18
Shrew/ Curmudgeon (t = -15.05, p < .001), Despondent (t = -11.62, p < .001), Vulnerable
(t = -11.47, p < .001).
Hypothesis 4a assumes a stronger relationship between message strategy and
negative stereotype activation than relational level and negative stereotype activation.
This hypothesis was supported. A MANOVA revealed a significant multivariate effect
for both relational level and message strategy (F(4,179) = 4.30, p = .002, Wilks’ =
.912; F(4,179) = 61.56, p < .001, Wilks’ = .42) respectively (Table 4). A Bartlett test of
sphericity ((9) = 438.12, p < .001) for this MANOVA was significant and revealed that
the multivariate analysis was appropriate to use. There was, also, a difference between
relational level and message strategy on 2 of the 4 negative superordinate categories. For
Severely Impaired the message strategy versus relational level on negative stereotype
activation was F = 123.90, p < .001; F = 8.21, p = .005, respectively, for
Shrew/Curmudgeon the message strategy versus relational level was F = 241.77, p <
.001; F = 14.78, p < .001, respectively, for Despondent the message strategy versus
relational level was F = 143.00, p < .001; F = 14.49, p < .001, respectively, and for
Vulnerable the message strategy versus relational level on negative stereotype activation
was F = 136.49, p < .001; F = 10.32, p = .002, respectively. Also, the effect size of
message strategy (eta
2
= .58), on negative stereotypes was much larger than relational
level (eta² = .09). According to Cohen (1988), these eta² levels indicate that message
strategy had a large effect on negative stereotype while relational level only had a
moderate effect on negative stereotype activation.
Hypothesis 4b assumes a stronger relationship between message strategy and
negative stereotype activation than age and negative stereotype activation. This
Convention