‘63

 

c: the 3343;?” 35

{SAN S

C
331‘

.9
.I
.0

535:5

’5'
Q

m .

3:

Ni
V

f
O

I: ”En!
L I}

D Q
:‘f‘

$64

LIBRARY

Michigan State
University

 

 

AN EMPIRICAL INVESTIGATIOBI OF ‘HE "vHOBFIAN" EYECTHESIS

USING ARTIFICIAL LIRGUIS TIC LAPERIAL

Paul Rosenthal

A THESIS

Submitted to
Kichigan State University
in partial fulfillment of the requirements
for the degree of

Department of Communication

1964

ABSTRACT
AN EMPIRICAL INVESTIGATION OF THE "WHORFIAN" HYPOTHESIS
USING ARTIFICIAL LINGUISTIC MATERIAL

by Paul Rosenthal

The present study was designed to test one aspect
of the "Whorfian" hypothesis, namely, that the lexical
aspect of language can influence the perceptions of its
speakers. The language variable manipulated was learning
of nonsense syllable names for shades of gray. The hypo-
thesis predicted improvement in discrimination between
the members of gray pairs as a function of learning a
nonsense syllable for each gray as opposed to learning
the same nonsense syllable name for both grays.

The study tests Whorf's contention that speakers of
a language in which few words described a broad referent
category would not be able to make distinctions within
that category as well as could persons whose language
contained many words which meaningfully broke down the
category. He cited the example of the Eskimos who have
seven distinct words within the category that is defined
in English by the one word, "snow." In this study, the
learning of two nonsense labels for the members of a
pair of gray shades may be likened to the Eskimos' abund-
ance of names of "snow." The learning of one nonsense

label for the members of a pair of gray shades corresponds

Paul Rosenthal

to the situation in English, where no distinctive names
break down the category of "snow."

The present design has certain advantages over other
designs that could be used to investigate the "Whorfian"
hypothesis. If we worked with English speaking subjects
and Eskimo subjects, any differences found in the ability
of these groups to perceive differences in "snow" might be
due to the difference in familiarity of these groups with
" snow." By enabling us to use subjects from one society,
this design makes possible the elimination of the extran-
eous variable, familiarity with the stimulus dimension.
These subjects should be approximately equally familiar
with the gray cues selected. Shades of gray were employed
to rule out the possibility of Ss'learning to selectively
respond to perceptually prominent attributes of the cues.
Each subject received both values of the linguistic var-
iable, thus controlling for inter subject variability.

The two gray pairs used in the study were selected in
a pilot study so that their members were approximately
equally difficult to discriminate when presented tachis-
toscOpically. However, we felt it necessary to control
for any remaining differences in discriminability. For
each pair of grays, one half the Ss were assigned the
identical label learning condition and one half the dif-

ferent label learning condition. Also, a counterbalance

Paul Bosenthal

was introduced to control for possible nonsense syllable-
cue interaction.

In general the procedure was made up of three phases:
pretest, training, and post-test. The pretest consisted
of tachistoscopic presentation of the gray pairs in order
to obtain a measure of discrimination accuracy when the
eXposure time for the grays was near the discrimination
threshold. After the pretest, learning trials were begun.
Each subject saw each of the four grays associated with a
nonsense syllable (one syllable for both members of one
pair and different syllables for the members of the other
pair) which he was instructed to anticipate when he saw
the gray shade. When three perfect trials through the
list of cue response associates was achieved, the subject
was dismissed. When he returned the following day he was
given training to a second criterion in which the gray cues
were more similar to those of the pretest. A post-test was
then administered which was a replication of the pretest.
Its purpose was the redetermination of discrimination
accuracy for the pairs of grays presented in the pretest.
It would indicate any differential change in accuracy that
might have resulted from the different learning conditions-

The test of the hypothesis, a learning condition by
pre-post interaction effect, was not significant. These

negative findings were compared to the positive results

Paul Rosenthal

of a similar study where selective attention to common or
disparate attributes of the cues could have been learned
during the training phase. It was concluded that selective
attention to common or distinct attributes of cues may be
one mechanism by which the degree of lexical differentiation
within a stimulus dimension influences SS' ability to dis-

criminate between stimuli along the dimension.

 

TAcLE OF CONTENTS

Page
List of Tables
I Means and Standard Deviations of Accuracy Scores 19
II ANOVA of Number of Correct Discriminations 20
List of Figures
I Gray Cues Used in the Study ll
II General Plan of the Design for a Single Subject 16
Introduction . . . . . . . . . . . . . . . . . . . l
I'IethOd. o o o o o o o o o o o o o o o o o o o 1.0
Results 0 O O O 0 O O O O O O O O O O O I O 0 l7
DiSCUSSion O O O O O O O O O O O O O O O O O O O 22
conClUSion O O O I O O O O O O O O O O O O O O O 26
Bibliography . . . . . . . . . . . . . . . . . . . 27
iAppend.iCeS O O O O O O O O 0 O O O O O O O O O O 28
A. Random Assignment of Nonsense Labels to Grays
to Accomplish Controls . . . . . . . . . . 28

B. Data Sheets Used in the Study . . . . . . . . 29

 

INTRODUCTION

There is a long tradition of scholastic inquiry con-
cerning an assumed relation between language and thought
or conception (Sapir, 1921; Esper, 1935; Khorf, 1956).
This study is an attempt to investigate empirically one
aspect of this so-called "Whorfian" or "Sapir-whorf"
hypothesis. The study involved constructing within Ss
two different languages, the first being limited in lex-
icon while the second was more versatile.

The "Whorfian" hypothesis maintains that language
functions not simply as a device for reporting experience,
but also as a means of defining eXperience for its speakers.
Whorf (1956) believed that language, in itself, served to
direct the perceptions and the habitual modes of analyzing
experience (world views) of individuals. More recent ad-
vocates of the hypothesis have taken the more defensible
position that language, along with certain universal bio-
logical, psychological, and social characteristics, all
serve to influence an individual's perceptual world.

Language in the present context is meant to include
both its structural (phonemic, morphemic, and syntactic)

and semantic or lexical aspects. The structural aspects

provide the frames into which utterances are cast. The
semantic aspect consists of a self-contained system of
meanings. Walker, Jenkins, and Sebeok (l95h, p. 193) pro-
vide an example of these two components of language and
the manner in which they may be related in a communication:
"If I am to communicate about colors, I must employ the
color terms available in the language (semantic or lexical
component), and languages will differ in how the color term-
inology 'carves up' the physical spectrum; I must also follow
the grammatical rules (structural component) by which lexical
items are compounded in the language."

whorf's writings contain references to both aSpects
of language as determinants of the world views of their
speakers. For example, working with the American Indian
language, HOpi, he reasoned that the world view of HOpi
individuals would differ from that of individuals of EurOp-
ean background, because their language uses a different
tense system. In regard to the semantic aspect of language,
he reasoned that speakers of a language in which one word
described a broad referent category would not be able to
make distinctions within that category as well as could
persons whose language contained words which meaningfully
broke down the category (the most frequently used example
is that of the Eskimo who has seven distinct words within

the category that is defined in English by one word "snow"),

- 3 -

When it comes to the explication of examples where
world view has been influenced by language, however, many
researchers in this area err by citing instances that can
be traced to other cultural phenomena beside language.
Other researchers unknowingly resort to circular reason-
ing in defense of the hypothesis (Walker, et. al., 1954,
p. 194). They state that people in different societies
have different world views or perceptions, because their
languages differ. "hen they are asked how it is that
they know that the world views do indeed differ, the re-
searchers reply that they know, because individuals in
these societies use language differently.

It is obvious that one requires independent measures
of language and perception or world view if one is to
test the "Whorfian" hypothesis. This can be achieved by
eliciting non-linguistic behavioral responses to linguis-
tic stimuli. It will then be possible to infer back from

these behavioral responses to the perceptions that pre-
ceded them.

Carroll and Casagrande (1958) have reported a study
‘that relates grammatical structure to perception. They
. noted that in Navaho, verbs which concern the handling
of objects invariably contain suffixes which refer to
the object being handled. Ihey hypothesized that Navahos
would tend to be more aware of the "shape" dimension than

would Lnglish speakers. To test this hypothesis, they

-4-

selected two groups of Navaho children, one predominantly
Navaho speaking and one predominantly English speaking.
Each subject was shown ten pairs of objects (colored wooden
blocks, sticks, and pieces of rope). Each pair differed
in tWo respects: color and size, color and shape, or size
and shape. With each pair, the researchers also presented
a third object similar to each member of the pair in one
of the three dimensions mentioned above, and had the subject
select the member of the pair which the third object seemed
most similar to. Navaho dominant children made their
selections according to "shape" more often than did English
dominant children, but the difference was not significant.
In regard to the semantic aspect of language, which
is more directly relevant to this study, there have been
several important contributions. Carroll and Casagrande
(1958) noted that in certain instances EOpi has one verb
to define activities which seem unrelated to English
speakers, because English has separate verbs denoting
these activities. They selected sets of three stimulus
figures so that two pictures in each set depicted acti-
vities defined by one verb in Hopi, while another com-
bination of two pictures depicted activities defined by
one verb in English.
English and Hopi speakers were asked to select which

two pictures of the set seemed most similar. The procedure

-5-

might be clarified with an example. Hopi makes no dis-
tinction between the "pouring" (intentionally) and the
"spilling" (accidentally) of objects, but it does have

a separate verb to denote "drOpping" of a single object.
The subjects were shown the following pictures: A- A
man pouring apples out of a basket, B- A man accidentally
drOpping a coin from his pocket, and C- A man spilling

a pitcher of milk. As hypothesized, English speakers
paired pictures B and C, while HOpi speakers tended to
pair pictures A and C. This was eXplained by the fact
that English speakers recognize the accidental component
of B and C, but the Navahos, having no easily available
verbal distinction between the accidental and purposeful
activities involved, tend to pair A and C.

Brown and Lenneberg(l95#) did a study that suggests
that the lexical aspect of language influences perception
as indexed through behavioral responses. A sample of
English speakers was shown a set of color patches and
asked to label them. It was found that certain colors
were given the same label more often than was the case
for certain other colors. These measures of the "cod-
ability" of different hues are said to "constitute an
empirical estimation of the way in which English lexical
codification.'ca1vesup' the stimulus dimension of wave

length of light."

-5-

In the second part of the study, a group of subjects
were shown sets of four color patches of different codabil-
ities and then asked to recognize these four patches in
a group of 120 patches. The subjects had significantly
greater success in recognizing the more codable of the
colors.

Fridja and Van de Geer (1961) tested the correlation
between "codability" of facial expressions and recogni-
tion. Ss were shown thirty facial expressions and res-
ponded with adjectival descriptions for each expression.
At a later time, 88 saw the facial expressions briefly
and then attempted to recognize each expression when pre-
sented in a test set of sixty. There was a significant
correlation between the codability of the facial expres-
sions and their recognizability.

There seem to be several difficulties in the inter-
pretation of these studies. In the Brown and Lenneberg
study, the variable "codability" might be due to other
cultural factors besides language. That is, colorsthat
occur more frequently in the society may come to acquire
more specific names. Thus, recognition might be traced
back to familiarity. Along this same line of reasoning,
it is possible that the subjects were not sufficiently
interested in their task to recognize any of the color

patches. They may simply have selected those colors that

-7-

were most familiar to them in their everyday environment.
Similar questions can be raised concerning the Frijda and
Van de Geer study.

Similar difficulties are encountered in the inter-
pretation of inter-linguistic designs such as those of
Carroll and Casagrande. In these designs researchers
select strikingly difficult features in two languages,
but it is hard to guarantee that the observed behavioral
correlates are not due to irrelevant cultural features
like differences in familiarity with the stimulus objects.

In this study, two pairs of gray color patches were
selected so that the members of each pair were approxi-
mately equally difficult to discriminate. SS learned
one nonsense syllable name for each of the two shades
of gray in one pair, and one name for both shades of
gray in the other pair. The hypothesis was that subjects
would show a significant pre—post increase in discrimina-
tion accuracy for the pair associated with different
labels. No such effect should appear for the pair asso-
ciated with one label.

Results obtained by Stellwagen (1963) indicate that
these predictions require that the members of each pair
be presented successively in the discrimination task at
an exposure period approximating the discrimination

threshold. Simultaneous presentation prevents the SS

-8-

from mediating to each gray separately, thus only success-
ive presentation was investigated.

Returning to the previously mentioned example of the
Eskimo who has seven words for what English speaking in-
dividuals call "snow," if we were to work with Eskimo
and English speaking subjects, we might find that the
Eskimo would perceive differences in the "snow continuum"
better than English speakers could. However, the difference
in perception might be due to differences in familiarity
(theEEkimo is certainly more familiar with snow than the
English speaker).

In the present study we controlled for the familiarity
of the subjects with the stimulus dimension used in the
determination of the dependent variable by building both
language treatments into subjects from the same society.

It seems reasonable to assume that these subjects have
approximately equal familiarity with particular grays,

even though the amount of familiarity may differ from

gray to gray. The application of both treatment conditions
to each subject controls for further inter-subject variabi-
lity in familiarity with the stimulus dimension.

Differences in the difficulty of discriminating be-
tween the pairs may result from the greater physical dis—
parity of one pair or from the fact the the members of

one pair are more familiar in our society than members of

-9-

the other pair. To control for these possible differ-
ences in discriminability, for each pair of grays we had
one-half the 83 learn different names and one-half the
subjects learn one name.

Discrimination of points on the "grayness" continuum
was chosen as the measure of the criterion variable,
because selective attention to common or disparate attri-
butes of gray patches is not possible. The effect of the
linguistic variable might be inflated if for the different
label condition 88 learned to respond selectively to dis-
parate characteristics of cues, or if for the same label
condition, Ss learned to selectively respond to percep-
tually similar characteristics of cues. Such an effect
might have been responsible for the positive results ob-
tained by Stellwagen (1963) in a study which differed
from the present study in that inkblots were used as cues.
With gray cues, 53 must respond to relations or ratios
between the stimuli. The fact that the grays have no
common English names rules out another attribute of cues
that may be attended to selectively. This study will pro-
vide evidence as to whether the lexical variable can facil-

itate discrimination when selective attention is eliminated.

 

lO

{ETHOD

Apparatus and Materials -

The exposure device employed allowed stimuli to be
presented at any of thiry durations ranging from thirty
msec. to 1,750 msec. An AC timer relayed current to a
rapid start, BZ-H Circline Fluorescent Lamp. A distance
of 46 inches separated Ss from the cues. Ss saw the cues
through the center of the circular stimulus lamp. During
all phases of the study, the intensity of light at the
stimulus card when the stimulus was being presented was
.7 candles/ft.2 The gray cues were mounted on 83 x 11
inch cards. The stimulus cards were placed on an auto-
matic card changer capable of presenting fifty cards.

The card changer could be programmed for different inter-
vals between presentation of stimuli.

The gray shades used were selected from the full
set of Color-View papers. All patches were from the same
production run. The gray pairs used consisted of shades
4 and 9, and 13 and 15.

The two pairs of gray patches were selected in a
pilot study so that the two shades within one pair were
as difficult to distinguish as the shades comprising the

other pair. This was accomplished by pairing various of

- 11 -

the Color-View papers and presenting the members of each
pair successively at six second intervals in the tachis-
tos00pe. Eash S was repeatedly shown these two "differ-
ent" pairs (4-9, 13-15) along with other pairs which con-
sisted of one of the four grays being shown twice (4-4,
9-9, 13-13, 15-15). Figure I presents the four gray cues
used in the study.

Figure I

Gray Cues Used in the Study

Gray 4 Gray 9

 

Gray 13 Gray 15

Subjects responded as to whether the grays in each
pair looked " same" or "different." The two pairs sel-
ected were those for which subjects had equal success in
discriuinating between the members over a range of eXpo-

sure tines.

-12...

The subje ctive equality of distances between vari-
ous grays was also checked by having subjects place the
grays along a linear continuum so that the distances be-
tween grays seemed to correspond to the differences in
brightness. Averaging the values given by a group of
subjects for each of the grays gives scale values for
these shades directly on a linear continuum of "gray-
ness." The method employed is one of the "Subjective
Estimate" scaling techniques described by Torgeson
(1960. pp- 61-93).

The nonsense syllables were chosen to be phoneti-
cally distinctive. The syllables selected were: buv,
faw, ged, hib, hoy, mep, vos, and zir. Six pairs of
nonsense syllables were selected from the above syllables
so that the members differed in the vowel and consonants
used. For the gray pair that was to be labeled differently,
one member selected at random was associated with the
same syllable that was associated with each of the grays
in the other pair of grays. The other member of the "dif-
ferent" pair was associated with the second nonsense syl-
lable.

We have already mentioned the manner in which possible
differences in the discriminability of the gray pairs was
controlled. A counterbalance was also arranged to control
for the possibility that characteristics of cues interact
with characteristics of nonsense syllables. If an S was

to learn mep, mep, mep, ged, another S was assigned ged,

- 13 -

ged, ged, mep. Appendix A illustrates the assignment of

nonsence syllables to grays employing these controls.

Procedure -

The first step in the procedure was the determina-
tion of the S's threshold for discrininating between the
grays in the two pairs. The subject was shown the two
pairs of grays that were, in fact, different; he was also
shown two pairs whose members are identical. As in the
pilot study, the subject responded as to whether the grays
in each pair looked "same" or "different." Also as before,
the grays are presented at six second intervals, using
the tachistoscope.

The procedure was repeated a number of times at dif-
ferent exposure speeds until a duration was found which
resulted in slightly better than chance success discrim-
ination. i'he order of presentation of the pairs was
changed after every few trials, as was the component gray
in each of the identical pairs. The grays appeared as
upright rectangles (5/16" x 1-1/4") without background.
Approximately one minute was taken between trials to re-
stack the cards on which the grays were mounted.

The discrimination threshold thus determined was
used throughout the pretest. The pretest procedure was
essentially the same as that of the threshold determina-\-

tion. There were six trial runs of four pairs each, with

- 14 -

two "identical" pairs and two "different" pairs. Pre-
sentation order was systematically varied as shown in
Appendix B. The subject was allowed to continue in the
study if he was successful on at least 14 of the 24
pairs, and yet did not achieve complete success or com-
plete lack of success for the six presentations of either
"different" pair. These criteria indicate that the sub-
ject sees some differences, but still has room for im-
prove ment.

During the training phase, each of the four grays
was associated with the appropriate nonsense label. This
was accomplished by presenting each gray tachistoscopically
followed by the name. The grays and syllables were pre-
sented for 1,750 msec., with the gray shades appearing
aslarge rectangles (3" x 3%") against a light background
which, in turn, was bordered by black. The subject anti-
cipated the name after seeing the gray. After reaching
the criterion of three perfect trials through three random
orders of the list of four cue-response associates, the
subject was released and told to return at the same hour
on the following day.

On the second day, the subject was presented with
three new random orders of cue-resaonse associates (Appen-
dix 3 gives the presentation orders for the two days

Then he reached the criterion of three successive errorless

- 15 _

trials, he was asked to anticipate the labels without
being able to see them after the grays. when this
criterion was achieved, the grays were then presented

so as to appear as narrow upright rectangles as in the
pretest. This use of criteria of increasing difficulty
enabled the subject to master the task of learning labels
for grays under the difficult but necessary condition
that the grays appear in perceptually similar situations
in the discrimination and learning phases.

The post-test was then administered. It consisted
of a redetermination of discrimination accuracy for the
pairs of grays presented in the pretest. It was, in
fact, an exact replication of the pretest. Figure II
presents the general plan of the study for a single hypo-

thetical subject.

Subjects - Twenty-four 38 were used. They were under-
graduate students taking introductory psychology courses

at Michigan State University.

16

General Plan of the Design for a Single Subject

Pretest Measurement

of Discrimination

Accuracy for Pairs
of Visual Cues

Learning
Conditions

Post Test Measure-
ment of Discrimina-
tion Accuracy

 

Grays 4-9 (Different)

Grays 13-15 (Same)

Control Gray Pairs
4'49 9‘99 13-13!
15-15

Gray 4 - Hep

Gray 9 - Ged

Gray 13- Ged

Gray 15 - Ged

Grays 4-9

Grays 13-15

Control Gray Pairs
4-4, 9-9. 13-13,
15-15

 

 

17

As mentioned previously, a crude discrimination
threshold for the two gray pairs was determined for each
subject. The mean threshold was 646 msec. with a stand-
ard deviation of 214 msec.

The mean trials required to reach learning crit-
erion I on the first day was 6.13. The mean trials re-
quired to reach the three criteria on the second day was
17.67.

Comparison of the mean trials on the second day to
reach learning criterion I (background) and learning crit-
erion III (no background) indicated that subjects had more
difficulty learning phrases for the grays when the grays
were exposed without background, as opposed to when they
were exposed as larger areas against a light background.
The mean trials to achieve criterion I was 4.58, as op-
posed to 9.79 for criterion III.

Comparisons were made of errors in the "different"
and "same" learning conditians. Under the criterion where
the Ss saw the grays against a light background, the mean
terrors for the "same name" pair was 1.29 with a standard
deviation of 2.66. For the "different name" pair, the
mean was 2.71 with a standard deviation of 2.70. A test

of the difference between means was significant at the

.05 level. When the grays appeared without background,
the mean and standard deviation for the "same" pair were
1.46 and 1.89 respectively, as opposed to a mean of 4.29
and standard deviation of 3.58 for the "different" pair.
Again, the difference in means was significant (at the
.01 level).

Means and standard deviations of pre and post-test
discrimination accuracy scores are presented in Table I.
Inspection indicates that increases in discrimination
accuracy scores appeared for the different label condition
in both counterbalances. For the same label learning con-
dition, subjects in one counterbalance showed a decrease
in discrimination accuracy, while subjects in the other
counterbalance showed a pre to post-test increase. Sur-
prisingly, the largest increase in discrimination accuracy

appeared in the same label learning condition.

19

TABLE I

Means and Standard Deviations of the Accuracy
Scores; Cell N = 12; A Hean of 3.00 = Chance.

 

Counter— Counter-
balance I (Al) balance II (A2)

 

 

£33 (01) X 3.50 3.00

so 1.09 1.35

Same (Bl) POST (02) X 3.08 3.58
SD 1.56 1.50

PAS (01) X 3.08 3.75

SD 1.50 1.14

Different (32) X 3.42 3.83
r03? (02) so 1.31 1.34

 

To test the significance of the main effect means (pre
vs. post, same vs. different learning conditions, and non-
sense syllable counterbalance) and their interactions, the
accuracy scores were subjected to the AXBXCXS analysis of
variance described by Winer (1962, p. 319). A summary of
this analysis is presented in Table II. hone of the three
treatment effects was significant. The learning condition
by pre-post interaction, a direct test of the hypothesis,
was not significant. Similarly, the other overall inter-

action effects were not significant.

20
TABLE II

ANOVA of Number of Correct Discriminations

 

 

Source of Variation DE Mean Square F
Between Ss 23
Counterbalance (A) 1 1.76 .94
Error (Bet.) 22 ' 1.87
Within Ss 72
Same - Diff. (B) 1 ‘ 1.26 .39
s x A 1 1.76 .55
B X SS (Nn A) Error (W1) 22 3.21
Pre- Post (C) l .51 .53
C x A 1 .84 .87
C x 33 (wn A) Error (W2) 22 .97
B X C l .09 .07
B X C X A l 2.34 1.89
B X C x 88 (En A) Error (w3). 22 1.24

 

Counterbalances had been constructed to control for
the possibility that nonsense syllables and cues may
interact in a complex manner. Einer (1962, p. 208) main-
tains that individual comparisons should be made irres-

pective of the outcome of the overall r test, if these

comparisons were built into the design because of some

--—_‘-

 

-21-

theoretical concern of the study. Therefore, individual
tests of significance of pre-post changes, and pre-post
by learning condition interaction within each counter-
balance were justified. The (wz) and (W3) error terms
from the ANOVA were used to test pre-post changes and
learning condition by pre-post interaction respectively.
None of the significance tests were significant.
Inspection of the raw data indicated no differences
in the counterbalances constructed to test for differ-

ences in discriminability between the two gray pairs.

 

DISCUSSION

In this study the hypothesis was that Ss would show
an improvement in accuracy of discrimination from pre-
test to post-test after learning different as opposed to
identical nonsense syllable names for two shades of gray.
When an S has learned one name for two shades of gray,
each of the grays will elicit the same mediational res-
ponse and identical response produced stimulation. SS
can discriminate between the grays only in terms of the
physical difference between the stimuli (percentage of
light reflected).

In the instance where different names are learned
for the two stimuli, Ss have an additional cue to make
the discrimination task easier. Each gray elicits a
distinct mediational response, and distinct response pro-
duced stimulation. Es can discriminate between the grays
in terms of the original physical difference, and also in
terms of the differences in response produced stimulation.
Thus, learning different names should lead to greater per-
ceived distinctiveness in cues. However, the hypothesis
was not confirmed by the data. A possible eXplanation

for the negative findings is discussed on the following

page.

- 23 -

The relevance of this study to the "Whorfian" hypo-
thesis might be questioned on the ground that the study
does not deal with existing natural languages. However,
by artificially manipulating the lexical variable, both
of its values could be built into SS from a single cul-
tural background. This eliminated variability due to
differences in familiarity with the stimulus dimension
under investigation, as these SS should have approximately
equal familiarity with the brightness continuum. This
same variable complicates the interpretation of inter-
linguistic designs. Building both treatments into each S
served to control for between S variability.

By having one-half the SS associate different labels
with the members of one pair, and having the other one-
half of the SS associate one label with this pair, we
were able to control for differences in discriminability
between the pairs. Varying familiarity of cues along the
stimulus dimension within a culture might cause these
differences in discriminability. Different familiarities
of stimuli could have been responsible for the results
obtained by Brown and Lenneberg (1954) and Frijda and
Van de Gear (1961).

Comparison of the negative results of this study
with positive results obtained by Stellwagen (1963) in

a related study seems to throw some light on a mechanism

by which linguistic variables influence perception.
Stellwagen's study was similar to the present study,

but differed in that inxblots were used as cues rather
than grays. Learning different labels for inkblcts did
result in increased accuracy of discrimination, although
the interaction between nonsense labels and cues appeared
to be an important factor. That is, differences were
found for subjects who learned ged, ged, ged, mep for the
cues, as OppOSed to subjects who learned mep, mep, mep,
ged.

The inkblots differ in an important respect from the
grays used in this study. In the different label condi-
tion of the inkblot study, SS could learn to selectively
respond to some difference in shape in one particular area
on the blots. This selective attention to one attribute
of the stimuli would lead to a response of "different"
when 'the blots were presented near the discrimination
threshold. Similarly, Ss instructed to learn a common
label for two blots might respond selectively to some
similarity in configuration made more noticeable by the
identical labels. This would lead to a response of "same"
in the discrimination task. Also, the irregularly Shaped
inkblots might tend to elicit verbal associations by
virtue of accidental resemblances in shape to physical
objects. These associations would confound the effect

of the learned nonsense syllable labels.

- 25 _

Using grays in the measurement of the linguistic
variable eliminates the possibility of selective attention
to common and different attributes. Gray shades have no
physical attributes which may be selected out as Similar
or different. Ss can only respond to discrepancies in
brightness between the grays. Also, grays do not elicit
any verbal associations or common names as inkblots might.

Frijda and Van de Geer (1961) believe that the in-
fluence of language codes on cognition is greatest when
the stimuli are not distinct or "outstandingly" different.
In such cases, SS will group stimuli in terms of the lin-
guistic categories, and tend to disregard the ambiguous
aspects of the stimuli themselves. However, the inkblot
study and the present study seem to indicate that dis-
tinctiveness of stimuli can be an important factor. The
gray Shades lacked "outstanding" physical differences, and
distinctive labels were not sufficient cues to enable SS
to discriminate between the grays..

It seems that under the conditions of these two studies,
lexical codes will not affect perception unless there are
originally certain physical differences between the stim-
uli. Further research in various experimental situations
is required to determine to what degree the stimuli must
be distinct if linguistic codes are to influence percep-

tion.

 

26

CONCLUSION

The negative finding of this study where selective
attention is ruled out and the positive finding in a
study where selective attention was a distinct pos-

sibility, lead to the following conclusion:

In this experimental situation,selective
attention to common or distinct attributes of
cues seems to be one mechanism by which the
degree of lexical differentiation within a
stimulus dimension influence Ss' ability to
discriminate between stimuli along the dim-

ension.

 

10.

BIBLIOGEAPHY

Brown, B.w., and Lenneberg, E.H. "A Study in Language
and Cognition," J. of Abnorm. and Soc. Esvchol.,

49: 454-1162, 195E. '—

 

Carroll, J.B., and Casagrande, J.B. "The Function of
Language Classes in Behavior," in Readings in Social
Psychology. Maccoby, Newcomb, and Hartley (Eds.),
New York: Holt, Binehart, and Einston, 1958, 18-31.

Esper, E.A. "Language," in A Handbook of Social Psychology.
Murchison (id.), korcester, Aass.: Clark University
Press, 1935, 417-450.

Frijda, Nico E., and Van de Geer, John P. "Codability and
‘Recognition: An Experiment with Facial Expressions,"
Acta Psychologica, 18: 360-367, 1961.

 

Sapir, B. Language; An Introduction to the Study of
Speech. New York: Harcourt, Brace, 1921.

Stellwagen, W.T. Technical Report G-l9359: "Accuracy of
discrimination between visual stimuli as a function of
learning distinctive responses to the stimuli," 1963.

Torgeson, W.S. Theory and Methods of Sampling. New York:
Wiley and sons, 1960, 61-930

Walker, D.E., Jenkins, J.J., and Sebeok, T.A., "Language,
Culture, and Cognition," J. g: Abnorm. gpd Soc. Psychol.,
Tsycholinguistics Supplement, Osgood and Sebeok TEds.),
49: 192-903, 1954.

Nhorf, B.L. Language, Thought, and Reality, with an intro-
duction by J.B. Carroll. Cambridge, Mass.: Technology
Press, 1956.

Miner, B.J. Statistical Principles in Experimental Design.
New York: McGraw-Hill, 1962.

28

APPENDIX A

NGYSENSE LABEL ASSIGNMENTS TO GRAYS

There were six combinations of nonsense syllables
employed: hoy-zir, ged-vos, hib-buv, ged-faw, vos-buv.
For each of the six combinations, four subjects learned
different cue-nonsense label arrangements. In this man-
ner, the controls for discriminability differences be-
tween the gray pairs and for nonsense syllable-gray
interaction effects were carried out.

The following is the assignment of nonsense syllable
labels to grays for one of the six nonsense label com-
binations and can serve as an example for the other five

nonsense label combinations:

 

 

 

 

GrayeSyll. Gray-Syll. GrayrSyll. Tray-Syll.
51: u- zir Sg: 4- hoy S3: 4- boy Sn: 4- zir
9- zir 9- hoy 9- zir 9- hoy
13- hoy 13- zir 13- zir 13- hoy

15- zir 15- hoy 15- zir l5— hoy

 

APPENDIX B

Data Sheets - (contain order of presentation of
gray pairs in pre and post-tests, and order of pre-

sentation of grays in learning trials).

29

TEST TRIALS

Diff Same Diff

m
a
(D

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Switch Setting

 

 

 

 

 

 

rasrssr
1 4-4 4-9 13-13 13-15
2 13-15 15-15 4-9 9-9
3 13-13 4-9 13-15 9-9
4 13-15 4-4 15-15 4-9
5 4-9 15-15 13-15 13-13
6 4-4 13-15 9-9 4-9

 

 

 

 

 

 

Y Correct:

Diff

31

LS

'ARNING TRIA

L“
3—4

‘RDER I

 

 

 

 

 

 

 

 

 

 

 

 

e
S
a
r
h
TI
liLvlll
4.1...I 1|. iTl .I ll ill III III II. .I LI. (.1 IF. Ill
WNTInwlil Tl'TITllTi_III'T|LTIL
3.1
RIM-l: 4] LI TI 1 I III III A." j Ill-I .' Ila! ll:-
111 1

ein!!! III II II F Fl lT' 1T I'A IJ J III?
WhIIITII'l.rLrlul,-ILIIL~JIIIIIIIS
inlll III rll Tll TI .Igrl I?! Ilr .L r LI .II LT! Lr
B
"U llll rlu if r I: ll «All. 1T IlT II. III T I. '4
HMHFIUIv [III I II I I.
an fil * I1 I. fil .l fil a
\NTI. :1 II II T: ll sl. II T. T. fl.al ll
.1 /

'1! III II II LII T! L lull lLv .L
R 1 a. nr 1

O/ 3 5 U. 5 3 :4. Q/ 5 3 0/ MW
.1. l l _.l_ l l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Same Diff

Number of errors:

32

TRIALS

m
I
VJ

51:11:. '

LT:
«b.4114

ORDER II

00

 

 

 

 

 

 

 

 

 

 

 

8
Cu
a
r
h
P
“Nail! .I I I .4I I JI I L
D1.
Q.ArI I I I I I l .I I ll: —
RiI T 7 ~11
m
R
U.
.irl. II 1. 1. 11 IL II. II II I1 _
a T. e
m
I L a
.IT I T I r . I IA 1T 7 Cu
B
_.U.TII TIII III I I II I LII I Ill WM
3 o
n
Du f
1 O
u a r
in J M
NH.
3 .4 5 5 3 4 5 :4 9 3
.1 1. .1 11 .1 11

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FOST- PES‘
1 4-4 4-9 13-13 13-15
2 13-15 15-15 “-9 9-9
3 13-13 4-9 13-15 9-9
4 13-15 4-4 15-15 4-9
5 4-9 15-15 13-15 13-13
6 4-4 13-15 9-9 4-9
#Correct: Same /6 Diff\ /6

33

mow USE ow

Ream b

B

11?“

 

"'11'11‘1’111111:11311111111111!

68 926