Professional Documents
Culture Documents
Postman - Bruner 1948 - Personal
Postman - Bruner 1948 - Personal
IN PERCEPTION*
BY LEO POSTMAN
Indiana University
JEROME S. BRUNER
Harvard University
AND
ELLIOTT McGINNIES
University of Alabama
HAT one sees, what one ob- tribute to perceptual selection over and
50 ^> t .11
50 OI5
•*•"•' -*n
30
,-•• /
\
ll .17 50
50
fi
\
N,
S
/
/
11
(i
V
\ / Ii y
h'
30 t=.
\
? \ 035 10
\
1
_,
.23 30 ' \
30
ot
""^^ \
\ \
* A
^
1
MJ
t
10 .055 ^
h EC AesSoc Pol R«3l i }
10 i 30
JG RM c
1fh EC AesSoc Pol R91 .01 Th EC Aes Soc Pd Rel
T
50
K
r"" x
*--^
V*
^ s O3
50 035 ^ 50
N
.10
v OS
30
^
J\ s
^* .055
HC
i h EC AesSoc Pol Rel •an
30
i
i k
7
t- —1 ' —,
k,^
.14
/ X
S^ / \
10
^ \
/ ^ 075 50
I
> ~~
,09
in
10
I"
/
.18
JK \, SG
T h EC AesSoc Pol Re>l
30 V /^
// /N 13
. T h EC AesSoc Pol Rel
/£
bU 04 I0
t
^ 17
"
er«
50
50 .05
RB
-.065
Th EC AesSoc Pol Ret
50 01 50 .018
\ I \
\
50 /
V
\ .015 50 / '
.025 50 \ ( .08
/
/
\
> >/ CN / \ , '\
ft
/
/
k ^•* .025 30 •\\s /
/
\\ flCK "*O
/ V \
t
,\ \ \ \
.10
30
J
V \
\
/
/
\ J K "*
^1
^
\ •
/
<
/, .22 50 I 5O
• /
.13
V, /
; \ i
\
!/ .24
t
\
\ j
s
\\
\
o. y\ n /
/
t
3O ^
\\ /;/ Ji
^i
\
i
30
\
\ '/
\ /
/! \ \\
^\
j
l*% *^O
10
DH
> .28 10
DL
1A |/-\
EN
\ it 17
1'h EC AesSot Pol Rel Th EC AesSoc Pol Rel Th EC AesSoc Pol Rel
I / \
/,
^
/
\ /
/
K,,
30 ^k t
^7 .07 30 .05 30 .02
s?\
/'
\
\
\ i \
//('
^
\ /
t "
i 1
f
10 .11 10 09 10 04
Ssi. HG IV
T h EC AesSoc Pol R<il T h EC AesSoc Pol ftH 1h EC AesSoc Pol Rel
50 .06 50 .09 50
^ ^.1 i 02
1 ><
30 >f-—
*-—•
V>
/
^
y~ \
%
/ ,/\V
Y\
g* ~^ ^
.10 30 \ .13 30 \\ .04
\
\ < / S,
\
^ \ \\
10 .14 If 17 10 r .06
JA JC JE
Th EC Aes Soc Pol Rel
scores are plotted against the left-hand ordinate. Average recognition times for the words
representing these values are plotted against the right-hand ordinate. Solid lines represent
value scores, dotted lines represent times of recognition.
LEO POSTMAN, JEROME S. BRUNER AND ELLIOTT MCGINNIES
ject in random order. Each word was RESULTS OF THE EXPERIMENT
exposed three times for .01 second. Analysis of Recognition Thresholds*
If the subject failed to recognize the Is time of recognition significantly in-
word, three exposures were then given fluenced by the value which a given
at .02, .03 second, etc., at exposure times stimulus word represents? Each sub-
increasing in even steps of .01 second ject's value profile was compared with
.until recognition occurred. A full rec- his "time-of-recognition profile." The
ord was kept of all the subject's pre- value profile is a type of psychograph
recognition responses. Subjects were on which the subject's scores in the six
TABLE 2
CHI-SQUARE TEST OF SIGNIFICANCE OF ASSOCIATION BETWEEN VALUE PREFERENCE AND
TIME OF RECOGNITION
(Theoretical Frequencies are in Italics.)
VALUE SCORES
Above Mean Below Mean
Above
Mean 156 216
TIME OF i8i .40 790.55 372
RECOGNITION
Below
Mean 283 245
257.56 270.45 528
439 46i
y2=n.87 P = < .01
.100
.090
H
.080
o
V
d>
c
o
.070 -
.060 •
6
Value Rank
FIG. 2. AVERAGE TIMES OF RECOGNITION FOR,THE WORDS REPRESENTING THE Six VALUES
OF THE ALLPORT-VERNON STUDY ARRANGED IN RANK ORDER
correspondence between the two profiles Statistical analysis confirms this im-
(e.g., the profiles of RB and IV). Such pression. The value scores of each sub-
striking relationships are not, of course, ject were classified as falling above or
the rule. One isolated case (JC) shows below the population mean (30) for
what appears to be a reversal, high- the Allport-Vernon test. His time-of-
value words requiring, on the whole, a recognition scores were similarly di-
longer exposure time than less-valued vided into those falling below or above
148 LEO POSTMAN, JEROME S. BRUNER AND ELLIOTT McGiNNiEs
his own mean time of recognition. graphically in Figure 2. The signifi-
Combining the results for all subjects cance of the difference between the
into a two-by-two contingency table mean times of recognition of stimulus
(Table 2), a chi-square test of independ- words was tested for all possible com-
ence was performed. The obtained binations of value ranks. As Table 4
chi-square value of 11.87 indicates, at shows, the words symbolizing the sub-
a high level of confidence, that the as- jects' highest ranking value are recog-
sociation between value orientation and nized at exposure times significantly
time of recognition is not random.8 shorter than those required for words
An analysis in terms of a two-by-two symbolizing their lowest ranking value.
table, though useful, can do little more A comparison of the highest ranking
than indicate that a general relation- and second lowest value (Ranks i and
ship does exist. For purposes of more 5) yields a similar result. All other dif-
TABLE 3 ferences fail to reach statistical signifi-
cance although they are predominantly
MEAN TIMES OF RECOGNITION AS A FUNCTION
OF INDIVIDUAL VALUE RANKS REPRE-
in the expected direction.
SENTED BY THE STIMULUS WORDS The great majority of subjects, then,
conform to a general pattern. The
MEAN TIME op
higher the value represented by a word,
VALUE RANK RECOGNITION IN SECONDS the more rapidly is it lively to be rec-
ognized.
I .075 Analysis of Attempted Solutions.
2 .082 Statistical analysis shows that value acts
3 .082
.089 as a sensitizer, lowers the perceptual
4
5 .098 threshold. But value orientation does
6 .097 more than that. It is an active, selective
disposition which in many subtle ways
detailed analysis, each subject's value affects the hypotheses and attempts at
scores were, therefore, ranked from solution which precede the actual rec-
highest (Rank i) to lowest (Rank 6). ognition of a stimulus word. Much can
For the group as a whole, the average be learped about the role of value as
time of recognition was computed for an organizing factor in perception from
each of the six ranks. Note that the an analysis of pre-solution behavior.
analysis here is in terms of ran\ of value Each subject's perceptual behavior
rather than in terms of specific value forms an individualized pattern and our
areas. That is to say, Rank i could be preceding analysis of group data inevi-
any one of the six values for a given tably sacrifices a great deal of highly
subject, and so on down for the re- suggestive information about individual
maining ranks. The mean times of rec- "styles" of perceiving. As a first ap-
proximation to a more intensive inves-
ognition for the six value ranks are pre-
tigation of perceptual behavior, we have
sented numerically in Table 3 and
examined carefully and sought to clas-
8
Our findings are congruent with the re- sify individual pre-solution responses.
sults of earlier research carried out by A. G. Our effort has been to find categories
Woolbert as reported by Cantril and Allport (6).
Woolbert found that subjects perceived preferen- of classification which might throw into
tially those items in a dummy newspaper which relief the directive influence of value
were most closely related to their dominant orientation on perception.
values as measured by the Allport-Vernon Study
of Values. The following categories for the
PERSONAL VALUES AS SELECTIVE FACTORS IN PERCEPTION 149
analysis of pre-solution responses or 3. Structural responses: Under this
hypotheses have emerged: heading fall the very frequent incorrect
i. Covaluant responses: This category hypotheses based on the structural char-
comprises responses which can be acteristics of the stimulus word. An
unambiguously classified as represent- illustrative sequence of hypotheses given
TABLE 4
SIGNIFICANCE OF DIFFERENCES BETWEEN MEAN RECOGNITION TIMES FOR ALL
COMBINATIONS OF VALUE RANKS
Entries in the Table Represent Values of t, and P (in italics).
.83
2
.40
.80 .07
3
>.4o >-9°
6
2.42 1.54 1.67 .80 .09
<.02 >./0 .10 >-4<> >.9o
ing the same value area as the stimulus by one subject in response to the word
word. The subject who saw the word loving was: movies, mowing, moving,
Easter when the stimulus word was lowing, and finally loving. A frequent
sacred illustrates the covaluant category. stimulus-bound, structural hypothesis
2. Contravaluant responses: In some was the response turkey for theory.
cases, the words reported in the pre- 4, Nonsense responses: Two types of
solution period were opposite in mean- responses are included here: (a) non-
ing to the stimulus word or served to sense words, such as linone for income,
derogate it. An instance is provided or weelby for wealthy; and (b) partial
by a subject who saw scornful upon responses in which the subject's hy-
presentation of the stimulus word help- pothesis consisted of an enumeration of
ful. Or revenge instead of blessed. parts of a word or individual letters.
150 LEO POSTMAN, JEROME S. BRUNER AND ELLIOTT MC&NNIES
5. Unrelated responses: This is our urements, we nonetheless present the
residual category. All responses which results of our classification as the sim-
could not be related to the stimulus plest and most convenient description
word in terms of any of the above cate- of general trends.
gories were provisionally classified as That several of our categories did
unrelated. Responses such as upper and discriminate between pre-solution re-
carol to a word like useful .may serve sponses to high- and low-value words
as an illustration. We do not for a may be taken as a presumptive demon-
moment believe that they are haphaz- stration of their validity. Covaluant
ard responses. The fact that this cate- hypotheses occur with significantly
TABLE 5
MEAN FREQUENCY PER WORD OF DIFFERENT PRE-SOLUTION HYPOTHESES FOR HIGH-VALUE
AND LOW-VALUE WORDS
gory turned out to be the most numer- higher frequency in response to high-
ous is a commentary on the inadequacy value words than they do in response to
of existing analytic categories in the low-value words. A complementary
study of pre-solution behavior in per- finding is that both contravaluant and
ception. nonsense hypotheses , appear more
Table 5 represents the mean fre- prominently among responses to low-
quency with which each of these kinds value words. There is a similar tend-
of pre-solution hypotheses occurred per ency for structural hypotheses to be
stimulus word in the subjects' high- associated more frequently with high-
value (Ranks i, 2, and 3) and low- value words, though the difference falls
value (Ranks 4, 5, and 6) areas. Table short of statistical significance. Our
5 also shows the significance of the residual category, unrelated hypotheses,
differences in the mean frequency of favors neither high- nor low-value stim-
the various response categories when ulus words, nor is there any particular
high- and low-value areas are compared. reason why it should.
We are ready to grant at the outset
that the categories of classification used THE ROLE OF VALUE ORIENTATION IN
in the analysis of pre-solution hypothe- PERCEPTUAL SELECTION
ses are tentative. Their reliability has Selection is one of the three basic
,not as yet been demonstrated. The cate- adaptive processes that operate in per-
gories, moreover, are not always mu- ception. Inextricably linked with selec-
tually exclusive. Without claiming any tion are accentuation and fixation. Once
high degree of precision in our meas- selected, a percept may be accentuated,
PERSONAL VALUES AS SELECTIVE FACTORS IN PERCEPTION 151
i.e., certain of its features ^may be em- which value orientation becomes, a de-
phasized (3, 4, 5). Fixation denotes the terminant of selection.
persistence and preferential retention Our results lead us to propose three
50
High-Value Words
40 Low-Value Words
30
e
a
!!
10
Type of Hypothesis
FIG 3. MEAN FREQUENCY WITH WHICH VARIOUS TYPES OF PRE-RECOGNITION HYPOTHESES
WERE GIVEN IN RESPONSE TO HIGH-VALUE AND LOW-VALUE WORDS