disorder in autistic subjects beyond general developmental delay. The lack of a similar finding for the lower functioning autistic group (i.e., their rate of nondevelopmental error fell between the higher autistic and the mentally handicapped groups) may be due to the brevity of most of their utterances (i.e., their MLU was an average of 4.01 morphemes reducing the possibility of any but the simplest syntactic errors). Both autistic groups made more semantic errors than the normal children, again affirming the presence of a language abnormality in autism not solely accounted for by general intellectual level or even general "language level" as measured with either Wechsler VIQ or PPVT-R scores. The results of the mentally handicapped group fell between those of the lower autistic and the normal groups. Correlational analyses showed a positive relationship between higher Raven's IQs and increased semantic errors for the combined autistic group. In contrast, the mentally handicapped group showed a negative relationship between verbal IQ and semantic errors. Several hypotheses proposed to account for unusual use of words and phrases require consideration. One possibility was that normally developing children would mark their uncertainty about the appropriateness of a word or phrase, indicating a sense of social risk. If the source of language disorder in autistic speakers was primarily social and cognitive (Baron-Cohen et al., 1985), we might expect fewer such markings in the autistic population. This was not the case. In fact, none of the diagnostic groups showed any trend toward marking "risky" situations (i.e., where a word or phrase was used inappropriately). This is not to say that social skill deficits do not contribute to language disorder in autistic children, but rather that they did not account for within-utterance idiosyncratic language any more for the autistic children than they did for the mentally handicapped or normal groups. It should also be noted that even normally developing children may not always spontaneously mark socially risky utterances. One normally developing 10-year-old girl, in attempting to describe the types of ravioli (e.g., beet, carrot) in a new restaurant, used the word "outlinings" without indicating its dubious appropriateness, until her mother's smile led her to admit that she did not know the correct term. Perhaps, the autistic child does not receive such feedback because listeners are reluctant to react negatively to a disordered child's efforts to communicate. Or, the autistic child may be unable to use the sometimes subtle social signals regarding risky communications given in everyday interaction. Another social possibility suggested to account for the abnormal utterances was the failure to establish joint reference in a conversation. Un-
Neologisms and Idiosyncratic Language in Autism
127
fortunately, in our sampling situations, topics were generally introduced by the examiner. Thus, there were insufficient opportunities to study our subjects' abilities to establish and maintain joint reference. Investigations using less structured situations may reveal this to be an area of deficit. A final hypothesis was that unusual use of words and phrases in autism might be due to different cognitive "social meanings" assigned to words (Rutter, 1983), perhaps even at the level of basic concepts. However, in our sample, the referent was usually readily identifiable and one to which competent speakers would also refer. Unconventional categorization, when it occurs, may be more common in very young children (Le Couteur et al., 1989). In considering what factors may operate differentially among the diagnostic groups, we suggest that this may be, at least partially, explained by the notion of developmental asynchrony between linguistic domains (TagerFlusberg, 1987). The notion of developmental asynchrony is not new or restricted to language development. Studies of nonverbal social behavior (Mundy, Sigman, Ungerer, & Sherman, 1986; Sigman, Mundy, Sherman, & Ungerer, 1986) have also documented asynchronous development. In our sample, in the area of developmental syntax, the expected progression of diagnostic groups was found with normal subjects making least errors, followed by the high autistic, the mentally handicapped, and the lower autistic groups. In both other linguistic domains, nondevelopmental syntax and semantics, the group hierarchy was very different. Here, the normal subjects again made the least errors, but they were followed by the mentally handicapped, then the lower autistic, and finally the higher autistic group. On these measures, the highest functioning autistic subjects performed more like the lower autistic group than their language skill-matched (normal) controls. In developmental syntax, as IQ score decreased, errors increased. In nondevelopmental syntax and semantics, the groups with the highest IQs made both the least and the most errors. Asynchrony was further substantiated by the positive relationship between sentence length and the frequency of nondevelopmental syntax and semantic errors in the autistic group. There was no such relationship between MLU and developmental syntax errors. Even though the higher autistic children sometimes looked like nonautistic mentally handicapped children and sometimes like the normally developing children in their frequency of errors, the communicative impact was often quite different. Examples are provided in Table III. A sentence like "It makes me want to go as deep as economical with it," produced by a nonretarded autistic adolescent, sounds more peculiar than "He's gone running to there the horse is," a sentence produced by a mentally handicapped youngster of the same age. This difference may, partially at least, be ac-
128
Volden and Lord
counted for by the autistic child's longer average utterance length. In some cases, the higher autistic group's unique errors in nondevelopmental syntax or their sophisticated, but incorrect, lexical selection (e.g., using "economical" instead of "possible") may also contribute to the communicative effect. Ungerer (1989) proposed that the search for a specific deficit underlying the social dysfunction of autism may have failed because the source of the dysfunction, if a single source exists, may change over time. Our findings suggest that this conceptualization of a dynamic relationship between social and communicative domains can also be applied to language development at an utterance by utterance level. No single cognitive or social source or identifiable combination of factors was responsible for the linguistic errors made. Instead, the variation within and between individuals and groups and across utterances suggests that simple comparisons among single developmental domains may be insufficient to explain the language disorder associated with autism. Examination of dynamic relationships within and among linguistic, social, and cognitive domains may lead to more useful explanations of a communicative disorder where the whole may well be greater than the sum of its parts. In summary, neologisms and unusual words and phrases were produced by almost all autistic speakers and, to a lesser extent, by many mentally handicapped children. These errors occurred very rarely in normally developing subjects of equivalent chronological and language age. Individual words used in error by children from all diagnostic groups were most often phonologically or conceptually similar to the intended English words. On the other hand, more autistic children produced utterances containing words that were neither phonologically nor conceptually related to the intended English word than children from other diagnostic groups. Few true neologisms were found, though all but two examples were produced by autistic children. Within individuals, different relationships among various aspects of language, such as utterance length, syntactic complexity, and semantic complexity may result in different sorts of unusual uses, from true neologisms to partial neologisms to idiosyncratic phrases. Investigation of the nature of these relationships may provide important answers to questions about the nature and source of language disorder in autism.
REFERENCES American PsychiatricAssociation. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author.
Neologisms and Idiosyncratic Language in Autism
129
Baltaxe, C. A. M., & Simmons, J. Q. (1983). Communication deficits in the adolescent and adult autistic. Seminars in Speech and Language, 4, 27-42. Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a 'theory of mind'? Cognition, 21, 37-46. Bartak, L., Rutter, M., & Cox, A. (1975). A comparative study of infantile autism and specific developmental receptive language disorder: I: The children. British Journal of Psychiatry, 126, 127-145. Bartolucci, G., Pierce, S., Streiner, D., & Tolkln-Eppel, P. (1976). Phonological investigation of verbal autistic and mentally retarded subjects. Journal of Autism and Childhood Schizophrenia, 6, 303-315. Brown, R. (1973). A first language. Cambridge, MA: Harvard University Press. Cantwell, D. P., Baker, L., & Rutter, M. (1978). A comparative study of infantile autism and specific developmental receptive language disorder: IV. Analysis of syntax and language function. Journal of Child Psychology and Psychiatry, 19, 351-363. Cohen, D. J., Caparulo, B. K., Gold, J. R., Waldo, M. C., Shaywitz, B., Ruttenberg, B. A., & Rimland, B. (1978). Agreement in diagnosis: Clinical assessment and behavior rating scales for pervasively disturbed children. Journal of the American Academy of Child Psychiatry, 17, 589-603. Fein, D., Humes, M., Kaplan, E., Lucci, D., Waterhouse, L. (1984). The question of left hemisphere dysfunction in infantile autism. Psychological Bulletin, 95, 258-281. Fein, D., Skoff, B., & Mirsky, A. F. (1981). Clinical correlates of brain dysfunction in autistic children. Journal of Autism and Developmental Disorders, 11, 303-316. Howlin, P. (1984). The acquisition of grammatical morphemes in autistic children: A critique and replication of the findings of Bartolucci, Pierce, and Streiner, 1980. Journal of Autism and Developmental Disorders, 14, 127-136. Kanner, L. (1946). Irrelevant and metaphorical language. American Journal of Psychiatry, 103, 242-246. Le Couteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Holdgrafer, M., & McLennan, J. D. (1989). Autism Diagnostic Interview: A semi-structured interview for parents and caregivers of autistic persons. Journal of Autism and Developmental Disorders, 19, 363-387. Lord, C., Rutter, M., Goode, S., Heemsbergen, J., Jordan, H., Mawhood, L., Schopler, E. (1989). Autism Diagnostic Observation Schedule: A standardized observation of communicative and social behavior. Journal of Autism and Developmental Disorders, 19, 185212. Miller, J., & Chapman, R. (1983). Systematic analysis of language transcripts. Madison: University of Wisconsin, Waisman Center. Mundy, P., Sigman, M., Ungerer, J., & Sherman, T. (1986). Defining the social deficits of autism: The contribution of nonverbal communication measures. Journal of Child Psychology and Psychiatry, 27, 657-669. Rutter, M. (1965). Speech disorders in a series of autistic children. In A. Franklin (Ed.), Children with communication problems (pp. 39-47). London: Pitman. Rutter, M. (1978). Language disorder and development. In M. Rutter & E. Schopler (Eds.), Autism: A reappraisal of concepts & treatment (pp. 85-104). New York: Plenum Press. Rutter, M. (1983). Cognitive deficits in the pathogenesis of autism. Journal of Child Psychology and Psychiatry, 24, 513-532. Schopler, E., Reichler, R. J., & Renner, B. R. (1986). The Childhood Autism Rating Scale (CARS) for diagnost& screenhlg and classification of autism. New York: Irvington Publishers. Sigman, M., Mundy, P., Sherman, T., & Ungerer, J. (1986). Social interactions of autistic, mentally retarded and normal children and their caregivers. Journal of ChiM Psychology and Psychiatry, 27, 647-656. Tager-Flusberg, H. (1981). On the nature of linguistic functioning in early infantile autism. Journal of Autism and Developmental Disorders, 11, 45-56.
130
Volden and Lord
Tager-Flusberg, H. (1985). Psycholinguistic approaches to language and communication in autism. In E. Sehopler & G. Mesibov (Eds.), Communication problems in autism (pp. 69-87). New York: Plenum Press. Tager-Flusberg, H. (1987). On the nature of a language acquisition disorder: The example of autism. In F. Kessel (Ed.), The development of language and language researchers (Essays in honor of Roger Brown) (pp. 249-267). Hillsdale, NJ: Erlbaum. Ungerer, J. A. (1989). The early development of autistic children: Implications for defining primary deficits. In G. Dawson (Ed.), Autism: Nature, diagnosis and treatment (pp. 75-91). New York: Guilford. Volkmar, F. R. (1987). Diagnostic issues in the pervasive developmental disorders. Journal of Child Psychology and Psychiatry, 28, 365-369. Waterhouse, L., & Fein, D. (1982). Language skills in developmentally disabled children. Brain and Language, 15, 307-333. Wechsler, D. (1974). Wechsler Intelligence Scale for Children-Revised. New York: Psychological Corp. Wechsler, D. (1981). WechslerAdult Intelligence Scale-Revised. New York: Psychological Corp. World Health Organization. (1987). ICD-IO 1986 draft of chapter 5 categories FOO-F99. Mental, behavioural and developmental disorders. Geneva: Author.