Abstract
In this article, two hypotheses were tested to explain attitudinal ratings like SURPRISE, SUGGESTION, REMINDER etc. of four rising nuclear contours observed in a Dutch question corpus and described as (a) H*L H%, (b) H* H%, (c) L*H H% and (d) L* H%. According to one hypothesis, the middle tones in (a) and (c) should be parcelled out, such that their absence produces contours (b) and (d), respectively, predicting communality of meaning within (a, c) that excludes (b, d). That is, (b, d) could be viewed as undershot variants of (a, c), with undershoot expressing a shade of meaning different from that of the fully realised pitch accents. This hypothesis was not confirmed by the data, though. The other hypothesis was based on the concept of the ‘frequency code’, which associ ates high/rising F₀ values with non-assertive behaviour and appeared to provide a much better model to predict listeners’ ratings.