This article addresses a seemingly trivial question-''Based on theolog
ical measurements, when are two samples different beyond experimental
error?''-and proceeds to give a non trivial answer. The preferred answ
er requires the application of t statistics to multiple measurements o
n each sample. The larger the number of multiple measurements made, th
e lower is the uncertainty in conclusively discriminating between two
samples. However, such multiple measurements on every sample are seldo
m feasible in real life and then it becomes necessary to resort to the
use of Z statistics. A stepwise procedure for the application of both
, t and Z statistics is given. Knowledge of the measurement precision
is a prerequisite for scientifically answering this seemingly trivial
question. Hence, the precision of capillary rheometry measurements (ap
parent viscosity, extrudate swell ratio, and entrance pressure loss) i
s discussed in detail. The precision of these capillary rheometry meas
urements improves as the magnitude of the variable measured (force/pre
ssure or swell) increases. This implies that the precision of capillar
y rheometry measurements can be improved by lowering the experimental
temperature. Extensive numerical data on various high-density polyethy
lenes generated using two different capillary rheometers are given and
are compared with the ASTM round-robin data. (C) 1995 John Wiley & So
ns, Inc.