The Kiddie Formal Thought Disorder Rating Scale:

Clinical Assessment, Reliability, and Validity

ROCHELLE CAPLAN, M.D. , DONALD GUTHRIE, PH.D., BARBARA FISH, M.D. ,

PETER E. TANGUAY, M.D. , AND GABRIELA DAVID-LANDO, M.SC.

Abstract. The Kiddie Formal Thought DisorderStory Game and the KiddieFormalThought DisorderScale
wereadministeredto schizophrenic, schizotypal, and normal children,aged 5 to 13 years. The story game elicited
more elaborate speech samples than did a structured clinical interview focused on psychotic symptomatology.
The sum of illogical thinking and loose associations was a reliable (kappa = 0.77), sensitive (79 %), and specific
(90%) indicator of schizophrenia in this sample. It also demonstrated significant developmental changes in the
schizophrenic and normal subjects. Incoherence and poverty of content of speech were infrequently rated in both
schizophrenic and normal subjects J. Am. Acad. Child Adolesc. Psychiatry, 1989, 28, 3:408-416. Key Words:
formal thought disorder,schizophrenia, middlechildhood, reliability, validity.

Formal thought disorder (FTD) is one of the DSM-Ill The dearth of studies on FTD in childhood schizophrenia
primary inclusionary criteria for the diagnosis of schizophre- may stem from a number of difficulties involved in the clinical
nia. A DSM-Ill diagnosis ofFTD is based on the presence of diagnosis of FTD in childhood. First, there are no guidelines
illogical thinking, loose associations, incoherence, and poverty in the existing literature for applying the four DSM-Ill signs
of content of speech. The clinical importance of thought of FTD to children of different ages. Child psychiatrists must,
disorder in adult schizophrenia is reflected in the large number therefore, rely on their own sense of the norms of children's
of clinical (Andreasen, 1979; Harrow and Quinlan, 1977; speech to identify FTD in children and to distinguish it from
Johnston and Holzman, 1979), linguistic (Chaika, 1974; the speech of normal young children.
Rochester and Martin, 1979), and information processing This lack of developmental guidelines poses a significant
studies (Asamow et al., 1986; Chapman and Chapman, 1973; problem to clinicians because young normal children are often
Neuchterlein et al., 1986; Sacuzzo and Braff, 1986) and in unaware of an adult listener's needs and assume that the adult
the diverse approaches used to study this phenomenon. makes logical (Piaget, 1959) and linguistic (Maratsos, 1976)
In contrast to the wealth of studies on FTD in adults, there connections for them. Children, however, become more
has been little research on FTD in early and middle childhood. skilled at presenting their thoughts to the listener in a logical
The research that does exist consists primarily of clinical and cohesive way during the latter part of middle childhood
studies in which adult FTD criteria were used in children with (Olson and Nickerson, 1978; Romaine, 1984). In the absence
schizophrenia (Arboleda and Holzman, 1985; Cantor et al., of guidelines for differentiating between ITD and immature
1982; Fish and Ritvo, 1979; Green et al., 1984; Kolvin, 1971), speech, clinicians could overdiagnose or underdiagnose FTD.
pervasive developmental disorder (Arboleda and Holzman, The second problem involves the relative paucity of spon-
1985), and infantile autism (Fish and Ritvo, 1979; Shapiro taneous or expansive speech in middle childhood. Whereas
and Huebner, 1976). None of these studies, however, have adults talk spontaneously in "paragraphs"-thus providing
specified how these adult criteria were operationalized for use the listener with speech samples long enough for rating FTD-
with children. Using Johnston and Holzman's (1979) young children use only one to two utterances to express
Thought Disorder Index, Arboleda and Holzman (1985) dem- themselves. A technique for assessing FTD in children should,
onstrated the importance of controlling for the level of cog- therefore, guide the rater in measuring FTD within short units
nitive development when assessing thought disorder in chil- of speech .
dren under 10 years of age. A th ird obstacle to identifying FTD in children is that
children under the age of 9 years may resist answering inter-
Accepted August 11, 1988. view questions, particularly those that probe for psychotic
From the Division of Child Psychiatry, Neuropsychiatric Institute. symptomatology. As a result , their answers tend to be brief
UCLA . Reprint requests to Dr. Caplan. Neuropsychiatric Institute. and unelaborated, often limited to a simple "Yes," "No," or
760 Westwood Plaza, Los Angeles, CA 90024.
This study was supported by National Institute of Mental Health "I don 't know ." In contrast to the more elaborate speech
Research Scientist Development Award KOl-MH00538 . by the Child- samples of adult patients, the speech samples elicited in
hood Psychoses Clinical Research Center A ward MH-30 897, and by psychiatric interviews of children may, therefore, be inade-
the UCLA Women's Hospital Auxiliary. quate for determining the presence of FTD.
The authors acknowledgethe help ofJody Kescek. Marilyn Fiedler. Three studies are presented in this paper, with four goals.
Eileen Sadeh. and the staff and children of Walden and Humphrey
A venue Elementary Schools. Technical assistance was given by Anne The first goal was to develop a reliable instrument, the Kidd ie
Brothers, Amy Mo. Diane Putney. R.N.. Diane Greene. and Chenga Formal Thought Disorder Scale (K-FTDS), that would oper-
Buanausi . Sondra Perdu. Ph.D., and Dolores Adams, M.Sc.. also ationalize the four DSM-lll signs for rating FTD in middle
provided statistical consultation. Tracey Sherman. Ph.D.. Robert childhood schizophrenia (Study I). The second goal was to
Asarnow, Ph.D.. and Judy Fay. Ph.D., provided valuable editorial
comments and discussion. design a clinical and research procedure, the Kiddie Formal
0890-8567/89/2803-0408$02.00/0© 1989 by the American Acad- Thought Disorder Story Game (SG), for the collection of
emy of Child and Adolescent Psychiatry. adequate speech samples from schizophrenic children in this
408
 TESTING THE K-fTDS 409
age group (Study II). The third goal was to determine the selected from the previously described larger sample of normal
diagnostic validity of the K-FrDS (Study III). Finally, the children based on relatively high K-FrDS ratings. This en-
fourth goal was to assess how the child's developmental level sured an adequate base rate of FrD in the sample studied
affected his/her K-FrDS scores (Discussion). (Shrout et aI., 1987).
The subjects in the reliability study were divided into two
Method cohorts, Group I and Group II, each rated by a different pair
Subjects of raters. Table I presents diagnostic and demographic infor-
mation on these subjects.
The subjects included in these studies were drawn from a
sample of 17 schizophrenic, 4 schizotypal, and 44 normal Procedures
children, aged 5 to 13 years old. The schizophrenic and
The Kiddie Formal Thought Disorder Story Game (SG).
schizotypal children were recruited from the University of
T~e. SG,.developed by R.C., is modeled after Gardner's (1971)
California, Los Angeles (UCLA) Neuropsychiatric Institute
chmcal instrument for therapeutic communication with chil-
Inpatient and Outpatient Children's Services and also from a
dr~n. It includes three parts. In the first and last parts, the
Los Angeles school for the emotionally disturbed.
child hears a recorded story, retells the tale, and is then asked
All schizophrenic and schizotypal children, hereafter re-
a series of standard open-ended questions about the story. In
ferred to as the target subjects, were diagnosed by the Diag-
the second part, the child is asked to make up a story on one
nostic Unit of UCLA's Childhood Psychoses Clinical Re-
of four topics.
search Center. This assured that the diagnostic process was
The topics of the stories in the SG (a dream about a friendly
done independently of the research team and that the raters
ghost, an ostracized little boy, the Incredible Hulk, a witch, a
had no knowledge of the diagnoses of the children. The target
good or a bad child, an unhappy child) were chosen because
subjects were diagnosed with the Interview for Childhood
of their potential for eliciting pathological thought content in
Disorders and Schizophrenia (lCDS) (Russell et aI., 1987),
children. It was hoped that while actively engaged in thinking
which is derived from the K-SADS (Puig-Antich and Cham-
about pathological thought content, the nonthreatening and
bers, 1978) and the Diagnostic Interview for Children and
indirect interview technique used in the SG would allow the
Adolescents (Herjanic and Campbell, 1977). This reliable
psychotic child to be more expansive and to reveal more FrD
(kappa = 0.89) structured interview includes questions that
in his/her speech.
ensure completeness in assessing schizophrenia and schizoty-
To ensure standardization, the stories and the instructions
pal personality disorder.
to the child were audiotaped. The standard questions that
The normal subjects were recruited from two Los Angeles
followed the stories, however, were not prerecorded. This
schools and from the community via advertisements in a local
enabled the interviewer to elicit additional speech from the
newspaper. Children with a history of psychiatric, neurologi-
child by means of open-ended questions.
cal, and/or language disorders were excluded from the study.
The first author (R.C.) or a psychiatric nurse administered
The average IQ of the target and normal subjects was 90
the 20- to 25-minute long SG. The SG and guidelines for
and 113, respectively. Seventy-six percent of the target and
interviewer training can be obtained from R.C. Videotapes of
90% of the normal subjects were from middle-class families
the entire story game were used for rating via the K-FrDS.
(Hollingshead II and III). The socioeconomic status of the
The Kiddie Formal Thought Disorder Scale (K-FTDS).
remaining target subjects was Hollingshead I (3%), IV (18%),
Table 2 presents an outline, with examples, of the scale's four
and V (3%). The remaining normal subjects were classified
signs: illogical thinking (ILL), loose associations (LA), inco-
as Hollingshead IV (10%). Most of the target subjects were
herence (INC), and poverty of content of thought (POC). The
Anglos (81%) as compared to 54% of the normal group. Two
K-FrDS definitions of the signs were based on the Research
target subjects (6%) were from Afro-Asian and four (13%)
Diagnostic Criteria (ROC) (Spitzer et aI., 1975) and on An-
were from Hispanic families. The remaining normal subjects
dreasen's Thought, Language, and Communication Scale
were English speaking from Hispanic (27%), Afro-American
(10%), and Asian (5%) families.
TABLE I. Description ofSubjects in the Kiddie Formal Thought
Forty-two percent of the target subjects were receiving Disorder Scale (K-FTDS) Reliability Study
neuroleptic medication at the time of the study and 68% of
these children were also hospitalized. Most of the nonmedi- Diagnosis" Sex
Age
cated children were outpatients (80%) and had been medica- Subjects N SZ SPD N (yrs.) M F IQh
tion-free for a minimum of 2 weeks before their participation By diagnosis
in this study. Targets 21 17 4 o 10.1 16 5 84
The procedures were explained to the child and parent(s) Normals 7 0 o 7 8.6 5 2 105
and their written informed consent was obtained. By raters
Group I'" 16 8 I 7 9.1 13 3 97
Study I: Reliability of the Kiddie Formal Thought Group Il" 1210 2 0 10.3 93 87
Disorder Scale (K-FTDS) Total groups, I and II 28 18 3 7 9.3 22 6 94
METHODS a SZ = sc~izophrenia, SPD = schizotypal personality disorder, N
Subjects = normal children,
h IQ scores based on the WISe-R.
The reliability study comprised 17 schizophrenic, 4 schi- c All subjects were divided into two cohorts, Group I and Group II
zotypal, and 7 normal subjects. The normal subjects were (see text).
410 CAPLAN ET AL.

(Andreasen, 1979). R.C. has operationalized these signs for TABLE 2. Synopsis of the Kiddie Formal Thought DisorderRating
use with children so that incoherence can be rated with one Scale (K·FTDSj
utterance while illogical thinking, loose associations, and pov- ITO Sign Definition Example
erty of content of speech are rated with a minimum of two Illogical (ILL) Inappropriate and "I left my hat in
utterances. An utterance is defined as a noun and a verbal immature use of her room be-
clause followed by a pause. The operationalization guidelines causal utterances . cause her
for the K-FfDS were derived from studies on the develop- name is
ment of children's conversation skills (Mc'Tear, 1985; Ochs Mary."
and SchiefTelin, 1979; Romaine, 1984) and from clinical Unfounded and inap- "Sometimes I'll
experience with schizophrenic and other psychotic children. propriate reasoning go to bed and
The K-FfDS definitions were, however, based on "adult" in noncausal utter- when I'm
norms. Thus, within the constraints of children's nonexpan- ances. done laugh-
ing, I start
sive speech, the criteria for the K-FfD signs were used on the
wheezing and
child's speech samples in the way they would be used for an that's when I
adult. In the K-FfDS the four FTD signs are rated as mutually relax."
exclusive. Utterances in which "I don't like
Incomplete utterances, such as word filters and false starts , statements are si- that story,
and signs of immature discourse, such as inappropriate use of multaneously made but I liked it
the definite article, the indefinite article , demonstratives, and and refuted. as a story."
pronouns are distinguished from FTD. They are coded sepa- Incoherence (INC) The contents of an Interviewer:
rately for a study on the pragmatic/discourse skills of schizo- utterance are not "What hap-
understood by the pened next in
phrenic children.
listener because of your story?"
Illogical Thinking (IL L) scrambled syntax. Child : "The
day no
To rate ILL the rater needs to determine whether, in adult witches no
terms, the facts, causes , and conclusions of the child's utter- day goes."
ances are logically sound. ILL is evaluated in three conditions Loose associations The child changes the Interviewer:
(Table 2). In the first condition, the child's ability to use (LA) topic of conversa- "Why do you
causal linguistic constructs appropriately is assessed when he/ tion to a new unre- think that's a
lated topic without reason not to
she uses utterances introduced by "because," "so that," or "if'
preparing the lis- like Tim ?"
(Emerson and Gekonski, 1980; Olson and Nickerson, 1978). tener for the topic Child: "And I
The second condition is one in which the child presents the change. call my mom
listener with an explanation or reasoning that is clearly un- sweetie."
founded. In the third condition, the child contradicts him or Poverty of content of In the presence of at "I suppose . . .
herself within one to two utterances by simultaneously mak- speech (POC) least two utter- What?
ing and refuting statements without informing the listener of ances, the child Maybe .. .
the apparent contradiction. does not elaborate Well yes, I
on the topic. see. I suppose
Loose Associations (LA) that's all."

LA is rated when the child makes an utterance that is ofT-

topic without having previously prepared the listener for the
topic change (Table 2). The K-FTDS provides the rater with Poverty a/Content (POC)
guidelines on how to determine whether the child's speech is The DSM-1I1 definit ion of pac stipulates that despite an
on-topic or off-topic in the three sections of the SG. During adequate length of speech , the patient provides the listener
the Question and Answer sections of the SG, the interviewer with insufficient information. Based on our pilot work , ade-
introduces the topic of conversation with each standardized quate length was defined as being two or more utterances in
question. Because of the open ended structure of these ques- response to a given question . If the child elaborated and
tions, the K-FTDS instructs the rater how to track the topic developed the topic of conversation, this was not rated as
changes during the ensuing con versation. POCo If, however, he/she merely rephrased the question or
For the remaining sections of the SG-the child 's recall of his own first statement, without providing the listener with a
the recorded stories and his/her own narrative-the initial new aspect, this was rated as pac (Table 2). Due to the fact
topic of the story is considered to be the ongoing topic of that the K-FTDS signs are mutually exclusive , POC is not
conversation. For example, if while telling a story about a rated if the impoverished content results from ILL, LA, or
witch the child digresses and tells a story about magic ponies, INC.
this would be rated as LA.
Coding the K-FTDS
Incoherence (INC)
Two pairs of trained raters with no knowledge of the child's
An utterance is rated as INC if the rater cannot comprehend diagnosis participated in the reliability study. The first pair
the contents of an utterance because of scrambled syntax included two undergraduate students with no previous clinical
(Table 2). and research experience. They rated the tapes of the children
 TESTING THE K-FfDS 411

included in Group I (Table 1). The two remaining raters were similar in the target (0.76) and normal (0.80) subjects. Because
experienced in using clinical and behavioral research methods of a low LA base rate in the normal children, kappa was not
and they coded the tapes of subjects included in Group II meaningful. The kappa values for Group I's inexperienced
(Table I). undergraduate student raters were consistent with those of
The raters scored the frequency with which each K-FrOS Group II's experienced research assistants.
sign occurred during the SG to obtain the ILL, LA, INC, and The videotaped and transcribed SG speech samples dem-
POC scores. The sum of these scores was the Total FrO score onstrated similar KSU M , KILL, and K LA values for the video-
for each subject. To control for the variable amount of speech tapes and transcriptions (Table 3). The kappa values for the
elicited from different children, the K-FrOS score per utter- audiotaped speech samples, however, were significantly lower
ance was computed by dividing the ILL, LA, INC, POC, and than those found for the videotaped (t = 2.6, df = 29, p <
Total FrO scores by the number of utterances made by each 0.03) recording method.
child.
The chance corrected kappa reliability statistic (Reiss, Study II: Effectiveness of the Kiddie Formal Thought
1973), was calculated for the interrater agreement on the Disorder Story Game (SG)
presence or absence of K-FrOS ratings throughout the SG of
all 28 subjects included in the reliability study. The possible To examine the SG effectiveness in eliciting good speech
influence of diagnosis, the raters' previous research experi- samples from children, the speech samples obtained were
ence, and the recording technique on interrater agreement compared with those elicited by an open-ended clinical inter-
was assessed by comparing the kappa values for the target and view that focused specifically on psychotic symptomatology.
normal subject, for Group I and Group II, and for videotaped,
METHODS
audiotaped, and transcribed speech samples of the SG. Each
of the three recording methods was examined on two target Subjects and Procedures
and one normal child, with a total of six target and three
normal children. The Children's Schizophrenia Interview (CSI), developed
by B.F. and R.C., is an open-ended, structured interview that
RESULTS focuses on psychotic symptomatology. It was used for this
Table 3 presents the chance corrected reliability (kappa) of study as a source of speech samples assumed to be similar to,
the ILL, LA, and Total FrO scores within and across diag- and, therefore, representative of typical speech samples ob-
nostic groups, sets of raters, method of recording the speech tained from children by traditional psychiatric interviews.
samples, and interview. Due to a low base rate, the kappa The SG and the CSI were administered to a subsample of
values for INC and POC were not statistically meaningful 14 schizophrenic, 3 schizotypal, and 15 normal children, aged
(Shrout et al., 1987). INC and POC were each rated in only 5 to 13 years. Children participating in this study were seen
one child. on two occasions. At the first meeting the children partici-
ILL and LA were reliably rated in the 28 children included pated in the SG, a play session, and other cognitive testing.
in the reliability study with kappa values of 0.78 and 0.71, At the second meeting, held no later than 1 week after the
respectively. The overall K-FrOS kappa for both ILL and SG, the same interviewer conducted the CSI, which lasted 45
LA, referred to as K SU M , was 0.77. to 60 minutes. The first 20-minute segment of the videotaped
When calculated by diagnostic group (Table 3), KILL was CSI was compared with the entire SG videotape. The average
talking time of the target and normal subjects during the SG
was 15.3 (SO 3.9) minutes.
TABLE 3. K-FTDS Kappa Values across Diagnostic Groups. Raters, Three measures were used to evaluate the speech yield from
Recording Method. and Interview Procedures the SG and the CSI: (1) The number of utterances made by
Kappa Values the child was obtained during a 5-minute segment of the SG
and the CSI. (2) The quality of utterances was judged as either
Assessment Procedure N KILL SEa KLA SE KS U M SE good or poor based on how well the child elaborated on the
Diagnostic group topics of conversation and on the presence of numerous
Allsubjects 28 0.78 0.04 0.71 0.01 0.77 0.04 "Yes," "No," "I don't know," "When will we be done?" type
Target" 21 0.76 0.04 0.75 0.01 0.78 0.02 responses. (3) The child's compliance during the SG and the
Normal' 7 0.80 0.07 CSI was judged as good or poor.
Raters The interrater agreement on the utterance rate was calcu-
Inexperienced (Group I) 16 0.87 0.07 0.77 0.01 0.77 0.05 lated by means of the correlation coefficient (0.98). Percent
Experienced (Group II) 12 0.75 0.02 0.66 0.04 0.77 0.05 agreement was used to compute the interrater agreement for
Recording method
Video 3 0.87 0.09 0.71 0.03 0.83 0.09 the non parametric global ratings of speech quality (0.85) and
Transcription 30.86 0.16 0.91 0.18 0.86 0.15 compliance (0.88). The interrater agreement for these meas-
Audio 3 0.66 0.09 0.66 0.11 0.64 0.04 ures was not kappa corrected because of the low base rate of
Interview poor quality of speech and poor compliance (Shrout et al.,
SG 28 0.78 0.04 0.71 0.01 0.77 0.04 1987).
CSI II 0.44 0.1 0.48 0.05 0.45 0.07 The effectiveness of the SG and the CSI in eliciting good
a Standarderror. speech samples from children was also measured by compar-
"Schizophrenic and schizotypal children. ing the K-FrOS scores and K-FrDS kappa values obtained
.. KLAand KS U M not computed because of a lowbaserate (see text). with these two interview procedures. The K-FrDS chance
412 CAPLAN ET AL.

corrected interrater agreement on the CSI speech samples was TABLE 5. Distribution ofthe Target and Normal Children in the
examined on seven schizophrenic and four normal children. Validity Study by Age and Sex
Age Groups Sex
RESULTS
4.4-7 7-9.6 9.7-12.5 IQ
Table 4 presents data on the number of utterances, quality N yrs. yrs. yrs. M F (WISC-R)
of speech, compliance, and K-FrDS scores obtained from the
SG and the CSI. A repeated measures analysis of variance of Target
Schizophrenic 16 0 4 12 13 3 90
the number of utterances suggested no difference between the Schizotypal 4 0 I 2 2
3 86
SG and the CSI by diagnostic group or across all subjects. Normal"
More than 80% of the target and normal children produced CA matches 20 0 7 13 15 5 113
speech samples of good quality and were compliant during MA matches 20 5" 8 7 15 5 113
the SG (Table 4). Only one target subject refused to participate " Eleven children were common to both CA and MA matches.
in the story game. A chi-square analysis of dichotomous "The target subjects were older than 7 years. Five target subjects
variables for repeated measures (Guthrie, 1981) demonstrated were matched with normal matches between 4.4 and 7 years based
that the SG produced significantly better speech quality (X = on their MAs.
17.55, df= I, p < 0.0001) and compliance (X = 3.89, df= I,
p < 0.05) than the CSI for the combined target and normal
groups. The chi squares for the group and group by test tal age (MA) (Table 5). Eight of these target subjects were also
procedures, however, were not significant for either quality of included in the reliability study. Normative data was also
speech or compliance. collected from an additional 13 children, five of whom were
A repeated measure t test of the K-FrDS scores obtained aged 4.4 to 6.9 years, three of whom were 7 to 9.5 years, and
with the SG and the CSI demonstrated significantly lower (p five of whom were 9.6 to 13.5 years.
< 0.(03) K-FrDS scores on the CSI (Table 4). The KSU M , To compute the diagnostic and developmental validity of
KILL, and K LA of the CSI speech samples were significantly the K-FrDS, separate two-way analyses of variance were
lower (p < 0.001) than those obtained when using the SG conducted with the K-FrDS scores (LLL, LA, and Total
speech samples (Table 3). FrD) as dependent variables, and diagnostic group, chrono-
logical age group, and mental age group as classification
Study III: Validity of the Kiddie Formal Thought Disorder variables. Data were analyzed following logarithmic transfor-
Rating Scale (K-FTDS) mation. Actual values are reported, but significance tests were
This study was based on two hypotheses. The first, a diag- based on transformed values.
nostic hypothesis, predicted that K-FrDS scores would dif- The specificity and sensitivity of the K-FrDS were calcu-
ferentiate schizophrenic and schizotypal children from nor- lated by computing the number of false positive and false
mal children. The second, a developmental hypothesis, pro- negative K-FrD scores in the target and normal groups. A
posed that there would be age-related changes in the simple additive model for two strata was used for age correc-
schizophrenic, schizotypal, and normal children's K-FrDS tion of the K-FrDS scores. The average K-FrDS score for
scores and that the youngest schizophrenic children would normal children with a MA greater than age 7 years was used
have the highest FrD scores. as a baseline. The correction term was calculated by subtract-
ing the baseline from the average K-FrDS score of the
METHODS younger normal children. The K-FrDS scores of both the
Subjects and Procedures normal and target children were then adjusted by this differ-
ence in averages.
The SG and K-FrDS was administered to a subsample of
16 schizophrenic and four schizotypal children, matched with
RESULTS
29 normal children by sex, chronological age (CA), and men-
Table 6 presents the group average ILL, LA, and Total
FrD scores for the target children, the normal chronological
TABLE 4. Amount ofSpeech. Speech Yield. and FTD Scores age matches, and the normal mental age matches.
Obtained with the Kiddie Formal Thought Disorder Story Game When the ILL scores (Table 6) of the target subjects were
and the Children's Schizophrenia Interview
compared to those of the CA normal matches, the age x
diagnostic group interaction was not significant (F = 0.65, df
Speech Yield
K-ITDS = 2,39, p < 0.4). Significant main effects were found for both
No. of
UtterancesQuality Compliance Score diagnosis (F = 5.55, df = 1,39, p < 0.03) and age (F = 3.97,
(Total df= 1,39, P < 0.06). The two-way ANaYA of the target and
Interview N X SD Good Poor Good Poor ITD) MA matched normal subjects demonstrated a main effect for
SG both diagnosis (F = 5.47, df= 1,39, p < 0.03) and age (F =
Targets 17 13.3 4.4 88% 12% 82% 18% 0.44 6.85, df = 2,39, p < 0.006) with no diagnosis x age group
Normals 15 11.5 4.3 93% 7% 87% 13% 0.25 interaction (F = 0.21, df = 2,39, p < 0.81).
CSI The normal children had an average LA score and standard
Targets 17 11.94 3.6 76% 24% 53% 47% 0.26 deviation of zero (Table 6). Nonparametric statistics were
Normals 15 11.8 3.5 87% 13% 80% 20% 0.15 used to compare the incidence of LA scores above zero in the
 TESTING THE K-FTOS 413
TABLE 6. Illogical Thinking, Loose Associations. and Total FTD Scores in the Target and Normal Children Matched by
Chronological (CA) and Mental Age (MA)
CA Matches" MA Matches"
Target Normal Target Normal
Age Group X SO X SO X SO X SO
Illogical thinking
4.4-7 yrs." 0.32 0.11 0.21 0.21
7-9.6 yrs. 0.29 0.14 0.16 0.25 0.22 0.19 0.09 0.09
9.7-12.5 yrs. 0.15 0.11 0.06 0.06 0.11 0.06 0.05 0.03
All ages 0.21 0.15 0.11 0.19 0.21 0.15 0.10 0.13
Loose associations"
4.4-7 yrs. 0.20 0.22 0.0 0.0
7-9.6 yrs. 0.15 0.18 0.0 0.0 0.08 0.07 0.007 0.01
9.7-12.5 yrs. 0.10 0.11 0.002 0.007 0.11 0.14 0.0 0.0
All ages 0.13 0.14 0.001 0.004 0.13 0.14 0.003 0.01
Total rrn-
4.4-7 yrs. 0.57 0.20 0.2 0.22
7-9.6 yrs. 0.4 0.23 0.12 0.25 0.30 0.16 0.09 0.10
9.7-12.5yrs. 0.24 0.10 0.07 0.06 0.22 12 0.05 0.03
All ages 0.44 0.32 0.14 0.30 0.44 0.32 0.12 0.18
a See Table 5 for sample sizes.
"One target subject with a MA of 6.7 was matched to a normal child who was 7.2 years. This pair was included in the 7- to 9.6-year age
group for the data analyses.
..Significance tests between the diagnostic groups were not computed because of the zero base rate of LA in the 7- to 9.6-year-old CA
matches, and in the 4.4- to 7- and 9.7- to 12.5-year-old MA matches.
d Sum of the illogical thinking and loose associations scores.

target and normal groups. LA occurred in 75% of the target TABLE 7. K-FTDS Scores ofNormal Subjects
subjects versus II % of the normal subjects (x 1 = 20.41, df =
K-FrOS Signs
38, p < 0.0001) and was rated 15 times more frequently than
in the normal children. Illogical Loose
The total FrD scores of the target and normal CA matches Thinking Associations Total FrO
demonstrated no interaction between diagnosis and age (F = Age Groups N X SO X SO X SO
1.35, df = 1,39, p < 0.26) with a significant main effect for
4.4-7 yrs. 14 0.25 (0.23) 0.02 (0.03) 0.26 (0.23)
both diagnosis (F = 19.01, df> 1,39, p < 0.0004) and age (F
7-9.5 yrs. II 0.06 (0.08) 0.00 (0.00) 0.06 (0.08)
= 5.92, df= 1,39, P < 0.02). In the case of the normal MA
9.6-13.5 yrs. 17 0.06 (0.06) 0.00 (0.00) 0.06 (0.06)
matches, there was also no diagnosis x age interaction (F = All ages 42 0.11 (0.15) 0.006 (0.02) 0.12 (0.16)
1.69, df = 2,39, p < 0.21). There were, however, significant
effects for diagnosis (F = 34.8, df = 1,39, p < 0.00 I) and for
age (F = 8.19, df = 2,39, p < 0.003). occurred at age 7 years. Developmental changes were not
From the diagnostic perspective, the data indicated that the found in the K-FrDS scores of the CA matched normal
K-FrDS scores differed statistically between the target group children because the children in this group were older than 7
and both groups of normal subjects. In terms of differentiating years (Table 6).
between schizotypal and schizophrenic children, however, two When examined by age, the youngest target subjects had
of the four schizotypal subjects had an average Total FrD the highest ILL (F = 3.5, df = 2,19, p < 0.05) and Total FrD
score (0.05) similar to those of the normal subjects; the average scores (F = 7.51, df= 2,19, p < 0.004) (Table 6). Although
score (0.55) of the other two schizotypal children was in the the LA scores of the youngest target subjects were higher than
pathological range (see below). those of the older target children, these differences were not
Table 7 presents the normative data on the K-FrDS scores. statistically significant. Similar to the normal children, age
Significant differences occurred across age groups in the ILL seven years was the developmental cutoff point for the K-
(F = 7.36, df = 2,39, p < 0.002) and Total FrD (F = 8.56, FrDS scores of the target subjects. The developmental differ-
df = 2,39, p < 0.0009) scores of the normal subjects. The LA ences in the ILL scores of the 7- to 9.6-year-old target subjects
scores across age groups could not be compared because this was accounted for by scores of those children who had a
K-FrD sign occurred rarely in the normal children under 7 mental age of less than 7 years.
years and not at all in those above 7 years. Because of the The sensitivity and specificity of the ILL, LA, and Total
infrequency of LA in the older normal groups, there was a FrD scores are shown in Figure 1. The diagnostic validity of
similarity in the ILL and Total FrD scores of the middle and the Total FrD score was clearly superior to that of its two
old normal children. The cutoff point for the developmental components, ILL and LA. Prior to age correction, the sensi-
differences in the K-FrDS scores of the normal children tivity and specificity of the Total FrD scores was 86% with a
414 CAPLAN ET AL.

ILLOGICAL THINKING LOOSE ASSOCIATIONS

o .2 .4 .6 o .2 .4 .6 .8
I I I I I I I I I
I _ TARGET .I: i_, _ I •
•
CA-CONTROL~I
o

if--
i~o
::----_-------
MA-CONTROL em " ---r---,-r-.-----,--,-----.----,

100f;
TOTAL
o ~
Se
I f .I
FTO o .2 .4 .6 .8 1.0 1.2 1.4
I I I I I I I I
TARGET
CA 08~00
...........
•• • ••o • o •
CONTROL 0 8g
o 0
000:00 0
MA 00<:ar0
CONTROL 00
00
00

I i
FIG. I. Sensitivity and specificity of the Total Formal Thought Disorder. Illogical Thinking. and Loose Association Scores. Se = sensitivity; Sp
= specificity; 0 = normal children who were used as both chronological and mental age matches.

cutoff point of 0.2. After the age correction. Total ITO had children from normal children independent of the effects of
a sensitivity and specificity of79% and 90%, respectively. An chronological and mental age. The sum of the Total ITO,
LA score of 0 was highly specific (91%). This measure, how- was a sensitive (79%) and specific (90%) measure of ITO in
ever, was not sensitive enough (46%) to diagnose all of the children. The presence of LA in middle childhood was partic-
target subjects , a quarter (5 of 20) of whom also had an LA ularly suggestive ofa diagnosis of schizophrenia. The K-ITOS
score of O. The age corrected sensitivity and specificity of ILL scores of the schizotypal children spanned the interval be-
was 74% and 76%, respectively, for a cut point of 0.1. tween the normal and the schizophrenic groups.
Fourth, significant developmental changes were found in
Discussion the ILL scores and Total ITO scores of the target and normal
subjects. The youngest target subjects had the highest K-ITOS
The results of these three studies have demonstrated the scores. Similar to other cognitive skills, age 7 was the cutoff
following. First, two of the four K-ITOS signs, ILL and LA, point for the developmental changes in the K-ITOS scores
were reliably assessed with both videotapes and transcriptions (Piaget, 1928; Piaget, 1959).
in the schizophrenic, schizotypal , and normal children. The The low frequency of INC in this sample is comparable to
two remaining K-ITOS signs, INC and POC, had a low base that described in adult schizophrenics (Andreasen and Olsen,
rate. The interrater reliability for the sum of the ILL and LA 1982). The low base rate of POC, however, could stem from
scores was consistent across diagnostic groups and across a developmental or instrument-based cause. From the devel-
raters independent of their prior clinical and research training. opmental perspective, the speech of children in middle child-
Second, the story game was a better technique than a standard hood tends to be nondiscursive and unelaborated. This char-
psychiatric interview for eliciting speech for the assessment of acteristic of children's speech was observed in the preliminary
ITO in middle childhood. work and was a guide in operationalizing the criterion for an
Third, the ILL and LA scores differentiated schizophrenic "adequate amount of speech" to two or more utterances. This
 TESTING THE K-FTDS 415

explanation for the infrequency of pac is supported by the Hispanic and Asian children in the normal groups and Afro-
observation that the only child in this study with pac was a Asian children in the target groups, there is not yet enough
very bright 12.9-year-old normal boy with excellent verbal evidence for the role of ethnicity on children's verbal com-
skills and with a tendency to be very expansive in his speech. petence (Ochs, 1987, personal communication).
In terms of the K-FrDS instrument, the scale's four signs The findings also suggest that ILL and LA are valid meas-
are mutually exclusive. It is, therefore, possible that pac was ures of FrO in middle childhood and that the scale also has
"underrated" because its definition could not be applied to diagnostic validity in younger children. In fact, an age correc-
utterances that had already been rated as ILL, LA, or INC. tion of the K-FTDS scores of the young normal children
Finally, .the infrequency of both INC and pac does not improved the instrument's specificity to 90%. Thus, if used
appear to be related to neuroleptic medication for three with children under 7 years, the K-FrDS could prevent
reasons. First, only 42% of the sample received neuroleptics overdiagnosis of immature speech as schizophrenic FTD.
at the time of the study. Second, there were no significant Together with the good and consistent interrater reliability
differences between the K-FrDS scores of the medicated and for ILL and LA, the validity findings demonstrate that the K-
non medicated patients. Third, Spohn et al. ( 1986) have shown FTDS is a good clinical research instrument for studying FTD
that neuroleptics are probably less likely to mitigate the oc- in middle childhood. Although currently a research instru-
currence and severity of negative signs, such as pac in adult ment, the K-FTDS uses a clinical technique that could be
schizophrenics. readily adapted for regular clinical use.
The results of the reliability study suggest that prior psychi-
atric training or research experience does not affect the inter- References
rater reliability.
The CSI was administered to the children after they were Andreasen, N. C. (1979), Thought, language, and communication
disorders. Arch. Gen. Psychiatry, 36: 1315-1323.
familiar with the interviewer and were possibly feeling less - - Olsen, S. (1982), Negative vs. positive schizophrenia. Arch. Gen.
anxiety than during the SG. Despite the fact that the story Psychiatry, 39:789-794.
game was administered at the child's first meeting with the Arboleda, C. & Holzman, P. S. (1985), Thought disorder in children
interviewer, it elicited superior speech samples. The results of at risk for psychosis. Arch. Gen. Psychiatry, 42: 1004-1013.
Asarnow, R. F., Sherman, T. L. & Strandburg, R. (1986), The search
the SG efficacy study are better understood if viewed in terms
for a psychobiological substrate of childhood onset schizophrenia.
of the different techniques used in the SG and the CSI. Both J. Am. Acad. Child Psychiatry, 26:601-614.
techniques focused on pathological thought content. The CSI, Bernstein, B. (1964), Elaborated and restricted codes: their social
however, approached symptoms with direct questions, origins and some consequences. In: The Ethnography of Commu-
whereas the SG used an indirect method of questioning by nication. ed. J. Gumper & D. Hymes. American Anthropology,
66:55-69.
focusing on the characters in the stories. The results of this Chaika, E. O. (1974), A linguist looks at "schizophrenic" language.
study indicate that the indirect or "projective" approach, as Brain Lang., 1:257-276.
used in the TDI (Johnston and Holzman, 1979), appears to Chapman, L. J. & Chapman, J. P. (1973), Disordered Thought in
be far superior than direct questioning for eliciting speech Schizophrenia. New Jersey: Prentice Hall.
Cantor, S., Evans, J., Pearce, J. & Pezzot-Pearce, T. (1982), Childhood
suitable for FrO rating in children. schizophrenia: present but not accounted for. Am. J. Psychiatry,
Three factors, one cognitive and two social, could have 139:758-763.
influenced the diagnostic differences between the K-FrDS Emerson, H. F. & Gekonski, W. L. (1980), Development and com-
scores of the target and normal subjects. The cognitive factor prehension of sentences with "because" or "if." J. Exp. Child
was the difference in the IQ scores of the target and normal Psychol., 29:202-224.
Fish, B. & Ritvo, E. R. (1979), Psychoses of childhood. In: Basic
subjects (Table 5). To further examine the role of IQ on the Handbook ofChild Psychiatry, ed. J.D. Noshpitz. New York: Basic
K-FrDS scores, the authors are currently conducting a project Books, v. II, pp. 249-304.
on the cognitive functioning of thought disordered schizo- Fleiss, J. (1973), Statistical Methodsfor Rates and Proportions. New
phrenic children. Preliminary findings suggest that even York: Wiley.
Gardner, R. A. (1971), Therapeutic Communication with Children.
though there is a reciprocal relationship between K-FrDS The Mutual Storytelling Technique. New York: Science House.
ratings and IQ, the K-FrDS significantly differentiates schiz- Green, W. H., Campbell, M., Hardesty, A. S. et al. (1984), A com-
ophrenic from normal children after controlling for the effect parison of schizophrenic and autistic children. J. Am. Acad. Child
of IQ. This suggests that the normal children had lower K- Psychiatry, 4:399-409.
FrDS scores because they did not have a diagnosis of schiz- Guthrie, D. (1981), Analysis of dichotomous variables in repeated
measures experiments. Psychol. Bull., 90: 189-195.
ophrenia or schizotypal personality disorder, and not because Hairno, S. & Holzman, P. S. (1979), Thought disorder in schizophren-
they had higher IQ scores. ics and normal controls: social class and race differences. J. Consult.
The two remaining potential confounding factors were the Clin. Psychol., 47:963-967.
greater proportion of middle-class families and the more Harrow, M. & Quinlan, D. (1977), Is disordered thinking unique to
schizophrenia? Arch. Gen. Psychiatry, 34: 15-21.
heterogenous ethnic composition of the normal group. Middle Herjanic, B. & Campbell, W. (1977), Differentiating psychiatrically
class families emphasize good conversation skills and these disturbed children on the basis ofa structured interview. J. Abnorm.
could have contributed to the normal children's lower K- Child Psychol., 5: 127-134.
FTDS scores (Bernstein, 1964; Labov, 1966; Piaget, 1959). Johnston M. H. & Holzman, P. S. (1979), Assessing Schizophrenic
Haimo and Holzman (1979) found significant class differ- Thinking: A Clinical and Research Instrument for Measuring
Thought Disorder. San Francisco: Jossey-Bass Publishers.
ences in the Thought Disorder Index (TDI) scores of adult Kolvin, I. (1971), Studies in childhood psychoses. Br. J. Psychiatry,
schizophrenics and normal controls. 118:381-384.
Despite the presence of a greater proportion of bilingual Labov, W. (1966), The Social Stratification of English in New York
416 CAPLAN ET AL.

City. Washington, DC: Center for Applied Linguistics. Discourse ofSchizophrenic Speakers. New York : Plenum Press.
Maratsos, M. P. (1976), The Use of Definite and Indefinite Reference Romaine, S. (1984), The Language ofChildren and Adolescents. New
in Young Children. Cambridge: Cambridge University Press. York: Basil Blackwell.
McTear, M. F. (1985) , Children 's Conversation . Oxford, Basil: Black- Russell, A., Bott, L. & Sammons, C. (1987), The phenomenology of
well Publisher. schizotypal disorder ofchildhood: A schizophrenia spectrum disor-
Neuchterlein, K. H., Edell, W. S., Norris, M. & Dawson, M. E. der'! Paper presented at the 34th Annual Meeting of the American
(1986) , Attentional vulnerability ind icators, thought disorder, and Academy of Child Psychiatry, Washington, DC.
negative symptoms. Schizophr. Bull. , 12:408-426. Sacuzzo, D. P. & BraIT, D. L. (1986) , Information-processing abnor-
Ochs, E. & SchieITelin, B. B. (1979), Developmental Pragmatics. New malities. Schizophr. Bull. , 12:448-459.
York : Academic Press. Shapiro, T. & Huebner, H. F. (1976), Speech patterns offive psychotic
Olson, D. R. & Nickerson, N. (1978) , Language development through children. J. Am. Acad. Child Psychiatry, 15:278-293.
the school years. In: Children's Language, ed. K.E. Nelson. New Shrout, P. E., Spitzer, R. L. & Reiss, J. L. (1987), Quantification of
York : Gardner Press, v. I, pp. 117-170. agreement in psychiatric diagnosis revisited . Arch. Gen. Psychiatry,
Piaget, J. (1928), Judgment and Reasoning in the Child. New York : 44;172-177.
Harcourt Brace. Spitzer, R. L., Endicott, J. & Robins, E. (1975), Research Diagnostic
Piaget, J. (1959), The Language and Thought ofthe Child. New York : Criteria (RDC) for a Selected Group ofFunctional Disorders, 2nd .
Humanities Press. Ed. New York: Biometric Research, New York State Psychiatric
Puig-Antich, J. & Chambers, W. (1978) , The Schedule for Affective Institute.
Disorders and Schizophrenia for School-Age Children (Kiddie- Spohn, H. E., Lolafaye, C, Larson, J., Mittleman, F., Spray, J. &
SA DS). New York: New York State Psychiatric Institute. Hayes, K. (1986), Episodic and residual thought pathology in
Rochester, S. R. & Martin, J. R. (1979), Crazy Talk: A Study ofthe chronic schizophrenics. Schizophr. Bull. , 12:394-407.