Input via the keyboard can be slow and tedious for many computer
users, but the problem is particularly severe for those with motor
disabilities.  Strategies that reduce the number of keystrokes
required can help limit the problems these users face.  In a
programming enviroment, word prediction has emerged as a highly
successful strategy for saving keystrokes.  The use of statistical
measures such as the recency and repetitiveness of words can be
used to guide the prediction process to a 40% keystroke saving. 
However, these measures ignore valuable information about the
program structure.  The goal of this study was to test whether the
use of the syntax of a programming language can effectively assist
these statistical prediction strategies.  The results show that
inclusion of syntactic information of the Pascal programming
language can account for a further 3% increase in the number of
keystrokes saved, as well as a 9% increase in the accuracy of
predictions.  This result was obtained by running a collection of
Pascal programs on two separate applications that simulate text
input in a predictive Pascal program editor.  One of these
simulators uses only statistical prediction, whereas the other
includes the syntactical approach.  The average savings returned
by the two simulators were compared by performing a paired sample
means t-test.