Section 14.9 Two Within-Subjects Variables and One Between-Subjects Variable 489Table 14.10 Expected mean squares
Source df E(MS)
Between subjects an 21
A(groups) a 21
Ss w/in groups a(n 2 1)
Within subjects na(bc 2 1)
Bb 21
AB (a 2 1)(b 2 1)
B 3 Ss w/in groups a(b 2 1)(n 2 1)
Cc 21
AC (a 2 1)(c 2 1)
C 3 Ss w/in groups a(c 2 1)(n 2 1)
BC (b 2 1)(c 2 1)
ABC (a 2 1) (b 2 1)(c 2 1)
BC 3 Ss w/in groups a(b 2 1)(c 2 1)(n 2 1)
Total N 21s^2 e 1 ns^2 bgps^2 e 1 ns^2 bgp 1 ns^2 abgs^2 e 1 ns^2 bgp 1 nau^2 bgs^2 e 1 bs^2 gps^2 e 1 bs^2 gp 1 nbu^2 ags^2 e 1 bs^2 gp 1 nabu^2 gs^2 e 1 cs^2 bps^2 e 1 cs^2 bp 1 ncu^2 abs^2 e 1 cs^2 bp 1 nacu^2 bs^2 e 1 bcs^2 ps^2 e 1 bcs^2 p 1 nbcu^2 aBefore we consider an example, we will examine the expected mean squares for this
design. These are presented in Table 14.10 for the case of the model in which all factors
other than subjects are fixed. (subjects are treated as a random factor.) From the expected
mean squares it is evident that we will have four error terms for this design. As before, the
is used to test the between-subjects effect. When it comes to the within-
subjects terms, however, Band the interaction of Bwith Aare tested by B 3 Ss within
groups; Cand its interaction with Aare tested by C 3 Ss within groups; and BCand its in-
teraction with Aare tested by BC 3 Ss within groups. Why this is necessary is apparent
from the expected mean squares.An Analysis of Data on Conditioned Suppression
Assume that a tiny “click” on your clock radio always slightly precedes your loud and intru-
sive alarm going off. Over time that click (psychologists would call it a “CS”) could come
to elicit the responses normally produced by the alarm (the “US”). Moreover, it is possible
that simply presenting the click might lead to the suppression of an ongoing behavior, even
if that click is not accompanied by the alarm. (If you were lying there reading, you might
pause in your reading.) In a laboratory investigation of how the click affects (suppresses)
ongoing behavior, Bouton and Swartzentruber (1985) investigated the degree to which a
tone, which had previously been paired with shock, would suppress the rate of an ongoing
bar-pressing response in rats. Suppression was measured by taking the ratio of the number
of bar presses during a 1-minute test period following the tone to the total number of bar
presses during both a baseline period and the test period. For all groups, behavior was
assessed in two Phases—a Shock Phase (shock accompanied the tone) and a No-shock Phase
(shock did not accompany the tone) repeated over a series of four Cycles of the experiment.
It may be easier to understand the design of the study if you first glance at the layout of
Table 14.11. During Phase I, Group A-Bwas placed in Box A. After a 1-minute baseline inter-
val, during which the animal bar-pressed for food, a tone was presented for 1 minute and was
followed by a mild shock. The degree of suppression of the bar-pressing response when the
tone was present (a normal fear response) was recorded. The animal was then placed in Box BMSSs w/in groups