IffiTHODS 0? STUDYING CONTROLLED ViTORD ASSOCIATIONS
MILDPJCD 7/E3T LORBTG
A DISSERTATION
Submitted to the Board of University Studies of the
Johns Hopkins University in conformity with
the requirements for the degree of
. Doctor of Philosophy
1916
n-^i^+c
w
ACKNOWLEDGIIEIIC
I wish to acknowledge my greatest obligation to
Dr. Knight Dunlap, whose supervision, criticisms, and
suggestions were of foremost importance in the success-
ful completion of this study, Lly thanks are due also
to Professor John B. Watson for his assistance in making
this research possible.
TaTple of Contents ,
I. History and Development of the Word Association Method,
II. Methods of Studying Controlled Yiford Associations.
1. Purpose of the Investigation.
2. Apparatus.
III. Experiment I . Unselected Kouns and Adjectives.
1. Preparation of Material and Llethod of Procedure.
2. Results.
3. Conclusions.
IV, Experiment II . Selected Kouns and Adjectives.
1. Preparation of Material and Llethod of Procedure,
2. Results.
3. Conclusions.
V. Experiment III . Repetition of Stimulus Words,
1. Ptirpose and Procedure.
2. Results.
3. Conclusions.
VI. Experiment IV . Double Associations.
1. Purpose and Procedure.
2. Results.
3. Conclusions.
VII. Sximmary of Conclusions.
VIII. Pr.rther Problems in the f/ord Association Llethod
IX. Tables of Results. (I to XIZ inclitsive)
X. Bibliography.
XI. Tables of Stimulus Words.
XII. Vita.
History and Pevelopement of the Word Association Method .
The first experimental work on the word association meth-
od was published in 1879 by Francis Galton (1), He prepared
a list of 75 words to each of v/hieh he obtained one or sev-
eral free associations, using himself as subject. The list
was repeated four times in different environments, and the
results convinced him that the associations are not governed
to any extent by rasnory. It took 660 seconds to form the
505 ideas, which persuaded him that the mental process he
was investigating could not be revery or continuous associa-
tion. Of the 505 associations only 289 were different, so
there was much less variety in the ideas than he had sup-
posed. He examined the relative number of associations com-
ing from different periods of life, and found that one-half
of them concerned events of youth. His associations fell in-
to three classes, (1) imagined soixnds of words, (2) sense im-
agery, (3) histrionic representations. His list of words he
found to be of three types, (1) those that allow of mental
imagery, (2) those that allow of histrionic interpretation,
and (3) abstract words. For Galton, histrionic ideas always
appear as the first association, verbal associations occur
about equally as first or second association, and imagery
usually comes second. Even at this early date Galton antici-
pated the application of the words association method to
psychoanalysis .
In 1883 Wujidt published an article on psychological meth-
ods (2) in which he proposed a statistical method of collect-
ing a large group of single associations for one subject, in-
vestigating their origin and getting the association time.
He suggested, too, a correlation of results between normal
and pathological subjects.
In the same year, Trautschold (3), a student of Wundt,
made an extended investigation of the word association re-
action, using the Hipp chronoscope and two Morse keys, which
were released with the finger simultaneously with the speak-
ing of stimulus and response word. He had four subjects and
obtained averages on 40 to 256 words for the (1) simple re-
action time, (2) the v/ord reaction time, that is, simple re-
action time to the apperception of a word, (3) the word dis-
crimination time, v/hich is (2) minus (1), (4) the association
reaction time, and (5) the association time, which is (4)
minus (2). It is this association time that he v;as most in-
terested in. Its average for all subjects v/as 0,727 seconds.
The association reaction time varied from 0,896 secondsto
1,154 seconds, for different su.bjects. Trautschold obtained
further results in one type of controlled association where
the response v/as required to be a subsuraption. The associa-
tion tifne here is higher than for free association, 0,874
seconds. He classified his 60 stimulus words into three
classes, concrete objects, words expressing conditions, and
abstracts, and found that the association time is increas-
ingly longer for each class. He further gives a classifi-
cation of the associations into inner and outer associations
with various subgroups.
Cattell, in 1886, workirxg in Wundt's laboratory, was the
next to take up the problem. His first paper (5) gives the
time it takes to see and name objects. He exposed through a
slit various objects on a revolving drum and got the reaction
times with the Hipp chronoscope. For letters the time is l/E
second. He finds that a slit width siich that two letters
are e>qposed at once shortens the time 1/40 second, and on up
to five letters v/hich shortens the time l/SOO second. The
time for words having no connection is 1/4 second per v/ord,
for v;ords making sentences 1/8 second. He gives results al-
so on the comparative time for different size letters, types
of script, foreign words, colors and pictures of objects.
Cattell's second paper (6) in the same year gives results
for simple reaction time to light and sound, and shov/s that
the reaction time is shortest for a concentrated state of at-
tention, greater for normal attention and longest for dis-
tracted attention, when the distraction is continuous adding
and the stimulus is light or an electric shock. These varia-
tions in reaction time are slight however, but sufficient for
him to argue for a theory of attention. After getting fur-
ther results on distraction Cattell concludes that (1) the
ordinary degrees of attention do not affect the raction time,
(2) perception and v/illing are not factors in the reaction
time, for they are not necessary for the motor center to act.
The willing is done before the stimulus occurs and merely
sets the brain paths in a state of readiness. The perception
time is the interval betv.-een sensation and perception. It
is not possible to add a perception to a simple reaction with-
out adding also a will act. Altering the perception does not
change the will time making it possible to study the percep-
tion time. The sul);iect v/as required to discriminate between
two black discs, one vdth a white spot. By subtracting the
simple reaction time from this time we get the perception
time, which was 61 sigma and 95 sigma for the two subjects.
He repeated this procedure for different colors, and got long-
er times. He foxmd that practice lowers the perception time.
The discrimination time for letters is longer than for colors,
and for words only slightly longer still, for pictures it is
about the same as for colors. Cattell studied the will time
by requiring the subject to react to tvi'o stimuli with the two
hands respectively. He found this perception time plus v/ill
time greater than for the perception time. This holds true
for speech reactions, and stimulus colors, pictures, letters
and words. He has results also on the time taken to see and
to name letters, words colors, and pictures. Attention has
little effect on this reaction time, there is a practice ef-
fect during two years work and fatigue is manifested through
an increase in the reaction time after the first two or three
reactions of a series. After this there is no further in-
crease. In 1887 Cattell gave further results on the associa-
tion of ideas (7). He gives the time required to name ob-
jects in a foreign language, for translating words of dif-
ferent lengths into English and German, and various kinds of
controlled associations, (1) where there is only one possi-
ble answer, (2) where the answer allows of some choice, (3)
where choice is freer than in (2), (4) v;here a Judgment must
he made. Cattell continued his work in 188S, collaborating
with Bryant (8). They give results for free association to
500 concrete nouns, proper noujis and abstract nouns, verbs
and adjectives, for the authors, and for 10 concrete nouns
and 10 abstract nouns for University graduates (men), Bryn
Mawr students and German symnasium students. The association
reaction time varied from a little more than one second to
seven seconds, according to the mental development of the
group, and the type of stimulus word. Abstract nouns gives
a longer time than concrete nouns. The authors next found
the average time per word to write continuous associations
for each word in three groups of nouns for 20 seconds, using
school, college and graduate students as sixbjects. The re-
action time decreases with growth and education. There is a
slightly faster rate as the class rank is higher. The number
of words written decreases as the series is continued, indi-
cating fatigue or decrease in attention. There are more as-
sociations to concrete than to abstract words. -There is a
summary of all the response words for all subjects for each
stimulus word, forming a small lexicon of the most frequent
associations. The associations are further classified as
objective and logical, with subgroups.
Scripture, 1892 (9), investigated the problem of mediate
association. The subject v;as first shown a series of cards
each having on it a word and some kind of a sign. Another
series was then given with the same signs and different words.
The first series was then repeated vrf.thout the signs, and to
each word the suh^ect was asked to name any v/ord in the ser-
ies oectirring to him. If the sign is the linl: then the two
words with the same sign should he associated. Scripture
found no mediate association in a great many cases, bxit out
of the 185 experiments there were 79 cases of mediate asso-
ciation, from which Scripture concludes that the phenomenon
is possible.
Mtbasterherg (10) repeated Scripture's method more ex-
tensively hut got entirely negative results.
Bourdon, 1893 (13), obtained results on the character
of the response word for certain types of controlled associa-
tions; (1) letters or letter couplets as stimuli requiring
word responses gave responses with a large amount of phonetic
and syllabic similarity, (E) stim.ulus letters requiring let-
ter responses gave rise to contiguous letters or to letters
having a phonetic similarity, in about an equal nrunber of cases,
(3) stimulus letters requiring colors as responses gave rise
to associations usually marked by phonetic similarity, (4)
free word associations were usually determined by meaning,
v/ith phonetic similarity playing only a chance effect. Names
of objects and acts are associated with words of the same
class, qualities with qualities and objects. Some people
give answers coming under fewer categories, which indicates
a logical mind.
Hov/e, 1893, (14), made observations on the problem of
mediate association. A repetition of Scripture's technic gave
negative results. Howe argues that free associatiorxs alone
can give data regarding the existence of mediate association.
BergstrOm, 1894, (15), studied the interference of dif-
ferent associations upon each other. The subject v/as required
to sort 80 cards into 10 piles, each containing 8 cards with
the same pictures. In the second trial the position of the
piles was changed giving an interference effect. BergstrOm
concludes that the interference effect of an association
hears a constant relation to the practice effect, and is, in
fact, equivalent to it.
Calkins, 1894 (16), investigated the relative signifi-
cance of frequency, vividness, recency and earliness in the
formation of associations. The method was to expose a color
and a number successively, and after an interval another col-
or and number. After the series was finished the subject was
shown the colors alone in altered order, and asked to \7rite
dov/n what number v/as suggested, if any. Devices of repeat-
ing the same colors with different numbers, the same color
with the same number, of having certain colors and munbers oc-
cur always at the beginning of the series, or at the end, were
made use of to secure data on the factors under investiga-
tion. In every ease the more frequent, recent or vivid as-
sociations formed the largest niuiiber of correct recalls of
numbers. The frequency factor was found more effective than
primacy, vividness than recency, and frequency than recency.
Jastrow, 1894, (17), made a study of the community and
association of ideas. Sixty-nine students were shown 10
words, one at a tiine, and told to vrrite the first five words
STiggested to them. This gave 69 lists of 50 words each. He
then classified the ractions into two groups, giving the num-
ber of different and the number of unique re^ onses, in each
of the five places in the association series. "The proportion
of different words, as also of unique or once-used words, in-
creases as the associations proceed." Data jn men and v;oraen
indicated that there is less coraraujaity of associations among
women, TAdiich the author thinks due to too limited data. A
few results are given on the frequency of certain associations.
The associations are also divided into nine categories and
compared.
Jerusalem, 1894 (18), made a report of a single case of
mediate association. A scene long since forgotten appeared
suddenly to the mind of the individual as he was deeply en-
grossed in his work. But the connecting link, at first ap-
parently unconscious, was found to be the unnoticed fragrance
of a flov/er in the room, that had been in the original scene.
Smith, 1894 (19), investigating mediate association, ob-
tained entirely negative results as had Mttnsterberg. Scrip-
ture's technic was used in a modified forra with no results.
T/undt, 1894 (21) comments on Jerusalem's case of mediate
association. He claims that the connecting link was not un-
conscious but only unnoticed; it was perceived but notapper-
ceived.
Aschaffenburg, 1896 (24), studied free associations in
normal individuals. He divided his associations into differ-
ent subgroups under internal, external, clang and mere reac-
tion associations. In the first experiment a word was given
for the subject to write 100 continuous associations. These
were classified as above. In the second experiment 100 single
free associations were obtained from each subject. Experi-
ments were also made similar to these, getting reaction times
with the lip key. The author found that external associations
occur more often than internal associations and the reaction
time is shorter. The reaction time varied from 1180 to 14E6
Sigma, and the author concludes that the reaction time does
not distinguish normal from abnormal. It was noted that the
same stimulus word frequently led to the same response in dif-
ferent individuals.
Aschaffenburg next investigated the effect on association
of extreme exhaustion (24), which was induced by working con-
tinuously all night vdthout food. The same procedure was fol-
lowed as in the previous experiment. He found that more as-
sociations of long standing occur, content becomes a weaker fac-
tor and clangs are frequent and appear first at the height of
the exhaustion. The number of repeated responses does not in-
crease during the night and the exhaustion lengthens the re-
action time only slightly. The clangs, words completions and
rhymes are a sign of the reaction degenerating into a pure
motor reaction. In 1904 (24), he obtained 12,900 reactions
and 6150 reaction times on 11 pure manic depressive cases,
using only one and two syllable stimulus words, and fovind that
during excitement the results are the same as for exhaustion.
Alcohol and hunger also give these same results. In the de-
10
pressive state there are no variations in content from the
normal, although the reaction time increases.
Calkins in 1896 (25), continued her work on association
along the same lines as before, to see the effect of sugges-
tion. She uses suggestion in the sense of potency of the dif-
ferent factors, recency, frequency, vividness and primacy to
form associations. iTrequency is shown to he the strongest
factor. A monograph (26) incoi^orates these results along
with an analysis of the association process.
Kraepelin, 1896 (27), enumerated the psychological meth-
ods adapted to psychiatry among which the association method
occupies the major position. He emphasizes the necessity of
a careful interpretation of lengthened reaction times in
pathological cases, and regrets that much of the previous
v;ork v/ith this method has "been useless. He reviews the work
of Cattell and Mtosterberg, and points out the importance of
securing data on the content of associations and their stabil-
ity, and on continuous associations.
Calkins, 1898 (30), investigated the tendency to combine
ideas which are unconnected in their presentation. She pre-
sented one series of pairs of words visually, one series oral-
ly, and one series of paired pictures of objects. [The sub-
ject was to v/rite an introspection in regard to the concomi-
tant imagery. She finds that in 30.8 percent of the 638
cases, the subject has combined the two concepts into one
image picture. Nearly one half of the subjects show the ten-
dency to combination in one third of all possible cases. She
also obtained results on associations with childhood exper-
11
iences as first suggested by Galton, The subjects were 90
V/ellesley students and 84 people in middle life or advanced
age. The stimulus v/ords were 15 objects connected with child-
hood. The results showed that even the older group had less
childhood associations (33.4%) than Galton (39^), while for
students there were only 14,7^o. The number of recent asso-
ciations form 30.95^ and 32,7^ of the total respectively for
the older and younger group; for Galton these came only to
157a. Calkins thinks vie live largely in the present which
accounts for the results.
Ziehen, 1898 (34), added another to the numberous class-
ifications of associations. He divides them into spontan-
eous and judgment associations with various subgroups.
Sommer, 1899 (35), discusses methods of classification
of associations and gives the characteristics of the associa-
tions for certain pathological types.
Wreschner, 1900 (38), investigated the associations in
a case of idiocy. He found that (1) the quality of the stim-
ulus word has great influence on the kind of association, and
the quality of the reaction is less the greater it is for the
stimulus, (2) the reactions are predominantly adjectives, (3)
content associations have longer reaction times than clang as-
sociations, (4) the poorer the quality of the association and
the less familiar the stimulus word, the longer the reaction
time, which holds for both clang and content associations, (5)
practise shortens the reaction time and improves the quality
of the association.
1£
Kellor, 1901 (40), investigated associations in white
and negro criminal and workhc-ase women, and compared them with
results for University students (women?). Three types of
tests were used, (l) free continuots associations Vifritten for
one minute in reponse to each of a group of words, (2) free
continous associations in response to direct stimulation of
the five senses by appropriate stimuli, (3) constrained asso-
ciations, and (4) free continuous written associations to cer-
tain abstract terras. All stimuli were well vdthin the range
of the low grade of education of the subiects. The author
first compared the different subjects in regard to the dis-
tribution of the associations into the various types of im-
agery. For all subjects visual imagery predominates. Since
the negro women could not write, the second experiment was de-
vised and given only to them and to the white women crimi-
nals and workhouse irjnates. Again visual imagery predominates,
The rate of association for students was double that for
criminals and the workhouse class. Comparison was not possi-
ble with the negroes. By the second method the rate is high-
er for negroes than for whites (exclusive of students), due
to the elementary form of the negro associations, and to the
fact that the reaction for negroes v/ere spoken, for whites
written. The difference in rate between stxidents and white
criminals and negroes is due to the fact of the latter 's de-
ficient mental training, and to their greater tendency to
suppression. In students the free associations tend to be
propressively away from the first word, or of mixed form; the
associations of criminals of poor ediication tend to refer more
often directly to the initial word. Data was secured from the
associations to abstract words on the ethical, â– educational, ^^nd
environmental conditions of the various classes. The author
thinks the differences are not so much those between criminals
and normals as between educated and uneducated. He fails to
emphasize however the importance as a causal factor of feeble-
mindedness T;hich must have existed to a large extent in the
criminal classes he worked with.
Thumb and Marbe, 1901 (41), found that the associations
are often reversible. This holds for adverbs, adjectives and
pronouns, but not for nujnbers to which the usual response is
a higher number. The associations to all these classes are
mostly words belonging to the same class. This holds true al-
so for words expressing family relationship.
Mayer and Orth, 1901 (42), made a qualitive study of as-
sociations and concluded from their results with free asso-
ciations that (1) the response word can occur with or without
the intermediate occurrence of a conscious experience between
it and the stimulus, (2) the reactions are quicker when no
conscious experience intervenes, (3) the reactions are quick-
er the fewer the intervening conscious experiences, and (5)
reactions are longer v/hen the intervening conscious experience
is a will act than when it is not. These conscious exper-
iences are in most cases without feeling tone, but its pres-
ence increases the reaction time.
Bourdon, 1902 (43), obtained results on the frequency of
identical response words to 100 stimulus words for 100 sub-
14
jects of the same grade of intelligence and education.
Schmidt, 190E, (44) obtained the seme resiilts with verbs
that ThTimh and Marbe had with other parts of speech. He made
a more extended investigation in this one group than had his
predecessors, using 14 conjugation forms of 30 verbs on 8 ten
year old boys. This gave 3360 reactions. He found that the
verb responses predominate and formed S9fo of all responses,
forms of the same verb being more frequent and of shorter re-
action time than of other verbs.
Claparede, 1903 (45), made aix analytic study of associa-
tion, including a bibliography of over 300 names. He dis-
cusses disputed problems in association, incliiding those under
laboratory investigation.
Fuhrmann, 1903 (46), found that in idiots as contrasted
with epileptics the ability to forni supraordinate and subor-
dinate associations decreases.
Pieron, 1903 (47), attempted to explain introspectively
the failiire of the experimentalists to verify mediate associa-
tion. He conceives of mediate association as the arousal of
medially associated images of subconscious ideas.
Bleuler, 1904 (51), in his introduction to the paper of
Jung and Riklin discussed the importance of the study of as-
sociation. All mental processes involve the association of
ideas so that a thorough Icnowledge of the associations is es-
sential to an understanding of other mental activities. Every
association is made up of a group of more or less connected
ideas, which is called a constellation, and every associative
15
aotivity is the product of the whole psychic being of the past
and the present, with all its experiences and strivings. It
is an index then of all mental activity, which we only need to
solve to i-mow the w?iOle man. Bleuler recommends the method
of single associations rather than continiio-us associations,
and claims already to diagnose with the association method,
dementia praecox, epilepsy, imhecility, hysteria, and hopes
soon to he able to differentiate types of paranoia.
Jung and Riklin, 1904 (53), made an extended investiga-
tion into the associations of normal subjects preliminary to
a study of pathological subjects. Four hundred stimulus
v/ords of one, two and three syllable adjectives, nouns, verb,
adverbs and numbers, in random order were used. The reaction