Academia.eduAcademia.edu

Figure 3 – uploaded by Paul Boersma

See full PDFdownloadDownload figure
If Middle Limburgian had the default phonetic interpretation of the link between segments and tones (§4.3), *laupan must have been pronounced as [l5wpan] or [lswpdn] depending on the intonation contour of the sentence. Aligned with H*L, the words in (16) sounded like [kléjind] , [bféjt], [feeqvers], [swpan], and [déswrh], with the usual restriction of not sounding tone on voiceless segments. Se a Ee Se FE ee ee Oe Aen: LMM nS, See Memes ees Mien eee meee mae Structurally, the diphthongs *ei, *oey, and *au can be interpreted as monomoraic if we regard them as the underlying VC sequences /ej/, /oey/ and /aw/, and if coda consonants are still non-moraic. Conversely, if the diphthongs are VC sequences, they side with other VC sequences in not requiring lengthening of the first part in open syllables, and in being interpreted by learner as equally long as the lengthened vowels in open syllables, i.e. as monomoraic. We see some example words in (16), which assumes, as before, default left-to-right tone assignment on moras.

Figure 3 If Middle Limburgian had the default phonetic interpretation of the link between segments and tones (§4.3), *laupan must have been pronounced as [l5wpan] or [lswpdn] depending on the intonation contour of the sentence. Aligned with H*L, the words in (16) sounded like [kléjind] , [bféjt], [feeqvers], [swpan], and [déswrh], with the usual restriction of not sounding tone on voiceless segments. Se a Ee Se FE ee ee Oe Aen: LMM nS, See Memes ees Mien eee meee mae Structurally, the diphthongs *ei, *oey, and *au can be interpreted as monomoraic if we regard them as the underlying VC sequences /ej/, /oey/ and /aw/, and if coda consonants are still non-moraic. Conversely, if the diphthongs are VC sequences, they side with other VC sequences in not requiring lengthening of the first part in open syllables, and in being interpreted by learner as equally long as the lengthened vowels in open syllables, i.e. as monomoraic. We see some example words in (16), which assumes, as before, default left-to-right tone assignment on moras.

Related Figures (7)

The recipe for creating these structures can thus be formulated as follows:
The recipe for creating these structures can thus be formulated as follows:
8 : . ‘ : ‘ Sentence-finally, the L would either have to be deleted or (more interestingly) incorporated somehow into the last mora (a case of tone crowding).
8 : . ‘ : ‘ Sentence-finally, the L would either have to be deleted or (more interestingly) incorporated somehow into the last mora (a case of tone crowding).
Foreshadowings. Given the diversion between non-high monophthongs on hand, and diphthongs and high vowels on the other hand, we can now already the one see the contours of an explanation for the behaviour of the words in §1 and §2: moderr fl5apon ‘large’ has a moving tone because it has an original long vowe *slaapoan), while breit ‘broad’ and lajk ‘long’ have level tones because they (MLt have ar original diphthong or short vowel. But we can also see that several issues still have to be settled: the apparent bimoraicity of modern breit and layk is related to a MLb process of open syllable lengthening (§5), and the fact that the moving tone in kniin ‘rabbits’ i: different from the level tone in kniin ‘rabbit’ is related to a later process of fina drop (87). schwe If the MLb long high vowels were monomoraic, there are still two possibilities for their phonological surface representation: they were phonologically simple short vowels (/i/, /y/, /u/), or they were, like the diphthongs, VC structures (/ij/, /yy/, /uw/). It is unlikely that *ii and *uu were short, since their later reflexes before /st/ are long (Rm. liiston ‘lists’, vutisten ‘fists’), whereas the later reflexes of the short vowels before /st/ are still short (Rm. kastan ‘cupboards’, kastan ‘cost’, kestan ‘chests’, koesten ‘coasts’). So it is likely that *ii and *uu were diphthongal VC structures (with non- moraic C). Liberman (1999) actually argues that *ii and *uu were the diphthongs /ij/ and /uw/ in the whole West-Germanic area. This would explain the fact that when later the lengthening of vowels in open syllables started to crowd the vowel space, the high vowels changed towards [aj] and [aw] independently in at least three unconnected areas (England, Brabant, and Bavaria, ultimately leading to the diphthongs in ‘wine’ and ‘house’ in standard English, Dutch, and German). The fact that none of the Scandinavian languages has developed this kind of diphthongs, despite the same lengthening of vowels in open syllables, should then be attributed to the idea that the Scandinavian high vowels were genuinely bimoraic /i:/ and /u:/. Thus, an explanation for the difference between Lithuanian and Limburgian is given in the structures in (17). hand, and diphthongs and high vowels on the other hand, we can now already see the
Foreshadowings. Given the diversion between non-high monophthongs on hand, and diphthongs and high vowels on the other hand, we can now already the one see the contours of an explanation for the behaviour of the words in §1 and §2: moderr fl5apon ‘large’ has a moving tone because it has an original long vowe *slaapoan), while breit ‘broad’ and lajk ‘long’ have level tones because they (MLt have ar original diphthong or short vowel. But we can also see that several issues still have to be settled: the apparent bimoraicity of modern breit and layk is related to a MLb process of open syllable lengthening (§5), and the fact that the moving tone in kniin ‘rabbits’ i: different from the level tone in kniin ‘rabbit’ is related to a later process of fina drop (87). schwe If the MLb long high vowels were monomoraic, there are still two possibilities for their phonological surface representation: they were phonologically simple short vowels (/i/, /y/, /u/), or they were, like the diphthongs, VC structures (/ij/, /yy/, /uw/). It is unlikely that *ii and *uu were short, since their later reflexes before /st/ are long (Rm. liiston ‘lists’, vutisten ‘fists’), whereas the later reflexes of the short vowels before /st/ are still short (Rm. kastan ‘cupboards’, kastan ‘cost’, kestan ‘chests’, koesten ‘coasts’). So it is likely that *ii and *uu were diphthongal VC structures (with non- moraic C). Liberman (1999) actually argues that *ii and *uu were the diphthongs /ij/ and /uw/ in the whole West-Germanic area. This would explain the fact that when later the lengthening of vowels in open syllables started to crowd the vowel space, the high vowels changed towards [aj] and [aw] independently in at least three unconnected areas (England, Brabant, and Bavaria, ultimately leading to the diphthongs in ‘wine’ and ‘house’ in standard English, Dutch, and German). The fact that none of the Scandinavian languages has developed this kind of diphthongs, despite the same lengthening of vowels in open syllables, should then be attributed to the idea that the Scandinavian high vowels were genuinely bimoraic /i:/ and /u:/. Thus, an explanation for the difference between Lithuanian and Limburgian is given in the structures in (17). hand, and diphthongs and high vowels on the other hand, we can now already see the
(42) Surface structures of words with and without final schwa, voiceless version In these graphs syllable boundaries are represented as periods. We know that the loss of schwa ultimately led to a change in phonologica structures across generations: original strdote ‘street’ and d5at ‘deed’ consisted o three and two moras, respectively, as shown in (7) and (14), while present-day Jtrso and d5at rhyme perfectly, i.e., they must consist of the same number of moras, which will assume to be two. As in the account of Open Syllabe Lengthening, the account o Schwa Drop will involve three elements: the phonological structure maintained by the first generation (cf. (8)), the variation in pronunciation allowed by this structure, and thi reinterpretation of selected overt forms as a new phonological structure by the nex generation (cf. (19)). The lexical tonal representations for generation | of the acute forn strjota ‘street’ and the circumflex form weéko ‘week’, aligned to an HL intonatiot contour, are given in (42), which also includes the structures for words without fina eephwa strdote ‘street’ and the circumflex form weeka ‘week’, aligned to an HL intonation
(42) Surface structures of words with and without final schwa, voiceless version In these graphs syllable boundaries are represented as periods. We know that the loss of schwa ultimately led to a change in phonologica structures across generations: original strdote ‘street’ and d5at ‘deed’ consisted o three and two moras, respectively, as shown in (7) and (14), while present-day Jtrso and d5at rhyme perfectly, i.e., they must consist of the same number of moras, which will assume to be two. As in the account of Open Syllabe Lengthening, the account o Schwa Drop will involve three elements: the phonological structure maintained by the first generation (cf. (8)), the variation in pronunciation allowed by this structure, and thi reinterpretation of selected overt forms as a new phonological structure by the nex generation (cf. (19)). The lexical tonal representations for generation | of the acute forn strjota ‘street’ and the circumflex form weéko ‘week’, aligned to an HL intonatiot contour, are given in (42), which also includes the structures for words without fina eephwa strdote ‘street’ and the circumflex form weeka ‘week’, aligned to an HL intonation
(43) Surface structures after schwa drop, voiceless version For str3Sot the change was well integrated in Limburgian phonotactics: the resulting rhyme was identical with the already existing rhyme in ddot. For weék, the change was also reasonably well integrated, because the vowel e& was already allowed ir monosyllabic words deriving from analogical lengthening (note that the MLb *ee hac become éa). Now if some children create the structures in (43), they may go on to learr at a later point that some of these words can have a schwa appended to them. In effect the structures have changed across the two generations from bisyllabic with variable schwa drop to monosyllabic with variable schwa extension (perhaps like modern Frenck séc ‘five’). It is likely that speakers with the structures in (43) produce forms wit schwa less often than people who still have the bisyllabic structures. The thirc generation, then, will hear even fewer forms with schwa than the second and is thus structure but are hardly audible in the overt form, can in some languages, like Japanes apparently easily be maintained by speakers and transferred to the next generation. But Limburgian is not Japanese. Learners of Japanese “only” have to know th word-final consonants are generally ruled out in their language, as is evidenced by tk full vowels that have to occur after voiced obstruents and sonorants. Limburgi< learners, however, had ample evidence that Limburgian did allow word-final consonan of all sorts. In a situation where the learner interprets [d55t] as the correspondir structure in (42), she also has to interpret [str35t3] in a way that makes sense in hi emergent phonology. While some learners may indeed come up with the bisyllab structure in (42), perhaps aided by some overt adult forms with full vowels, son learners will act differently. We have to understand that the transcription [str33t: suggests a final vowel, but that a voiceless vowel, especially a short one, often cann be distinguished from the glottal fricative [h], so that an almost equivalent transcriptic is [str5sth]. Since utterance-final plosives tend to be slightly aspirated anyway (by tl continuation of uninhibited expiration), so that d5ot will sound very much like [d33t° we have the ingredients for a merger, especially since word-final aspiration contrasts ai cross-linguistically extremely rare. Some children, therefore, will perceive [str33t3] ar [wéék3] as the monosyllabic structures in (43). become ea). Now if some children create the structures in (43), they may go on to learn sec ‘five’). It is likely that speakers with the structures in (43) produce forms with
(43) Surface structures after schwa drop, voiceless version For str3Sot the change was well integrated in Limburgian phonotactics: the resulting rhyme was identical with the already existing rhyme in ddot. For weék, the change was also reasonably well integrated, because the vowel e& was already allowed ir monosyllabic words deriving from analogical lengthening (note that the MLb *ee hac become éa). Now if some children create the structures in (43), they may go on to learr at a later point that some of these words can have a schwa appended to them. In effect the structures have changed across the two generations from bisyllabic with variable schwa drop to monosyllabic with variable schwa extension (perhaps like modern Frenck séc ‘five’). It is likely that speakers with the structures in (43) produce forms wit schwa less often than people who still have the bisyllabic structures. The thirc generation, then, will hear even fewer forms with schwa than the second and is thus structure but are hardly audible in the overt form, can in some languages, like Japanes apparently easily be maintained by speakers and transferred to the next generation. But Limburgian is not Japanese. Learners of Japanese “only” have to know th word-final consonants are generally ruled out in their language, as is evidenced by tk full vowels that have to occur after voiced obstruents and sonorants. Limburgi< learners, however, had ample evidence that Limburgian did allow word-final consonan of all sorts. In a situation where the learner interprets [d55t] as the correspondir structure in (42), she also has to interpret [str35t3] in a way that makes sense in hi emergent phonology. While some learners may indeed come up with the bisyllab structure in (42), perhaps aided by some overt adult forms with full vowels, son learners will act differently. We have to understand that the transcription [str33t: suggests a final vowel, but that a voiceless vowel, especially a short one, often cann be distinguished from the glottal fricative [h], so that an almost equivalent transcriptic is [str5sth]. Since utterance-final plosives tend to be slightly aspirated anyway (by tl continuation of uninhibited expiration), so that d5ot will sound very much like [d33t° we have the ingredients for a merger, especially since word-final aspiration contrasts ai cross-linguistically extremely rare. Some children, therefore, will perceive [str33t3] ar [wéék3] as the monosyllabic structures in (43). become ea). Now if some children create the structures in (43), they may go on to learn sec ‘five’). It is likely that speakers with the structures in (43) produce forms with
(46) Surface structures of words with and without final schwa, voiced version These structures are completely analogous to (42). Again, the tones are only associated with the heads of the tonal moras, and onset segments cannot head moras. As before, there cannot be more than two moras in a syllable, simply because no more than two tonal moras are needed to represent the surface structure. This means that the final n of fidon or wiin does not head a mora (this contrasts with cases like man ‘man’, in which the final n is the second tonal mora of the syllable).
(46) Surface structures of words with and without final schwa, voiced version These structures are completely analogous to (42). Again, the tones are only associated with the heads of the tonal moras, and onset segments cannot head moras. As before, there cannot be more than two moras in a syllable, simply because no more than two tonal moras are needed to represent the surface structure. This means that the final n of fidon or wiin does not head a mora (this contrasts with cases like man ‘man’, in which the final n is the second tonal mora of the syllable).
(54) Surface structures after schwa drop, obstruent version Present-day reflexes. In the areas that maintain a lexical tone contrast on short vowels before voiceless consonants, such as Moresnet (Jongen 1972), words like kat currently have a circumflex, i.e. the pitch movement occurs in the next syllable, and words like béd currently have an acute, i.e. the pitch movement occurs in the syllable itself (instead of between e and d, the pitch movement nowadays falls within € as a result of the second final devoicing and/or degemination, see §11). Some present-day forms are listed in (55) and (56).
(54) Surface structures after schwa drop, obstruent version Present-day reflexes. In the areas that maintain a lexical tone contrast on short vowels before voiceless consonants, such as Moresnet (Jongen 1972), words like kat currently have a circumflex, i.e. the pitch movement occurs in the next syllable, and words like béd currently have an acute, i.e. the pitch movement occurs in the syllable itself (instead of between e and d, the pitch movement nowadays falls within € as a result of the second final devoicing and/or degemination, see §11). Some present-day forms are listed in (55) and (56).
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved