Bitonal lexical pitch accents in the Limburgian dialect of Borgloon

Transcription

1 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon Jörg Peters Abstract Borgloon is one of the westernmost places in Belgian Limburg which has a word accent contrast, also known as the distinction between accent 1 and accent 2. To learn more about the tonal system of the local dialect two reading tasks were carried out. The first task examined how words which form tonal minimal pairs in the neighbouring dialects of Tongeren and Hasselt are realized in the dialect of Borgloon. The results attest substantial variation between the dialects, deriving both from variation in the distribution of accent 1 and accent 2 and from variation in the lexemes used. In the second task, tonal minimal pairs were embedded in different prosodic contexts. The results suggest that the dialect of Borgloon, like other Limburgian dialects, marks accent 2 by lexical tone, while accent 1 is lexically toneless. The Borgloon dialect differs, however, in marking accent 2 by a sequence of two lexical tones rather than by a single one. 1. Introduction 1 Borgloon is a small town in the Belgian province of Limburg, with a population of some 10,000. It lies 15 km east of the provincial boundary with Brabant and 7 km north of the Dutch-French language border. The dialect of Borgloon belongs to the West-Limburgian dialect group in Belgium (Goossens 1965). West Limburg forms the westernmost part of the South Low Franconian and Central Franconian tone accent area covering the Belgian and Dutch provinces of Limburg, the northeast of the Belgian province of Liège, the southwest of North Rhine-Westphalia, the northern Palatinate, and Luxemburg (see figure 1).

2 170 Jörg Peters Belgian Limburg Netherlands Venlo Roermond Dutch Limburg Borgloon Belgium Hasselt Tongeren Province of Liège Maastricht Cologne North Rhine-Westphalia Germany Rhine Luxemburg Trier France Palatinate Figure 1. The South Low Franconian and Central Franconian tone accent area (marked by dotted line). The Limburgian dialects have a word accent contrast comparable to the lexical tone contrast in Swedish and Norwegian. In Limburgian dialectology, this contrast is traditionally known as the distinction between stoottoon ( push tone ) and sleeptoon ( dragging tone ) (Grootaers 1910: 123), here referred to as accent 1 and accent 2, respectively. The contrast is used both to distinguish between lexemes, such as 1 ball (dance) and 2 ball (toy), and between grammatical forms of a single lexeme, such as 1 day-nom-pl and 2 day-nom-sg (superscripts mark accent class). 2 Previous studies on the neighbouring West-Limburgian dialects of Tongeren and Hasselt (Grootaers 1910; Grootaers and Grauls 1930; Heijmans 1999; Gussenhoven 2004: 245; Peters 2006) suggest that these dialects differ from the East-Limburgian dialects of Maasbracht, Venlo and Roermond (Hermans 1985, 1994; Bruce and Hermans 1999; Gussenhoven and van der Vliet 1999; Gussenhoven 2000) in at least two respects. First, in the eastern dialects a steep

3 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 171 fall of accent 1 contrasts with high-level or mid-rising pitch of accent 2, while in the western dialects a non-steep fall of accent 1 contrasts with a rise of accent 2 which reaches its peak only after the accented syllable. Second, both the phonetic realisation of the contrast and the distribution of the contrast over different syllable types suggest that in the eastern dialects the contrast is morabound, while in the western dialects moraic structure does not affect the realisation of the contrast. In both eastern and western dialects, the word accent contrast has been analyzed as a contrast between a lexical tone in accent 2 and no lexical tone in accent 1 (for an alternative analysis see Hermans 1985, 1994 and Bruce and Hermans 1999). In eastern dialects, accent 2 bears a lexical H tone, leading to a contrast between the tonal sequences HH and HL when combined with a high pitch accent of declarative contours (Gussenhoven and van der Vliet 1999; Gussenhoven 2000). In the western dialects, accent 2 bears a lexical L tone. As only one tone at a time can associate to the accented syllable in the latter dialects, the lexical L tone of accent 2 prevents the focal H tone from associating to the accented syllable, and thus delays the focal peak (Heijmans 1999; Peters 2006). The purpose of the present paper is to show that the dialect of Borgloon has a western-type word accent contrast but is still distinct from the dialects of Tongeren and Hasselt in marking accent 2 by a sequence of two lexical tones rather than by a single one. The organization of this article is as follows. Section 2 reports observations on how words which form tonal minimal pairs in neighbouring dialects are realized in the dialect of Borgloon (2.1) and on the interaction of the tonal contrast with sentence intonation (2.2). Section 3 proposes a tonal analysis. Section 4 concludes with a few remarks on the relevance of the dialect of Borgloon to a tonal typology of Limburgian dialects. 2. Data Two reading tasks were carried out. The purpose of the first task was to provide background information on the use of the accentual contrast and to make tonal minimal pairs available for further examination. A tonal minimal pair consists of a pair of segmentally identical word forms that differ by accent class only. The purpose of the second task was to determine how speakers realize these minimally different word forms in different prosodic contexts.

4 172 Jörg Peters 2.1. Tonal minimal pairs Despite its important role in political and cultural history, Borgloon is an almost blank area on the dialectological map of Belgian Limburg. Neither a dictionary nor a grammar is available for the local dialect. Information mainly comes from general dictionaries of Limburgian and from a few descriptions of the West-Limburgian district Lonerlands, to which Borgloon belongs (Stevens 1952). 3 To obtain a suitable set of target words for further examination, we checked tonal minimal pairs from the dialects of Tongeren and Hasselt for equivalents in the Borgloon dialect. Using the dictionaries of the Tongeren dialect by Stevens (1986) and of the Hasselt dialect by Staelens (1989) as our main sources, we selected 90 minimal pairs (56 from Tongeren, 25 from Hasselt, and 9 attested for both places). The tonal minimal pairs were inserted in non-final and final position of carrier sentences such as given in (1). (1) a. Mareike has said. b.. Mareike says. Combining each carrier sentence with each member of the 90 minimal pairs we obtained 360 test sentences. A 70-year-old male native speaker (M1) was presented with one sentence at a time in literary orthography adapted from Oris (2000) and asked to read it out in a natural fashion with at least one repetition. The results attest a substantial amount of variation between Borgloon, Tongeren, and Hasselt. The Borgloon speaker reproduced less than 20% of the minimal pairs from Tongeren and less than 30% of the minimal pairs from Hasselt as minimal pairs in his own dialect. In many cases, the Borgloon equivalents differed by vowel quality, especially when compared with minimal pairs from Hasselt, as illustrated in (2). (2) Tongeren Hasselt Borgloon day-pl day-sg b egg-dim oaken

5 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 173 About the same number of minimal pairs from Tongeren and Hasselt were reproduced by pairs of word forms differing both by tonal and non-tonal features, as illustrated in (3). (3) Tongeren Hasselt Borgloon a rabbit-pl rabbit-sg b tooth-dim aunt-dim c. 1 1 cupboard 2 2 candle d. 1 1 Easter 2 2 portion The Borgloon word forms in (3a, b) differ by both accent class and vowel quality, while the word forms in (3c, d) differ by both accent class and quantity. Stevens (1952) notes that speakers from Lonerlands lengthen short vowels before /s/ and (or//). The examples in (3c, d), however, suggest that lengthening before /s/ is not independent of accent class. Moreover, our speaker combined an accentual contrast with a quantity contrast before /l/ as well, as shown in (4). (4) Tongeren Hasselt Borgloon 1 1 handle-pl 2 2 handle-sg In some cases, the speaker did not preserve a phonological contrast at all. Examples are given in (5). (5) Tongeren Hasselt Borgloon a stone-sg stone-pl b. 1 2 to work 2 2 weather-dim Figure 2 summarizes the relative frequencies of the Borgloon counterparts to tonal minimal pairs from Tongeren and Hasselt. The bars show how many tonal minimal pairs from the neighbouring dialects the Borgloon speaker distinguished by accent only, by accent plus vowel quality, by accent plus quan-

6 174 Jörg Peters tity, etc. The variation attested for the dialect of Borgloon in figure 2 is remarkable given that Tongeren is only 9 kilometers away from Borgloon and Hasselt 14 kilometers % Accent Accent + vowel quality Source: Tongeren Accent + quantity Other contrast Source: Hasselt No contrast (both acc 1) No contrast (both acc 2) No equivalents Figure 2. Relative frequencies of Borgloon counterparts to tonal minimal pairs from Tongeren (N = 65) and Hasselt (N = 34), classified according to type of contrast Pitch contours Materials In the second reading task, we presented speakers with sentences varying by pragmatic condition ( declarative, interrogative (yesno questions), continuative ) and distance of the target word to the end of the intonational phrase (IP) (final vs. non-final). In the declarative condition, the target words occurred in nuclear and postnuclear final and non-final position, and in prenuclear position. In the interrogative and continuative conditions, the target words were restricted to the nuclear final and non-final position, yielding a total of nine conditions. As we found no phonologically relevant difference between the accentual contrast in prenuclear and nuclear position, we do not address the prenuclear condition separately.

7 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 175 We created minimally different pairs of test sentences with the help of the tonal minimal pairs in (6). 4 (6) a. 1 2 leg-pl leg-sg b. 1 2 ball (dance) ball (toy) c. 1 2 day-pl day-sg d. 1 2 egg-dim oaken Note that the stressed syllables of the words in (6) contain at least two sonorant moras, that is a long vowel, a diphthong or a short vowel plus a sonorant consonant (m, n, ). We further included near-minimal pairs whose accented syllables contain an ambisyllabic second mora (7a) or a single sonorant mora (7b). (7) a 1 2 inside to bind b. 1 2 tax dachshund The test set containing monosyllabic target words consisted of 72 test sentences (4 minimal pairs x 2 accent classes x 9 conditions). Leaving out the final condition for disyllabic target words, we obtained an additional set of 20 test sentences (2 minimal pairs x 2 accent classes x 5 conditions). We recorded data from two male speakers (M1 and M2) and one female speaker (F1). All speakers were native of the local dialect, with ages ranging between 65 and 85 years. They were presented with one experimental sentence at a time and asked to read it out in a natural fashion with at least one repetition. Most sentences where preceded by an introductory statement or a question. Introductory statements were read by the same speaker. Questions were read by a second native speaker. The resulting data base contained 794 utterances Nuclear non-final condition Figure 3 gives diagrammatic representations of typical pitch contours on accent 1 and accent 2 words in nuclear non-final position. In both declaratives and interrogatives, accent 2 differs from accent 1 in two respects. First, the peak occurs later in accent 2 words than in accent 1 words. While the peak of accent 1 words occurs on the second half of the accented syllable, the peak of accent 2 words occurs after the accented syllable, mostly on the last syllable before the next stress. Second, the rising movement to the pitch peak differs according to accent class. In accent 1 words the pitch rises

8 176 Jörg Peters gradually up to the peak on the second half of the accented syllable. The accented syllable of accent 2 words, in contrast, bears mid-level pitch or pitch rising from mid level. Continuatives pattern like declaratives and interrogatives except that the contour ends with high level pitch, as shown in the right panel of figure 3. Declarative/Interrogative Continuative * * Figure 3. Pitch contours in declarative, interrogative, and continuative nuclear non-final condition. Solid lines mark accent 1 and dotted lines accent 2. The boxes indicate the position of the nuclear syllable (*). Durational differences are ignored. There is an alternative realisation of accent 2 both in declaratives/ interrogatives and in continuatives. In this case the nuclear syllable bears falling-rising pitch rather than mid-level pitch or pitch rising from mid level, as illustrated by the dotted lines in figure 4. As a consequence, the pitch contours of declaratives and interrogatives show two peaks in accent 2. The first peak occurs near the beginning of the accented syllable and the second after the nuclear syllable. Even though the first peak sometimes moves into the vowel, the accented syllable is generally perceived with low or falling pitch rather than with high pitch. Figures 5 and 6 give F 0 contours for both realisations of accent 2 in declaratives (the target words are underlined). 5 Declarative/Interrogative Continuative * * Figure 4. Alternative realisation of accent 2 in declaratives, interrogative, and continuative nuclear non-final condition.

9 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon !" Time (sec) Figure 5. / 2 / in declarative nuclear non-final condition (they have with a ball (toy) dance-pp They danced with a ball ). Speaker M !" Time (sec) Figure 6. Second variant of / 2 / in declarative nuclear non-final condition. Speaker F1. 6 The second variant of accent 2 is less frequent than the first one and is missing in speaker M2. The other two speakers used both variants regardless of the target word chosen. In some instances of declaratives, the nuclear accent was found to be downstepped leading to lowered pitch and compressed pitch range in both accent 1 and accent 2 syllables. We also note that accent 2 syllables in nuclear non-final position are some 25% longer than accent 1 syllables, regardless of the pitch contour used. No durational difference was observed in the minimal pairs containing less than two tautosyllabic sonorant moras, that is in / 1 / / 2 / and / 1 / / 2 / (cf. section 2.2.5). Speaker M2 tended to replace /daks 2 / by /daas 2 /.

10 178 Jörg Peters Nuclear final condition Accent 1 in nuclear final position resembles accent 1 in nuclear non-final position, except that in declaratives and interrogatives the final falling pitch often fails to reach fully low level (see left panel of figure 7). In accent 2, lengthening of the accented syllable, which exceeds that of accent 2 syllables in nonfinal position, allows speakers to realize both mid-level pitch and a rise on the accented syllable, whereas the final falling movement is truncated. In continuatives, accent 1 syllables bear rising pitch, while the pitch contour on accent 2 syllables can hardly be distinguished from the pitch contour on accent 2 syllables in declaratives and interrogatives (see right panel of figure 7). Declarative/Interrogative Continuative * * Figure 7. Pitch contours of declaratives, interrogatives, and continuatives with the target word in nuclear final position. Durational differences ignored. As in the non-final condition, speakers M1 and F1 produced accent 2 alternatively with falling-rising pitch on the accented syllable. These contours are illustrated by the dotted lines in figure 8. Declarative/Interrogative Continuative * * Figure 8. Alternative realisation of accent 2 in nuclear final position of declaratives, interrogatives, and continuatives. Figures 9 and 10 give examples of both variants of accent 2 in declaratives.

11 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon "" Time (sec) Figure 9. / 2 / in declarative nuclear final condition (there=is a ball There is a ball (toy) ). Speaker M "" Time (sec) Figure 10. / 2 / in declarative nuclear final condition. Second variant. Speaker M1. As the final fall is partially truncated, neither the first nor the second variant of accent 2 in continuatives can clearly be distinguished from the respective variants of accent 2 in declaratives and interrogatives. All three speakers used another two variants of accent 2 in declaratives and interrogatives, which differ more clearly from the two variants mentioned. In both cases, the pitch peak occurs on the first half of the accented syllable rather than on the second half. In the first case, the peak precedes mid-level pitch (see figure 11). In the second case, the peak precedes a falling-rising movement (see figure 12).

12 180 Jörg Peters "" Time (sec) Figure 11. / 2 / in interrogative nuclear final condition (dance they with a ball Are they dancing with a ball (toy)?). Third variant. Speaker M "" Time (sec) Figure 12. / 2 / in interrogative nuclear final condition. Fourth variant. Speaker F1. Figure 13 summarizes the four realisations of accent 2 in declaratives and interrogatives. The diagrammatic representations suggest that the relation between variant c and d is comparable to the relation between variant a and b. In both cases, mid-level pitch alternates with falling-rising pitch. Similarly, the relation between variant a and c is comparable to the relation between b and d. In both cases, the (highest) peak moves from the second half to the first half of the accented syllable.

13 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 181 a b c * d * * * Figure 13. Four variants of accent 2 in nuclear final position of declaratives and interrogatives. As we shall see in section 3.2, all variants shown in figure 3 can be derived from the same tone sequence Post-nuclear conditions The accentual contrast is preserved on postnuclear words and thus is independent of the presence of a pitch accent. Postnuclear accent 2 syllables of declaratives and interrogatives are low pitched, while accent 1 words follow the course of the overall pitch contour, as illustrated by / 1 / and / 2 / in figures 14 and 15 (nuclear words are given in bold type). Closer inspection of the F 0 contour in figure 15 reveals that the presence of the accent 2 word not only adds a low pitch target to the accented syllable but also a high target to the following unstressed syllable. Replacing an accent 1 word by an accent 2 word in postnuclear position thus does not only add an elbow to the gradually falling pitch movement after the nuclear syllable but also a peak which replicates in miniature the postponed peak of nuclear accent 2 words. A similar implementation of the contrast was found in postnuclear final position. Monosyllabic accent 2 words in final position show a (downscaled) falling-rising movement, while accent 1 words again follow the course of the overall pitch contour.

14 182 Jörg Peters zin x blu m 1 d ## Time (sec) Figure 14. / 1 / in declarative postnuclear non-final condition (there are still flowers inside in the kitchen there are flowers left in the kitchen ). Speaker M $ x a d blu m 2 d ## Time (sec) Figure 15. / 2 / in declarative postnuclear non-final condition (I must still only the flowers bind in the kitchen Just the flowers are left for me to bind in the kitchen ). Speaker F1. Test sentences with target words in postnuclear position of continuatives were not included in the reading task. Data from the continuative nuclear nonfinal condition show, however, that accent 2 words in postnuclear position of continuatives likewise add a low and a high target. The falling-rising pitch on the stressed syllable of / 2 / in figure 16 illustrates this observation.

15 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon ds 2! z p Time (sec) Figure 16. / 2 / in continuative postnuclear condition (at first have they on a ball dance-pp... At first they danced at a ball...). Speaker F1. When more than one accent 2 word follows the nuclear word, each of these words adds a low and a high target, as illustrated by / 2 / and / 2 / in figure d 2! 1 2 $ Time (sec) Figure 17. / 2 / and / $ 2 / in continuative postnuclear condition (at first have her leg- PL pain do-pp... At first did her legs hurt...). Speaker F Distribution of the contrast The accentual contrast is confined to stressed syllables, but apparently there are no minimal requirements on the number of sonorant moras a syllable must contain to bear either accent 1 or accent 2. All three speakers made a tonal distinction between / 1 / and/ 2 /, whose accented syllables contain one tautosyllabic and one ambisyllabic sonorant mora each (see figures 14 and 15). Two speakers (M1 and F1) likewise distinguished tonally between / 1 / and / 2 /, which contain only one sonorant mora. Figures 18 and 19 illustrate such a contrast in nuclear non-final position of declaratives.

16 184 Jörg Peters ! Time (sec) Figure 18. / 1 / in declarative nuclear non-final condition (I have my tax yet not pay-pp I did not pay my tax yet ). Speaker M ! 2 %% Time (sec) Figure 19. / 2 / in declarative nuclear non-final condition (and I have my dachshund also yet not pay-pp And I did not yet pay my dachshund either ). Speaker M1. For / 2 /, falling-rising pitch on the accented syllable (see figure 4) is attested as well, as shown in figure 20. 7

17 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon &! 2 $ Time (sec) Figure 20. / 2 / in declarative nuclear non-final condition. Second variant (... for the dachshund must pay... must pay for the dachshund ). Speaker F1. 3. Phonological analysis Grootaers (1910) and Grootaers and Grauls (1930) characterized the accentual contrast in Tongeren and Hasselt as involving both a pitch contrast and an intensity contrast. More recent analyses by Heijmans (1999), Gussenhoven (2004: 245) and Peters (in press) propose purely tonal analyses, as no systematic differences in intensity were observed in their speakers. The same holds for the present data from Borgloon. We therefore propose a tonal analysis of the accentual contrast. We use the autosegmental-metrical model as a theoretical framework, that is we assume a tonal structure that is represented separately from the segmental string and consists of a linear sequence of local events. These events are lexical and postlexical tones, which associate with prosodic units or align with the edges of those units. Comprehensive introductions to this framework are given by Ladd (1996) and Gussenhoven (2004) Lexical pitch accents The analyses of the dialects of Tongeren and Hasselt by Heijmans (1999), Gussenhoven (2004), and Peters (in press) assume that accent 2 words come with a lexical low tone which is prelinked to the stressed syllable, whereas accent 1 words are lexically toneless. The accentual contrast is privative in the sense that only one of its members is specified for tone and represents a

18 186 Jörg Peters marked option. These analyses account for both a delayed peak on accent 2 syllables in nuclear position and a low target on accent 2 syllables in postnuclear position, where no pitch accent is present. The examples in (8) illustrate the Hasselt analysis for / / (cheese-dim) and / / (stocking-dim) in nuclear non-final position (after Peters [in press] lexical tones are given in bold type, } designates the final boundary of an IP). 8 (8) a. { 2 1 } L H* L no but I must also a few cheese-pl-dim have No but I also need some little cheeses b. { 2 2 } L*L H L no but I must also a few stocking-pl-dim have No but I also need some little stockings There is no reason to assume for Hasselt smaller tone-bearing units (TBUs) than the stressed syllable, as monomoraic and bimoraic syllables do not differ regarding their capacity to bear tones. Assuming the syllable rather than the mora to be the TBU has the further advantage that it motivates the observed peak delay in nuclear accent 2 words. When the syllable is the TBU, the nuclear syllable provides only a single docking site for tones to associate. In accent 2 words, this site is already occupied by the lexical L tone. Therefore, the postlexical H tone cannot associate to the nuclear syllable. The examples in (9) illustrate analogous analyses for 1 and 2 in postnuclear position (again, the nuclear word is given in bold type). (9) a. { 2 &' 1 } L L*L H L Pierre has for a hen care-pp Pierre cared for a hen

19 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 187 b. { 2 &' 2 } L L*L H L L Pierre has for them care-pp Pierre cared for them Both in the nuclear and in the postnuclear condition, the assumption of a lexical L tone for accent 2 accounts for the observation that accent 2 words always come with a low target. The postnuclear implementation of accent 2 is vital for arguing that the accentual contrast in Hasselt is a privative contrast. The similar melodies on accent 1 and accent 2 words in nuclear position allow for an alternative analysis, which reduces the accentual contrast to a difference in the timing of the accentual gesture. This timing difference could be attributed to a change in the internal structure of the pitch accent such that the accentual contrast is interpreted as the contrast between LH* in accent 1 and L*H in accent 2. The pitch difference in postnuclear position, however, cannot be captured in terms of a timing difference. The postnuclear realisation of the accentual contrast in Borgloon warrants a privative analysis for this dialect as well. But the Hasselt analysis is not transferable to Borgloon without adjustments. Assuming a lexical low tone for accent 2 and no tone for accent 1 would fail to account for the fact that accent 2 words add two additional targets to the pitch contour rather than one. Evidence comes from the realisation of the accentual contrast in postnuclear position. Both in non-final and final postnuclear position, accent 2 words are not only lower than accent 1 words, but the low pitch is also followed by a rise to a lower-scaled postnuclear pitch peak, as the comparison of / 2 / in figure 15 with /bn 1 / in figure 14 shows. This difference suggests that accent 2 words come with both a low tone associating to the stressed syllable and a floating high tone afterwards. Additional evidence for two tonal targets on accent 2 words is found in nuclear position. In the contour variants of accent 2 words shown in figure 4, a fall-rise precedes the rising movement to the final peak, which can hardly be accounted for by adding a single low tone. Again, adding a sequence of a low and a high tone accounts better for the contours observed, as shown in (10) (for the moment, postnuclear accent 2 will be ignored).

20 188 Jörg Peters (10) {! 2 2 } LHL H L they have with a ball danced they danced with a ball Another drawback of simply transferring the Hasselt analysis to Borgloon would be its failure to motivate the presence of mid-level plateaus in the first contour variants of accent 2 shown in figure 3. Assuming the tonal sequence LHLH instead with the second L associating to the accented syllable allows interpreting the mid-level plateau as an alternative phonetic implementation of the medial tone sequence HL. This interpretation gets further support by the observation that both variants of accent 2 represent only the endpoints of a continuum encompassing many intermediate realisations between falling and mid-level pitch on the nuclear syllable. Assuming accent 2 to be specified for L and H, at least three alternative analyses of accent 2 in nuclear position are possible. The lexical tone sequence LH might precede the pitch accent, as in (11a), or follow after it, as in (11b) and (11c), with only one tone associating to the accented syllable, as in (11a) and (11c), or two tones, as in (11b). (11) a. {! 2 2 } LHL* H L b. {! 2 2 } LH*L*H L c. {! 2 2 } LH L* H L

21 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 189 The realisation of accent 2 in postnuclear position suggests that lexical L is prelinked to the accented syllable of the accent 2 word. According to (11a), the leading tone of the pitch accent would displace the prelinked lexical tone from its docking site on the accented syllable. This behaviour would be unexpected for two reasons. First, in view of the dialect of Hasselt, where the lexical L tone prevents the postlexical H tone from associating to the nuclear syllable, we would not expect that a postlexical tone displaces a prelinked lexical tone from its docking site. Second, it is also unexpected that the leading tone of the postlexical pitch accent LH* rather than its starred tone (the tone that is designated for associating to the stressed syllable) competes with the lexical tone for the same docking site. 9 By representing accent 2 as in (11b) we avoid these problems, but (11b) has two other drawbacks. First, two tones would associate to the same syllable even in syllables containing a single sonorant mora (see section 2.2.5). In this respect, the dialect of Borgloon would differ from all other known Limburgian dialects, where the association of two tones to the same syllable is either avoided or restricted to syllables containing two sonorant moras. Second, the first high target on nuclear accent 2 syllables mostly occurs near the onset of the accented syllable (see section 2.2.2). This again is untypical for associated tones whose targets are usually located in the sonorant part of the rime. We therefore represent accent 2 as in (11c), that is we interpret the accentual contrast as a contrast between the presence and the absence of a bitonal lexical pitch accent L*H, which prevents the H tone of the postlexical pitch accent LH* from associating to the accented syllable Tonal system Based on the assumption of a bitonal lexical pitch accent that follows the postlexical pitch accent in nuclear position, most contour variants reported in section 2.2 can be characterized in a uniform way. We start with the assumption that the default pitch accent of the Borgloon dialect is LH*, which may be downstepped. H* accounts for the pitch peak that occurs on the accented syllable when specified for accent 1. The leading L tone accounts for the observation that usually a low target precedes the nuclear pitch peak, regardless of the preceding prosodic context. The final IP boundary is either specified by a low boundary tone (L ) or toneless (Ø ). Accordingly, two nuclear contours are available, LH*L and LH*Ø. In our data, LH*L (with or without downstep) was preferred in declaratives and interrogatives, whereas LH*Ø was confined to continuatives.

22 190 Jörg Peters If no accent 2 word follows after the nuclear syllable, the pitch of the LH*L contour gradually decreases until the end, while the LH*Ø contour comes with a high postnuclear plateau, as shown in (12). In the latter case, the second high target at the final IP boundary is specified by H*, which spreads rightwards (indicated by ). (12) a. L H* L b. L H* Ø Accent 2 words bear a lexical pitch accent L*H, with L being prelinked to the accented syllable. Accent 1 words are lexically toneless. As argued in section 3.1, the postlexical pitch accent LH* precedes the lexical pitch accent rather than follows it, resulting in the tonal sequence LHL*H in accent 2 words. The different sources of the first H and the second L do not prevent their phonetic targets from converging such that HL* is mostly implemented as mid-level pitch or halfway between falling and mid-level pitch. We account for this pattern by assuming an optional implementation rule that applies to HL* in the context L H (tones may be lexical or postlexical). We assume that the TBU is the stressed syllable, just as in Hasselt. There is no need to assume smaller TBUs such as the mora. On the one hand, no more than one docking site on the accented syllable is necessary to realize the accentual contrast. On the other hand, two of the three speakers were found to realize the accentual contrast on syllables containing a single sonorant mora and all speakers preserved the contrast on syllables whose second sonorant mora is ambisyllabic (see section 2.2.5). There is no difference between accent 1 words and unaccented words. Therefore, the distinction between accent 2 and accent 1, which refers to the traditional distinction between sleeptoon and stoottoon, can likewise be interpreted as a distinction between accent and no accent. For this reason, we did not mark accent 1 in previous examples except for illustrative purposes when referring to the tonal minimal pairs examined. We may also conclude that in stressed syllables containing a single sonorant mora the contrast is not neutralized, as in East-Limburgian dialects. In the dialect of Borgloon, all stressed syllables are either marked for accent (= accent 2) or unmarked. The examples in (13) illustrate the proposed analysis for the declarative nuclear non-final condition (for 13b and 13c compare figures 5 and 6, respectively). 10

23 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 191 (13) a. Accent 1 {! 1 2 } L LH* L H* L*H(L ) b. Accent 2, var. 1 {!" 2 2 } L LH* LHL*H L*H(L ) c. Accent 2, var. 2 {!" 2 2 } L LH* LHL* H L*H(L ) In all sentences, the accent 2 word / 2 / occurs in postnuclear final position. This word adds the lexical pitch accent L*H, which is downscaled when compared with the nuclear accent and leaves no time for realizing the final low boundary tone. (14) shows that continuative utterances can be represented in the same way, except that the final boundary tone is missing. The postnuclear falling-rising movement again can be attributed to the lexical pitch accent on postnuclear / 2 /. (14) a. Accent 1 {! 1 2 } L LH* L H* L*HØ b. Accent 2, var. 1 {!" 2 2 } L LH* LHL* H L*HØ c. Accent 2, var. 2 {!" 2 2 } L L H* LH L* H L*HØ

24 192 Jörg Peters (15) and (16) give analogous representations for the nuclear final condition of declaratives and continuatives, respectively (for 15b and c see figures 9 and 10, respectively). (15) a. Accent 1 {" 1 } L L H*L b. Accent 2, var. 1 {"" 2 } L LHL*H(L ) c. Accent 2, var. 2 {"" 2 } L LHL*H(L ) (16) a. Accent 1 { 2 %" 1 } L LH* L* H L H*Ø b. Accent 2, var. 1 { 2 "" 2 } L LH* L* H LHL*HØ c. Accent 2, var. 2 { 2 "" 2 } L LH* L* H LHL*HØ The variants of the declarative and interrogative nuclear final condition illustrated in figures 11 and 12 are the only variants that cannot be captured by the tonal sequence LHL*H. They can be derived from the variants in (15b) and (15c), however, when we assume that the postlexical H tone rather than the lexical L tone associates to the nuclear syllable, as shown in (17).

25 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 193 (17) a. { 2 "" 2 } L L* H LH*LH(L ) b. { 2 "" 2 } L L* H LH*LH(L ) The variation between (17a) and (17b) can be attributed to an optional implementation rule which implements LH in the context LH* L as midlevel pitch, as in (17a). Finally, (18) gives representations of the accentual contrast in postnuclear non-final position (cf. figures 14 and 15). (18) a. { zin nx 1 d!! 2 } L LH* L H* L*H(L ) b. { 2 $x a d 2 d!! 2 } L L* H L H* L H* L* H L*H(L ) In section 2.2, we mentioned durational differences between accent 1 and accent 2 syllables. As a rule, accent 2 syllables are longer than accent 1 syllables. Durational differences are larger in nuclear than in postnuclear position and in final than in non-final position in the IP. As accent 2 adds two tones to the tonal string, we may attribute the lengthening of accent 2 syllables to tonal crowding. In this way, we can also account for the different degrees of lengthening in nuclear and postnuclear position and in final and non-final position. The largest durational difference between accent 1 and accent 2 is found in nuclear final position of declaratives and interrogatives, where accent 2 syllables bear five tones (the last being the final boundary tone), whereas accent 1 syllables bear three tones.

26 194 Jörg Peters Alternatively, the lengthening of accent 2 syllables can be interpreted as a strategy to enlarge the contrast as it increases the distance between the pitch contours on accent 1 and accent 2 syllables. Accordingly, duration can be interpreted as an enhancing feature, which, unlike pitch, is not primary in the sense of directly expressing the phonological feature involved (Stevens and Keyser 1989; Gussenhoven and Peters 2004) Summary The tonal grammar of Borgloon can be summarized as in (19). (19) Basic grammar a. Nuclear pitch contours LH* L (declaratives and interrogatives) LH* Ø (continuatives) (with or without downstep of LH*) b. Lexical tone contrast Accent 1: lexically toneless Accent 2: L*H c. TBU = stressed syllable d. Phonetic implementation rules (optional): HL* mid-level pitch / L H LH mid-level pitch / LH* L Figure 21 summarizes the most frequent contour variants and their representations according to (19) on monosyllabic words in nuclear and postnuclear position. Figure 22 compares the third and fourth variant of accent 2 with accent 1 in nuclear final condition of declaratives and interrogatives (cf. figures 11 and 12). These are the only variants that presuppose a change of tonal structure in associating postlexical H rather than lexical L to the stressed syllable.

27 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 195 Declarative/Interrogative Continuative nuclear non-final * Acc 1 L H* L Acc 2 L H L* H L * Acc 1 L H* L Acc 2 L H L* H L * Acc 1 L H* L Acc 2 L H L* H L * Acc 1 L H * Ø Acc 2 L H L* H Ø nuclear final * Acc 1 L H* L Acc 2 L H L* H L * Acc 1 L H*Ø Acc 2 L H L*H Ø * Acc 1 L H* L Acc 2 L H L* H L * Acc 1 L H*Ø Acc 2 L H L* HØ postnuclear nonfinal * Acc 1 L H* L Acc 2 L H* L* H L postnuclear final * Acc 1 L H* L Acc 2 L H* L*H(L ) Figure 21. Realisation of accent 1 and accent 2 on monosyllabic words in different prosodic contexts. Durational differences are ignored.

28 196 Jörg Peters * Acc 1 L H* L Acc 2 L H* L H (L ) * Acc 1 L H* L Acc 2 L H* L H(L ) Figure 22. Alternative realisations of accent 1 and accent 2 in nuclear final condition. 4. Concluding remarks The dialect of Borgloon proved to be a typical West-Limburgian dialect by having a lexical tone contrast that is syllable-based rather than mora-based and realized with a lexical low tone rather than with a high tone on accent 2 syllables. In both respects, the dialect of Borgloon resembles the dialects of Hasselt and Tongeren 11 but differs from the East-Limburgian dialects of Maasbracht, Venlo, and Roermond. The dialect of Borgloon proved to be exceptional, however, by interpreting the contrast as an opposition between the presence and the absence of a bitonal sequence LH, which can be interpreted as a lexical pitch accent, with L associating to a stressed syllable and H being a trailing tone. The presence of two tonal targets can most clearly be seen in accent 2 words in postnuclear position, where no postlexical pitch accent interferes with lexical tones. The interpretation of the accentual contrast in Borgloon thus may extend previous typologies of Limburgian tonal dialects, which account for the tonal quality of the lexical tone and the order of lexical and postlexical tones but not for a variable number of lexical tones (Gussenhoven 1999; Gussenhoven and Bruce 1999). At first glance, the dialect of Borgloon looks very much like Stockholm Swedish as described by Bruce (1977, 1990). Bruce assumed for Stockholm Swedish bitonal lexical pitch accents, as we did for Borgloon. On the other hand, both dialects differ in important ways. First, according to Bruce s analysis both accent 1 and accent 2 bear a lexical pitch accent, which is HL* in accent 1 and H*L in accent 2, while in Borgloon the lexical pitch accent is restricted to accent 2. Second, the pitch contour on nuclear accent 2 words in Borgloon is more similar to Swedish accent 1 than to accent 2, as in Borgloon the lexical L tone rather than the preceding H tone associates to the accented syllable. This difference is shown by comparing the tonal representations in

29 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 197 (13c) and (20) (the representations in (20) are adopted from Gussenhoven (2004: 211), replacing association lines with star notation). (20) a. {med 1 anden} b. {med 2 anden} L H+L* H L L H*+LHL with the duck with the ghost Third, according to the given analyses the postponed peak in accent 1 and accent 2 of Stockholm Swedish and in accent 2 of the dialect of Borgloon is represented differently. In Stockholm Swedish, as in Tongeren and Hasselt, the postnuclear peak originates from focal H. For Borgloon, an alternative analysis was preferred suggesting that the lexical pitch accent follows the postlexical pitch accent. Accordingly, the postnuclear peak of accent 2 in Borgloon originates from a lexical H rather than from a focal H. It would therefore be an oversimplification to assume that the Borgloon dialect is more Scandinavian than other Limburgian dialects in all respects. But in any case, the tonal system of this dialect suggests that the typology of tonal dialects in Limburg needs to be extended by including a parameter that distinguishes between lexical pitch accents of different complexity. Notes 1. This work was carried out as part of the project Tonale Dialecten in het Nederlands, which was funded by the Vlaams-Nederlands Comité, a joint research foundation of the Fonds voor Wetenschappelijk Onderzoek Vlaanderen (FWO) and the Netherlands Organisation for Scientific Research (NWO). I thank my speakers for giving me their time and best efforts. I am also grateful to Ben Hermans and an anonymous reviewer for many helpful comments. 2. Traditional accounts mark accent class by diacritics that occur after the nucleus or sonorant rime. We place superscripts at the end of the accented syllable since the presence of the accentual contrast does not depend on segmental (or moraic) structure in the dialect of Borgloon (see sec. 3). 3. According to Xavier Staelens (Hasselt), the late J. Bellefroid prepared a dictionary of the Borgloon dialect in the 1990s but did not finish this work during his lifetime. Oris (2000) published a dictionary of the nearby village of Kortessem (Kotsove), which is also used by Borgloon speakers but lacks detailed phonological information. 4. An additional word pair included in the reading test, / to wash and / to grow, will be ignored as our speakers realized both words with accent 2.

30 198 Jörg Peters 5. Note that in the following illustrations we mark accent 1 only in target words (see also section 3.2). 6. Note that /in figure 6 is a variant of/figure 5 both meaning with.is also attested for Kortessem (Oris 2000). 7. The initial fall, which spans nearly 2.5 semitones, is too large to be attributed to a microprosodic influence. Moreover, from the presence of the voiced initial plosive we may expect lowering rather than rising of the initial pitch. 8. Note that we use the star to mark the association of a tone to a stressed syllable regardless of whether this tone is a lexical or postlexical tone. In (8b), the starred tone is the lexical tone, while the postlexical tones of the pitch accent LH remain unassociated. 9. Note that we use the star notation in two different ways. When we talk about pitch accents, such as LH*, without reference to its use in a given utterance, the star indicates that the starred tone is either prelinked to a stressed syllable, as in lexical pitch accents, or designated for associating to a stressed syllable. In tonal representations of a given utterance, as in (11c), we only mark those tones as starred that actually associate. In the case of (11c), postlexical LH derives from the default LH* pitch accent, but H* becomes unstarred H as it does not find a docking site to associate. 10. For ease of comparison, we uniformly write // rather than // for with. 11. The data reported by Heijmans (1999) do not provide conclusive evidence against a mora-based view for Tongeren, as pointed out by Gussenhoven (2004: 245f). At least in the older dialect, however, the accentual contrast has probably been syllable-based rather than mora-based. Evidence comes from the descriptions by Grootaers (1910, 1944) and Stevens (1986), who list numerous monomoraic accent 2 words including those minimally contrasting with monomoraic accent 1 words. References Bruce, Gösta 1977 Swedish word accents in sentence perspective. Lund: Gleerup Alignment and composition of tonal accents: Comments on Silverman and Pierrehumbert s paper. In Papers in Laboratory Phonology I: Between the grammar and physics of speech, John. Kingston, and Mary Beckman (eds.), Cambridge: CUP. Bruce, Gösta, and Ben Hermans 1999 Word tone in Germanic languages. In Word prosodic systems in the languages of Europe, H. van der Hulst (ed.), Berlin: Mouton de Gruyter. Goossens, Jan 1965 Die Gliederung des Südniederfränkischen. Rheinische Vierteljahresblätter 30: Grootaers, Ludovic J. J Het dialect van Tongeren. Eene phonetisch-historische studie [The dialect of Tongeren. A phonetic-historical study]. Leipzig: Harrassowitz.

31 Bitonal lexical pitch accents in the Limburgian dialect of Borgloon 199 Grootaers, Ludovic J. J., and Jan Grauls 1930 Klankleer van het Hasseltsch dialect. Leuven: de Vlaamsche Drukkerij. Grooaters, Ludovic J. J Bijdrage tot de Zuidlimburgsche tonologie. In Feestbundel H. J. van de Wijer, Deel II, Henri Draye (ed.), Leuven: Instituut voor Vlaamsche toponymie. Gussenhoven, Carlos 1999 Tone systems in Dutch Limburgian dialects. In Proceedings of the Symposium Cross-Linguistic Studies of Tonal Phenomena: Tonogenesis, Typology, and Related Topics, Shigeki Kaji (ed.), Tokyo: Tokyo University of Foreign Studies The lexical tone contrast of Roermond Dutch in optimality theory. In Prosody: Theory and experiment, Merle Horne (ed.), Dordrecht: Kluwer The phonology of tone and intonation. Cambridge: Cambridge University Press. Gussenhoven, Carlos, and Gösta Bruce 1999 Word prosody and intonation. In Word prosodic systems in the languages of Europe, Harry van der Hulst (ed.), Berlin: Mouton de Gruyter. Gussenhoven, Carlos, and Peter van der Vliet 1999 The phonology of tone and intonation in the Dutch dialect of Venlo. Journal of Linguistics 35: Gussenhoven, Carlos and Jörg Peters 2004 A tonal analysis of Cologne Schärfung. Phonology 21: Heijmans, Linda 1999 The representation of the Tongeren lexical tone contrast. Manuscript. LSA Linguistic Institute. Summer Hermans, Ben 1985 Het Limburgs en het Litouws als metrisch gebonden toontalen. Spektator 14, The composite nature of accent: With case studies of the Limburgian and Serbo-Croatian pitch accents. Tilburg: Katholieke Universiteit Brabant. Ladd, D. Robert 1996 Intonational phonology. Cambridge: Cambridge University Press. Oris, Jef 2000 Diksjenêr van Kotsove. Kortessem: Gemeentebestuur. Peters, Jörg 2006 The dialect of Hasselt. Journal of the International Phonetic Association 36: (in press) Tone and intonation in the dialect of Hasselt. Linguistics. Staelens, Xavier 1989 Dieksjenèèr van 't (H)essels. Nederlands Hasselts Woordenboek. Hasselt: de Langeman [3rd rev. edition].

32 200 Jörg Peters Stevens, André 1952 Structuur en historische ondergrond van het Haspengouws taallandschap. In: Het oude land van Loon. Hasselt: De Limburgse drukkerijen Túngërsë Diksjënéér. Woordenboek van het Tongers. Tongeren: Vanormelingen. Stevens, Kenneth, and Samuel J. Keyser 1989 Primary features and their enhancement in consonants. Language 65: