APPENDIX D
Manipulation Checks
Variance due to Stories:
Comparisons were made, using ANOVA, within balanced and within unbalanced categories, in order to determine whether the con-tents of the four stories that compose each of these two categories were contributing to the variance in the dependent variables, or vvhether the data from ali the balanced stories (or the unbalanced stories) could, if desired, be combined. It was the intent of the experimenter's procedures in the pilot study to make the stories equivalant in ali relevant respects except for the particular balance configurations.
To test the effects of story, four one-way ANOVAs of Ratings by Story, overall treatments, (done separately for balanced and unbalanced categories, and separately for consistency and for pleasantness) gave the results shown in Table D. 1 and D. 2.
From the Tables D. 1 and D. 2, we see that consistency ratings do not show any significant differences among the stories within the balanced category, or among those in the unbalanced category.
Subjects judge balanced stories as consistent, or unbalanced stories as inconsistent, regardless of the story content. But, pleasantness ratings do show higly significant story-by-story differences. If we recall that the content of the stories using the two different dependent variables is exactly the same9, and that the dependent variable question is identical except for the word "inconsistent" or "unpleasant", the above finding can be interpreted to indicate that the processes involved in judging pleasantness are different from the processes involved in judging consistency.
Story Differences:
In this section I affl trying to show that whatever differences I find betvveen stories when the response measure is pleasantness are not due to their configurations (i.e., the combination of + or—signs), but to some other aspect öf the wording in each story.
To clarify the nature of story differences, when pleasantness is the response measure, a series of t-tests have been performed among the ratings of the individual stories, and among groups of stories cor-responding to certain specific configurations.
9 The two maximum conditions bear exception to this; maximum/consistency questioıınaire and maximum /pleasantness questionnaire differ also in some other respects (see section on procedures).
117
Table D.l
ANOVA Consistency Ratings of Four Stories in each of the Balanced or Unbalanced Triads:
BALANCED CONFIGURATIONS
SOURCE SS DF MS F P
Between Stories 5.3 3 1.8 0.32 .999
Within Stories 4574.3 833 5.5
0.32 .999
Total 4579.6 836 5.5
0.32 .999
UNBALANCED CONFIGURATIONS
SOURCE SS DF MS F P
Between Stories 8.9 3 3.0 0.30 .999
Within Stories 8240.0 833 9.9
0.30 .999
Total 8248 .9 839 9.9
0.30 .999
Table D.2
ANOVA Pleasantness Ratings of Four Stories in each of the Balanced or Unbalanced Triads:
BALANCED CONFIGURATIONS
SOURCE SS DF MS F P
Betvveen Stories 91.4 3 30.5 4.50 .004 Within Stories 5656.6 835 6.7
4.50 .004
Total 5747.9 838 6.9
4.50 .004
UNBALANCED CONFIGURATIONS
SOURCE SS DF MS F P
Betvveen Stories 84.2 3 28 .1' 3.90 .009 Within Stories 6012.6 835 7.2
3.90 .009
Total 6096.8 838 7.3
3.90 .009
One such obvious configuration is balance versus unbalance, and we already saw that it involves a strong, clearcut differentiation.
Table D. 3 summarizes this finding. Apart from the mere fact of dis-tinguishing balanced from uıbalanced triads, we also see some addi-tional findings: for instance, the table of mean ratings for individual stories (Table 4.1) had indicated that balanced triads are uniformly judged to be consistent, and unbalanced ones inconsistent.
118
TABLE D. 3
Means and Matched T-test Results CONFİGURATİON
OF BALANCE
CONFİGURATİON OF UNBALANCE Al
Soap Seascape
A4 Opera
A5 Wagner
A8 River
e a
£ o a a o S
A2 African
Mask A3 Jukebox
A6 Dali
A7 Swamp P RESPONSE
MEASURE +
+ + + o. S o
a 3 o m 0
+ + - + +
illi
Consistency 8.3 8.4 8.5 8.3 8.4 3.8 4.0 3.8 3 .9 3.7 * Pleasantness 3.1 3.8 3.0 3.2 3.3 5.4 5.7 5.4 4.9 5.6 *
* p < .0005
As can be seen from Table D.3 even though the configuration effect is somewhat less distinct in the case of pleasantness ratings, by and large, balanced stories are judged to be rather unpleasant, vvhile the unbalanced ones are rated more towards the pleasant end, and this difference was significant.
However, in the case of pleasantness ratings, not ali the stories within the balanced or unbalanced category have received the same degree of pleasantness ratings. In fact, here we observe some, although minör, story differences when we individually compare the stories within a pair. With individual t-test comparisons the results of which are given in Table D. 4, once again consistency ratings do not signifi-cantly differ among the stories that compose a balanced category or an unbalanced category. (Out of the 12 paired t-tests, none of them came out significant).
In the case of pleasantness ratings, out of the 12 comparisons, 6 were observed to be significant. Of these 6 significant t-test com-parions, 3 had the following composition:
A4 v.s. A5; A4 v.s. A8; A l v.s. A4 The remaining 3 had the follovving composition:
A6 vs. A7; A2 vs. A6; A3 vs. A6.
It seems that story A4 is responsible for the significant t-differ-ences in the first 3 comparisons, and the story A 6 is associated with the last 3.
To see if such story differences are somehow connected with
"higher order" configuration differences, certain groups of stories in which story A4, and story A6 took part were formed, with respect to their configurational similarities, and average ratings of groups were compared (Table D. 5).
119
TABLE D.4
Matched T-Test Results (between individual stories) SIGNIFICANCE LEVEL
COMPARİSON CONSİSTENCY PLEASANTNESS
A 1 '—- 4 (Seascape)
\s. A 4 +
(Soap Opera)
ns* .002
A 5 + (Wagner)
vs. A 8 + -(River)
ns ns
A 3 - + +
(Jukebox) vs. A 6 (Dali)
ns .046
A 2 — + + (Afr, Mask)
vs. A 7 (Svvamp)
ns
\
ns
A 1 + -(Seascape)
vs. A 5 +
(Wagjıer)
ns ns
A 4 + (Soap Opera)
vs. A 8 + -(River)
ns .001
A 3 — + + (Jukebox)
vs. A 7 (Svvamp)
ns , ns
A 2 - + + (Afr. Mask)
vs. A 6 (Dali)
ns .001
A 1 — + -(Seascape)
vs. A 8 +
-(River) ns ns
A 4 + (Soap Opera)
vs. A 5 +
(Wagner)
ns ,0005
A 2 - + + (Afr. Mask)
vs. A 3 - + + (Jtıkebox)
ns ns
A 6 (Dali)
vs. A 7
-(Svvamp)
ns .001
*p < .05
Altogether, six comparisons with combined average ratings were made for each dependent measure, and none came out significant for consistency ratings. In the case of pleasantness ratings, the effect of story A6 ("Dali" story) disappeared and so failed to point to a negativity effect or a positivity preference. Among ali the combina-tions that story A6 was put in, I suspected a significant difference only in the A6 ( ) and A7 ( ) versus A2 (— + + ) and A 3 ( 1-+) comparison. Here, the individual effect of a story A6 might have joined with a preference for positivity (a configurational bias)
120
Table D. 5
Matched T-Test Results (betvveen story combinations)
COMPARISON
SIGNIFICANCE LEVEL COMPARISON CONSİSTENCY PLEASANTNESS A 4 + A 5 +
(Soap Opera) ' (Wagner)
& vs. &
A 1 — + - A 8 — + — (Seascape) (River)
ns .009
A 4 + A 5 + (Soap Opera) (Wagner)
<6 vs. &
A 8 — + — A l — + —
(River) (Seascape)
ns .006
A 4 + A 8 — + — (Soap Opera) (River)
& vs. &
A 5 + A ı _ + _
(Wagner) (Seascape)
ns .043
A 6 A 7 —
(Dali) (Svvamp)
& vs. &
A 3 — 4- + A 2 — + + (Jukebox) African Mask)
ns .004
A 6 A 7
(Dali) (Svvamp)
& vs. &
A 2 — + + A 3 — + + African. Mask) (Jukebox)
ns
A 6 A 2 - + + (Dali) (African Mask)
& Vs. &
Al A 3 - + +
(Svvamp) (Jukebox)
ns ns
to produce a significant difference. This did not materialize, and the individual effect of story A6 has practically disappeared.
Nevertheless the effect of story A4 remained significant. It is interesting to note, however, that it remained as an isolated effect since it appeared in every case of the 3 composite comparisons, and thus failed to point to a systematic "higher order" configuration effect. Story A 4 happens to be balanced, but there are 3 other balanced stories. It seems as if an affective process is responsible for this result, since it would be in line with the expectation that such processes have the tendency to get associated with concrete and minute details in the content (when the content is not sufficiently structured).
121
To recapitulate, closer examination of story differences as revealed by pleasantness shows that such differences are not due to the par-ticular configurations that make up "balanced"ness or "unbalanced"-ness, but are specifically connected with the content of a particular story. Therefore, combining the data from ali the balanced stories (or the unbalanced stories) when obtaining the results of this study does not seem to be unjustified in the light of the above manipulation checks.
/
APPENDIX E