This analysis was in response to a question from one of you:

For ndiff=1, nrel=4, consider serial position function:
For Subject 38 it is:
position <- c(399,509,490,627)

Also for ndiff=1, consider RT vs nrel (size of display)
For Subject 38 it is:
size <- c(382,416,497,517)

Let dur = mean duration of a matching comparison

Bamber's model says that the size effect should be linear,
with slope = half the match duration

If testing proceeds from left to right on all trials, then the position
effect should be linear with slope = match duration.

What are the two slopes?

lsfit(1:4,size)$coef[2]
 48.6 msec/test

If the model and the hypothesized testing order are correct,
then the position slope should be twice this, i.e., about 97
msec/position

lsfit(1:4,position)$coef[2]
 66.5 msec/position

But it is substantially smaller than 97.  This could happen if
testing did not proceed uniformly from left to right.