1. Are dialects changing into regiolects?
The question will be answered on the basis of the judgments we obtain in the web survey. On the basis of these judgments, the dialects will be clustered. Dialects are classified into different groups so that similar dialects are in the same group. For both judgments we will determine the natural number of groups (clusters). We will use the elbow criterion which says that the number of clusters should be chosen so that adding another cluster does not add significant information. If the percentage of variance explained by the clusters is plotted against the number of clusters, the first clusters will add much information (explain a lot of variance), but at some point the marginal gain will drop, giving an angle in the graph (the elbow) (Aldenderfer & Blashfield, 1984). We will also consider the L method, an efficient algorithm that finds the "knee" in a 'number of clusters vs. clustering evaluation metric' graph. The method was introduced by Salvador & Chan (2004). In this way we test our first hypothesis that dialect areas have been fused to larger and less sharply distinguished areas, namely regiolects. We may also test the hypothesis whether especially small dialect areas will fuse with larger ones.
2. Is the lexical level affected more strongly than the phonological and phonetic levels?
All of the recordings will be transcribed and digitized. The digitized transcriptions are the input for the computational procedures. The transcriptions will be used to calculate distances computationally. For the lexical level, we will use a simple binary measure – two forms are equal (0) or different (1) – or Goebl’s weighted similarity measure, a method in which the coincidence of rarely used forms counts more heavily than those of more frequent ones (Goebl (1984), p. 85; for application to Dutch lexical distances see Heeringa & Nerbonne (2006)). Lexical phonological, postlexical and purely phonetic differences are measured using Levenshtein distance, a string edit distance measure (for application to Dutch see Heeringa (2004) and Heeringa & Nerbonne (2006)).
For each linguistic level the measurements will be performed on the basis of the old male speakers and on the basis of the young female speakers separately. We will determine first whether the number of natural groups found on the basis of the latter measurements will be larger than the number of natural groups based on the first measurements. Second we test on which linguistic level the difference between the two classes is largest. In this way we test the second hypothesis that the lexical level will be affected more strongly than the phonological and phonetic levels.
Additionally we will compare the dialects to standard Dutch for each linguistic level. We expect that the recordings of the young female speakers will be closer to standard Dutch than those of the male speakers. Similar research was carried out by Heeringa & Nerbonne (2000) and Heeringa et al. (2000), but since we measure the degree of convergence per linguistic level, we are able to answer the question which linguistic level shows convergence most clearly. Per level we may test the hypothesis that the change from dialect to regiolect especially affects areas where dialects are relatively distant from the standard language.
3. Is the perception of the speakers affected? Has the speech production of the speakers changed more than the speaker’s perception?
These questions will be answered on the basis of the judgments which are obtained with the web survey. When both old males and young females listen to recordings of the same class (either old males or young females), we expect that the young female judgments will suggest fewer groups. This confirms our third hypothesis that the perception of the speakers has been changed from distinguishing dialects to distinguishing regiolects. Janson (1983) writes that 'for an individual in a situation of change, perception seems to lag behind production'. We will compare the contrasts in number and size of groups between the old males and young females at the perceptive level with the contrast we found at the production level, thus testing the hypothesis that perception lags behind production in the change from dialects to regiolects.