The results of two perception tests designed to shed light on the perception of rhoticity are reported. In the first, rhotic tokens of the words fort, stars and hurt were played to forty phonetically trained listeners in two stimulus conditions: unfiltered and filtered. In the unfiltered condition F3 and all spectral components above it up to the Nyquist frequency of 5.5kHz were present. In the filtered condition all components above F2 had been removed by lowpass filtering.
Contrary to received expectation, most listeners reported hearing stronger rhoticity in the filtered condition. This result is explained in terms of F1-F2 relations and the broad-band auditory integration hypothesis (Bladon 1983). In the second experiment, a non-rhotic token of the word nurse was played to twenty three phonetically trained listeners, a subset of the listeners for the first experiment, in an unfiltered and filtered condition. In the filtered condition F3 and all upper frequencies were once more removed. Again contrary to received expectation, and perhaps even more surprisingly, a clear majority of listeners heard the filtered token as rhotic. Taken together, the results of both experiments indicate that, far from inducing the perception of rhoticity, F3 may in fact have an inhibiting effect on it.