Garry,
M., Sharman, S. J., Wade, K. A., Hunt, M. J. & Smith, P. J. (2001).
Imagination inflation is a fact, not an artifact. Memory & Cognition,
29, 719-729.
This article is a reply to the one
by Pezdek
and Eddy (2001). In that article, Pezdek and Eddy argued that
findings of imagination inflation may simply reflect regression towards
the mean. Their argument boiled down to the fact that Garry et al.
originally restricted their analysis to those items that were given initially
low ratings on the LEI. Not surprisingly, Garry and crew didn't care
for that argument too awfully much and they gave the following reply.
Specific Criticisms of Pezdek
and Eddy
-
First thing Garry et al. do is fault
Pezdek and Eddy for their definition of imagination inflation. According
to this reply imagination inflation occurs when imagining an event changes
the rated confidence that the event happened. The term itself is
neutral as to whether this is a function of creating a false memory.
-
For the regression analysis, Garry
et al. make two arguments:
-
First they cite someone who says that
its generally a bad idea to use residualized change scores.
-
They also argue that in the specific
case of imagination inflation the approach will not be able to find an
effect if one exists.
-
Garry et al. acknowledge that regression
towards the mean is occurring. But the question is whether there
is an effect of imagination over and above what one would expect based
on regression toward the mean. To test this one needs to look at
how much the control items change from time 1 to time 2 and compare it
with how much the imagined items change from time 1 to time 2.
-
Garry et al. argue that using an ANOVA
with data like this is inappropriate because there tend to be violations
of the equal variance assumption. They then describe t test they
peroformed showing that the probability of giving a higher LEI rating at
time two was greater for items that were imagined than for items that weren't.
-
Garry et al. argue that Pezdek and
Eddy aren't really looking at regression towards the mean, but rather
regression towards the midpoint of the scale. The mean LEI scores
were actually lower than the midpoint of the scale and if regression is
occurring the proper break point is above and below that mean, not above
and below the midpoint of the scale.
So Garry et al. make the case that
imagination inflation is the real deal and the Pezdek and Eddy are just
stacking "turtles on top of turtles".