For viscoelastic fluids the use of a “pressure tap” to measure the normal stress on a channel wall can lead to large errors. In previous studies the true normal stress on the wall was theoretically predicted or indirectly measured and compared to the erroneous measurements. Here pressure transducers are flush mounted on a rectangular channel approximating infinite parallel plates so that the true stress on the wall can be measured experimentally. A transducer is recessed to create a hole; the reading thus obtained is compared with the true stress to determine the hole error. Solutions of 0.9, 2, and 3% Polyox (polyethylene oxide, MW=4,000,000) in water were studied. Shear rates reached 17,000 sec−1 and hole errors up to 6000 dynes/cm3 were found. In the region of creeping flow, the hole error is equal to −0.16 of the primary normal stress difference. Hole errors have been determined at Reynolds numbers from 10−4 to 103 and at the higher values cross zero and become positive, as they are for Newtonian fluids. A mechanism is presented which explains this behavior of viscoelastic fluids over the seven cycle range of Reynolds number.