A comparative study of dreams and related fantasies


Material Information

A comparative study of dreams and related fantasies
Physical Description:
vii, 118 leaves : ; 28 cm.
Blume, Ginger Elaine, 1948-
Publication Date:


Subjects / Keywords:
Fantasy   ( lcsh )
Dreams   ( lcsh )
Psychology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Psychology -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis--University of Florida.
Bibliography: leaves 111-116.
Statement of Responsibility:
by Ginger Elaine Blume.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000090499
oclc - 05802063
notis - AAK5891
System ID:

Full Text







Like the protagonist in my favorite fantasy story,

Alice's Adventures in Wonderland, by Lewis Carroll, I have

undergone major life transformations while exploring this

research. There were times when I felt much too small

for the huge task before me. There were times when I felt

lost and perplexed, as if lost in the Queen's maze of

rose hedges. There were times when talks with the statis-

tician seemed like conversations with the Mad Hatter.

There were times when I heard the White Rabbit's warning,

"I'm late, I'm late, for a very important date," and

experienced my own sense of urgency to submit the finalized

dissertation. And there were times when I felt sure I'd

never wake up from the dream.

Fortunately, sanity and clarity arose out of contacts

with Dr. Webb, my chairman. He was clearly the main effect

(p < .05) in the completion of this dissertation. His

faith in my own sense of timing was a highly significant

factor. And, of course, his technical expertise and

knowledge were invaluable. Other committee members, while

less actively involved, were no less highly supportive.

Many thanks to Dr. Ben Barger, Dr. Robert Hornberger,

Dr. Norman Markel, Dr. Vernon Van De Riet, and Dr. Marilyn G.


The "raters," David Cunningham and Anne Gallagher,

deserve recognition for their exceptional perseverance in

performing numerous tedious tasks. Similarly, the "judges"

volunteered an enormous amount of time and effort in

sorting and cataloguing fantasy protocols. When the

computer failed, their human effort transformed the data

into meaningful information. Therefore, special apprecia-

tion is extended to Michael Bonnet, Mariam Greenfield, Jan

Hembree, Judith Levy, Bernie Webb, and Marilyn G. Holly.

Jan Hembree, Paul Nelson, and Corie Fox skillfully

challenged my competitive spirit and spurred me on to

completion of this research. Bob Boylin, who entered the

scene near the final stages of my dissertation, proved to

be the best fantasy I had encountered.








Background . .
Research Findings .
Conceptualization and Hypotheses


Subjects .
Raters .
Judges .
Instrumentation .
Procedure .
Data Collection .
Protocols .
Sorting Procedures
Rating Procedures
Analyses ..



. 23
. 29
S 29
. 30
. 30
S 34
. 34
. 3,6
. 37

Hypothesis 1 .
Hypotheses 2 and 3 .
Hypotheses 4 and 5 .. ..
Hypothesis 6 .
Hypothesis 7 .
Hypothesis S .
Conclusions and Implications

S. 32
. 86
S. 91
. 93

S 95



. 21
. 5
. 22





REFERENCES ... . .. .. 111


Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of
the Requirements for the Degree of Doctor of Philosophy



Ginger Elaine Blume

June, 1979

Chairman: Wilse B. Webb
Co-Chairman: Vernon Van De Riet
Major Department: Psychology

An issue which warranted research attention was

whether the dream report was significantly different from

other forms of internally-generated fantasy reports.

Studies which had previously compared dreams and projective

fantasy materials had partially and unsystematically

explored this issue. No systematic examination of fantasy

mentation obtained from four fantasy sources within the

same group of subjects has been reported.

Four independently derived written fantasy samples from

each of 48 subjects were compared. Varying in degree of

structure, fantasy samples produced by subjects were

dreams, free fantasies, written samples of significant life

experiences, and stories in response to Thematic Appercep-

tion Test (TAT) cards. The subjective evaluations of a

heterogeneous group of six judges were employed as the

"criterion of comparison."

Overall, judges were able, at a greater than chance

level, to identify each of the four fantasy materials belong-

ing to a single subject. That is, on the individual level,

the relationship among content-related aspects of various

forms of fantasy were significantly related. This was

found to be true as well with respect to a structural

property (word count) of fantasy materials. These findings

supported the notion that dreams, free fantasies, written

samples, and TAT's are orderly, non-random events which

reflect a common source of fantasy rather than unique types

of fantasy.

Four forms of fantasy mentation were evaluated, not

only in relationship to the individual but also in rela-

tionship to two selected, independently defined group

differences. Each of the fantasy levels was found to be

significantly sensitive to sex membership and test-reported

anxiety level.

The results of this study supported the viewpoint that

various forms of fantasy mentation are "splinters" from a

common source. Each may be a more sensitive reflector of

particular individual characteristics, yet all are sensi-

tive to and capable of revealing individual and group differ-

ences. In summary, fantasies, whether originating during

a waking or sleeping state of consciousness, are more

similar than different and appear to emerge from the same

primary process pool. They are all equally useful as

clinical tools.





Human beings spend most of their time engaging in some

kind of mental activity, much of which consists not of

ordered thought, but of bits and snatches of inner experience:

wandering internal monologues, daydreams, fantasies, dreams,

reveries, and so on. All of these so-called "wool-

gathering" activities have received both clinical and artis-

tic but very little scientific attention, with the exception

of dreams. Indeed, the first book devoted entirely to the

topic of daydreaming was not published in the United States

until 1956, entitled Daydreaming,by Jerome L. Singer. Klinger

(1971) maintains in his book, Structure and Functions of

Fantasy, that "after millennia of philosophical inquiry into

mind, a half century of experimental analysis cf conscious-

ness, and a further century of experimental analysis of

behavior, very little is actually known about fantasy" ;p. 4)

Therefore, lacking a conceptualization and methodology for

studying the covert process of fantasy, past investigators

found themselves relying on "fantasy-like" processes, in an

attempt. to gain insight into these ephemeral and fleeting

thoughts. This included investigations of free play, studies

of psychotic language, Thematic Apperception Test stories,

and Rorschach inkblot responses, to name a few. These

"substitutes" for fantasy were sometimes confused for the

real thing, leading to misconceptions and ill-founded notions

regarding normal fantasy. In comparison, dreams of the

night have received significant scientific notice, particu-

larly beginning with the discovery in 1953 by Aserinsky and

Kleitman that dreaming could be objectively identified by

EEG patterns. The resulting research technology enabled

investigators to systematically and objectively explore

dreams and dreaming.

While historically the major psychologists have disa-

greed on the functional significance of dreams, they have all

agreed on the importance attributed to dreams. Freud, for

example, viewed dreaming as the preserver of sleep. The

dream was hypothesized to represent an unfulfilled, repressed

wish which was expressed in disguised form (Freud, 1958).

Jung saw the dream's purpose to be twofold. Activities of the

dream supposedly compensated for conscious orientation to

balance out the personality. In addition, some dreams

expressed universal tendencies about man's nature through

expression of the collective unconscious (Jung, 1933).

Adler viewed the dream as preparation for the subsequent

day's activity-a dress rehearsal for waking life. While

their proposed theories differ, their underlying assumptions

appear strikingly similar. They all assume that dreams are

meaningful, non-random events in that they bear some

relationship, whether overt or covert, to the waking life of

the dreamer.

Dominant psychological theories of the century have

unequivocally maintained that dreams and fantasies represent

a common psychological process. Freud (1953), in his paper

on the relation of the poet to daydreaming, treated these two

phenomena as though they were one and the same: "I cannot

pass over the relation of fantasies to dreams. Our nocturnal

dreams are nothing but such fantasies. Language, in its

unrivaled wisdom, long ago decided the question of the

essential nature of dreams by giving the name of 'daydreams'

to the airy creations of fantasy" (p. 67).

Psychoanalysts have since maintained that fantasies,

including dreams, daydreams, myths, and artistic productions

tap the same underlying psychological process (wish ful-

fillment). According to Freud, "every separate phantasy

contains the fulfillment of a wish, and improves on unsatis-

factory reality" (in Feshbach, 1955). Thus, unsatisfied wishes

were conceptualized by Freud and others as the driving

power behind fantasies. Freud and his successors have also

maintained that these various "splinters of fantasy" share

"primary process" as a common structural style.

Piaget (1962) has asserted that both the content and

the symbolic structure of children's dreams and symbolic

play are intimately related. On the basis of systematic and

naturalistic observation, Piaget concluded that all imagina-

tive thought was, in fact, interiorizedd play." Although

psychoanalytic theorists had arrived at the same conclusion,

that fantasy and play were a single process, they viewed

fantasy as a prior process which eventually led to motoric

expression in play (Rapaport, 1951), rather than the

reverse developmental order.

Adler also expressed the belief that dreams parallel

events of waking life (in Klinger, 1971). Apparently, how-

ever, Adler saw little need for distinguishing dreams from

imaginative thought, for he felt that they equally expressed

an individual's waking style of life (Jones, 1974). That

is, the Adlerian notion views dreams as working on solutions

to current problems derived from waking conscious experience.

Other theorists expressing a similar view include French,

Hall, and Fromm, to name only a few (Kramer & Whitman,


The notion that fantasy and dreams share the same

properties and serve similar functions is extensively shared

by various theorists. Until very recently, however, there

has been little impetus for empirically verifying the

similarities between fantasies and dreams, and few investi-

gations of the relationship have been attempted (Klinger,

1971; Singer, 1956). Furthermore, the few studies which

have compared fantasies and dreams have largely employed


structured stimuli such as TAT stories, Rorschach responses,

etc., rather than collecting spontaneous or free fantasies

in situ. This strong reliance on "substitute," and hence

artificial fantasy material, has probably introduced bias

into the interrelationships between normal fantasy and

dreams, as well as contributed to contradictory findings.

Research Findincs

Content variables related to dreams and projective

fantasy have been systematically investigated. Tomkins (1942)

reported that dream life found representation in imaginative

production of the Thematic Apperception Test (TAT), based

on comparison of major themes of the TAT and manifest content

of a single subject's dreams. Similarly, Sarason (1944),

employing 34 mentally defective subjects, compared TAT

stories with dreams, and found, in general, that dream

material and stories were similar. Sarason arrived at this

conclusion despite the fact that all major TAT themes were

not represented in the dream reports. A more comprehensive

study which confirmed Sarason's impressions, was carried

out by Gordon in 1953, usinq a Need-Press Analysis to com-

pare dream responses to TAT responses. Gordon took 23 psy-

chiatric patients and, on the basis of 42 content variables

(e.g. aggression, affiliation, sex, dependency, independence),

compared ratings between dreams and TAT's. Eleve. of

the 42 TAT-Dream correlations were found to be statistically

significant, and 32 were positive. Gordon concluded that a

definite and reliable relationship, although not a particu-

larly close one, existed between the type of content found

in subjects' dreams and their TAT stories. Although all of

these studies suffer from numerous methodological flaws,

they do indicate at least a moderate relationship between

thematic content of TAT stories and morning-after dream


In addition to content variables, TAT's and dreams

have been compared on the basis of structural variables.

Structural variables refer to the manner in which an indi-

vidual interprets a fantasy stimulus, i.e., how the response

is given as opposed to what is given (Kagan & Lesser, 1961).

Foulkes and Rechtschaffen (1964) found structural similari-

ties between waking TAT fantasies and nocturnal REM-sleep

mentation, but not between waking TAT fantasies and NREM-

sleep mentation. Also, they found little systematic rela-

ship in "instinctual content" between TAT stories and REM

reports. In a later study by Foulkes, Pivik, Steadman,

Spear, and Symonds (1967), with 32 boys ages 6 through 12,

they found no relationship between the structure of Children's

Apperception Test (CAT) scores and dream scores. These

authors concluded that "styles of mental approach are not

nearly so stable or consistent across the sleep-wakefulness

border for children as they are for adults" (p. 463).

Fiss, Klein, and Bokert (1966) have initiated a differ-

ent strategy for assessing the relationship between dream

states and waking TAT fantasies. To obtain fantasies, sub-

jects were awakened from REM period and were asked for

fantasies in response to TAT cards. They found that dream

reports and corresponding fantasies were strikingly similar

in content. They concluded that when a REM-dream is aborted,

a waking fantasy appears to take on some of the characteris-

tics of dream mentation. Hence, dreams and fantasies may

serve a similar purpose or function.

A series of related studies involving dream deprivation

have also attempted to show a functional relationship between

dreams and fantasies. A dream deprivation study reported by

Cartwright (1969) found that "on those nights when a secon-

daryprocess task was required of the subjects during the

deprivation awakenings,more REM compensation followed than

onnights when primary process material was elicited" (p. 365).

A study by Palmer (1963) had also reported a relationship

between waking fantasies (Rorschach scores) and total sleep

deprivation. Palmer found a tendency for a balance shift

toward the introversive side of the M:Sum C ratio on the

Rorschach, following total sleep deprivation. Similarly,

Lerner (1966) found a significant positive relationship be-

tween human movement responses (M) on the Holtzman inkblots

and REM-dream deprivation. A later study by Cartwright,

Gore, Hancock, McCarthy, and Weiner (1973) further confirmed

the previous studies which had found an increase in Rorschach

human movement responses and a shift to the introspective

side of the experience balance following REM deprivation.

All these studies imply a specific need for primary process

thinking which is unsatisfied by other secondary forms of


It can be observed that the results from the above-

mentioned studies, which have compared content and/or struc-

tural aspects of projective fantasies and dreams, are replete

with inconsistencies. For instance, the Foulkes and

Rechtschaffen (1964) and Foulkes et al. (1967) studies dif-

fered from the earlier reported studies with respect to

technique for collecting dream reports. The earlier studies

used "morning-after" dream reports rather than REM-EEG

"sleep-interruption" dream reports; an observation which may

aid in explaining some of the inconsistent research results.

In contrast to earlier studies, dreams obtained by the

sleep-interruption method have been found, as a rule, to

resemble TAT fantasies with respect to structural properties

but not with respect to thematic content (Klinger, 1971).

Additionally, studies comparing TAT and dreams have differed

both with respect to subject classification (i.e. some

studies have utilized normal subjects and others, psychiatric

subjects) and age of subjects. Also, Hall and Van de Castle

(1966) have called attention to the fact that researchers,

when comparing dreams and TAT's have typically imposed

on dream protocols scales which were originally constructed

solely for purposes of measuring TAT stories (Kagan & Lesser,

1961). Besides being a somewhat arbitrary and inappropriate

classification system with dreams, the TAT "need" scores

have also been found to be highly unstable. Thematic

Apperception Test repeat reliability has been shown to depend,

in part, on time elapsed between successive administrations,

stability of the personality being tested, and stability of

the individual's environment (Tomkins, 1942). The lack of

consistent research findings between dream reports and TAT

stories is better understood in light of the above observa-


In addition to these shortcomings, researchers have

mistakenly viewed projective fantasy material as providing

adequate representation for normal waking free fantasy. Both

Holt (1961) and Singer (1966), however, considered them to

represent dramatically divergent processes. Klinger (1971)

also viewed these two forms of fantasy as being highly dif-

ferent. He defined projective fantasy as ". a subject's

overt symbolic behavior when he is confronted with a more

or less standard, more or less ambiguous stimulus, and is

instructed to communicate to an examiner certain of his

responses to it ." (p. 90), whereas, free fantasy was

defined conceptually as ". spontaneous covert symbolic

activity that is not part of a perceptual scanning process or

of directed problem solving. It is spontaneous in the sense

that its production occurs without deliberate interference

from an examiner" (p. 90). It should be apparent from these

definitions that when we talk about normal waking fantasy

we are dealing with a broad range of cognitive processes and

products. The difference between free and projective fantasy

may have major psychological implications. Klinger (1971)

has concluded in his exhaustive study of fantasy that the

relevance of projective fantasy to free fantasy remains to

be established by empirical examination.

When other classes of fantasies, other than those ob-

tained by projective techniques, have beer compared to dream

mentation, consistent support for the notion that these two

forms of thought tap into an underlying psychic process is

revealed. Beck (1967) systematically examined the relation-

ship between daydreams and nightdreams with respect to the-

matic content and found that in eight out of ten patients

themes of daydreams appeared in most or all of the corre-

sponding nightdreams. Another study (Foulkes, Larson, Swanson,

& Rardin, 1969) involving a structural and content analysis

of male dreams of institutionalized and control subjects,

found a direct carryover of the wakefulness role a person

assumes to the role he assumes in his own dreams. Starker

(1974) compared daydreaming styles with nocturnal dreaming.

He hypothesized that the three daydreaming factors described

by Singer and Antrobus (1963) (i.e. conflictual or negative

daydreaming, positive daydreaming, and anxious, distractible

daydreaming) reflected fundamental qualities of fantasy life

which would also be reflected in sleeping fantasies.

Starker found profound and significant differences in dreaf

structure among subjects of differing daydream styles. In

a similar study, Amanat (1974) employed a content analysis

method, as well as a structural approach, in comparing day-

dreams and nightdreams of emotionally disturbed adolescents.

Elaborate steps were taken in this study to reduce the effects

of methodological shortcomings. Self-report daydreams were

compared to nightdreams, collected by the morning-after

method, for subjects in two psychiatric groups (inpatient

and outpatient) and one normal control group. Amanat found

positive correlations between content themes of "repetitive

daydreams" and nightdreams in each subject, with much

greater correlations occurring in the inpatient psychiatric

group than in the other two groups. There were also signifi-

cant differences between the contents of daydreams and night-

dreams in the psychiatric and control population groups, with

various degrees of positive correlations in each subject

and/or group of subjects. This study differed significantly

from previously reported studies in that repetitive or

reoccurring daydreams were introduced as a point of compari-

son with nightdreams.

Utilizing yet another form of fantasy, Cartwright (1966)

compared dream and drug-induced fantasy behavior. Dreams

were collected by use of the REM-EEG sleep interruption

technique, and drug-induced fantasy content was collected

during a 4-hour time period. During this time, subjects

were instructed to give a running verbal account of their

experience as it occurred. Despite evidence which suggested

that dream responsiveness and drug responsiveness formed two

independent patterns, Cartwright found convincing evidence

to show that drug-induced fantasies were,indeed, drawing

from the same source as normal dream content during wakeful-

ness. First, subjects experienced difficulty subjectively

differentiating the two states. Second, there was a dramatic

reduction in both the amount of time spent in REM and a

lower REM% the night following the drug session. And more

conclusively, blind matching of the dream content and halluci-

natory content records by a clinical psychologist was

correctly made at a very high level of significance. Also,

correctmatches were much more likely to occur when the collec-

tion sequence of the two experiences was Drug-Dream rather

than Dream-Drug. This study was particularly impressive

in view of the fact that an average of 12 days elapsed be-

tween collection of the two sets of data, ruling out the

possibility that correct matches were made on the basis of

superficial similarities.

Additional evidence supporting the contention that dreams

and fantasies tap into similar underlying psychic processes

has been provided by researchers investigating individual

differences. These studies have used other than projective

fantasies and have given rise to important findings.

Some investigators have focused not only on the rela-

tionship between dreams and fantasies, but on the potential

of these two phenomena to serve as accurate predictors of

subjects in predetermined criterion groups. Kramer, Baldridge,

Whitman, Ornstein, and Smith (1969) were interested in demon-

strating that dreams reflect differing personality configura-

tions and are capable of furnishing clues to areas of signifi-

cant differences among diagnostic groups. With respect to

man: :est content of recalled dream reports, Kramer et al.(1959)

found significant differences between two diagnostic groups

(paranoid schizophrenic and depressed) and a control group.

In a study by Segal (1974), fantasy material was used as

criterion for prediction of potential drug users. Seventy

percent of the non-drug users were correctly predicted as

to group membership. Gold and Robertson (1975) collected

night and day imagery of selected psychotic children and found

dream content to be different from that of normal children

of the same sex and age. These studies are compatible with,

and partially support, the notion that dreams and fantasies

reflect stable underlying personality differences.

In an important study, Cartwright, Monroe, and Palmer

(1967) examined the relationship between day and nighttime

behavior through a REM deprivation technique. Ten normal,

young, adult subjects were REM-deprived for three nights in

an effort to understand "where the dream goes" when an

individual is REM deprived. Cartwright et al. (1967)

hypothesized that dream-like mentation was ongoing but

ordinarily only reached conscious awareness during REM sleep

conditions. Further, when this REM state was artificially

limited, subjects' ability to productively utilize an

analogous state such as daydreaming, and thus avoid detri-

mental effects of REM deprivation, was hypothesized to be

contingent upon an individual subject's response style to

REM deprivation. These hypotheses were confirmed in this

study which also found at least three different patterns of

relationship between daytime fantasies and nighttime dreams

(disruption, substitution, and compensation).

Cartwright extended this notion of individual differ-

ences in a study with Ratzel (Cartwright & Ratzel, 1972).

A consistent relationship between subjects with high and low

REM-onset fantasy responses and their reactions to REM

deprivation was found. High REM-onset fantasizers showed

little or no post-deprivation test change on Rorschach,

Wechsler Adult Intelligence Scale (WAIS), and other test

scores, following REM loss, in marked contrast to low REM-

onset fantasizers. Cartwright and Ratzel noted that "those

who have a good deal of dream and fantasy material to report

at the REM-onset awakenings do not show as much compensation

after three nights of deprivation as those who had little or

none" (p. 365).

In the above two studies, the variable which seemed

most vital to determining subjects' reactions to REM depriva-

tion was the degree to which dream-like activity was limited

to REM periods during the night.

Several of the above-mentioned studies (Cartwright, 1966;

Cartwright et al., 1973; Cartwright et al., 1967; Cartwright

& Ratzel, 1972; Fiss et al., 1966; Lerner, 1966; Palmer,

1963) have attempted to unravel the functional relationship

of dreams and fantasies by examining the extent to which

dream-like activity was manifested in analogous waking states.

This has typically been observed by removing the dreaming

phase from sleep and monitoring the effects of the loss on

related states of consciousness. Cartwright (1969) has

recently pointed out that this subtraction strategy has taught

us a great deal about the REM state but very little about

dreaming in particular. There has been accumulated evidence

that when REM time is restricted there is pressure built up

for REM restoration (Cartwright, 1969). Thus, the REM

state serves some repetitive need. This in no way assures

any similar need quality to "the dream." In short, the

need for the REM state may be independent of the psychologi-

cal need for dream content. Cartwright (1969) has proposed

the following hypothesis: "They may both serve needed func-

tions which are independent of each other. REM time may

represent one convenient set of conditions for the dream to

take place. However, there may be other conditions, and if

the REM state is removed or reduced, the dream may become

dissociated from it and appear elsewhere" (p. 362). This

hypothesis was strongly supported by an earlier reported

study (Cartwright, 1966) which found a striking similarity

between the content of dreams and drug-induced fantasy

behavior during REM deprivation. Cartwright was implying

that the dream experience is capable of becoming dissociated

from the REM state, particularly when that state is in short

supply, as in REM deprivation conditions, etc. If this

were true, one would expect to find little, if any, signifi-

cant difference in the content and structure of dreams and

normal fantasies. In fact, a more recent study by Foulkes

and Fleisher (1975), using EEG and EOG recordings and sub-

jective criteria,has shown that subjects when in an awake

and relaxed state experienced hallucinatory and/or regres-

sive mentation, independent of the sleep state.

In the attempt to establish an underlying relationship

between various forms of waking and nighttime mentation,

studies have largely been focused upon dreams and projec-

tive fantasies and, more recently, upon dreams and free

fantasies. However, some attempts have also been made to

compare dreams with dreams within individuals and to compare

daydreams with projective fantasies.

The results of studies which have examined dreams in

sequence also imply some underlying ego process which is

stable over time and is consistently reflected in dreams of an

individual. Dement and Wolpert (1958) examined the rela-

tionship of four or more dreams occurring across the same

night with respect to manifest content. They found consider-

able relatedness among dreams occurring on the same night,

even to the extent of all dreams being interrelated by

expressing variations of the same theme or a continuing de-

velopment of the initial action. This finding was supported

by Offenkrantz and Rechtschaffen's study (1963) which found

not only continuities of sequential dreams, but also that

similar elements in the manifest contents of dreams con-

sistently reoccurred in the same ordinal position of sequences

on different nights. Another study (Trosman, Rechtschaffen,

Offendrantz, & Wolpert, 1960) rarely noted any direct rela-

tionship among the manifest content of sequential dreams of

a night. However, they did note the appearance of unique

elements at similar points in a dream sequence, suggesting

the organization of manifest dream contents into regular

patterns. A tendency for a cyclic relationship in latent con-

tent of sequential dreams was also reported. Kramer,

Whitman, Baldridge, and Lansky (1964), in an examination of

the interrelationship of dreams of a night, clinically and

experimentally confirmed the presence of a sequential pattern

of dream content, with a repetitious dream pattern represent-

ing a pattern which was on an opposite end of a continuum

from sequential dreams. A more recent study by Kramer,

Czaya, Arand, and Roth (1974) also confirmed previous findings

that the psychological dream experience has a developmental

course across the REM periods of a night. This trend was

supported in a study by Kramer et al. (1964) by use of

experimenter ratings, subject ratings, and ratings of mani-

fest content. Findings from the above-reported studies are

consistent with Freud's assumption that "dreams of a single

night are part of a single whole" (1958, pp. 333-334) and

are capable of revealing individual differences.

A relationship has been demonstrated between projective

fantasies (TAT and Rorschach) and free fantasies (daydreams),

but the specific nature of this relationship is equivocal.

For example, Page (1957), employing a questionnaire method

of collecting daydreams, compared the frequency of reported

daydreaming to a number of Rorschach variables. He found

that the number of human movement responses proved to be the

only scoring category significantly associated with daydream

frequency. In an earlier study, Page (1956) had also com-

pared the frequency of reported daydreaming to abbreviated

TAT responses. He found the TAT with respect to productivity

and emotional tone did not differentiate groups of high and

low frequency daydreamers. There was a tendency, however,

for "richer" TAT themes to be produced by low productive

daydreamers. An analysis of the relationship between com-

parable daydreaming and TAT themes did suggest an inverse

relationship between everyday fantasy and elicited TAT

fantasy. These studies were discrepant and suggest the need

for additional research.

To date, tremendous gaps remain in our understanding

of the relationship between dreams, normal fantasies, and

fantasy-like phenomena. Until quite recently, investigators

have lacked sophisticated methodology for examining fantasy

in its more natural form (i.e. an ongoing process) and have

tended to examine it in an unnatural, encapsulated form only,

as seen in previously reviewed studies which typically relied

upon TAT responses, etc. According to Klinger (1969), the

topic of free fantasy has consequently been virtually neg-

lected by researchers during the past forty years. Investi-

gators who have attempted to get at free fantasy in its more

natural form have generally requested subjects to describe

a recent daydream in detail, or to estimate the frequency of

various kinds of daydreams (Klinger, 1971). Whatever the

method, techniques for verbally reporting fantasies carried

liabilities as well as advantages. Fantasies are often

nonverbal (as are dreams) and hence proved difficult to

verbalize or clearly express. Methods which required recall

of daydreams that took place more than a few seconds earlier

also took a risk, since vivid fantasies were rapidly for-

gotten (Klinger, 1971). However, some promising techniques

have been developed. One was "thought-sampling"-i.e., re-

questing reports of fantasies as they occurred in time, as

an ongoing process (Klinger, 1971). Another was an approach

developed by Bertini, Lewis, and Witkin (1964) who

originally designed procedures for use in studying hypnagogic

and related phenomena, utilizing subjects "thinking out

loud." Subjects were requested to verbalize their thoughts

continuously as they occurred in time. Other investigators,

such as Cameron, Frank, Lifter, and Morrissey (1968), have

asked subjects to report recalled thoughts prior to an

experimental interruption. Gottschalk and Gleser (1969)

have asked subjects to talk for 5 to 10 minutes about a

meaningful and significant life experience. Klinger (1974)

has more recently experimented with various methods of

thought-sampling. Similar to the approach of Bertini et

al. (1964), Klinger has required subjects to verbalize

continuously for a period of 5 to 30 minutes. Rychlak (1973)

has also experimented with free imagery associations with

high school students during half-hour sessions. Foulkes and

Fleisher (1975), employing techniques developed in dream

research, have asked subjects to report, upon an auditory

signal which appears at randomly selected times from 1 to

9 minutes, their very last, presignal experience. Kripke

and Sonnenschein (1973) have also employed sleep research

methods for collecting free fantasy. In one study, each of

10 subjects was isolated in a room and was required to write

on a card a summary of "what he was thinking" during

successive 5-minute intervals.

These research techniques for tapping ongoing fantasies

and thoughts have led to a better understanding of the

relationship between waking and sleeping states. Earlier

studies had shown that these states were not discrete but

continuous. For example, Cartwright (1966) found evidence

that dream mentation was not restricted to the sleep state

but could enter consciousness during the waking state as

well. Foulkes and Vogel (1965) had also demonstrated that

the processes seen in dreams were by no means confined to

REM periods but could be seen in waking fantasy, hypnagogic

imagery, etc. That dream-like states are not confined to

REM periods of sleep has also been supported in studies

such as that of Fiss et al. (1966) which showed a "carry-

over" effect from dreams to waking fantasies (TAT's).

Demonstrably, waking and sleeping states are continuous not

only with respect to psychological processes, but also with

respect to various rhythmic biological and physiological

processes. Kripke and Sonnenschein (1973) found evidence

of a significant 90-minute rhythm in daytime fantasy

characteristic of REM nighttime cycles of EEG activity. An

excellent article by Globus (1966), reviewing findings on

cycles in real time, supported the notion that the REM state

is essentially a time-locked psychophysiologic process

which periodically emerges and does not have a fundamental

relationship to sleep or other special conditions.

Apparently, then, normal fantasy is susceptible to

many of the same effects of activation as are dreams, is

seemingly continuous with dreaming in time, shares many of

the same properties as dreams, but in different proportions,

and may serve at least some of the same functions. The

intimation is that fantasy and dreams are continuous (i.e.,

aspects of a unitary continuing fantasy process) and are

structurally and functionally closely related. However,

Singer (1976) has pointed out that we need more extensive

samples of ongoing daytime thoughts before we can be certain

that nighttime mentation does not have unique properties

with respect to waking mentation.

Conceptualization and Hypotheses

The validity of generalizing from dream material to

fantasy production, and vice versa, depends on the proposi-

tion that the two psychological processes are highly related

and that for at least some purposes the one may stand as an

analogue for the other. To date, this assumption has not

received adequate empirical verification, particularly with

respect to spontaneous fantasy. Although many research

studies have attempted to provide answers to this intriguing

question, none have allowed for adequate comparisons between

studies, due to lack of consistency in choice of subject

populations, techniques, criterion variables, definition of

terms, categories of analysis, etc. Despite these drawbacks,

there is at least general agreement that dreams and fantasies

bear some consistent relationship to each other. The kinship

of dreams with fantasy is reflected in a-number of associa-

tions which viewed as a whole, seem persuasive, although not

conclusive. Therefore, this study attempts to more closely

explore the relationship between dreams, spontaneous fantasies,

and structured fantasies.

Specifically, the questions posed in this research are

as follows:

1. Are dreams, free fantasies, written samples, and
TAT's meaningful to the clinician? That is, do
each of the fantasy materials sensitively reflect
individual as well as group differences?

2. Do all forms of fantasy material provide similar
information regarding the individual and the
group? If not, which materials are more useful?

3. Are fantasy materials (i.e. projective instruments
designed to tap primary process mentation) only
as good as the judge (examiner) utilizing them?
Or can it be demonstrated that "fantasy expres-
sions" are reliable and valid indices of person-
ality differences as well?

In order to answer these questions, eight hypotheses

were generated and tested. Each hypothesis, along with a

brief rationale, is stated below.

Many theorists, including Freud, Adler, and Jung, have

assumed that dreams are meaningful, non-random psychological

events. Hence, dreams are expected to reflection the

individual level,distinct differences between individuals.

Likewise, other splinters of fantasy expression are expected

to be sensitive discriminators of individual differences.

Such differences should be sufficiently characteristic of

the individual to be distinguishable from person to person.

Therefore, the following relationship is expected to be


Hol: Written productions arising from four different
forms of fantasy mentation (dreams, free fan-
tasies, written samples, and TAT's) of an in-
dividual have sufficient coherence to be
recognizable as originating from the same

On the group level it has already been demonstrated that

differential psychological groups (i.e. normals versus

psychiatrically disturbed individuals) and individuals of

differing sexual membership display consistent differences

with respect to dream content (Gold & Robertson, 1975;

Kramer, Roth, & Cisco, 1976; Segal, 1974; to name a few).

That other forms of fantasy expression would similarly

reflect group differences has not been fully explored.

Therefore, in this study, utilizing four different forms of

fantasy materials, two criterion variables were chosen for

differentiating subjects into known groups.

First, sex, which is one of the most highly researched

variables was chosen. Previous research has demonstrated

that male-female differences in dream content is an accepted

fact which has been most clearly and consistently demonstrated

in non-laboratory dream studies (Winget & Kramer, in press).

Since dreams are only one form of fantasy mentation, we

expect other forms of fantasy to similarly reflect group

differences, particularly with respect to sex.

Secondly, anxiety level was chosen as a relevant vari-

able for differentiating subjects in terms of group member-

ship. Anxiety was chosen as a sensitive discriminator,

given its central role in the understanding of psychological

processes and its traditional acceptance as an index of

emotional disturbance. In fact, in both the behavioral and

medical sciences, anxiety has been viewed as a central

explanatory concept in a majority of contemporary theories

of personality and psychopathology. In this study, anxiety

was operationally defined with respect to responses on the

Self-Evaluation Questionnaire, STAI Form X-2, by Spielberger,

Gorsuch and Lushene (1970).

By use of the above two criterion variables, sex and

anxiety, the following four hypotheses are proposed:

Ho2: Written productions arising from four different
forms of fantasy mentation can be accurately
and reliably sorted by judges into either of two
discrete domains (male and female) according to
subject's stated sex.

Ho3: Written productions arising from four different
forms of fantasy mentation can be accurately
and reliably sorted by judges into one of three dis-
crete domains (low, medium, or high) according
to subject's obtained anxiety scores.

Ho4: Judges' ability to successfully sort written
fantasy productions according to sex is similar
across the four forms of fantasy materials.

Ho-: Judges' ability to successfully sort written fantasy
productions according to anxiety level is similar
across the four forms of fantasy materials.

Historically, psychologists and personality theorists

have viewed primary process mentation to be unique to the

individual and reflective of underlying psychological

conflicts which are specific to the individual. Therefore,

each level of fantasy material would be expected to be a

reliable and valid index of personality differences. Hence,

each fantasy level may reveal individual differences to

judges. Therefore, subjects are expected to consistently

reveal themselves to judges across dreams, free fantasies,

written samples, and TAT's. The following relationship is

expected to be observed:

Ho6: Subjects reveal their sexual identity consis-
tently to judges across each of the four fantasy
productions. A similar finding is expected to
be observed with respect to subject's anxiety

Research findings have been stated previously which

have shown that the content of dreams is significantly

related to the content of several other forms of fantasy

mentation. Given this consistency in content (what a person

says) across fantasy levels, a similar consistency is

expected in the structure (how a person says what he says)

of fantasy mentation. Therefore, the following relation-

ship is hypothesized to occur:

Ho7: Written productions from each of the four fantasy
materials are correlated at a greater than chance
level, according to structural analysis scores.

Previous research has also demonstrated that the con-

tent of fantasy mentation is sensitive to group differences.

Similarly, the structure of fantasy mentation is expected

to be sensitive to group differences on at least some occa-

sions. In particular, a significant and strong relationship

between anxiety and quantity of verbal output has been

documented repeatedly in the research (Balken & Masserman,

1940; Benton, Hartman, & Sarason, 1955; Westrope, 1953).

Therefore, if each of the forms of fantasy materials is to

serve as a sensitive index to group differences, the follow-

ing relationship is expected to be observed:

Ho : Structural analysis scores vary in relationship
to subjects' anxiety levels on each of the
four forms of fantasy productions.




Twenty-four male and 24 female volunteer subjects

were obtained from a population of undergraduate psychology

students at the University of Florida. With regard to

age, college students were expected to be a relatively

homogeneous group. However, any subject not within the age

range of 17-25 was omitted from the study. With regard to

sleeping conditions, college students were also expected

to be a relatively homogeneous group. That is, under-

graduate college students typically reside under similar

living conditions (i.e. dorms or shared apartments).


Three raters were employed in this study. Two were

undergraduate honor students in psychology and one was the

principal experimenter. Both undergraduate raters were in

their junior year of college and had had some previous

experience as assistants in research. The experimenter

had a master's degree in clinical psychology at the time

this research was conducted.


Six judges were selected for use in this study. The

judges represented a relatively heterogeneous group, varying

along several dimensions: age, sex, clinical experience,

academic credentials, and familiarity with fantasy materials.

Judge 1 was a female professor of philosophy with

several years' training in Jungian psychology. Judge 2 was

a female master's degree level English teacher who at the

time of this study was self-employed as a potter. Judge 3

was a male graduate student in experimental psychology who

was quite sophisticated in quantitative sleep research.

Judge 4 was a male graduate research professor in experimen-

tal psychology with a well-known reputation in the area of

experimental sleep research. Judge 5 was a female master's

degree level graduate student in clinical psychology with an

avid interest in the use of projective techniques. And

finally, Judge 6 was a female professor in clinical psychology

who excelled both as a clinician and as a sensitive inter-

preter of projective materials.


A measure of reported trait anxiety was obtained from

each subject in response to Spielberger, Gorsuch, and

Lushene's (1970) Self-Evaluation Questionnaire, STAI Form

X-2, (see Appendix A). This scale consists of 20 statements

requesting persons to describe how they "generally" feel.

Specifically, subjects were asked to report the frequency

with which they experienced a symptom of trait anxiety.

Spielberger et al. (1970) have reported a test-retest

reliability coefficient on the STAI A-Trait Scale of r = .86

for males and r = .76 for females over a period of 20 days.

It has also been reported with respect to validity that the

correlations between the STAI A-Trait Scale, Cattell's IPAT

Anxiety Scale, and the Taylor Manifest Anxiety Scale are

moderately high for college students (between r = .80 and

r = .85). Given the evidence for strong convergent validity

of the trait anxiety scale, it was determined that this single

measure of trait anxiety was adequate for rating a subject's

general level of anxiety.


Data Collection

The STAI A-Trait Scale was administered to all subjects

during the first day of data collection. Subsequently,

subjects' test scores were rank ordered and subjects were

divided into three equal groups (high, medium, and low

anxiety). The STAI A-Trait Scale was re-administered in the

final stage of the study. A test-retest procedure was

employed to assure the stability of subjects' membership in

one of three previously designated anxiety groups. This

procedure, allowing 30 days between initial and subsequent

testing, revealed reliability coefficients of r = .85 for

both males and females. This was highly consistent with

the previously reported findings of Spielberger et al.

(1970) and lent further support to the reliability of

instrumentation employed in this study.

Dreams, free fantasies, and structured fantasies were

collected from each subject. To eliminate possible "sequenc-

ing" effects of data collection, data were collected in a

randomized Latin-Square fashion. That is, data were

gathered such that each of the four fantasy productions

collected occurred once in the first, second, third, and

fourth position of data collection. This design provided

for four different sequence possibilities of four fantasy

productions each. For example, if

A = Home dreams
B = Five minute "written samples"
C = Free fantasies
D = TAT fantasies,

then data were collected in the following order:


A group of 12 randomly selected subjects was assigned to

each of the above four sequences.

Dreams were collected by use of a morning-after tech-

nique. Subjects were asked to report each morning upon

awakening their recalled dreams on a prepared questionnaire

form (see Appendix B). Dreams were collected in this manner

for 14 consecutive days from each subject.

Spontaneous fantasies were collected by two methods.

One method, employing the basic format described by Kripke

and Sonnenschein (1973), involved collecting fantasies in

the dream laboratory. Subjects were asked to lie down and

relax in a dimly lit, quiet room. At randomly selected

intervals, ranging from 3 to 7 minutes, a buzzer sounded, at

which time subjects were asked to write on an index card

"what was going through their mind" just prior to the experi-

mental interruption. A sufficient time interval occurred

between each experimental interruption to allow subjects to

record their responses. Five samples of waking free fanta-

sies were collected in this manner over approximately a 30-

35 minute period. Oral instructions, which were delivered

to subjects in this phase of the study are presented in

Appendix C. Since people have been reported to produce more

fantasy-like materials during early morning hours when their

body temperature is lowest, periods of data collection

occurred at randomly dispersed times throughout the morning


Another form of spontaneous fantasies, although slightly

more structured, was collected by a second method, previously

described by Gottschalk and Gleser (1969). Subjects were

asked to produce three 5-minute "written samples" of a mean-

ingful life experience. These samples were collected in the

sleep laboratory (Appendix D).

A single investigator, using the aforementioned tech-

niques in small group settings, collected the two forms of

spontaneous fantasies. Both forms of spontaneous fantasies

were employed in order to obtain an adequate and comprehen-

sive sample of internally-evoked, spontaneous fantasy menta-


Highly structured fantasies were obtained from each

subject in response to 10 Thematic Apperception Test cards:

numbers 1, 2, 4, 6BM, 6GF, 7BM, 9GF, 12M, 13MF, and 16 (the

blank card). These particular TAT cards were chosen based

on their capacity to elicit a wide range of responses.*

Responses to all cards were obtained during a single session

with a single examiner overseeing the testing session.

Standard instructions (Murray, 1943) were given verbally and

were also written at the top of each response sheet. To

secure somewhat uniform and comparable samples, subjects

were given only one sheet of paper per TAT card, and were

asked to compose a complete story related to each test

stimuli (see Appendix E). Subjects were tested collectively

in small groups of four.

All dreams and fantasy productions, both spontaneous and

structured, were obtained in a written rather than a verbal

mode. Although fantasy material has been more commonly

reported in a verbal mode, it was believed by the investi-

gator that "consistency in reporting" was a crucial factor

in this study. Gottschalk and Gelser (1969) have warned that

researchers' inadequate understanding of how verbal and

*M. Kramer, personal communication (1976).

written methods of data collection affect data outcome,

strongly warrants employing a single method of data collec-

tion within a single study. Since home dreams were, by

necessity, written, the written method was chosen for this



Dream reports (varying numbers), 3 written samples

of life experiences, 5 free fantasies, and 10 Thematic

Apperception Test reports were available from each of the

48 subjects.

The first and last dream report, the first and last

written sample life experience report, and five free fantasies,

and the responses to TAT cards 1 and 16 were typed on 3 x 5

index cards. These reports were transferred verbatim with

minor exceptions onto index cards. On the recommendation of

Cartwright, highly unique stylistic components were edited

from the final prepared protocols. All cues indicating

subjects' identity and sex were also omitted from each pro-


Sorting Procedure

A judges' sorting or cataloguing procedure was

employed in evaluating a majority of the collected data.

The traditional content analysis approaches for detec-

ing individual and group consistencies across fantasy

*R. Cartwright, personal communication (1976).

materials inevitably lent bias toward a particular fantasy

material. Therefore, this approach was rejected and an

alternative procedure was chosen to avoid artificially

assigning aspects of written fantasy productions into pre-

determined categories. Such categories were commonly

designed for analyses of particular fantasy materials

(i.e. TAT classifications) and significantly biased end

results when applied to other forms of fantasy productions.

The alternative procedure chosen eliminated the need for

pre-determined criterion for evaluating subject consistency.

That is, judges were merely requested to sort or catalogue

protocols which appeared to originate from a single subject,

in a single unit. This method, in no way, was more or less

applicable to a particular form of fantasy material employed.

Fantasy protocols evaluated and sorted by judges were

prepared as previously described under "protocols." The

retained and edited dreams from each subject were stapled

together and treated as a single unit to be evaluated by

judges. An identical procedure was followed for each form

of fantasy samples (i.e. free fantasies, written samples,

and TAT's). Judges were requested to determine the four pro-

tocol units which belonged to a single subject (one protocol

unit from dreams, free fantasies, written samples, and TAT's

each) from a subgroup of four subjects' protocols. Sub-

group membership was structured such that sex of the four

subjects was held constant, irrespective of anxiety level.

Judges were asked to make the above determination,on a second

occasion, with four different subjects' protocols. On this

occasion, subgroup membership required that anxiety level of

the four subjects be held constant while the sex of sub-

jects varied. Instructions to the judges are provided in

Appendix F, under "task i."

Judges were also required to analyze all subjects'

protocols according to two pre-determined criteria. That is,

judges were requested to subjectively sort dreams, free

fantasies, written samples, and TAT's respectively into

three groups according to subjects' anxiety level (low,

medium, or high). Judges similarly carried out a second sort-

ing procedure with regard to subjects' sex membership (male

or female). Refer to Appendix F, "tasks 2 and 3" for specific

instructions to judges. Protocols from all 48 subjects were

sorted by judges in this phase of the study.

Rating Procedures

The raters' task was to provide simple word counts for

each subject's fantasy protocols. These word counts served

as an observable and quantifiable index of a structural

property of written fantasies. Raters met for a total of

five hours to discuss a reliable system for counting words.

A general guideline for obtaining consistent word counts

for dreams was provided (see Appendix G) at these sessions,

and several "practice" protocols were examined conjointly


by the three raters. The general guidelines utilized were

abstracted from a scoring system previously devised by

Foulkes and Shepherd (1970).


The analyses related to each hypothesis are specified in

detail by the hypotheses in the Results chapter.



Hol: Written productions arising from four different
forms of fantasy mentation (dreams, free fantasies,
written samples, and TAT's) of an individual
have sufficient coherence to be recognizable as
originating from the same individual.

Six judges were initially provided with 16 fantasy

protocols, presented in stacks of four dreams, four free

fantasies, four written samples, and four TAT's. Judges

were asked to sort these protocols into four complete sets

according to subjects. Hence, each set was to contain a

protocol from each of the four fantasy materials believed to

belong to a single subject. This was repeated for a second

set of four subjects. Overall, each judge arrived at eight

sets of protocols believed to belong to eight different


The results of judges' blind sortings were examined

with respect to each individual judge's ability. First,

each judge's performance was evaluated in terms of the

number of correct matches obtained on each of the eight

sets sorted. The number of matches obtained was defined in

the following manner: "No match" was defined as all four

protocols within a single set belonging to four different

subjects, "one match" as two protocols within a set belong-

ing to a single subject, "two matches" as three protocols

within a set belonging to a single subject, and "three

matches" as all four protocols within a set belonging to a

single subject.

Based on the above criteria, judges' performance in

deriving sets of protocols containing zero to three pos-

sible matches was evaluated according to combinatorial

probabilities. It was determined that for each set of pro-

tocols, with zero to three correct matches, the following

probabilities and resultant levels of statistical signifi-

cance applied:

Zero matches = 6/64 = .0937
One match = 45/64 = .7000
Two matches = 12/64 = .1875
Three matches = 1/64 = .0156.

Hence, only in those cases when all four levels of fantasy

material were correctly matched did a judge perform at a

greater than chance level. There were two judges (Judges 1

and 3) capable of performing this task at a greater than

chance level (p = .0156) on at least one set of protocols.

In fact, Judge 3 performed this task at a statistical level

of significance on five sets while Judge 1 demonstrated a

similarly high level of performance on two sets (Table 1).

In addition to examining the number of significant

matches obtained by each judge on each of the finally

derived sets, judges' performance was evaluated according

to an overall measure of ability. Summing the number of


N CN4 m CN N 0 -

r- o C M 0o

N i-q i-q N CN r-q

1-4 N N 0N

- -l4 N -4 N -








-l 4




i-l ^A re n

4a) -

0 0)




matches obtained on all eight sets for each judge provided

an adequate and overall measure of each judge's ability.

Each judge's score is recorded in Table 2 under the column

entitled "sum of matches." The sum of matches theoretically

ranged from 0 to 24 matches. Cummulative probabilities

were determined for each number of possible matches (refer

to Table 3) and significance levels calculated. The sig-

nificance level for each obtained sum of matches can also

be observed in Table 2. These figures indicate wide differ-

ences in individual judges. Judge 3 obtained 19 matches at

a highly significant level (p = .00002), followed by

Judge 1 who obtained 15 matches at the .004 level, and

Judge 5 who obtained 14 matches at the .013 level. Judges

2, 4, and 6, however, showed little evidence of appropriate


The observation that 50% of the judges' overall level

of high performance occurred at a greater than chance level

seemed impressive, but inconclusive. To determine how

likely this outcome would have occurred by chance alone,

the judges' performance as a group was examined according

to previously determined probabilities (see Table 3). Using

the formula p = [1 (.999979) ], given a total of six

judges, it was observed that the probability of one judge

out of six obtaining 19 matches or more by chance alone was

highly rare (p = .00126). In this study, not only did one

judge actually obtain 19 matches (Judge 3), but two

4ic N -K
* 0 4*

o o 0 CTI 1-1 o
0 0 0 C 0 CO
O co O r O co

CN i in I I

Mr (N r- (n (N

I (N

-H N

Hl N m L o









-4 Cn
H Ci o

* C
r00 0


I >1
LZ -4
Z oeo OH

* *


V) 1

I qa

Table 3

Cumulative Probability

Total Matches Probability Probability
Equal: Equals: Equals:



additional judges (Judges 1 and 5) succeeded in obtaining

15 and 14 matches at a greater than chance level (p =

.00432 and p = .01319, respectively). Judges were clearly

engaging in non-random sorting.

The results for Hypothesis 1 suggested that subjects

consistently revealed themselves across various forms of

fantasy material such that one-half of the judges (three

out of six) discriminated one individual from another at a

highly significant level. That fantasies from each of the

four fantasy levels of one subject could be successfully

distinguished by judges from those of another subject

attested to the presence of some form of coherence across

levels. Dreams, free fantasies, written samples, and TAT's

all appeared to be orderly, non-random, meaningful events.

In conclusion, these results are compatible with Hypothesis

1 and lend adequate support for its acceptance.

Ho2: Written productions arising from four differ-
ent forms of fantasy mentation can be accurately
and reliably sorted by judges into either of
two discrete domains (male or female) accord-
ing to subject's stated sex.

The six judges were required to subjectively sort all

dreams, free fantasies, written samples, and TAT's,respec-

tively, into two categories according to subjects' sex

membership (male or female). For each of the four fantasy

productions, both the most accurate (valid) and the most

reliable among a set of six judges were determined according

to their ability to sort subjects by sex. Judges' ability

was evaluated by use of a computer program specifically

designed to assess rater agreement and rater bias for

qualitative (nominal and cardinal) data (Chicchetti, Aivano,

and Vitale, in press).

Judges were evaluated for accuracy by comparing their

performance on each of the four sorting tasks (one sort per

fantasy production) to a criterion or "expert judge."

Judge 7 represented the correct response or criterion in

all cases throughout this study. In Table 4 judges are

listed in rank order according to ability to accurately

classify 48 subjects by sex on each of the four fantasy pro-

ductions. Note by this method, an observed agreement

(expressed as a percentage) between a judge and the expert

judge of at least 67% yielded a p value of .02 or greater.

With respect to dreams, three out of six judges (Judges

3, 4, and 5) were able to determine subjects' sex at a

greater than chance level (p < .02). On free fantasies,

four out of six judges (Judges 2, 3, 4, and 5) engaged in

non-random sorting- Similarly, on TAT's and written saimles,

five out of six judges (all but Judge 4) were successful in

determining subjects' sex at a greater than chance level.

In Table 5 these results are summarized for each judge

per fantasy level. Judges' sorting of sexual identity on

written fantasy productions was significantly related to the

actual sex of subjects by 50% of the judges on dreams, 67%

of the judges on free fantasies, 83% of the judges on written

Ranking of Judges

Table 4

According to Ability to Accurately
Subjects by Sex on Each of
the Fantasy Levels

Judge Observed Expected Z of p
Pair Agreement Agreement Kappa Value








Free Fantasies







Written Samples





































































Table 4--(continued)

Judge Observed Expected Z of p
Pair Agreement Agreement Kappa Value

6,7 .67 .50 2.30940 0.02092*

4,7 .63 .50 1.73205 0.08326


3,7 .79 .50 4.04145 0.00005*

2,7 .71 .50 2.88675 0.00389*

5,7 .71 .50 2.88575 0.00389*

6,7 .71 .50 2.88775 0.00389*

1,7 .67 .50 2.30940 0.02092*

4,7 .58 .50 1.15470 0.24821

*p = < .02.

Table 5

Summary or Judges' Accuracy in Cataloguing Subjects
by Sex per Fantasy Level (Percentage Agreement)

Fantasy Level 1 2 3 4 5 6

Dreams 58 54 67* 71* 75* 63

Free fantasies 63 71* 83* 71* 71* 63

Written samples 75* 67* 79* 63 67* 67*

TAT's 67* 71* 79* 58 71* 71*

*p < .02.

samples and TAT's. Also, two judges (Judges 3 and 5) signif-

icantly differentiated sex with all four levels of fantasy.

Hence, there is substantial evidence to suggest that written

productions arising from dreams, free fantasies, written

samples, and TAT's each can be validly sorted by judges

according to subjects' sex. This finding lends strong

support to the first half of Hypothesis 2.

Judges' ability was also examined with respect to

reliability. Based on a total of seven judges (six plus an

"expert judge"), a program from Cicchetti's statistical

package was used to compare the relative levels of all

possible 21 interjudge agreement pairings. This was

accomplished utilizing a system presented and described fully

by Aivano, Cicchetti, and Levine (in press), which compared

each judge, relative to all others, using ranking systems

based on levels of chance-corrected agreement. Chance-

corrected agreement was measured by z of kappa values.

First, ranks for the 21 comparisons were summed to obtain

a composite rank which reflected each judge's reliability,

relative to the remaining judges. By comparing the com-

posite rank scores of judges in Table 6, those judges with

the lowest scores were identified as the most reliable and

those with the highest scores as the least reliable. For

example, on the dream sort, Judge 5 was the most consistent

with other judges and Judge 2 the least consistent. In

comparison with data presented in Table 4, it was noted

that Judge 5 was also the most accurate and Judge 2 also

Table 6

Sex Sort: Ranking of Judges by Composite Scores
for Each Fantasy Level


Composite Rank Sum


Free Fantasies



Written Samples


3 41
6 61
1 65
2 65
5 73
4 111

the least accurate. On free fantasies, Judge 3 was the

most consistent as well as the most accurate and Judge 6 the

least consistent and the least accurate. Again, on

written samples, Judge 3 was the most consistent and the

most accurate, while Judge 6 was the least consistent and

next to least accurate. And finally, on TAT's,Judge 3 was

the most consistent and the most accurate and Judge 4 the

least consistent and the least accurate. On the above

sorting task, judges' reliability and validity tended to be

strongly associated. However, this tendency was only

documented for judges at the extremes of the continuum.

A second method of evaluating judges' reliability took

into consideration not only the "relative" position of

each judge, but also the "absolute" level of z values for

each judge. That is, judges were ranked according to the

number of significant (p < .05) z of kappa values among the

21 comparisons associated with judges. The results of this

ranking system are presented in Table 7. As can be seen

from comparing Tables 6 and 7, the rank orderings of judges

produced by the two ranking systems are quite similar

although not identical. By omitting those judges with no

significant z values and including only those judges with

at least two significant z values, this second method was

useful in determining those judges who demonstrated a high

degree of inter-judge reliability. According to this

criterion, five out of six judges (83%) could be considered

Table 7

Sex Sort:
Ranking of Judges by the Number of Significant Z
Values for Each of the Fantasy Levels


p < .05 Level




Free Fantasies



Written Samples

Total: 31

Table 7- (continued)

p < .05 Level


Total: 15


Note: "Total" represents a total of the number of times
each judge was significantly paired with another judge.

reliable when sorting dreams, four out of six (67%) when

sorting free fantasies, six out of six (100%) when sorting

written samples, and five out of six (83%) when sorting

TAT's. These findings lend substantial support to the

hypothesis that judges can reliably sort subjects accord-

ing to sex on each of the four fantasy productions.

The results presented in Table 7 also suggested that

the various fantasy materials were differentially sorted

as to sex by judges as a group. Clearly, written samples

were more reliably sorted by judges than any other fantasy

material. While sorting written samples, there were 31

occasions when judges were significantly (p < .05) paired

with other judges. This was in marked contrast to 15

occasions on TAT's, 13 occasions on dreams, and 10 occa-

sions on free fantasies.

In summary, the above reported findings with respect

to accuracy and reliability support Hypothesis 2 and

Hypothesis 2 is accepted.

Ho3: Written productions arising from four different
forms of fantasy mentation can be accurately
and reliably sorted by judges into one of three
discrete domains (low, medium, or high anxiety)
according to subjects' obtained anxiety scores.

The six judges were again required to subjectively

sort dreams, free fantasies, written samples, and TAT's.

However, on this occasion, the criterion variable for

sorting was anxiety rather than sex. This hypothesis was

examined by use of a computer program identical to that

used in testing Hypothesis 2. However, in testing Hypothesis

3, a "weighted" kappa statistic rather than an unweightedd"

kappa was the statistic of choice. This decision was based

on the fact that judges were asked to sort anxiety into one

of three categories (low, medium, or high),ordering anxiety,

in this case, along a continuous-ordinal scale. Cicchetti

(1976) and others (Cohen, 1968; Fleiss, Cohen, & Everitt,

1969) have described the rationale for using weighted kappa

with continuous-ordinal data rather than more conventional

approaches such as Kendall's tau or Pearson's product-

moment correlation. Unlike these conventional approaches,

weighted kappa measures the degree to which the paired

values of two variables are "identical" (not just "propor-

tional") and allows credit for partial rater agreement, etc.

Thus, in this study, a linear rater agreement "weighted"

system developed by Cicchetti (1976) was used to assess

rater agreement in sorting anxiety. This formula by

Cicchetti determined the weights for complete rater agree-

ment (always equal to 1), partial rater agreement (in this

case, equal to 0.5), and total rater disagreement (always

equal to 0). Specifically, on the three category "anxiety"

scale (low, medium, and high), judges' confusion between

both low and medium and medium and high received a weight

of 0.5, whereas judges' confusion between low and high

received a weight of 0 since this latter case represented

the maximal possible confusion. In summary, this system

took into consideration the observation that judge confusion

between low and high categorization was more clinically

serious than between low and medium categorization or be-

tween high and medium categorization.

As in Hypothesis 2, judges were evaluated both with

respect to accuracy and reliability. In Table 8, judges

are listed according to ability to accurately classify sub-

jects by anxiety on each of the four fantasy productions.

On all occasions of statistical significance p s .02.

With respect to dreams, four out of six judges (Judges

1, 4, 5, and 6) were able to determine subjects' anxiety at

a greater than chance level. On free fantasies, three out

of six judges (Judges 1, 5, and 6) were able to determine

subjects' anxiety at a statistical level of significance.

On written samples and TAT's, only two out of six judges

(Judges 3 and 5) were capable of determining subjects'

anxiety at a greater than chance level.

In Table 9 these results are summarized for each judge

per fantasy level. Judges' sorting of anxiety on written

productions was significantly related to the obtained anxiety

rating of subjects by 67% of the judges on dreams, 50% on

free fantasies, and 33% on written samples and TAT's. One

judge (Judge 5) significantly differentiated anxiety with

all four forms of fantasy. Furthermore, the observation

that subjects' anxiety was sorted by at least two judges

at a statistically significant level on each of the four

Table 8

Ranking of Judges According to Ability to Accurately
Classify Subjects by Anxiety on Each
of the Fantasy Levels

Judge Observed Expected Z of
Pair Agreement Agreement Kappa p value








Free Fantasies







Written Samples





































































Table 8--(continued)

Judge Observed Expected Z of
Pair Agreement Agreement Kappa p value

4,7 .60 .56 0.95852 0.33780

2,7 .58 .56 0.54773 0.58388


3,7 .69 .56 2.60169 0.00928*

5,7 .67 .56 2.19090 0.02846*

6,7 .65 .56 1.78010 0.07506

4,7 .63 .56 1.36931 0.17090

1,7 .60 .56 0.95852 0.33780

2,7 .60 .56 0.95852 0.33780

S .02.

Table 9

Summary of Judges' Accuracy in Cataloguing Subjects
by Anxiety per Fantasy Level
(Percentage Agreement)

Fantasy Level 1 2 3 4 5 6

Dreams 67* 65 58 67* 69* 71*

Free fantasies 73* 56 58 63 69* 67*

Written samples 63 58 69* 60 67* 63

TAT's 60 60 69* 63 67* 65

*p < .02.

fantasy productions lends adequate support for acceptance

of the first half of Hypothesis 3.

Judges reliability in sorting subjects according to

anxiety was determined by two methods previously described

with regard to Hypothesis 2. In brief, judges were

initially ranked according to their composite rank sums

(refer to Table 10). The results of these rankings showed,

for example, that Judge 1 was the most consistent with

other judges and Judge 3 the least consistent in sorting

dreams. Comparing these results with data presented in

Table 8, it was observed that Judge 1, although not the most

accurate, was one of the more accurate judges, and Judge 3

was the least accurate. A similar trend was noted in comparing

results from Table 10 with those from Table 8 for each of

the fantasy materials. For example, in sorting free fan-

tasies, Judge 1 was both the most consistent and the most

accurate and Judge 2 both the least consistent and least

accurate. In sorting written samples, the most consistent

judge (Judge 6) was not one of the more accurate judges,

whereas the least consistent judge (Judge 2) was also the

least accurate. And finally, on TAT's, a similar rela-

tionship was observed. The most consistent judge (Judge 1)

was not one of the more accurate judges, whereas the least

accurate judge (again Judge 2) was also the least consistent.

With respect to sorting anxiety, judges' reliability

and validity were strongly related for judges with lower

Table 10
Anxiety Sort: Ranking of Judges by Composite Score
for Each of the Fantasy Levels

Composite Rank Sum


Free Fantasies

Written Samples





rankings. It was observed that with judges of higher

reliability rankings there was a tendency (although not

complete correspondence) for a similar relationship to


Judges' reliability in sorting subjects according to

anxiety was additionally assessed by ranking judges accord-

ing to the number of significant (p L .05) z of kappa

values. The results of this ranking system are presented in

Table 11. In comparing results from Table 10 with Table 11,

it was observed that the rank orderings of judges produced

by the two systems were quite similar. By use of this

second system, a subset of reliable judges who demonstrated

reliability at a statistically significant level was derived.

According to the same criterion used in choosing reliable

judges in the sex sort, six out of six (100%) judges were

observed to be reliable when sorting dreams according to

anxiety, five out of six (83%) when sorting free fantasies

and written samples, and six out of six (100%) when sort-

ing TAT's. These findings lent substantial support to the

hypothesis that judges can reliably sort subjects accord-

ing to anxiety on each of the four fantasy productions.

The results presented in Table 11 also suggest that

the various fantasy materials were differentially sorted

as to anxiety by judges as a whole. Clearly, written

samples were least reliably sorted by judges compared to

other fantasy material. Recall that written samples were


Table 11

Anxiety Sort: Ranking of Judges by the Number of
Significant z Values for Each of
the Fantasy Levels

Judge p < .05 Level


1 6
6 6
2 5
4 5
5 5
3 3

Total: 30

Free Fantasies

1 5
6 5
5 5
3 4
4 4
2 0

Total: 23

Written Samples

5 3
6 3
1 2
3 1
4 1
2 0
Total: 10

Table 11- (continued)

p < .05 Level


Total: 26


Note: "Total" represents a total of the number of
times each judge was significantly paired with another

~ I~

more reliably sorted than other fantasy material when

judges sorted subjects according to sex. All other fantasy

materials were similarly sorted as to anxiety with 30

significant (p < .05) judges' pairings obtained on dreams,

23 on. free-fantasies and 26.on TAT's.

In summary, the above reported findings with respect

to accuracy and reliability support Hypothesis 3 and

Hypothesis 3 is accepted.

Ho4: Judges' ability to successfully sort written
fantasy productions according to sex is similar
across the four forms of fantasy materials.

Hypothesis 4 was evaluated from both a "clinical" and a

"research" perspective.

By use of the clinical or individual judge approach,

each judge's sorting ability on one fantasy level was com-

pared relative to his own ability on another fantasy level.

Based on judges' z of kappa values (refer to Table 4),

Spearman rank correlation coefficients were computed com-

paring judges' consistency in accurately sorting sex across

the four fantasy levels. The results are presented in

section one of Table 12 under "r for accuracy." The

absence of significant correlations indicated that judges

were unable to consistently sort subjects according to

actual sex membership across the various fantasy levels.

That is, by use of the clinical approach, judges were found

to display a differential ability in utilizing dreams, free

fantasies, written samples, and TAT's when sorting subjects






4-M (

(0 (4

2 U) m

>E >
i en m V

X 0U0

vi 000

(n0 1 r- 4 m o o




I l l


0) >

e > E-4
a ) Ai
cn 3 )

> ) E-i

n c,
M 4-1 (U~

) Uen
4- E-1
4- 9 E -

n en


-^*H a,
4 4) n0

iW 24

0) a) -4
s:i 44 :




3 E-


01 U3


-i >i
e en

0 0)





m )

a) -


L41 a






by sex. Their sorting ability was dependent upon the

fantasy material being sorted. The results displayed

a judge/fantasy material interaction and provided non-

substantial support for Hypothesis 4.

However, when judges' ability to utilize each of the

fantasy levels was viewed as a whole, utilizing the tradi-

tional "group" research approach, very different results

emerged. Based on Table 5, a complex chi square test was

employed to compare judges' scores across fantasy levels.

No significant differences were obtained when comparing

judges' ability as a group to accurately perceive subjects'

sex on any of the fantasy levels (X2 = 8.45 with X2

In contrast to the clinical or individual judge ap-

proach the research or group approach,indicated that

judges as a unit were highly consistent in sorting ability,

irrespective of fantasy material. While the above reported

findings are contradictory and vary with the assessment

approach employed, there is evidence that consistency does

exist across levels. Hypothesis 4 is therefore accepted.

Ho5: Judges' ability to successfully sort written
fantasy productions according to anxiety level
is similar across the four forms of fantasy

As in Hypothesis 4, Hypothesis 5 was evaluated from

both a clinical and a research perspective. Spearman rank

correlation coefficients were computed, based on previously

reported z of kappa values (refer to Table 8). The results

of this analysis are presented in the second section of

Table 12 under "rs for accuracy." The absence of signifi-

cant correlations indicated that judges were unable to

consistently sort subjects according to actual anxiety

level across the various fantasy levels. That is, given

a clinical approach in which each judge was evaluated for

internal consistency, it was found that judges displayed a

differential ability in utilizing dreams, free fantasies,

written samples, and TAT's when sorting subjects by anxiety.

Thus, as when sorting by sex, judges were unable to

reliably sort subjects by anxiety across the various fantasy

levels. An observed judge/fantasy material interaction was

non-supportive of Hypothesis 5.

However, different results were again obtained when a

research "group" analysis was conducted. Based on Table 9,

a complex chi square test was employed to compare

judges' scores across fantasy levels. No significant differ-

ences were observed between judges' ability as a "group" to

accurately perceive subjects' anxiety on any of the four

fantasy levels (X2 = 3.86, with X2 = 25).
Not unlike the sex sort results, the anxiety sort

results were contradictory, based upon the approach employed.

The clinical approach suggested differential judges' ability

while the research approach suggested consistent judges'

ability across fantasy levels. Given evidence that judge

consistency is possible, Hypothesis 5 is accepted.

Ho6: Subjects reveal their sexual identity consis-
tently to judges across each of the four fantasy
productions. A similar finding is expected to
be observed with respect to subjects' anxiety

In evaluating the above hypothesis, "subject consis-

tency" was defined in terms of "the percentage of judges

who accurately perceived each subject's classification"

on each fantasy level. Judge's accuracy on each subject was

obtained from information provided by the computer program

designed by Cicchetti et al. (in press). This program,

which was previously described, essentially provided des-

criptive data regarding each judge's classification of each

subject. By summing the number of judges who correctly

classified each subject and dividing by the number of total

judges, a percentage of accurate agreement was obtained for

each subject.

By use of the above criterion and in order to obtain

a measure of subject consistency, subjects were assigned

ranks (1 through 48) on each of the fantasy levels and

Kendall's rank order correlation coefficients were computed.

The results of this analysis are presented in the right-hand

column of Table 13 for both sex and anxiety sortings. With

respect to sex, subject consistency was observed on free

fantasies compared to written samples at a greater than

chance level (p = .04). However, this relationship was

C 0 0 (N r-4 r-


S0 I I 0












a) 0

-4 < E-l



a) a) -
c OH3

a4 a)Q 4

r4 0 R
0 o-H
Q) )-


o I co 0
O* **
I I I e

S (N



14 01


a) -r


31 3


0 03




E-4a ()


O (1U

a) 4-)
3 -H

e4 01
W a







4-4 Q(


) -H
fa 3

not a particularly strong one (T = .20). On the remaining

five comparisons of fantasy material, subjects did not

reveal themselves consistently to judges. Similarly, with

respect to anxiety, subject consistency was observed on

only one out of six comparisons. That is, subjects were

consistent across dreams and free fantasies at a greater

than chance level (p = .05). However, this relationship,

while significant, was not impressively strong (T = .19).

These findings suggested that with rare exception a differ-

ent subset of subjects consistently revealed themselves

(regardless of characteristic being observed by judges) on

one level of fantasy mentation as compared to another level.

Indeed, a number of negative correlations were observed,

suggesting that a given subset of subjects blatantly

revealing themselves to judges on one fantasy level might

similarly be particularly unrevealing to judges on another

level. These findings are incompatible with the hypothe-

sized relationship and Hypothesis 6 is rejected.

In viewing these results, one might question whether

the subjects, relative to each other, were indeed incon-

sistently expressing themselves across levels or whether

judges' ability to perceive subject characteristics was

highly inconsistent across levels.

A re-analysis of the data from a group "research"

perspective was found to be instrumental in clarifying this

question. The judges' scores on accurately perceiving

subjects (obtained in the first part of Hypothesis 6) for

dreams, free fantasies, written samples, and TAT's, respec-

tively, were summed and averaged. These mean percentage

scores of judges' overall ability as a group are presented

in Table 14. An analysis of variance, by use of a ran-

domized block design, was employed to compare the obtained

means. On both the sex and anxiety sort, no significant

differences were obtained between judges' ability as a whole

to accurately perceive subjects on any of the four fantasy


These latter findings, together with those initially

stated in Hypothesis 6 indicated that while judges as a

group were highly consistent in ability on each of the

fantasy levels, they were not necessarily consistent on

the same subjects on each level. Therefore, it was

reasonable to conclude that subjects were indeed inconsis-

tently revealing themselves across levels with respect to

specified sorting criteria.

Ho7: Written productions from each of the four fantasy
materials are correlated at a greater than chance
level according to structural analysis scores.

A simple word count, representing subjects' produc-

tivity on each of the fantasy materials, served as the

structural measure to be examined on the fantasy productions.

This was only one of many possible structural properties

which could have been examined in this study. Given the

sheer volume of fantasy material and lacking a reliable














i t7O





m a

U) ) L 2
m S

la a,

4J 4 0

Q h 3 &-


*H >
4-4 (I)



*M 1

0 0

S 4 a,
(0 (0

(0 ) a, E--

aC f]- 3 E-4

co n m 0

4) m co Ln m N
4 CO I 'T 1 CN


co r- co Ln L~-
Ln >D 0 r- N
CT) UD r- 0l d

m ( I< (


statistical program to evaluate more complex properties

(i.e., grammatical classes such as nouns, verbs, etc.),

word counts were considered sufficient and easily obtainable.

During preliminary stages of this research, several statis-

tical packages, including the General Inquirer by Stone,

Dunphy, Smith, and Ogilvie (1966), in conjunction with the

Harvard III Psychosociological Dictionary, were unsuccess-

fully employed. Major computer deficits were discovered in

the program and it was subsequently dropped under the advice

of Philip J. Stone, program originator at Harvard University.*

Word counts were obtained by three trained raters.

These counts were made according to a set of guidelines out-

lined in Appendix G. Each rater scored one-third of the

protocols. In addition, to ensure that all raters were

obtaining word counts in a uniform manner, all three raters

scored protocols from a common set of 12 subjects. A

discrepancy of less than five words was found on all but

three protocols. Each of the three discrepancies occurred

on dream protocols, with 14, 12, and 6 word count differ-

ences. Clearly, dreams proved most problematic for raters

in obtaining consistent word counts. This finding was

accounted for by the significant amount of non-dream or

extraneous material often included in the dream protocols

by subjects. In summary, raters appeared to be highly

*P. Stone, personal communication (1976).

consistent with each other in deriving word counts on the

various fantasy protocols.

Pearson product-moment correlation coefficients,

appropriate for interval-ratio data, were calculated in

comparing word counts for each of the forms of fantasy

material. The results of this analysis are presented in

Table 15. As anticipated, most (83%) of the comparisons

showed differing forms of fantasy materials to be signifi-

cantly correlated according to the previously determined

structural property. The only non-significant pairing

occurred between dreams and written samples. When the

fantasy materials were ordered along a continuum from highly

unstructured fantasies (dreams and free fantasies) to highly

structured fantasies (written samples and TAT's), it was

observed (see Table 15) that the strength of association,

in most cases, was related to how closely the fantasy

materials laid along this proposed continuum. That is, the

two most structured fantasies displayed the strongest and

most significant relationship (r001 = .54) while the two

most unstructured fantasies displayed the next strongest

and most significant relationship (r01 = .34).

In sum, the findings indicated a consistency in the

structural properties of subjects' fantasy expressions for

free fantasies compared with TAT's, written samples, and

dreams; TAT's compared with written samples, dreams, and

free fantasies; written samples compared with free fantasies

Table 15

Pearson Product-Moment Correlation Coefficients,
Comparing Each of the Four Fantasy Levels
According to Structural Analysis Scores

TAT's Samples Dreams

Free fantasies (.24) (.25) (.34)
.05 .04 .01

TAT's (.54) (.31)
.001 .02

Written samples (.05)


Note: Numbers in parentheses represent correlation
coefficients. Numbers beneath represent significance
levels for each cell.

and TAT's; and dreams compared with free fantasies and

TAT's. There appeared to be a stability to the ordering

of the structural property across various fantasy materials.

These results lend substantial support to Hypothesis 7

and its acceptance.

Hog: Structural analysis scores vary in relation-
ship to subjects' anxiety levels on each of the
four forms of fantasy productions.

A one-way analysis of variance test was employed to

determine if the three anxiety groups differed significantly

with respect to word counts on any of the fantasy materials.

Only on the TAT was a significant (p < .05) difference

obtained between groups. The mean structural analysis scores

for each of the anxiety levels on the TAT's were as follows:

Low anxiety group = 114.4
Medium anxiety group = 93.8
High anxiety group = 147.4

Applying the LSD multiple comparison test to determine which

groups differed significantly from each other, it was found

that all three groups differed from each other. Subjects in

the medium anxiety group had the lowest written output,

while subjects in the high anxiety group had the highest

written output.

In conclusion, structural analysis scores varied in

relationship to subjects' anxiety levels on only one of the

four forms of fantasy materials. Significant group differ-

ences were obtained on only the most structured fantasy

material (TAT's) gathered from subjects. Therefore,


insufficient support for Hypothesis 8 is found and

Hypothesis 8 is rejected.



This study attempted to establish by empirical examina-

tion the relevance of projective fantasy to various forms

of "free" fantasy. Historically, dreams have been regarded

as the royal road to the unconscious. However, whether

the dream report significantly differs from other verbal

fantasy reports has not been empirically established. No

previous study has simultaneously compared four forms of

fantasy reports, varying from highly structured to highly

unstructured samples. As a review of the literature

revealed, studies typically focused on comparisons between

dream reports and projective fantasy test results. Such

studies generally reported a greater similarity than differ-

ence between dreams and projectives (Klinger, 1971). A

weakness in the area of fantasy research has been the

neglect of free fantasies (more representative of ongoing

daydreaming) as they compare to dreams and projective


The "criterion for comparison" was a critical variable

in this and related research. In Chapter II, the rationale

for choosing a subjective criterion over an objective

content criterion was proposed. In short, while use of

judges' subjective cataloguing procedure had major limita-

tions (particularly with respect to generalizability), it

allowed for an answer to the preliminary questions:

"Are various forms of fantasy expression more similar than

dissimilar, and are they all sensitive discriminators of

individual differences?" Indeed, given the multitude of

possible content categories from which to choose, the

likelihood of choosing categories which would be equally

sensitive discriminators for all forms of fantasy expression

seemed improbable. Therefore, the subjective evaluation

of judges was chosen as the least biased criterion for

investigating these initial,yet critical, questions.

The decision to accept a subjective criterion received

support when the results of Hypotheses 1 and 6 were viewed

conjointly. The results of Hypothesis 1 suggested that

subjects displayed consistencies which were easily detected

by at least 50% of the judges, across four forms of fantasy


The results of Hypothesis 6 suggested that individual

subjects displayed themselves inconsistently across

fantasy forms with respect to previously specified

criteria (sex and anxiety). Therefore, it does not appear

advisable to "artificially" choose objective criteria if an

unbiased estimate of similarities across fantasy levels is

to be observed. That is, subject consistency may exist

across levels, but the categories chosen may not be sensi-

tive reflectors of such similarities. It can also be

seen from Hypothesis 6 that a major discrepancy in per-

ceiving subject consistency across fantasy levels may be

observed, depending upon the category being observed

(i.e. sex versus anxiety). According to Kramer, Roth,

and Palmer (1976), studies which have failed to discover

significant correlations between dreams and TAT's have all

utilized very limited categories for comparison. These

authors warned that an adequate study of the relationship

between various forms of fantasy productions must make an

adequate number of content and/or structural comparisons.

Perhaps the subjective, global approach, as used in this

study, offers the widest range of comparisons which is

possible without an exhaustive computerized content analysis

system. Additional studies employing a much larger sample

of judges should prove informative.

Hypothesis 1

The results of Hypothesis 1 are impressive. We know

that subject consistency does exist across various forms of

fantasy; we do not know on what basis these consistencies

exist, however. Future studies may find the latter

question a major challenge.

It might be argued that judges' ability to correctly

catalogue protocols into complete "sets" as predicted in

Hypothesis 1 may have been affected by several uncontrolled

factors. That is, judges' ability may have been affected

by unique characteristics of particular subjects employed

in this phase of the study. To determine whether there was

a significant difference in the total number of matches

obtained for each subject, Friedman's test (which is a chi

square approximation) was employed. This test was calcu-

lated based on rank orderings of the number of matches

obtained per subject across all judges. Friedman's test

was calculated separately for judges' tasks 1 and 2. The

results of this analysis for both tasks revealed no signifi-
cant difference between subjects (on task 1, X = 2.6,and
2 2
on task 2, X = 6.66 with X.05,3 = 7.82).
r .05,3
Given that chi square was approaching significance on

task 2, it was not surprising to note that the widest range

of discrepancies observed occurred between subjects

included in task 2. Recall, on task 2, that the anxiety

level of subjects was held constant while the sex of sub-

jects varied. On this task there was a tendency for judges to

obtain fewer matches on some subjects as opposed to

others. Five matches were obtained on subject 21 (male)

whereas 12 matches were obtained on subject 29 (female).

That the sex of subjects may be an influential factor in

judges' ability to recognize individual consistency across

fantasy levels was only slightly suggested and is in need

of further investigation. However, given the above results,

it must be concluded that judge's ability to successfully

catalogue various fantasy productions belonging to a single

subject was not significantly related to the sex or anxiety

of the particular subject being judged.

A second factor considered to possibly affect judges'

ability to correctly catalogue fantasy protocols into

complete sets was the stimulus materials employed. There-

fore, the relationship between the four fantasy productions

and judges' ability to match protocols of a single subject

across fantasy levels was explored. To examine a possible

fantasy level effect, the total number of tir3s a particular

fantasy material remained unmatched by a judge within a

"set" was determined. This was determined for all judges.

The resultant figures were rank ordered and Friedman's test

was calculated to assess whether any one fantasy production

was more difficult to integrate into a set than another.

Overall, dreams remained unmatched on 25 occasions, TAT's

on 19, written samples on 16, and free fantasies on 14.

However, the results of Friedman's test revealed no signifi-
cant difference (X = 4.85 with 3 d.f. at the 0.18 level

of significance) between fantasy productions,which remained

unmatched within sets, and a possible "fantasy level effect"

was rejected.

Given the limited number of judges and the limited

number of subjects utilized in this exploratory research, it

is proposed that an examination of observed trends (despite

an absence of statistical significance) would be revealing.

For example, it was previously noted in Hypothesis 1 that a

tendency existed for dreams to be least often integrated

into a set of four fantasy protocols. Recall that dreams

were originally derived from a "sleeping state of con-

sciousness" while free fantasies, written samples, and

TAT's were derived from a "waking state of consciousness."

Perhaps the presence of fantasies derived from both waking

and sleeping states complicated (although not significantly)

the judges' original task.

Judges' tasks may also have been complicated by the

variable of "setting." Winget and Kramer (in press) found

that the content of dream reports collected in various

settings (i.e. at home, in the laboratory, in the hospital,

etc.) varied depending upon the setting. Recall that.in this

study all fantasy reports were recorded and collected in

the same setting (the sleep laboratory) except for dreams.

Dreams were recorded at home upon awakening and were sub-

sequently collected at the end of two weeks at the sleep

laboratory. Perhaps the contrast in setting between dreams

and all other forms of fantasy was a factor which affected

the potential similarities to be discovered by judges.

Previous research studies suggesting such "contamination due

to setting" have been based on comparative studies

utilizing specified content categories. It is difficult

to assess the effects of setting upon fantasy reports when

subjective judgments are the comparison criterion.

There was a third and critical factor to be considered

in interpreting the results of Hypothesis 1. It could be

argued that judges were correctly distinguishing individual

fantasies based on recognizable differences in verbal style

rather than on consistent psychological differences in

individual fantasy productions. While this was possible,

it was unlikely. During the preparation of each protocol

for judges' cataloguing tasks, careful attention was given

to editing proper names, places, and other unique verbal

qualities. Furthermore, this experimenter did not readily

observe a wide variation in verbal styles of subjects which

would have easily served as clues to individual identities.

And finally, an earlier study by Kramer et al. (1976) empiri-

cally demonstrated that judges of varying abilities and

training were capable of distinguishing a single person's

dreams of one night from those of another night. Such

demonstration strongly argued against verbal style as the

sole distinguishing feature of a subject's individuality

on at least one form of fantasy mentation.

Hypotheses 2 and 3

In comparing the results from the sex sort (refer to

the first half of Hypothesis 2) with those obtained in the

anxiety sort (refer to first half of Hypothesis 3), it

remained to be explained why judges tended to experience

more difficulty accurately sorting anxiety than sex on at

least three of the four fantasy materials. It is proposed

that judges were required to make "finer" discrimination

when sorting subjects by anxiety (i.e. sort according to

three related categories) than when sorting subjects by

sex (i.e. sort according to two discrete categories). Hence,

a feasible explanation emerges with respect to the anxiety

sort. Perhaps a particular anxiety category (most likely

the middle category) was more difficult for judges to

accurately discriminate than others.

In order to determine whether judges had, in fact, found

it more difficult to accurately sort one of the anxiety

categories as opposed to another, a statistical procedure

for assessing "specific" category reliabilities was per-

formed (Cicchetti, Fontana, & Dowds, 1977). In this study,

each specific agreement index (expressed as a kappa value)

defined how much agreement there was for each category of

anxiety in terms of the average frequency with which judges

utilized each category. The results of this analysis are

presented in Table 16, comparing each judge with Judge 7

(the correct response) for each of the four fantasy produc-

tions. The kappa values range from 0 to 1, with a higher

value of kappa representing greater judge agreement. With

the exclusion of Judge 3 on dreams, free fantasies and TAT's,

all other judges experienced less judge agreement on the


3Cf ,









(U r-


S:= 5 =
zess ,qi

5 5C


,' I? ^


,_j- >


t Cf
i '

C '< is

i' j .
' j t

'DC~S 'C n

e'c- -SC-

-z= --fZ



* C,

'- '


~tJ*= Cr
n~N i

= = 6

3.:^ V- r3 3s -r ^^
-z= -E=

C 0 !MtN'C ^~ -^ f .

I- N *| -'
5- NNr -

5 5 5 -

3x= 3fC cC


',2B f
\z.Y ~



cu~ NOt OC~
'" N~ ~=- CI:


medium anxiety category, as compared to the low or high

anxiety categories. This was true regardless of fantasy

material being sorted. In short, on 88% (21 out of 24)

of the sorts, the level of judges' accuracy on the middle

category was less than obtained on either category at the

extreme. This seemed to suggest that judges' accuracy

might have increased had judges only been required to sort

subjects according to two anxiety categories (low and high)

rather than the required three. In conclusion, the

results, as shown in Table 16, indicated a judge/anxiety

level interaction which could account, in part, for judges'

increasing difficulty sorting subjects according to anxiety

versus sex classification.

As further evidence of judges' difficulty in making

discrimination, it was observed -that when sorting by sex,

judges tended to sort more accurately on structured fan-

tasies versus unstructured ones. An opposite trend was

observed on the anxiety sort.

Specifically, when sorting subjects by sex, the use of

written samples alone or TAT's alone provided sufficient

information for all but one out of six judges to successfully

sort subjects at at least the 0.02 level of significance

(Table 4). The same judge (Judge 4) was unsuccessful on

both of the above fantasy levels.

In contrast, when sorting subjects by anxiety, the use

of dreams alone provided a majority of the judges (4 out of

6) with sufficient information to successfully sort sub-

jects at at least the 0.02 level of significance (Table 8).

Additional unstructured fantasy material (free fantasies)

did not increase the number of judges capable of such

accurate sorting. One judge (Judge 2) was unable to suc-

cessfully sort subjects at a statistical level of signifi-

cance on all four forms of fantasy material, regardless of

degree of structure. Another judge (Judge 3) was able to

obtain successful sorts at a statistical level of signifi-

cance only on the more highly structured fantasy materials

(written samples and TAT's). Therefore, in situations

requiring complete accuracy of judges' evaluations of a

"latent" quality (i.e. anxiety), a combination of struc-

tured and unstructured fantasy protocols from each subject

would be necessary. If the experimenter were simply

interested in obtaining a significant number of accurate

evaluations from a large number of judges, a single, unstruc-

tured fantasy material (i.e. dreams) from each subject would

be sufficient.

In summary, when subjectively evaluating subjects with

respect to a quality of a more "manifest" nature (i.e. sex

membership), utilization of a single, structured fantasy

material (i.e. TAT) would be sufficient. However, in

evaluating subjects with respect to a more ambiguous

"latent" quality (i.e. anxiety), unstructured

fantasy material (i.e. dreams) would be more informative to

a greater number of judges.

This latter finding is not surprising when related to

past clinical knowledge. Clinicians (typically interested

in more latent or underlying aspects of the personality)

have known, based on experience, that individual differ-

ences are more readily observed in conjunction with an un-

structured, ambiguous situation or stimuli. Such stimuli,

like the Rorschach inkblots or dreams, allow more opportunity

for individual expression to be observed.

Hypotheses 4 and 5

Viewed from two different perspectives (the clinical and

the research), different conclusions for Hypotheses 4 and 5

emerged. Both perspectives were considered important to

report, particularly in a preliminary study such as this.

Given the research approach, caution in generalizing from

such a limited sample of judges is clearly warranted.

Interested in discovering which, if any, fantasy level is

more valuable as a clinical tool, both approaches illumi-

nated relevant issues. For example, judges functioning

as individual clinicians appeared to find particular fan-

tasy levels to be more transparent of subject characteris-

tics than others. Judges functioning as a research group

tended to find little difference in amount of accurate

information they could cull from fantasy material available.

Results of Hypotheses 4 and 5 suggested a judge/

fantasy material interaction when each individual judge's

performance was observed. Perhaps a judge/sorting task

interaction existed as well. To test this possibility, a

comparison of judge's ability, relative to other judges,

to consistently sort sex versus anxiety within the same

fantasy material was carried out. That is, judges' scores

for accuracy on the sex sort (presented in Table 5) were

ranked and compared with judges' scores for accuracy on the

anxiety sort (presented in Table 9) which were also ranked.

These comparisons were made within each of the four fantasy

materials. The above comparisons were made using Spearman

rank correlation coefficients. This same procedure was

also carried out, based on judges' scores for consistency

(previously reported in Tables 7 and 11) which were ranked.

The results of both sets of correlations are reported in the

last section of Table 12. According to both criteria for

ranking judges, none of the correlations were found to be

statistically significant. These results suggest a judge/

sorting task interaction. Thus, a judge's skill at sorting

sex is not significantly related to his skill at sorting

anxiety, given the same level of fantasy material. In fact,

with respect to the second criterion, an opposite trend

seemed to emerge. That is, for dreams, free fantasies, and

written samples, judges' consistency in sorting sex was

negatively correlated with ability to sort anxiety. These

findings, however, were not statistically significant.

In conclusion, utilizing the individual clinical

approach (each judge relative to himself), judges not only

displayed differential ability in utilizing various forms

of fantasy material, but also displayed a differential

ability in sorting subjects according to two categories

(sex and anxiety) within each fantasy level.

It is important to note that a judge/sorting task inter-

action was observed with variables (sex and anxiety) which

were reflective of "trait" qualities of subjects. Recall

that anxiety scores were stable (before and after data

collection scores remained significantly correlated) over

time and sensitive to enduring psychological qualities of

subjects. Similarly, sex membership is considered a "trait"

variable. The extent to which judges are sensitive to

"state" variables in subjects across fantasy levels requires

future exploration.

Hypothesis 6

Results of Hypothesis 6 indicated that individual sub-

jects inconsistently revealed themselves to judges across

various fantasy levels. On the other hand, judges' ability

as a group to accurately sort subjects according to

specified criteria was observed to be unaffected by fantasy

level. These results emphasize the importance of evaluating

judges' performance within an appropriate context. That is,

it is impressive to observe that judges as a group were

able to consistently categorize subjects according to two

objective criteria, despite individual subject's fluctua-

tions in self-disclosure across fantasy levels.

In evaluating the ability of judges as a whole in the

second half of Hypothesis 6, it was noted (see Table 14) that

the obtained means for the anxiety sorts were strikingly

lower than the obtained means for the sex sorts. To inves-

tigate this observation further, several t tests were cal-

culated, comparing means on the sex sort versus means on

the anxiety sort for dreams, free fantasies, written samples,

and TAT's, respectively. The resultant t values along with

their significance levels can be found in Table 17. Clearly,

the differences are highly significant in all cases. Judges'

overall ability on the sex sort significantly exceeded their

overall ability on the anxiety sort.

The above-reported finding is compatible with the

observed results of Hypotheses 2 and 3 (refer to discussion

section, Hypotheses 2 and 3). In short, judges not only

had difficulty evaluating subjects in the medium anxiety

category, but they also had significantly greater difficulty

evaluating subjects by anxiety than by sex, as a general

rule. While judges were consistent in perceiving subject

characteristics across fantasy levels, their performance

level varied, depending upon the sorting criterion.