19/16/03 9/16/03 Speech Perception in Phonology Pidato Persepsi di Fonologi Edward Flemming Edward Flemming Stanford University Stanford University 1. 1. Pengantar Gagasan bahwa sifat persepsi ujaran berperan dalam membentuk fonologi tidaklah baru. There is a substantial literature that proposes and tests perceptual explanations for Ada literatur substansial yang mengusulkan dan menguji penjelasan persepsi untuk phonological patterns, eg Liljencrants and Lindblom (1972), Lindblom (1986), Ohala fonologi pola, misalnya Liljencrants dan Lindblom (1972), Lindblom (1986), Ohala (1981, 1990, 1993). (1981, 1990, 1993). Most of this research addresses general tendencies in phonological Sebagian besar penelitian ini membahas kecenderungan umum dalam fonologi patterning, eg the cross-linguistic tendency for front vowels to be unrounded, which pola, misalnya kecenderungan lintas-linguistik untuk vokal depan untuk unrounded, yang leaves the problem of going from the general to the particular: particular languages may meninggalkan masalah pergi dari umum ke khusus: bahasa tertentu mungkin violate the general tendencies, as in the case of a language with front rounded vowels. melanggar kecenderungan umum, seperti dalam kasus bahasa dengan vokal bulat depan. Recently phonologists have begun to tackle this problem, incorporating principles that Baru-baru ini phonologists telah mulai mengatasi masalah ini, menggabungkan prinsip-prinsip yang invoke properties of human speech perception into models that derive generalizations memanggil sifat persepsi suara manusia ke dalam model yang berasal generalisasi about phonological systems, but also allow for analyses of individual languages. tentang sistem fonologis, tetapi juga memungkinkan untuk analisis bahasa individu. This Ini area of research has proven very productive, and there is now substantial evidence for the daerah penelitian telah terbukti sangat produktif, dan sekarang ada bukti substansial untuk importance of perceptual considerations in phonological theory, but there is little pentingnya pertimbangan persepsi dalam teori fonologi, tapi ada sedikit agreement on the proper formalization of the influence of speech perception on kesepakatan tentang formalisasi yang tepat dari pengaruh persepsi pidato phonology. fonologi. This issue is the organizing theme of the chapter Masalah ini adalah tema pengorganisasian bab 1 1 . . A key element in the development of this research has been Optimality Theory (OT, Sebuah elemen kunci dalam pengembangan penelitian ini telah optimalitas Teori (OT, Prince and Smolensky 1993), which offers a framework for constructing analyses of Pangeran dan Smolensky 1993), yang menawarkan kerangka kerja untuk membangun analisis individual languages out of constraints expressing general preferences of the kind bahasa individu keluar dari kendala mengekspresikan preferensi umum semacam itu

identified in the works cited above. diidentifikasi dalam karya-karya yang disebutkan di atas. In OT terms, the central question addressed here is: Dalam istilah PL, pertanyaan sentral dibahas di sini adalah: “What is the form of the constraints imposed on phonology by speech perception?” We "Apa bentuk kendala yang dikenakan pada fonologi oleh persepsi ujaran?" Kami will review the main types of evidence that have been used to argue for perceptual akan meninjau jenis utama dari bukti yang telah digunakan untuk berdebat untuk persepsi constraints in phonology to clarify exactly what kind of constraints they motivate. kendala dalam fonologi untuk memperjelas jenis apa kendala yang mereka memotivasi. In the Dalam 1 1 I would like to thank David Pisoni, Robert E. Remez and Donca Steriade for helpful comments on this Saya ingin mengucapkan terima kasih David Pisoni, Robert E. Remez dan Donca Steriade untuk komentar ini chapter. bab.

22 2 process, we will also examine the kinds of experimental evidence that have been adduced proses, kita juga akan memeriksa jenis bukti eksperimental yang telah dikemukakan in formulating analyses. dalam merumuskan analisis. 2. 2. Dispersion and Enhancement Dispersi dan Peningkatan The most direct evidence for perceptual constraints in phonology comes from Bukti yang paling langsung untuk kendala persepsi dalam fonologi berasal dari generalizations about inventories of phonological contrasts. generalisasi tentang persediaan kontras fonologi. Phonetic descriptions Fonetik deskripsi distinguish hundreds of sound types, but a typical language has only about 30 contrasting membedakan ratusan jenis suara, tapi bahasa yang khas hanya memiliki sekitar 30 kontras sounds (Maddieson 1984:7). suara (Maddieson 1984:7). These inventories of contrasting sounds are far from being a Ini persediaan suara kontras jauh dari menjadi random sample of the set of attested speech sounds, rather the observed inventories are sampel acak dari himpunan dibuktikan suara pidato, bukan persediaan diamati subject to many restrictions, some of which can be explained in terms of perceptual tunduk pada pembatasan banyak, beberapa di antaranya dapat dijelaskan dalam hal persepsi constraints. kendala. One well established example involves preferences for particular vowel qualities. Salah satu contoh mapan melibatkan preferensi untuk kualitas vokal tertentu. There is a strong cross-linguistic preference for vowels to be front unrounded or back Ada preferensi lintas-linguistik yang kuat untuk menjadi vokal depan atau belakang unrounded rounded unless they are low vowels, as in the common vowel inventories illustrated in dibulatkan kecuali mereka adalah vokal rendah, seperti dalam persediaan vokal umum digambarkan dalam

figure 1 (low vowels are typically described as central or back and unrounded). Angka 1 (vokal rendah biasanya digambarkan sebagai pusat atau kembali dan unrounded). In Di Maddieson's (1984) survey of a genetically diverse sample of languages, 94% of front Maddieson s (1984) survei sampel genetik beragam bahasa, 94% dari depan vowels are unrounded and 93.5% of back vowels are rounded. vokal unrounded dan 93,5% dari vokal kembali dibulatkan. Where a language does Dimana bahasa yang tidak have front rounded, central, or back unrounded vowels, these appear in addition to front telah bulat depan, tengah, atau belakang unrounded vokal, ini muncul di samping depan unrounded and back rounded vowels. unrounded dan kembali bulat vokal. Figure 1. Gambar 1. Two common vowel inventories. Dua persediaan vokal umum. It is hard to imagine any articulatory basis for this relationship between backness and Sulit membayangkan ada dasar artikulatoris untuk hubungan antara backness dan rounding. pembulatan. The tongue and lips are articulatorily relatively independent, so it would appear Lidah dan bibir articulatorily relatif independen, sehingga akan muncul i i u u a sebuah i i u u e e a sebuah o o

Page 3 Page 33 3 to be as easy to round the lips with the tongue body forward as with it retracted. untuk menjadi seperti mudah untuk putaran bibir dengan lidah tubuh ke depan seperti itu ditarik. On the Di other hand there is a straightforward perceptual account of the covariation of backness sisi lain ada rekening persepsi langsung dari covariation dari backness and rounding. dan pembulatan. The primary perceptual dimensions of vowel quality correspond well to the Dimensi persepsi kualitas vokal utama sesuai dengan baik ke frequencies of the first two formants (Delattre, Liberman, Cooper, and Gerstman 1952, frekuensi dari dua forman pertama (Delattre, Liberman, Cooper, dan Gerstman 1952, Plomp 1975, Shepard 1972). Plomp 1975, Shepard 1972). Front and back vowels are differentiated primarily by the Depan dan belakang vokal dibedakan terutama oleh frequency of the second formant (F2), with front vowels having a high F2 and back frekuensi forman kedua (F2), dengan vokal depan memiliki F2 tinggi dan kembali vowels having a low F2. vokal memiliki F2 rendah. Lip-rounding generally lowers F2, so the ordering of front and Lip-pembulatan pada umumnya menurunkan F2, sehingga Urutan depan dan back, rounded and unrounded vowels, and central vowels in terms of F2 is shown in kembali, bulat dan unrounded vokal, dan vokal pusat dalam hal F2 ditampilkan dalam figure 2. Angka 2. Thus the maximally distinct F2 contrast is between front unrounded and back Dengan demikian kontras F2 maksimal yang berbeda antara depan dan belakang unrounded

rounded vowels (Liljencrants and Lindblom 1972, Stevens, Keyser, and Kawasaki 1986). dibulatkan vokal (Liljencrants dan Lindblom 1972, Stevens, Keyser, dan Kawasaki 1986). Maximally distinct contrasts are preferred because they are less likely to be confused by Kontras maksimal berbeda yang disukai karena mereka kurang mungkin bingung dengan listeners. pendengar. Figure 2. Gambar 2. The ordering of vowel qualities on the F2 dimension. Urutan kualitas vokal pada dimensi F2. The general preference for maximally distinct contrasts follows from the Preferensi umum untuk kontras maksimal berbeda berikut dari functionalist hypothesis that phonological systems are well adapted for communication. fungsionalis hipotesis bahwa sistem fonologis beradaptasi dengan baik untuk komunikasi. Efficient communication depends on fast, accurate perception of speech sounds, and Komunikasi yang efisien tergantung pada cepat, akurat persepsi dari suara pidato, dan listeners are faster and more accurate in identifying the category to which a stimulus pendengar yang lebih cepat dan lebih akurat dalam mengidentifikasi kategori yang stimulus belongs if the stimulus is more distinct from contrasting categories (eg Ashby, Boynton termasuk jika stimulus lebih berbeda dari kategori kontras (misalnya Ashby, Boynton and Lee 1994, Kellogg 1931, Pisoni and Tash 1974, Podgorny and Garner 1979). dan Lee 1994, Kellogg 1931, Pisoni dan Tash 1974, Podgorny dan Garner 1979). We will Kami akan see that the principle of maximization of distinctiveness is the key perceptual constraint melihat bahwa prinsip maksimalisasi kekhasan adalah kendala persepsi kunci on phonology. pada fonologi. Evidence for this principle has been discussed under a variety of labels. Bukti untuk prinsip ini telah dibahas dalam berbagai label. Lindblom Lindblom and Engstrand (1989) refer to the tendency to maximize the perceptual distinctiveness of dan Engstrand (1989) mengacu pada kecenderungan untuk memaksimalkan kekhasan persepsi dari contrasting speech sounds as 'dispersion', invoking the notion of separation in perceptual kontras pidato suara sebagai 'dispersi', menyerukan gagasan pemisahan dalam persepsi i i y y µ μ û û F2 F2

Page 4 Page 44 4 space. ruang. Similar phenomena have been discussed by Stevens, Keyser, and Kawasaki (1986) Fenomena serupa telah dibahas oleh Stevens, Keyser, dan Kawasaki (1986) under the rubric of 'enhancement'. di bawah rubrik 'peningkatan'. They observe that distinctive features are often Mereka mengamati bahwa fitur khas sering accompanied by 'redundant' features that 'strengthen the acoustic representation of disertai dengan 'berlebihan' fitur 'yang memperkuat representasi akustik

distinctive features and contribute additional properties which help the listener to khas fitur dan berkontribusi sifat tambahan yang membantu pendengar untuk perceive the distinction' (p.426). merasakan perbedaan '(p.426). The relationship between [back] and [round] in vowels Hubungan antara [kembali] dan [putaran] di vokal is treated as one of enhancement: [round] enhances distinctive [back]. diperlakukan sebagai salah satu perangkat tambahan: [round] meningkatkan khas [kembali]. So enhancement Jadi peningkatan essentially involves combining feature differences so as to maximize the perceptual dasarnya melibatkan menggabungkan perbedaan fitur sehingga memaksimalkan persepsi distinctiveness of contrasts. kekhasan kontras. Consequently instances of enhancement also provide Akibatnya kasus peningkatan juga menyediakan evidence for maximization of distinctiveness. bukti untuk memaksimalkan kekhasan. Other work providing evidence for Pekerjaan lain memberikan bukti untuk dispersion/enhancement includes Ohala (1985:225f.), Diehl (1991), and Flemming dispersi / peningkatan meliputi Ohala (1985:225 f.), Diehl (1991), dan Flemming (2002:53-56). (2002:53-56). We will review two further cases here to illustrate the range of phenomena Kami akan meninjau dua kasus lebih lanjut di sini untuk menggambarkan berbagai fenomena involved. terlibat. Another example discussed by Stevens et al (1986) is the enhancement of frication Contoh lain dibahas oleh Stevens et al (1986) adalah peningkatan frication contrasts. kontras. Fricatives are distinguished from other sound types by the presence of Frikatif dibedakan dari jenis suara lain dengan kehadiran significant turbulence noise, generated by forcing a jet of air through a narrow turbulensi yang signifikan kebisingan, yang dihasilkan dengan memaksa jet udara melalui sempit constriction. penyempitan. The distinctiveness of this manner contrast can thus be enhanced by Kekhasan ini kontras dengan cara demikian dapat ditingkatkan dengan increasing the intensity of turbulence noise in the fricative. meningkatkan intensitas kebisingan turbulensi di frikatif tersebut. This is achieved by directing Hal ini dicapai dengan mengarahkan the jet of air against an obstacle downstream, as in the coronal sibilant [s], where a jet of jet udara melawan hilir kendala, seperti dalam bunyi berdesis koronal [s], di mana jet air is directed against the upper teeth (Stevens et al 1986:439, Shadle1991). udara ditujukan terhadap gigi atas (Stevens et al 1986:439, Shadle1991). The greater Semakin besar distinctness of such sibilant fricatives from non-fricatives can explain their cross- Kejelasan frikatif bunyi berdesis tersebut dari non-frikatif dapat menjelaskan cross-mereka linguistic prevalence: in Maddieson's (1984) survey, 83% of languages have some kind prevalensi linguistik: di (1984) survei Maddieson itu, 83% dari bahasa memiliki semacam of [s], and if a language has only one fricative it is usually an [s] sound (84%). dari [s], dan jika bahasa hanya memiliki satu frikatif itu biasanya [s] suara (84%). Maximization of the distinctiveness of contrasts between sibilants has been argued to Maksimalisasi kekhasan kontras antara sibilants telah menyatakan explain an otherwise puzzling observation about the realization of post-alveolar menjelaskan pengamatan sebaliknya membingungkan tentang realisasi pasca-alveolar

fricatives: in English and French, the post-alveolar fricative [S] is accompanied by lip frikatif: dalam bahasa Inggris dan Perancis, yang frikatif pasca-alveolar [S] disertai dengan lip protrusion (Ladefoged and Maddieson 1996:148). tonjolan (Ladefoged dan Maddieson 1996:148). There is no articulatory basis for this Tidak ada dasar untuk ini artikulatoris pattern, but it plausibly serves to make post-alveolar [S0] more distinct from the anterior Pola, tetapi masuk akal berfungsi untuk membuat pasca-alveolar [S0] lebih jelas dari anterior

Page 5 Halaman 55 5 sibilant [s]. berdesis [s]. These sounds are differentiated by the frequency of the first peak in the noise Suara ini dibedakan oleh frekuensi puncak pertama dalam kebisingan spectrum. spektrum. This peak is at the resonant frequency of the cavity in front of the constriction, Puncak ini adalah pada frekuensi resonansi dari rongga di depan penyempitan, and so is lower in post-alveolar fricatives, since they have a larger front cavity than dan begitu rendah pasca-alveolar frikatif, karena mereka memiliki rongga depan lebih besar dari dentals and alveolars. dentals dan alveolars. Protruding the lips increases this difference by further enlarging the Menonjol bibir meningkatkan perbedaan ini dengan lebih memperbesar front cavity (Ladefoged and Maddieson 1996:149). depan rongga (Ladefoged dan Maddieson 1996:149). Polish provides an interesting variant Polandia menyediakan varian yang menarik of this pattern. dari pola ini. There are three contrasting sibilants, dental [s5], alveopalatal [˛], and Ada tiga sibilants kontras, gigi [s5], alveopalatal [˛], dan retroflex (apical post-alveolar) [ß], and the retroflex is produced with lip protrusion retroflks (apikal pasca-alveolar) [ß], dan retroflks yang diproduksi dengan tonjolan bibir (Puppel, Nawrocka-Fisiak, and Krassowska 1977:157). (Puppel, Nawrocka-Fisiak, dan Krassowska 1977:157). This is the expected pattern given Ini adalah pola yang diharapkan diberikan the goal of maximizing distinctiveness because the retroflex has the lowest front cavity tujuan memaksimalkan kekhasan karena retroflks memiliki rongga depan terendah resonance due to the space below the tongue blade. karena ruang bawah pisau lidah resonansi. Lowering this resonance further by Menurunkan resonansi ini lebih lanjut dengan protruding the lips makes the retroflex more distinct from the other sibilants (Flemming menonjol bibir membuat retroflks lebih berbeda dari sibilants lain (Flemming 2002:55f.). 2002:55 f.). 1.1 Phonological analyses of dispersion effects 1.1 fonologi analisis efek dispersi There have been two basic approaches to the analysis of dispersion effects: (i) Ada dua pendekatan dasar untuk analisis efek dispersi: (i) analyses that incorporate a preference for maximally distinct contrasts into phonological analisis yang menggabungkan preferensi untuk kontras maksimal berbeda dalam fonologi

theory, and (ii) analyses that employ standard markedness constraints. teori, dan (ii) analisis yang mempekerjakan kendala markedness standar. The latter Yang terakhir approach is in a sense the default option, since it employs only the standard apparatus of Pendekatan dalam arti pilihan default, karena hanya mempekerjakan aparat standar phonological theory (as outlined in the next section), but we will see that dispersion fonologis teori (seperti yang dijelaskan dalam bagian berikutnya), tetapi kita akan melihat bahwa dispersi effects provide strong evidence for the distinctiveness constraints posited in the former Efek memberikan bukti yang kuat untuk kendala kekhasan mengemukakan di bekas approach. Pendekatan. We will first provide a brief overview of Optimality Theory (OT, Prince and Pertama-tama kita akan memberikan gambaran singkat mengenai Teori optimalitas (PL, Prince dan Smolensky 1993) and its suitability as a framework for formalizing the influence of Smolensky 1993) dan kesesuaian sebagai kerangka kerja untuk memformalkan pengaruh speech perception on phonology, then we will turn to the particular proposals for pidato persepsi tentang fonologi, maka kita akan beralih ke proposal tertentu untuk formalizing perceptual constraints memformalkan kendala persepsi 2 2 . . Although the discussion will focus on analyses Meskipun diskusi akan fokus pada analisis formulated in OT, the issues raised are relevant to any analysis of these phenomena. dirumuskan dalam PL, isu yang diangkat relevan untuk setiap analisis dari fenomena ini. 2 2 For more detailed introductions to OT, see Kager (1998) and McCarthy (2001). Untuk perkenalan lebih rinci untuk PL, lihat Kager (1998) dan McCarthy (2001).

Page 6 Halaman 66 6 1.2 Optimality Theory 1,2 optimalitas Teori In its basic form, an OT grammar maps input underlying forms onto their surface Dalam bentuk dasarnya, sebuah tata bahasa masukan peta PL bentuk dasar ke permukaan mereka realizations. realisasi. For example, in Russian obstruents are devoiced in word-final position, so Misalnya, di hambat Rusia devoiced dalam kata-akhir posisi, sehingga the morpheme /sad/ 'garden' is pronounced [sat] (the underlying voiced final stop morfem / sedih / 'taman' diucapkan [sat] (pemberhentian terakhir mendasari bersuara surfaces when a vowel-initial suffix is added, as in the dative singular [sadu]). permukaan ketika akhiran vokal-awal ditambahkan, seperti dalam bentuk tunggal dativ [Sadu]). In OT, the Dalam PL, yang mapping between input and output is divided into two components: a mapping from an pemetaan antara input dan output dibagi menjadi dua komponen: sebuah pemetaan dari input form to a set of candidate outputs, and an evaluation function which selects the best form input untuk satu set output calon, dan fungsi evaluasi yang memilih yang terbaik member of the candidate set as the actual output. anggota calon ditetapkan sebagai output yang sebenarnya. The optimality of candidate outputs is Optimalitas output kandidat

determined by reference to a ranked set of constraints. ditentukan dengan mengacu pada satu set peringkat kendala. Standard OT posits two basic types of constraints: constraints that evaluate the well- Standar PL mengemukakan dua tipe dasar kendala: kendala yang mengevaluasi-baik formedness of the candidate outputs—markedness constraints—and constraints that formedness calon output-markedness kendala-kendala dan yang require the output to be as similar to the input as possible—faithfulness constraints. membutuhkan output menjadi seperti mirip dengan masukan mungkin kesetiaan-kendala. These Ini two types of constraints are liable to conflict – satisfying markedness constraints often dua jenis kendala bertanggung jawab untuk konflik - kendala markedness memuaskan sering requires altering the input, which necessarily violates some faithfulness constraint. membutuhkan mengubah masukan, yang tentu melanggar beberapa kendala kesetiaan. For Untuk example, a simple-minded analysis of the Russian facts above posits a markedness Sebagai contoh, sebuah analisis sederhana-berpikiran dari fakta-fakta di atas Rusia memposisikan markedness suatu constraint forbidding word-final voiced obstruents, *F kendala melarang kata-akhir hambat bersuara, * F INAL Inal V V OICED OICED O O BSTRUENT BSTRUENT . . The fully Sepenuhnya faithful realization of [sad] violates this constraint, but devoicing the final stop, as in [sat] realisasi setia [menyedihkan] melanggar kendala ini, tapi KF halte akhir, seperti dalam [duduk] violates the faithfulness constraint I melanggar kendala kesetiaan saya DENT Dent [ [ VOICE VOICE ] which requires that voicing ] Yang mensyaratkan bahwa menyuarakan specifications of input segments should be unchanged in the output. spesifikasi segmen masukan harus berubah dalam output. Conflicts between constraints are resolved by reference to a ranking of the Konflik antara kendala diselesaikan dengan mengacu pada ranking dari constraints: the higher ranked constraint prevails. kendala: kendala peringkat yang lebih tinggi berlaku. So in Russian, Jadi dalam bahasa Rusia, *F * F INAL Inal V V OICED OICED O O BSTRUENT BSTRUENT

must outrank I Saya harus mengungguli DENT Dent [ [ VOICE VOICE ] (written: ] (Tertulis: *F * F INAL Inal V V OICED OICED O O BSTRUENT BSTRUENT >> I Saya >> DENT Dent [ [ VOICE VOICE ]) since the voicing of an input stop is ]) Sejak menyuarakan dari berhenti input changed in order to satisfy the former constraint. diubah untuk memenuhi kendala mantan. If this ranking were reversed the Jika peringkat ini telah membalikkan candidate [sad] would win. Calon [menyedihkan] akan menang. OT analyses are typically illustrated using tableaux, as in (1). Analisis PL biasanya digambarkan menggunakan Tableaux, seperti pada (1). The input form is Bentuk input shown in the top left cell while the candidate outputs are listed below it in the first ditampilkan dalam sel kiri atas sedangkan output kandidat yang tercantum di bawah ini dalam pertama

Page 7 Page 77 7 column. kolom. The constraints are listed in the top row, with higher-ranked constraints on the Kendala tercantum di baris atas, dengan tinggi-peringkat kendala pada left. kiri. If a candidate violates a constraint, a mark (*) is placed at the intersection of the Jika seorang kandidat melanggar kendala, tanda (*) ditempatkan di persimpangan constraint column and the candidate row. kendala kolom dan baris kandidat. In (1), candidate (a), [sad] violates Dalam (1), calon (a), [sad] melanggar *F * F INAL Inal V V OICED OICED O O BSTRUENT BSTRUENT , so a mark is placed under that constraint in row (a). , Sehingga tanda ditempatkan di bawah bahwa kendala berturut-turut (a). Candidate (b), [sat], satisfies this constraint, so [sad] is eliminated (indicated by the Calon (b), [duduk], memenuhi kendala ini, sehingga [sedih] dihilangkan (ditunjukkan oleh exclamation point after the mark), and [sat] is the optimal output (indicated by the seru setelah tanda), dan [duduk] adalah output yang optimal (ditunjukkan oleh

'pointing hand' in the first column). 'Menunjuk tangan' di kolom pertama). Note that it is not necessary to satisfy all the Perhatikan bahwa tidak perlu untuk memenuhi semua constraints in order to be the optimal candidate – candidate (b), [sat], is optimal although kendala untuk menjadi calon yang optimal - calon (b), [duduk], optimal meskipun it violates I itu melanggar I DENT Dent [ [ VOICE VOICE ]. ]. Indeed, since constraints frequently conflict, it is not usually Memang, karena kendala yang sering konflik, biasanya tidak possible to satisfy them all. mungkin untuk memuaskan mereka semua. (1) (1) /sad/ / Sedih / *F * F INAL Inal V V OICED OICED O O BS BS I Saya DENT Dent [ [ VOICE VOICE ] ] a. a. sad sedih *! *! b. b. + + sat duduk * * One of the key strengths of OT is the way in which it relates the analysis of the Salah satu kekuatan kunci dari PL adalah cara yang berkaitan dengan analisis typology of languages to the analyses of individual languages. tipologi bahasa dengan analisis bahasa individu. The two are connected by Keduanya dihubungkan oleh the hypothesis that all phonological grammars are constructed from the same set of hipotesis bahwa semua tata bahasa fonologis dibangun dari set yang sama constraints, but differ in the ranking of those constraints. kendala, namun berbeda dalam peringkat tersebut kendala. Typological universals can then Universal tipologis kemudian dapat be derived from the nature of the universal set of constraints, while the patterns of akan berasal dari sifat dari himpunan universal kendala, sedangkan pola individual languages are hypothesized to derive from particular rankings of these bahasa individu yang diduga berasal dari peringkat tertentu ini constraints. kendala.

This provides a suitable framework for formalizing the preference for distinct Ini menyediakan kerangka kerja yang cocok untuk meresmikan preferensi untuk berbeda contrasts because any such preference is a universal tendency which may be violated to a kontras karena preferensi tersebut adalah kecenderungan universal yang dapat dilanggar dengan greater or lesser extent as a result of conflicting constraints. besar atau lebih kecil sejauh sebagai akibat dari kendala bertentangan. For example, languages like Misalnya, bahasa seperti French and German have front rounded vowels in addition to front unrounded and back Perancis dan Jerman memiliki vokal bulat depan di samping depan dan belakang unrounded rounded vowels. dibulatkan vokal. That is, these languages eschew maximally distinct F2 contrasts in Artinya, bahasa-bahasa menjauhkan diri kontras F2 maksimal berbeda dalam favour of distinguishing more contrastive vowels. mendukung membedakan vokal lebih kontrastif. Conversely, grounding constraints in Sebaliknya, landasan kendala dalam basic considerations of communicative efficiency, and the nature of human speech dasar pertimbangan efisiensi komunikatif, dan sifat suara manusia

Page 8 Halaman 88 8 perception provides a basis for the universality of those constraints: if a constraint is Persepsi memberikan dasar bagi universalitas kendala tersebut: jika kendala adalah based on universal properties of communication and perception, it is unsurprising that it berdasarkan sifat universal komunikasi dan persepsi, maka tidak mengherankan jika hal itu is operative in all languages. yang bekerja dalam semua bahasa. 1.3 The constraints that motivate dispersion 1.3 Kendala yang memotivasi dispersi Two kinds of constraints have been proposed in the analysis of dispersion Dua jenis kendala telah diusulkan dalam analisis dispersi phenomena: basic segmental markedness constraints, and constraints on the fenomena: dasar kendala markedness segmental, dan kendala pada distinctiveness of contrasts. kekhasan kontras. A basic markedness constraint in OT prohibits some Sebuah kendala markedness dasar dalam PL melarang beberapa representational structure, such as a syllable without an onset, or a segment which has the representasional struktur, seperti suku kata tanpa awal, atau segmen yang memiliki feature combination [-sonorant, +voice]. Kombinasi fitur [-sonoran, suara +]. A number of researchers have suggested that Sejumlah peneliti telah menyarankan bahwa constraints of this form can be motivated by perceptual considerations (eg Hume 1998, kendala formulir ini dapat didorong oleh pertimbangan persepsi (misalnya Hume 1998, Côté 2000). Côté 2000). Certainly, the most common analysis of the preference for peripheral vowels Tentu saja, analisis yang paling umum dari preferensi untuk vokal perifer (ie front unrounded and back rounded vowels) has been to propose constraints against (Depan dan belakang yaitu unrounded dibulatkan vokal) telah mengusulkan kendala terhadap

other types of vowels, as in (2) (eg Calabrese 1988) lain jenis vokal, seperti dalam (2) (misalnya Calabrese 1988) 3 3 . . (2) *[-back, +round] (2) * [-kembali, + bulat] *[+back, -round] * [+ Kembali, sepanjang] Ranking these constraints above faithfulness to [back] or [round] yields a language Peringkat kendala-kendala di atas kesetiaan [kembali] atau [putaran] menghasilkan bahasa without non-peripheral vowels because inputs containing these vowels will not be tanpa non-perifer vokal karena masukan yang mengandung vokal tidak akan realized faithfully (3-4) menyadari setia (3-4) 4 4 . . (3) /y/ (3) / y / *[-back, * [-Kembali, +round] + Putaran] *[+back, * [+ Kembali, -round] -Putaran] I Saya DENT Dent [round] [Round] a. a. + + i i * * b. b. y y *! *! 3 3 It is common to specify central vowels as [+back, -round], in which cases these constraints are sufficient. Hal ini umum untuk menentukan vokal pusat sebagai [+ kembali, sepanjang], di mana kasus ini kendala yang cukup. If central vowels are distinguished from back unrounded vowels, a constraint against this class of vowels is Jika vokal pusat dibedakan dari vokal kembali unrounded, kendala terhadap kelas ini adalah vokal required also. diperlukan juga. 4 4 The dotted line between the top two constraints in (5-6) indicates that the relative ranking of these Garis putus-putus antara dua kendala teratas di (5-6) menunjukkan bahwa peringkat relatif dari constraints cannot be determined – that is, either ranking yields the desired result. kendala tidak dapat ditentukan - yaitu, baik hasil peringkat hasil yang diinginkan.

Page 9 Halaman 99 9 (4) (4)

/µ/ / Μ / *[-back, * [-Kembali, +round] + Putaran] *[+back, * [+ Kembali, -round] -Putaran] I Saya DENT Dent [round] [Round] a. a. + + u u * * b. b. µ μ *! *! Although these constraints can derive languages in which back and round co-vary Meskipun kendala dapat memperoleh bahasa yang kembali dan bulat co-bervariasi appropriately, they do not follow directly from the perceptual considerations behind tepat, mereka tidak mengikuti secara langsung dari pertimbangan persepsi belakang Liljencrants and Lindblom's (1972) analysis. Liljencrants dan Lindblom itu (1972) analisis. It was suggested that in order to facilitate Disarankan bahwa untuk memfasilitasi speech perception, contrasting sounds should be maximally distinct. persepsi ujaran, suara kontras harus maksimal berbeda. This explanation Penjelasan ini implies a dispreference for F2 contrasts involving non-peripheral vowels because they are menyiratkan dispreference untuk kontras F2 melibatkan non-perifer vokal karena mereka less distinct than contrasts between front unrounded and back rounded vowels. kurang jelas dibandingkan kontras antara depan unrounded dan vokal kembali membulat. The Itu constraints in (2) do not mention contrasts – they simply prohibit front rounded, central, kendala dalam (2) tidak menyebutkan kontras - mereka hanya melarang depan bulat, central, and back unrounded vowels. dan kembali unrounded vokal. Liljencrants and Lindblom's proposal is implemented more Proposal Liljencrants dan Lindblom yang diimplementasikan lebih directly by constraints that penalize less distinct contrasts (distinctiveness constraints), langsung oleh kendala yang menghukum kontras kurang jelas (kendala kekhasan), eg a constraint ranking along the lines shown in (5), where *XY means that words misalnya peringkat kendala di sepanjang garis yang ditunjukkan dalam (5), di mana * XY berarti bahwa kata-kata should not be minimally differentiated by the contrast between sounds X and Y (more tidak boleh minimal dibedakan oleh kontras antara suara X dan Y (lebih general formulations are discussed below). formulasi umum dibahas di bawah). (5) *y÷µ >> *i÷µ, *y÷u >> *i÷u (5) * y ÷ μ >> * i ÷ μ, * y ÷ u >> * i ÷ u The crucial difference between these two proposals is that the analysis based on Perbedaan penting antara kedua usulan ini adalah bahwa analisis yang didasarkan pada distinctiveness constraints predicts that non-peripheral vowels should be unproblematic kendala kekhasan memprediksi bahwa non-perifer vokal harus bermasalah

as long as they do not enter into front-back (F2) contrasts, whereas the constraints in (2) selama mereka tidak masuk ke depan-belakang (F2) kontras, sedangkan kendala dalam (2) ban these sound types regardless of what they contrast with. melarang jenis suara terlepas dari apa yang mereka kontras dengan. For example, a back Sebagai contoh, kembali unrounded vowel presents no particular perceptual difficulties if the listener knows that it vokal unrounded tidak menimbulkan kesulitan persepsi tertentu jika pendengar tahu bahwa itu is the only vowel that can appear in the context. adalah vokal-satunya yang dapat muncul dalam konteks. It does not violate *i÷µ or any other Ini tidak melanggar * i μ ÷ atau lainnya distinctiveness constraint because there is no contrast, but it would violate *[+back, - kendala kekhasan karena tidak ada kontras, tetapi akan melanggar * [+ kembali, - round]. putaran].

Page 10 Halaman 1010 10 In general, the reasoning outlined above motivates constraints based on the Secara umum, alasan yang diuraikan di atas memotivasi kendala berdasarkan distinctiveness of contrasts between sounds, not on the sounds themselves. kekhasan kontras antara suara, bukan pada suara itu sendiri. Basic Dasar markedness constraints as in (2) apply to individual sounds, not contrasts, and so cannot kendala markedness seperti dalam (2) berlaku untuk suara individu, tidak kontras, sehingga tidak bisa be motivated in this way. termotivasi dengan cara ini. More importantly, there is empirical evidence that phonology is Lebih penting lagi, ada bukti empiris bahwa fonologi adalah in fact subject to constraints on the distinctiveness of contrasts: the markedness of a pada kenyataannya tunduk pada kendala pada kekhasan kontras: yang markedness dari sound depends on the contrasts that it enters into. suara tergantung pada kontras yang masuk ke dalam. Before evaluating this evidence, it is useful to place distinctiveness constraints in the Sebelum mengevaluasi bukti ini, hal ini berguna untuk menempatkan kendala kekhasan dalam context of a specific model. konteks model tertentu. The most developed proposal is the dispersion theory of Proposal yang paling berkembang adalah teori dispersi contrast (Flemming 1995, 2001, 2002, Ní Chiosáin and Padgett 2001), which builds on Sebaliknya (Flemming 1995, 2001, 2002, NI Chiosáin dan Padgett 2001), yang dibangun di atas ideas from Lindblom's Theory of Adaptive Dispersion. ide dari Teori Lindblom tentang Dispersi Adaptive. In this model, the preference to Dalam model ini, preferensi untuk maximize the distinctiveness of contrasts is opposed by two other goals: maximization of memaksimalkan kekhasan kontras ditentang oleh dua gol lainnya: maksimalisasi the number of contrasts permitted in any given context, and minimization of articulatory jumlah kontras diizinkan dalam konteks tertentu, dan meminimalkan artikulatoris

effort. usaha. Increasing the number of contrasting sounds makes more efficient communication Peningkatan jumlah suara kontras membuat komunikasi lebih efisien possible by increasing the information content of each sound, since it allows a single mungkin dengan meningkatkan kandungan informasi dari suara masing-masing, karena memungkinkan satu segment to differentiate more words. segmen untuk membedakan kata-kata yang lebih banyak. This goal conflicts with maximizing distinctiveness Ini konflik tujuan dengan kekhasan memaksimalkan because fitting more contrasts into the finite space of possible speech sounds implies that karena pas kontras lebih ke dalam ruang terbatas suara pidato mungkin menyiratkan bahwa the sounds must be closer together. suara harus lebih dekat bersama-sama. Avoiding effortful articulations further restricts the Menghindari artikulasi effortful lanjut membatasi possibilities for realizing distinct contrasts, so this principle also conflicts with kemungkinan untuk mewujudkan kontras yang berbeda, sehingga prinsip ini juga bertentangan dengan maximization of distinctiveness maksimalisasi kekhasan 5 5 . . Thus selecting a set of contrasts that best satisfies these Jadi memilih satu set kontras yang paling memenuhi ini three goals involves finding an optimal balance between them (cf. Lindblom 1986). tiga gol melibatkan menemukan keseimbangan optimal antara mereka (lih. Lindblom 1986). This Ini optimization is modeled within the framework of OT. optimasi dimodelkan dalam kerangka PL. 5 5 Articulatory effort is not necessarily equivalent to energy expenditure. Upaya artikulatoris belum tentu setara dengan pengeluaran energi. Although this is usually assumed to Meskipun hal ini biasanya diasumsikan be an important component of articulatory effort (eg Lindblom 1983, Nelson 1983, Kirchner 1998), there menjadi komponen penting dari upaya artikulatoris (misalnya Lindblom 1983, Nelson 1983, Kirchner 1998), ada may also be costs associated with precision, for example. mungkin juga biaya yang terkait dengan presisi, misalnya. The aspect of effort that is most relevant in the Aspek usaha yang paling relevan dalam examples discussed here relates to the smoothness of movements – movements are hypothesized to be more contoh yang dibahas di sini berkaitan dengan kelancaran gerakan - gerakan yang diduga lebih difficult if they involve abrupt changes in direction. sulit jika mereka melibatkan perubahan mendadak dalam arah. It has been observed that humans generally employ Telah diamati bahwa manusia umumnya menggunakan smooth trajectories in speech production (Perkell 1997:357) and in arm movements (eg Flash and Hogan mulus lintasan dalam produksi ujaran (Perkell 1997:357) dan gerakan lengan (misalnya Flash dan Hogan

1985, Uno, Kawato and Suzuki 1989). 1985, Uno, Kawato dan Suzuki 1989). This preference has been attributed to minimization of energy Preferensi ini telah dikaitkan dengan meminimalkan energi expenditure (Nelson 1983), but it has also been analyzed in terms of minimizing error in the face of noise pengeluaran (Nelson 1983), tetapi juga telah dianalisis dalam hal meminimalkan kesalahan dalam menghadapi kebisingan internal to the motor control system (Wolpert and Harris 1998). internal untuk sistem kontrol motor (Wolpert dan Harris 1998).

Page 11 Halaman 1111 11 The preference to maximize the distinctiveness of contrasts is implemented in terms Preferensi untuk memaksimalkan kekhasan kontras diimplementasikan dalam of a ranked set of constraints requiring a specified minimum perceptual distance between satu set peringkat kendala yang membutuhkan jarak minimum antara persepsi tertentu contrasting forms (6). kontras bentuk (6). Sounds are represented as located in a multi-dimensional Suara yang direpresentasikan sebagai terletak di multi-dimensi perceptual space where closer sounds are more confusable. persepsi ruang di mana suara lebih dekat lebih confusable. For example, (7) shows the Sebagai contoh, (7) menunjukkan assumed location of high vowels on the dimension corresponding to F2 frequency, diasumsikan lokasi vokal tinggi pada dimensi yang sesuai dengan frekuensi F2, measured in arbitrary units. diukur dalam unit sewenang-wenang. Assuming for simplicity that these vowels differ on this Dengan asumsi untuk kesederhanaan bahwa vokal berbeda ini dimension only, it can be seen that the contrast [iu] involves a distance of 4, and thus dimensi saja, dapat dilihat bahwa kontras [iu] melibatkan jarak 4, dan dengan demikian satisfies all the M memenuhi semua M INDIST INDIST constraints in (6), while [iy] involves a distance of only 1, and kendala dalam (6), sedangkan [iy] melibatkan jarak hanya 1, dan thus violates M dengan demikian melanggar M INDIST INDIST = 2 and all lower-ranked constraints. = 2 dan semua rendah-peringkat kendala. In other words, the less Dengan kata lain, semakin sedikit distinct a contrast is, the greater the violation. berbeda kontras, semakin besar pelanggaran. (6) M (6) M INDIST INDIST = 1 >> M = 1 >> M INDIST INDIST = 2 >>... = 2 >> ... >> M M >> INDIST INDIST = 4 = 4 (7) F2: (7) F2: 5 5

4 4 3 3 2 2 1 1 i i y y µ μ u u The preference to maximize the number of contrasts is implemented as a positive Preferensi untuk memaksimalkan jumlah kontras diimplementasikan sebagai positif constraint, M kendala, M AXIMIZE CONTRASTS AXIMIZE kontras , which is satisfied by the largest inventory of contrasts. , Yang puas dengan persediaan terbesar dari kontras. The conflict between these two types of constraints is illustrated in (10-11) with the Konflik antara kedua jenis kendala diilustrasikan dalam (10-11) dengan simple example of selecting a set of contrasting high vowels. contoh sederhana memilih satu set kontras vokal tinggi. The balance between Keseimbangan antara maximizing distinctiveness and maximizing the number of contrasts is determined by memaksimalkan kekhasan dan memaksimalkan jumlah kontras ditentukan oleh position of M posisi M AXIMIZE CONTRASTS AXIMIZE kontras in the hierarchy of M dalam hirarki M INDIST INDIST constraints. kendala. In (8), Dalam (8), M M INDIST INDIST =3 outranks M = 3 outranks M AXIMIZE CONTRASTS AXIMIZE kontras , so the largest inventory, (d), is eliminated, , Sehingga persediaan terbesar, (d), dihilangkan, because it is does not satisfy M karena tidak memenuhi M INDIST INDIST =3. = 3. The most distinct inventory (a), containing front Persediaan yang paling berbeda (a), yang mengandung depan unrounded and back rounded vowels, best satisfies the M vokal unrounded dan kembali bulat, terbaik memuaskan M INDIST INDIST constraints, and hence is kendala, dan karenanya the winner. pemenang. Contrasts involving back unrounded vowels (b), or front rounded vowels (c) Kontras melibatkan kembali vokal unrounded (b), atau vokal bulat depan (c) are less distinct, and therefore lose to candidate (a). kurang jelas, dan karena itu kehilangan kepada calon (a).

Page 12 Page 1212 12

(8) (8) M M INDIST INDIST = 2 = 2 M M INDIST INDIST = 3 = 3 M M AXIMIZE AXIMIZE CONTRASTS Kontras M M INDIST INDIST = 4 = 4 a. a. + + i÷u i ÷ u b. b. i÷µ i ÷ μ ¸* ¸ * c. c. y÷u y ÷ u *! *! d. d. i÷ˆ÷u i ÷ ÷ u *! *! ** ** In (9), M Dalam (9), M AXIMIZE CONTRASTS AXIMIZE kontras ranks above M peringkat atas M INDIST INDIST =3 – ie the number of contrasts = 3 - yaitu jumlah kontras is more important. lebih penting. So the winning candidate is (d) which fits in three contrasting vowels Jadi calon yang menang adalah (d) yang cocok dalam tiga vokal kontras while satisfying the higher-ranked constraint M sementara memenuhi M kendala yang lebih tinggi-peringkat INDIST INDIST =2. = 2. (9) (9) M M INDIST INDIST = 2 = 2 M M AXIMIZE AXIMIZE CONTRASTS Kontras M M INDIST INDIST

= 3 = 3 M M INDIST INDIST = 4 = 4 a. a. i÷u! i ÷ u! b. b. i÷µ i ÷ μ ¸! ¸! * * c. c. y÷u! y ÷ u! * * d. d. + i÷ˆ÷u + I ÷ ÷ u * * ** ** Effort minimization is assumed to play a negligible role in the selection of F2 Upaya minimisasi diasumsikan memainkan peran diabaikan dalam pemilihan F2 contrasts in most contexts kontras dalam konteks yang paling 6 6 , but in other cases it may play a role in explaining why , Namun dalam kasus lain mungkin memainkan peran dalam menjelaskan mengapa languages do not avail themselves of maximally distinct contrasts. bahasa tidak menyediakan sendiri kontras maksimal yang berbeda. Another consequence of effort minimization is that difficult articulations should only Konsekuensi lain dari upaya minimisasi adalah bahwa artikulasi sulit hanya harus be employed in order to realize more distinct contrasts, so where contrasts are dipekerjakan dalam rangka mewujudkan kontras lebih jelas, jadi di mana kontras neutralized, considerations of effort minimization are likely to be dominant. dinetralkan, pertimbangan minimisasi usaha cenderung menjadi dominan. This leads to Hal ini menyebabkan the prediction that preferred vowel qualities should depend on contrastive status: in F2 prediksi bahwa kualitas vokal disukai harus tergantung pada status kontrastif: di F2 contrasts, front unrounded and back rounded vowels are preferred (8), but if all vowel F2 kontras, depan unrounded dan vokal kembali bulat lebih disukai (8), tetapi jika semua F2 vokal contrasts are neutralized, backness and rounding of vowels should be governed by effort kontras yang dinetralkan, backness dan pembulatan vokal harus diatur oleh usaha minimization. minimisasi. On the other hand, the basic markedness constraints in (2) are insensitive Di sisi lain, kendala markedness dasar dalam (2) tidak peka 6 6 Effort becomes a more significant factor where vowel duration is very short, and in the environment of Upaya menjadi faktor yang lebih signifikan di mana durasi vokal sangat pendek, dan di lingkungan

consonants that place strong constraints on F2, such as palatalized and velarized consonants (Flemming konsonan yang menempatkan kendala kuat pada F2, seperti konsonan palatalized dan velarized (Flemming 2001). 2001).

Page 13 Halaman 1313 13 to contrastive status, and consequently predict that peripheral vowels should be preferred status kontrastif, dan akibatnya memprediksi bahwa vokal perifer harus diutamakan in all circumstances. dalam segala situasi. Flemming (2001) discusses two test cases in which all F2 contrasts are neutralized: Flemming (2001) membahas kasus uji dua di mana semua kontras F2 dinetralkan: 'vertical' vowel inventories, and fully neutralizing vowel reduction in unstressed Persediaan vokal 'vertikal', dan sepenuhnya menetralisir pengurangan vokal dalam bertekanan syllables, as in English reduction to 'schwa'. suku kata, seperti dalam pengurangan Inggris untuk 'schwa'. Both cases conform to the predictions of the Kedua kasus sesuai dengan prediksi dispersion-theoretic analysis: backness and rounding of vowels assimilate to adjacent dispersi-teori analisis: backness dan pembulatan vokal berasimilasi dengan berdekatan consonants, often yielding central or centralized vowel qualities which would be highly konsonan, seringkali menghasilkan kualitas vokal pusat atau terpusat yang akan sangat marked in F2 contrasts, but are favoured by effort minimization. ditandai kontras F2, tetapi disukai oleh minimalisasi usaha. The best-known examples of 'vertical' vowel inventories, lacking F2 contrasts, are Contoh yang paling terkenal dari 'vertikal' persediaan vokal, kurang kontras F2, yang found in Northwest Caucasian languages such as Kabardian and Shapsug (Colarusso ditemukan dalam bahasa Northwest Kaukasia seperti Kabardian dan Shapsug (Colarusso 1988, 1992, Kuipers 1960, Smeets 1984). 1988, 1992, 1960 Kuipers, Smeets 1984). These languages are often described as having Bahasa ini sering digambarkan sebagai memiliki only central vowels, but this is a claim about the underlying vowel inventory posited as hanya pusat vokal, tapi ini adalah klaim tentang persediaan vokal yang mendasari mengemukakan sebagai part of a derivational analysis, not an observation about the surface vowels. bagian dari analisis derivatif, bukan sebuah pengamatan tentang vokal permukaan. On the Di surface, these languages have a system of five normal length vowels [i, e, a, o, u] permukaan, bahasa ini memiliki sistem lima vokal panjang normal [i, e, a, o, u] (Kuipers 1960:23f., Smeets 1984:123), and 'vertical' system of two extra short vowels, (Kuipers 1960:23 f, Smeets 1984:123.), Dan 'vertikal' sistem dua vokal pendek ekstra, which can be transcribed broadly as [, ] yang dapat ditranskripsi secara luas sebagai [,] 7 7 . . However, the precise backness and rounding Namun, tepat backness dan pembulatan of these vowels depends on context. ini vokal tergantung pada konteks. They are realized as a smooth transition between the Mereka menyadari sebagai transisi yang mulus antara

lip and tongue positions of the preceding and following consonants, deviating only to bibir dan lidah posisi konsonan sebelumnya dan berikutnya, menyimpang hanya untuk realize the required vowel height (Colarusso 1988:307). menyadari ketinggian vokal yang diperlukan (Colarusso 1988:307). An unrelated vertical vowel Sebuah vokal vertikal terkait language, Marshallese, is similar (Bender 1968, Choi 1992). bahasa, Marshall, mirip (Bender 1968, Choi 1992). The transitional vowel Vokal transisi qualities result from assimilation in backness and rounding to preceding and following kualitas hasil dari asimilasi dalam backness dan pembulatan untuk sebelum dan sesudahnya consonants, which is plausibly the least effort production strategy. konsonan, yang masuk akal strategi usaha produksi sedikit. The resulting vowel Vokal yang dihasilkan qualities are often central, back unrounded, front rounded, or short diphthongs involving kualitas sering sentral, kembali unrounded, depan bulat, atau diftong pendek melibatkan these qualities – all vowel types which would be highly marked in the presence of F2 kualitas ini - jenis vokal semua yang akan sangat ditandai dengan adanya F2 7 7 Kuipers actually transcribes the Kabardian high vowel as [], the mid-vowel as [a], and the 'long' low Kuipers sebenarnya mentranskripsi vokal tinggi Kabardian sebagai [], pertengahan-vokal sebagai [a], dan 'panjang' low vowel as [a#], and Colarusso (1988) follows him in this, but their descriptions, Colarusso's phonetic vokal sebagai [a #], dan Colarusso (1988) mengikutinya dalam hal ini, tapi deskripsi mereka, fonetik Colarusso s transcriptions, and acoustic data in Choi (1991) all indicate that the vowels are actually high and mid transkripsi, dan data akustik di Choi (1991) semua menunjukkan bahwa vokal sebenarnya tinggi dan mid respectively. masing.

Page 14 Halaman 1414 14 contrasts. kontras. There are no vertical vowel inventories containing the peripheral vowels that Tidak ada persediaan vokal vertikal yang berisi vokal perifer yang are predicted by the basic markedness constraints in (2) – ie there are no inventories diperkirakan oleh kendala markedness dasar dalam (2) - yaitu tidak ada persediaan such as [i, e, a] or [u, o, a] seperti [i, e, a] atau [u, o, a] 8 8 . . Neutralization of F2 contrasts is also observed in languages such as English where Netralisasi kontras F2 juga diamati dalam bahasa seperti bahasa Inggris di mana all vowel quality distinctions are neutralized to a 'schwa' vowel in some unstressed perbedaan kualitas vokal semua dinetralkan dengan vokal 'schwa' di beberapa bertekanan syllables. suku kata. This process is also found in Southern Italian dialects (Maiden 1995) and Dutch Proses ini juga ditemukan dalam dialek Italia selatan (Maiden 1995) dan Belanda (Booij 1995). (Booij 1995). Phonetic studies of schwa in Dutch (van Bergem 1994) and English Fonetik studi schwa dalam bahasa Belanda (van Bergem 1994) dan Inggris

(Kondo 1994) indicate that this vowel is comparable to a vertical vowel in that F2 is an (Kondo 1994) menunjukkan bahwa vokal ini sebanding dengan vokal vertikal dalam F2 adalah sebuah almost linear interpolation between values determined by the preceding and following hampir linier interpolasi antara nilai-nilai ditentukan oleh sebelum dan sesudahnya contexts. konteks. Again, schwa is a marked vowel where there are quality contrasts – it is often Sekali lagi, schwa adalah vokal ditandai di mana ada kontras berkualitas - sering excluded from those positions – but it is the unmarked vowel where all quality contrasts dikecualikan dari posisi tersebut - tetapi vokal ditandai dimana kualitas semua kontras are neutralized. dinetralkan. Basic markedness constraints predict that markedness should not depend Kendala markedness dasar memprediksi markedness yang tidak harus bergantung on contrastive status, so we should expect one of the peripheral vowels, [i, u] or [a], to be status kontrastif, jadi kita harus mengharapkan salah satu vokal perifer, [i, u] atau [a], untuk menjadi the sole vowel in neutralization contexts. vokal tunggal dalam konteks netralisasi. Distinctiveness constraints and basic markedness constraints are also differentiated Kekhasan kendala dan kendala markedness dasar juga dibedakan by predictions concerning enhancement. oleh prediksi mengenai peningkatan. Distinctiveness constraints predict that Kendala kekhasan memprediksi bahwa enhancement should only apply to contrasts, since enhancement is analyzed as a peningkatan seharusnya hanya berlaku untuk kontras, karena peningkatan dianalisis sebagai consequence of constraints on the distinctiveness of contrasts. konsekuensi dari kendala pada kekhasan kontras. This is inherent in Stevens Hal ini melekat dalam Stevens et al (1986) conception of enhancement, but it is not predicted by analyses in terms of et al (1986) konsepsi peningkatan, tetapi tidak diprediksi oleh analisis dari segi basic markedness constraints, because the latter are insensitive to contrast. dasar markedness kendala, karena yang terakhir tidak sensitif terhadap kontras. Evidence on Bukti this point comes from enhancement of stop voicing contrasts (Flemming 2001). titik ini berasal dari peningkatan kontras menyuarakan berhenti (Flemming 2001). Stevens Stevens et al (1986:439) argue that pre-nasalization can serve as an enhancement of stop voicing. et al (1986:439) berpendapat bahwa pra-nasalisasi dapat berfungsi sebagai peningkatan berhenti menyuarakan. One of the cues that distinguishes voiced stops from voiceless stops is the presence of Salah satu isyarat yang membedakan berhenti disuarakan dari berhenti bersuara adalah adanya voicing during the closure, as opposed to the silence of a voiceless stop closure (Stevens menyuarakan selama penutupan tersebut, yang bertentangan dengan keheningan penutupan berhenti bersuara (Stevens 8 8 It might be suggested that vertical vowels are phonologically unspecified for [back] and [round] rather Ini mungkin menyarankan bahwa vokal vertikal fonologis ditentukan untuk [kembali] dan [putaran] lebih

than being specified for the marked vowel qualities described here (Choi 1992). daripada yang ditentukan untuk kualitas vokal ditandai dijelaskan di sini (Choi 1992). However, such unspecified Namun, seperti ditentukan vowels only occur in the absence of F2 contrasts, so this would imply an even more dramatic change from vokal hanya terjadi karena tidak adanya kontras F2, jadi ini akan menyiratkan perubahan yang lebih dramatis dari a preference for peripheral vowels in F2 contrasts to a preference for otherwise unattested unspecified preferensi untuk vokal perifer dalam kontras F2 untuk preferensi untuk sebaliknya unattested ditentukan vowels where there is no contrast. vokal di mana tidak ada kontras.

Page 15 Halaman 1515 15 and Blumstein 1981), so the distinctiveness of this contrast can be increased by dan Blumstein 1981), sehingga kekhasan kontras ini dapat ditingkatkan dengan increasing the intensity of voicing. meningkatkan intensitas menyuarakan. This can be achieved by lowering the velum during Hal ini dapat dicapai dengan menurunkan velum selama the early part of the stop closure, yielding a pre-nasalized stop. bagian awal dari penutupan berhenti, berhenti menghasilkan pra-ternasal. It is generally difficult to Hal ini umumnya sulit untuk sustain voicing during a stop because air pressure builds up behind the closure, and when mempertahankan menyuarakan selama berhenti karena tekanan udara menumpuk di belakang penutupan, dan ketika oral pressure approaches sub-glottal pressure, airflow through the glottis ceases, and Tekanan lisan pendekatan sub-glotal tekanan, aliran udara melalui glottis berhenti, dan voicing ceases (Ohala 1983, Westbury and Keating 1986). menyuarakan berhenti (Ohala 1983, Westbury dan Keating 1986). Lowering the velum during Menurunkan velum selama the stop closure allows air to be vented through the nose, slowing the build up of oral penutupan berhenti memungkinkan udara untuk dibuang melalui hidung, memperlambat membangun oral pressure, and thus facilitating voicing. tekanan, dan dengan demikian memfasilitasi menyuarakan. In addition, voicing during an oral stop is radiated Selain itu, menyuarakan selama berhenti lisan dipancarkan only through the neck and face, resulting in a low intensity acoustic signal, whereas hanya melalui leher dan wajah, sehingga sinyal akustik intensitas rendah, sedangkan lowering the velum allows sound to be radiated from the nose, resulting in greater menurunkan velum memungkinkan suara yang akan terpancar dari hidung, sehingga lebih intensity. intensitas. Pre-nasalization serves as an enhancement of stop voicing contrasts in Mixtec Pra-nasalisasi berfungsi sebagai peningkatan kontras berhenti menyuarakan di Mixtec (Iverson and Salmons 1996), Southern Barasano (Smith and Smith 1971), Guaraní (Iverson dan Salmons 1996), Southern Barasano (Smith dan Smith 1971), Guaraní

(Gregores and Suárez 1967), and a variety of other languages discussed by Herbert (Gregores dan Suárez 1967), dan berbagai bahasa lain dibahas oleh Herbert (1986:16ff.) – that is, voiceless stops are contrasted with pre-nasalized stops rather than (1986:16 ff.) - Yaitu, berhenti bersuara dikontraskan dengan pra-ternasal berhenti daripada plain voiced stops. polos menyuarakan berhenti. But voiced stops are never enhanced by prenasalization where they do Tapi berhenti bersuara tidak pernah ditingkatkan oleh prenasalization di mana mereka melakukan not contrast with voiceless stops. tidak kontras dengan berhenti bersuara. Non-contrastive voiced stops can arise through inter- Non-kontrastif berhenti bersuara dapat timbul melalui inter- vocalic voicing, a pattern where voiced stops are found between vowels ([ada], not vokal menyuarakan, pola mana berhenti bersuara yang ditemukan antara vokal ([ADA], tidak *[ata], but only voiceless stops occur elsewhere ([ta], not *[da]). * [Ata], tetapi hanya berhenti bersuara terjadi di tempat lain ([ta], bukan * [da]). However we do not find Namun kita tidak menemukan intervocalic prenasalization of stops (ie prenasalized stops between vowels, but only intervocalic prenasalization berhenti (berhenti prenasalized yaitu antara vokal, tapi hanya voiceless stops elsewhere) bersuara berhenti di tempat lain) 9 9 . . These generalizations are very difficult to account for with simple markedness Ini generalisasi sangat sulit untuk menjelaskan dengan sederhana markedness constraints. kendala. The existence of languages which have pre-nasalized stops but not plain Keberadaan bahasa yang memiliki pra-ternasal berhenti tetapi tidak biasa voiced stops shows that some markedness constraint must favour pre-nasalized stops over berhenti menyuarakan menunjukkan bahwa beberapa kendala markedness harus mendukung pra-ternasal berhenti selama voiced stops, eg P bersuara berhenti, misalnya P RENASALIZE RENASALIZE 'voiced stops should be prenasalized'. 'Berhenti menyuarakan harus prenasalized'. Then a language Kemudian bahasa 9 9 See Kingston and Diehl (1994) for a related argument that voicing-dependent perturbations of F Lihat Kingston dan Diehl (1994) untuk sebuah argumen terkait yang menyuarakan tergantung gangguan dari F o o adjacent berdekatan to stops are active enhancements of stop voicing contrasts, so these effects are reduced or absent where untuk berhenti adalah perangkat aktif menyuarakan kontras berhenti, sehingga efek ini berkurang atau tidak ada di mana there is no voicing contrast. tidak ada kontras menyuarakan.

Page 16 Halaman 1616 16

with voiceless stops and prenasalized stops (like Mixtec) would be derived by ranking dengan berhenti bersuara dan berhenti prenasalized (seperti Mixtec) akan diturunkan dengan peringkat this constraint above faithfulness to [nasal] (10) so any voiced stops in the input are ini kendala di atas kesetiaan [hidung] (10) sehingga setiap berhenti bersuara di input adalah replaced by prenasalized stops. digantikan oleh berhenti prenasalized. (10) P (10) P RENASALIZE RENASALIZE >> I Saya >> DENT Dent [nasal] [Hidung] However, this ranking derives prenasalization of voiced stops even where voicing is Namun, peringkat ini berasal prenasalization berhenti bersuara bahkan di mana menyuarakan adalah not contrastive. tidak kontrastif. For example, if intervocalic voicing of stops follows from ranking a Misalnya, jika intervocalic menyuarakan berhenti berikut dari peringkat satu constraint against voiceless stops occurring between vowels (*VTV) above faithfulness kendala terhadap berhenti bersuara terjadi antara vokal (* VTV) atas kesetiaan to voicing (11), then this ranking can be combined with the prenasalization ranking in untuk menyuarakan (11), maka peringkat ini dapat dikombinasikan dengan peringkat prenasalization di (10) to derive the unattested pattern of intervocalic prenasalization, as shown in (12). (10) untuk mendapatkan pola unattested dari prenasalization intervocalic, seperti pada (12). (11) *VTV >> I (11) * VTV >> I DENT Dent [voice] [Suara] (12) (12) /ata/ / Ata / *VTV * VTV P P RE RE - - NASALIZE Membuat sengau I Saya DENT Dent [nasal] [Hidung] I Saya DENT Dent [voice] [Suara] a. a. ata ata *! *! b. b. ada ADA *! *!

* * c. c. + a<da + A <da * * * * This consequence is avoided if the constraint P Konsekuensi ini dihindari jika P kendala RENASALIZE RENASALIZE is replaced by constraints diganti dengan kendala favouring maximally distinct voicing contrasts, eg *TD >> *T- mendukung kontras menyuarakan maksimal yang berbeda, misalnya * TD >> * T- N N D (where T, D, and D (di mana T, D, dan N N D D represent voiceless, voiced and prenasalized stops, respectively). mewakili bersuara, berhenti bersuara dan prenasalized, masing-masing). These distinctiveness Kekhasan ini constraints only apply to contrasts, so prenasalization of voiced stops is correctly kendala hanya berlaku untuk kontras, sehingga prenasalization berhenti bersuara dengan benar predicted to occur only where there are voicing contrasts. diperkirakan hanya terjadi di mana ada yang menyuarakan kontras. Elsewhere voiced stops are Di tempat lain berhenti menyuarakan adalah preferred over prenasalized stops because voiced stops are simpler articulatorily. lebih disukai daripada berhenti prenasalized karena berhenti bersuara lebih sederhana articulatorily. These, and other examples discussed in Flemming (2001), indicate that phonology Ini, dan contoh-contoh lain yang dibahas dalam Flemming (2001), menunjukkan bahwa fonologi includes distinctiveness constraints, as we would expect if considerations of ease of termasuk kendala kekhasan, seperti yang kita harapkan jika pertimbangan kemudahan perception influence phonology. Persepsi pengaruh fonologi. Basic markedness constraints do not follow from Kendala markedness dasar tidak mengikuti dari perceptual considerations and cannot account for dispersion effects because dispersion persepsi pertimbangan dan tidak dapat menjelaskan efek dispersi karena dispersi

Page 17 Halaman 1717 17 applies only to contrasts while basic markedness constraints are indifferent to the hanya berlaku untuk sementara kontras kendala markedness dasar acuh tak acuh terhadap contrastive status of a sound. kontrastif status suara. 3. 3. Licensing by Cue Perizinan oleh Cue A second source of evidence for perceptual constraints is the typology of contextual neutralization. netralisasi. Contextual neutralization is a pattern of distribution in which a contrast is permitted in some environments, but is suspended in others. For example stop voicing contrasts may be permitted before sonorants ([ b a] vs. [ p a], [ b la] vs. [ p la]), but not before obstruents ([a p ta], *[a b ta]). In a situation like this, the voicing contrast is said to be

neutralized before obstruents. Steriade (1995, 1999) observes that different types of contrast have different characteristic environments of neutralization. For example the following are well-attested patterns of distribution for three types of contrasts, following Steriade (1999) 10 10 : : (13) (i) Obstruent voicing contrasts are permitted only before sonorants (eg German, Lithuanian, Russian, Sanskrit). (ii) Major place contrasts (labial vs. coronal vs. dorsal) are permitted only before vowels (eg Japanese, Luganda, Selayarese). (iii) Retroflexion contrasts (retroflex vs. apical alveolar) are permitted only after vowels (eg Gooniyandi, Miriwung, Walmatjari). Steriade argues that the general characterization of these diverse contexts of neutralization makes crucial reference to perceptual distinctiveness: in each case, the contrasts are neutralized first in environments where 'the cues to the relevant contrast would be diminished or obtainable only at the cost of additional articulatory maneuvers' (Steriade 1997:1). Contrasts differ in their distribution of cues so they are subject to different patterns of neutralization. This is dubbed the 'Licensing by Cue' hypothesis – 10 10 References: German, Lithuanian, Russian, Sanskrit: Steriade (1997), Japanese: Itô (1989), Luganda: Tucker (1962), Selayarese: Mithun and Basri (1986), Gooniyandi: McGregor (1990), Miriwung: Hamilton (1996), Walmatjari: Hudson and Richards (1969).

Page 18 Halaman 1818 18 the presence of a contrast in a particular environment is licensed by the availability of perceptual cues to that contrast. For example, the distribution of obstruent voicing contrasts (12i) is analyzed in these terms by Steriade (1997). One of the primary cues to obstruent voicing distinctions is Voice Onset Time (VOT), the lag between the release of the obstruent constriction and the onset of voicing (Lisker and Abramson 1970). Steriade observes that this cue is generally only available where a voiced sonorant follows, and so is absent before obstruents, and in word-final position before pause. Voicing contrasts in these environments can only be realized by cues such as voicing during the consonant constriction, consonant duration, and duration of the preceding vowel, which are hypothesized to be weaker cues than VOT. So according to this analysis, languages like Russian and German disallow voicing contrasts in precisely the environments where a key cue to the contrast, VOT, is unavailable. Given the importance of VOT as a cue to obstruent voicing, it is very plausible that voicing is less confusable before sonorants than before obstruents or word-finally, but there is surprisingly little direct evidence on this point. titik. Studies of voicing perception generally have not directly compared perception of voicing in different contexts. Similar factors have been argued to explain restrictions on the distribution of major

place contrasts (labial vs. coronal vs. dorsal). These contrasts preferentially occur where there is a following vowel, or, failing that, a following approximant. A number of studies have shown that major place distinctions are less confusable in pre-vocalic position than in pre-consonantal or pre-pausal position (Redford and Diehl 1999, Wright 2001). Fujimura et al (1978) and Ohala (1990) have also shown that release cues to major place contrasts dominate over closure cues in stimuli that have been edited so that these cues conflict. konflik. This difference in distinctiveness appears to have multiple causes. The greater Semakin besar distinctiveness of prevocalic stops may be attributed to the presence of the release burst which provides cues to place, in addition to the formant transition cues that are also

Page 19 Halaman 1919 19 available in post-vocalic position (Dorman et al 1977). Consonants clusters are often articulatorily overlapped so the constriction of a second consonant is formed before the constriction of the first consonant is released. Where the second consonant is an obstruent, this results in the loss or attenuation of the release burst of the first consonant (Henderson and Repp 1982). Another factor that has been suggested to contribute to the greater distinctiveness of prevocalic place contrasts is the nature of the peripheral auditory system (Wright 1996, 2001). 2001). Auditory nerve fibres respond most strongly to rapid rises from low intensity within their frequency band, and the transition from a consonant to a vowel often involves rapid onsets of this kind, especially where the consonant is an obstruent (Delgutte and Kiang 1984, Greenberg 1995). This effectively amplifies release formant transitions and stop bursts. As noted by Ohala (1990:261f.), experiments by Fujimura et al (1978) support an auditory-perceptual basis for the greater distinctiveness of onset consonants: they found that in stimuli with conflicting cues to place, release cues dominated closure cues, even when the stimuli were played backwards – ie the release cues were reversed closure transitions. However, Redford and Diehl (1999) also found that the formant transitions of onsets were more distinctly articulated than word-final consonants, so production differences may play a role in explaining the observed difference in distinctiveness. The patterns of distribution of obstruent voicing and major place contrasts are broadly similar in that both preferentially occur before sonorants, but there are differences of detail, some of which follow from differences in the nature of the cues to these two types of contrast. When obstruent voicing contrasts are permitted before sonorant consonants, they are allowed before all sonorants, whereas major place contrasts are usually subject to further restrictions. For example, many languages, including English, do not allow coronal stops before coronal laterals, although labials and velars contrast in this environment: plan, clan, *tlan (Kawasaki 1982:14).

Page 20 Halaman 2020 20 The insensitivity of voicing contrasts to the nature of a following sonorant is expected given that the primary cue to voicing is VOT. The realization of VOT depends

only on the presence of a voiced sonorant of sufficient duration; place of articulation, nasality, and laterality make little difference. On the other hand, primary cues to stop place contrasts are the release burst and formant transitions. Approximants and vowels allow the realization of both, but simply realizing a burst and formant transitions is not adequate to support contrast: the burst and/or formant transitions must be distinct for contrasting places of articulation. The distinctiveness of these cues can be affected by coarticulation with the following vowel or approximant. Kawasaki (1982:157f.) and Flemming (2002:132ff.) argue that these factors underlie the restrictions on coronal stops before laterals. That is, coarticulation effects make the burst and formant transitions of coronals insufficiently distinct from velars in this context. konteks. The lateral constrains the position of the tongue tip and body, so the formant transitions in coronal-lateral and velar-lateral clusters are very similar, while a labial is generally distinguished by lower F2 due to lip constriction (Kawasaki 1982:67ff., Olive, Greenwood, and Coleman 1993:284). The coronal and velar closures are at or behind the location of the lateral constriction, so in both cases frication noise is generated at this lateral constriction at release, resulting in acoustically similar bursts. A more striking example of how distribution of contrasts differs depending on the nature of the cues involved comes from the comparison between major place contrasts and retroflexion contrasts (Steriade 1995, 2001). The contrast between retroflex and apical alveolar consonants is found in many Australian and Dravidian languages. It is Sekarang commonly restricted to positions following a vowel, so it is neutralized word-initially and following consonants (Steriade 1995). This is in sharp distinction from most other place contrasts, which, as we have seen, occur preferentially before vowels. Steriade argues that this difference follows from differences in the distribution of cues to these types of contrasts. Retroflexes are distinguished from apical alveolars by a low third formant at

Page 21 Halaman 2121 21 closure (Stevens and Blumstein 1975). However, the tongue tip moves forward during the closure of a retroflex and is released at the alveolar ridge, so these sounds are articulatorily and acoustically very similar at release (Dave 1977, Butcher 1995, Anderson 1997, Spajic, Ladefoged and Bhaskararao 1994). Closure transitions are only available where the consonant is preceded by a vowel, hence this cue is missing in other environments, making the contrast less distinct (Anderson 1997). So the retroflexion contrast differs from other place contrasts in that it is realized most distinctly on a preceding vowel rather than a following vowel, but given this difference we can see that all place contrasts are liable to neutralize in environments where it would be difficult to make them distinct. It should be noted that the patterns of distribution described for major place and obstruent voicing contrasts have often been analyzed as involving neutralization of contrasts in the coda of syllables (eg Itô 1989, Vennemann 1972). Steriade (1997, 1999) and Côté (2000) present detailed arguments in favour of the 'Licensing by Cue' analysis. In the present context, the important weaknesses of a coda-neutralization account are that it does not extend to retroflexion contrasts, which are neutralized in word-initial and post-

consonantal onsets, but are permitted in codas and intervocalic onsets, and that it cannot relate the patterns of distribution to the nature of the features involved. While the analyses sketched above indicate that considerations of distinctiveness play a central role in accounting for the distribution of contrasts, it is clear that other constraints are important also. For example, stop bursts will only be absent before obstruents if some constraint requires the stop closure to overlap with the following consonant. konsonan. One general phenomenon that implicates additional constraints is word-final neutralization. netralisasi. For example, in German, obstruent voicing is neutralized preceding obstruents, and in word-final position. For words spoken in isolation, these are both environments in which VOT cues are unavailable, because there is no following sonorant, but in phrase-medial position, a word-final obstruent might be followed by a sonorant,

Page 22 Halaman 2222 22 allowing the realization of VOT differences. If contrast is governed strictly by the availability of cues, the voicing contrast should be permitted in this context, but in German, and many similar languages, voicing is neutralized in word-final position, regardless of phrasal context. So the analysis in terms of licensing by cue must be supplemented by additional constraints relating to morphosyntactic structure. Steriade (1997) analyzes this pattern as resulting from generalization of the citation form of words. kata. That is, there is a general preference to give words a uniform pronunciation in all contexts, and this is modeled on the pronunciation of the word spoken in isolation. This Ini analysis is formalized in terms of Output-Output Correspondence constraints (Benua 1997, Kenstowicz 1996, Steriade 2000). A comparable distinction between word-internal and cross-word sequences must be made in syllabification-based analyses in order to block syllabification of a word-final consonant as an onset to a following vowel-initial word. kata. 3.1 Formalizing Licensing by Cue Steriade (1997, 1999) formalizes the Licensing by Cue hypothesis in terms of constraints on the distinctiveness of contrasts. Although the specifics are rather different from dispersion theory, the general conception is very similar, so the same constraints motivated above in the analysis of enhancement can be used to analyze patterns of contextual neutralization (Flemming 2002:40ff.) 11 11 . . 3.1.1 Distinctiveness constraints In the case of obstruent voicing, we will assume that there is a perceptual dimension corresponding to VOT, which takes a value of 0 for voiced and 1 for voiceless 11 11 Steriade proposes constraints of the form *avoice/C that penalize obstruent voicing contrasts in a

particular context, C. These constraints are ranked according to the richness of cues to voicing available in that context. These constraints are replaced here by M INDIST constraints that refer directly to the cues that differentiate contrasting obstruents. This allows for variability in the cues realized in a given context, depending on the production strategy adopted (cf. Steriade 1997, Koontz-Garboden 2002).

Page 23 Halaman 2323 23 obstruents 12 12 . . Languages that restrict voicing contrasts to pre-sonorant positions require a VOT difference for the contrast to be adequately distinct. In other words, Dengan kata lain, M M INDIST =VOT:1 ranks above M AXIMIZE CONTRASTS . . This sets a threshold for minimum distinctiveness that can be met in pre-sonorant position, so a voicing contrast is permitted in that environment (14). (14)i . . _V M M INDIST = VOT:1 M M AXIMIZE CONTRASTS *[+voice, -son] a. a. F F dV-tV 44 44 * * b. b. dV dV 4! 4! * * c. c. tV 4! 4!

Pre-pausally, no VOT difference is possible, and a difference in closure voicing alone is insufficient, so a voicing contrast is not permitted (15, # indicates a word boundary). Given that there is no contrast, obstruents are realized with the least-effort laryngeal state. In pre-pausal position, this is voiceless, given the difficulties involved in maintaining vocal fold vibration during an obstruent (cf. §2.3). This preference is formalized as a constraint against voiced obstruents, *[+voice, -sonorant]. (15)i . . V_# M M INDIST = VOT:1 M M AXIMIZE CONTRASTS *[+voice, -son] a. a. Vd#-Vt# *! *! 44 44 * * b. b. Vd# 4 4 *! *! c. c. F F Vt# 4 4 Voicing contrasts are also neutralized before obstruents, because VOT differences cannot be realized in this position either (17). However, in this case, the neutralized stop is voiced, assimilating to the following obstruent (in this case [g]). A plausible analysis of this pattern is that it is especially difficult to initiate voicing during an obstruent – due to hysteresis effects it is easier to maintain voicing from a sonorant into a following 12 12 In fact there are two basic types of obstruent 'voicing' contrasts: fully voiced vs. voiceless unaspirated, and voiceless unaspirated vs. aspirated, so a more general analysis requires three levels of VOT (Flemming 2002). 2002).

Page 24 Halaman 2424 24 obstruent than it is to initiate voicing during an obstruent following a voiceless sound (Westbury and Keating 1986). Thus we can posit the constraint in (16), named *TD for brevity, universally ranked above *[+voice, -son].

(16) *TD: *[-voice][+voice, -sonorant] This constraint forces assimilation in obstruent sequences, as shown in the following tableau: (17) V_gV M M INDIST = VOT:1 M M AXIMIZE CONTRASTS *TD *[+voice, -son] a. a. VdgV-VtgV *! *! 44 44 * * * * b. b. + + VdgV 4 4 *! *! c. c. VtgV 4 4 *! *! Languages with broader distribution of obstruent voicing rank M AXIMIZE CONTRASTS above M INDIST =VOT:1, tolerating less distinct contrasts in order to realize more contrasts. But no language will prefer less distinct contrasts over more distinct contrasts of a similar type. Neutralization of retroflexion is analyzed along similar lines: a M INDIST constraint requiring an F3 difference is ranked above M AXIMIZE CONTRASTS , so the contrast is neutralized where it is not possible to realize this cue. 3.1.2 Faithfulness constraints An alternative approach to formulating the perceptual constraints that account for these generalizations about the distribution of contrasts makes use of faithfulness constraints (Steriade 1995, 2001, Jun 1995, Boersma 1998). This is a natural move since faithfulness constraints play a central role in the regulation of contrasts in standard OT. Essentially, a faithfulness constraint like I

DENT Dent F, where F is a feature, favours preserving underlying differences – if the input contains [+F], the output should contain [+F], if the input contains [-F], the output should contain [-F]. So if I DENT Dent F is satisfied, an

Page 25 Halaman 2525 25 underlying difference between [+F] and [-F] is preserved on the surface, and the language has a contrast in F. Perceptual factors are introduced by distinguishing I DENT Dent F constraints for different contexts, then ranking them according to the distinctiveness of an F contrast in that context. konteks. For example, we might posit the ranking of I DENT Dent [voice] constraints in (18). (18)I DENT Dent [voice]/ _ [+son] >> I DENT Dent [voice]/ _ # >> I DENT Dent [voice]/ _ [-son] The distribution of voicing contrasts is then determined by the position of a constraint against voiced obstruents, *[+voice, -son]. For example, the ranking in (19) derives neutralization everywhere except before sonorants (the German pattern) 13 13 . . If Jika *[+voice, -son] is ranked lower, then the contrast is permitted in more positions, but again contrasts are permitted first in more distinct environments. These constraints Kendala-kendala predict that neutralization always yields voiceless obstruents, so an additional constraint, such as *TD, is required to derive assimilation to following obstruents. (19)I DENT Dent [vce]/ _ [+son] >> *[+voice, -son] >> I DENT Dent [voi]/ _ # >> I DENT Dent [voi]/ _ [-son] This approach works elegantly in simple cases, but it has some limitations that make it incapable of providing a comprehensive account of perceptual effects. The fundamental Fundamental limitation of faithfulness constraints is that they can only block change between input and

output, they cannot motivate change. This is problematic because there are various phenomena that have been argued to be perceptually-motivated which crucially involve unfaithfulness to input forms, including the dispersion phenomena discussed in §2. For Untuk example, a language with only the peripheral vowels [i, e, a, o, u] must unfaithfully map non-peripheral input vowels such as [y, µ] onto one of these vowels. Ranking Peringkat I Saya DENT Dent [round] low in the constraint hierarchy, for example, makes it relatively acceptable 13 13 This analysis is structurally very similar to the one proposed in Lombardi (1995), but Lombardi employs a faithfulness constraint specific to pre-sonorant onsets.

Page 26 Halaman 2626 26 to realize [y, µ] as [i] and [u] respectively, but it does not favour these realizations. Unfaithful mappings can only be motivated by markedness constraints, and as we have seen above, the markedness constraints that best account for this pattern are distinctiveness constraints implementing a preference for maximally distinct F2 contrasts. The same applies to other cases of dispersion and enhancement. For example, Misalnya, enhancement of voicing contrasts by pre-nasalizing voiced stops (§2.3) implies unfaithful realization of input voiced stops as pre-nasalized stops, which must be motivated by a markedness constraint. More generally, perceptually-ranked featural faithfulness constraints can only account for patterns of neutralization, but arguably neutralization is just one way of avoiding an otherwise indistinct contrast. That is, an indistinct contrast may be avoided by giving up the contrast (neutralization), or by making the contrast more distinct (enhancement). We have seen that distinctiveness constraints can be used to derive both patterns, but perceptually-ranked faithfulness constraints can only derive neutralization. This limitation applies not only to the analysis of segment-internal enhancements of the kind just discussed, but also to modification of the environment of a contrast (cf. Côté 2000:175f., Hume and Johnson 2001:8f.). For example, it has been suggested that vowel epenthesis is often motivated by the need to make consonant contrasts more distinct (eg Wright 1996:40, Côté 2000). One such pattern is epenthesis into clusters of three consonants, exemplified from Yawelmani Yokuts (Newman 1944, Kisseberth 1970) in (20). dalam (20). Similar patterns are observed in Cairene Arabic (Broselow 1976) and Lenakel (Lynch 1978). (20) /pa?t+mi/ ‡ ‡ [pa?itmi] 'having fought' cf./pa?t+al/ ‡ ‡ [pa?tal]

'might fight' /lihm+mi/ ‡ ‡ [lihimmi] 'having run' cf./lihm+al/ ‡ ‡ [lihmal] 'might run'

Page 27 Halaman 2727 27 Côté (2000) analyzes this pattern in terms of the markedness constraint CV: 'A consonant is adjacent to a vowel' – that is, epenthesis applies to ensure that every consonant is adjacent to a vowel, which is not the case in a triconsonantal cluster. Formally, epenthesis is derived by ranking CV above D EP EP V, the faithfulness constraint that is violated by inserting a vowel 14 14 (see Kager 1998:107ff. for a similar analysis based on syllabification constraints). As Côté argues, it is perceptually desirable for consonants to be adjacent to a vowel because many consonantal contrasts are best realized in this position. As noted above, formant transitions are important place cues that are best realized on a vowel. The Itu contrast between presence and absence of a consonant is also more distinct adjacent to a vowel because the change in constriction between consonant and vowel results in salient spectral discontinuities (Liu 1996, Ohala 1980, Stevens 1998:245f.). The nature of the Sifat spectral change, eg the rate and magnitude of change in different frequency bands, may also provide cues to consonant manner (Stevens 1985, Liu 1996) 15 15 . . This analysis cannot be implemented in terms of perceptually-ranked faithfulness constraints. kendala. Ranking constraints against consonant deletion (M AX AX C) according to the strength of the cues to the presence of a consonant can only allow deletion of poorly-cued consonants, it cannot motivate epenthesis to improve the cues to a consonant. The Itu unfaithful insertion of a vowel can only be motivated by a markedness constraint violated by triconsonantal clusters, such as CV. Perceptually-ranked faithfulness constraints favour perceptually minimal changes between input and output. This arrangement predicts that indistinct contrasts are more likely to be lost because they can be neutralized by perceptually minimal changes, but it 14 14 D D

EP EP V must also be outranked by M AX AX C, the offending consonant is deleted rather than being rescued by vowel epenthesis. Deletion in triconsonantal clusters is observed in a number of languages, eg Korean (Kim and Shibatani 1976). 15 15 It should be noted that these kinds of considerations properly motivate constraints requiring consonant contrasts to be realized adjacent to vowels (ie distinctiveness constraints), as discussed at length in §2.3. The limitations of basic markedness constraints are addressed further in §3.1.3.

Page 28 Halaman 2828 28 does not implement a general preference for distinct contrasts, and so cannot account for the observed range of perceptually-motivated phenomena. However, there is evidence Namun, ada bukti that perceptually minimal change between input and output is preferred in alternations (ie contextual variation in the realization of morphemes) (Steriade forthcoming), so perceptual ranking of faithfulness constraints may be motivated on independent grounds. 3.1.3 Sound change via misperception The limitations of perceptually-ranked faithfulness constraints are shared by some theories that locate perceptual constraints in the process of sound change rather than in synchronic grammars (eg Blevins and Garrett 1998, Ohala 1990). According to these Menurut para accounts indistinct contrasts appear to be dispreferred in languages because they are more likely to be lost over time through misperception on the part of language users. For Untuk example, Ohala (1990) argues that consonants often assimilate in place to a following consonant (eg anka > aNka) because the unassimilated cluster is easily misperceived as the assimilated cluster. This is related to the observation above that post-vocalic major place contrasts are relatively indistinct, but according to Ohala this pattern results from '“innocent” misapprehension' on the part of listeners, so no dispreference for indistinct contrasts needs to be encoded in grammars. Sound change through misperception, like perceptually-ranked faithfulness constraints, can only hope to account for neutralization, not dispersion or enhancement. For example, at least some cases in which stop voicing contrasts are enhanced by prenasalization of voiced stops (§2.3) seem to have arisen via a sound change from earlier voiced stops to prenasalized stops (Herbert 1986:16ff.). This change cannot be attributed to misperception, rather prenasalization seems to be a strategy that speakers have hit upon to make stop voicing contrasts more distinct, so a preference for distinct contrasts is necessary to account for this pattern. In general, a mechanism of sound change via misperception only predicts that less distinct contrasts are more likely to be

Page 29 Halaman 29

29 29 lost, it cannot account for cases in which speakers appear to take measures to increase the distinctiveness of contrasts – ie dispersion and enhancement phenomena (cf. Steriade 2001:233ff. for a similar argument). Relating sound change directly to patterns of misperception also incorrectly predicts some unattested sound changes. For example, a study of vowel confusions in French (Robert-Ribes et al 1998) found that [i] is confused with [y] much more frequently than it is confused with [u]. This difference in distinctiveness is expected, and is hypothesized to lie behind the cross-linguistic preference for contrasts like [iu] over front rounding contrasts like [iy]. An 'innocent misapprehension' model might attribute this preference to the greater tendency for [iy] contrasts to be lost through misperception. However, the Namun, study found that [i] is misidentified as [y] at about the same rate as the converse misidentification of [y] as [i]. So if sound changes arise from misperception, we would expect a change [i] > [y] to be as likely as [y] > [i], but while the latter change is well attested (eg in Old English (Lass and Anderson 1975:286ff.) and Greek (Newton 1972:19)), unconditioned rounding of front vowels is unattested. Significantly, Secara signifikan, unrounding front vowels increases the distinctiveness of front-back contrasts, while the unattested change would reduce distinctiveness without any compensatory benefits. 3.1.4 Basic markedness constraints A final approach to formalizing the perceptual constraints responsible for contextual neutralization is to use basic markedness constraints. For example, in the analysis of voicing neutralization reviewed above, Steriade (1997) proposes a distinctiveness constraint against obstruent voicing contrasts appearing where there is no following sonorant. The closest equivalent basic markedness constraint would be a constraint requiring voiced obstruents to be followed by sonorants (cf. Lombardi 1995, 1999). Constraints of this kind are widely used in the analysis of contextual neutralization (McCarthy 2002:87), but usually without appealing to any perceptual motivation.

Page 30 Halaman 3030 30 However, some researchers have used basic markedness constraints to formalize perceptually motivated constraints (eg Côté 2000, Hume 1998). We saw in section 2.3 that basic markedness constraints are inadequate for the analysis of dispersion effects, and are difficult to motivate on perceptual grounds because perceptual considerations disfavour indistinct contrasts, not individual sounds. Similar Mirip difficulties face the use of basic markedness constraints in the analysis of Licensing by Cue effects. A basic constraint on obstruent voicing must ban [+voice] or [-voice] rather than the contrast between them. This is not only perceptually unmotivated, it leads to empirical difficulties. For example, it is common for the result of neutralization to be phonetically distinct from either of the sounds that occurs in positions of contrast (cf. Trubetzkoy 1939:71-3). This is the case in the neutralization of retroflexion contrasts, for example. contoh. Butcher (1995) studied several Australian languages that contrast retroflex and

apical alveolar consonants, and found that neutralization of this contrast in word-initial position yields an intermediate consonant, generally postalveolar (unlike apical alveolars), but apical rather than sub-laminal (the contrastive retroflexes are sub-laminal). This intermediate status is reflected in uncertainty among Australianists as to the appropriate transcription for these sounds (Butcher 1995, Steriade 1995). If the Jika distinction between retroflexes and apical alveolars is treated as binary (eg [+/- anterior]), then formulating a constraint against either retroflexes or apical alveolars in word initial position predicts that the other articulation should be favored in neutralization, which is not accurate, since an intermediate articulation is actually observed. diamati. If we make a three-way distinction between apical alveolars, retroflexes, and an intermediate articulation, then it is possible to formulate constraints against either extreme appearing in word-initial position, but it would also be necessary to prevent the intermediate place from surfacing in environments of contrast. These problems are avoided if we recognize that it is the contrast between retroflexes and apical alveolars that is problematic in word-initial position. In the absence of contrast, the intermediate

Page 31 Halaman 3131 31 articulation is preferred as less effortful than a sub-laminal retroflex, but more distinct from laminal coronals than an apical alveolar. 4. 4. Conclusions Kesimpulan The evidence reviewed here leads to the conclusions that (i) speech perception does play a role in shaping phonological patterns, and (ii) the relevant constraints are constraints on the distinctiveness of contrasts. We have examined two types of phonological patterns that have been related to the perceptual properties of speech sounds: dispersion/enhancement and contextual neutralization. netralisasi. Both phenomena can be analyzed in terms of a preference for more distinct contrasts and the converse dispreference for indistinct contrasts. Dispersion of contrastive sounds in perceptual space is a direct consequence of maximization of distinctiveness, while enhancement phenomena simply reflect the fact that greater distinctiveness is often achieved by covarying physiologically unrelated articulations such as tongue body backness and lip rounding. Contextual neutralization also follows from the preference for distinct contrasts given the fact that the distinctiveness of a contrast type varies according to context. For example, obstruent voicing contrasts are more distinct before a sonorant than in other environments, so some languages only allow voicing contrasts before sonorants, neutralizing the contrast elsewhere. So the two patterns are fundamentally similar: a language with front unrounded and back rounded vowels avoids the less distinct contrasts between front rounded and back rounded vowels, and a language that only allows obstruent voicing contrasts before sonorants avoids the less distinct contrasts involving obstruent voicing in other contexts. Alternative analyses in terms of basic markedness constraints, perceptually-ranked faithfulness constraints or sound change through misperception cannot provide adequate accounts of the full range of perceptually-based phonological phenomena.

Page 32 Halaman 3232 32 5. 5. References Referensi Anderson, Victoria (1997). The perception of coronals in Western Arrernte. Proceedings of Eurospeech '97: Fifth Conference on Speech Communication and Technology 1, 389-392. Ashby, F. Gregory, Geoffrey Boynton, and W. William Lee (1994). Categorization Kategorisasi response time with multidimensional stimuli. Perception and Psychophysics 55, 11- 27. 27. Bender, Byron W. (1968). Marshallese phonology. Oceanic Linguistics 7, 16-35. Benua, Laura (1997). Transderivational Identity . Ph.D. Ph.D. dissertation, University of Massachusetts, Amherst. Blevins, Juliette, and Andrew Garrett (1998). The origins of consonant-vowel metathesis. Language 74, 508-556. Boersma, Paul (1998). Functional Phonology . PhD dissertation, University of Amsterdam. Amsterdam. Booij, Geert (1995). The Phonology of Dutch . Oxford University Press, Oxford. Oxford University Press, Oxford. Broselow, Ellen I. (1976). The Phonology of Egyptian Arabic . PhD dissertation, PhD disertasi, University of Massachusetts, Amherst. University of Massachusetts, Amherst. Butcher, Andrew (1995). The phonetics of neutralization: the case of Australian coronals. Jack Windsor Lewis (ed.) Studies in General and English Phonetics: Essays in Honour of Professor JD O'Connor . Routledge, New York, 10-38. Calabrese, Andrea (1988). Towards a Theory of Phonological Alphabets . Ph.D. Ph.D. dissertation, MIT. disertasi, MIT. Choi, John D. (1992). Phonetic Underspecification and Target Interpolation: An Acoustic Study of Marshallese Vowel Allophony (UCLA Working Papers in Phonetics 82) . Ph.D. Ph.D. dissertation, University of California, Los Angeles. Colarusso, John (1988). The Northwest Caucasian Languages: A Phonological Survey . Garland Publishing, New York. Colarusso, John (1992). A Grammar of the Kabardian Language. University of Calgary Press, Calgary. Côté, Marie-Hélène (2000). Consonant Cluster Phonotactics: A Perceptual Approach . Ph.D. Ph.D. dissertation, MIT [distributed as MIT Working Papers in Linguistics and Philosophy ] Dave, Radhekant (1977). Retroflex and dental consonants in Gujarati. A palatographic and acoustic study. Annual Report of the Institute of Phonetics, University of Copenhagen 11: 27-156. Delattre, Pierre C., Alvin M. Liberman, Franklin S. Cooper, and Louis J. Gerstman (1952). (1952). An experimental study of the acoustic determinants of vowel color: observations on one- and two-formant vowels synthesized from spectrographic patterns. Word 8, 195-210. Delgutte, Bertrand, and Nelson YS Kiang (1984). Speech coding in the auditory nerve IV: Sounds with consonant-like dynamic characteristics. Journal of the Acoustical

Society of America 75, 897-907. Diehl, Randy L. (1991). The role of phonetics in the study of language. Phonetica , 48, 120-134. 120-134.

Page 33 Page 3333 33 Dorman, MF, Studdert-Kennedy, M., & Raphael, LJ (1977). Stop-consonant recognition: Release bursts and formant transitions as functionally equivalent, context-dependent cues. , 22 , 109-122. Flash, Tamar, and Neville Hogan (1985). The coordination of arm movements: An experimentally confirmed mathematical model. Journal of Neuroscience 5, 1688- 1703. Flemming, Edward (1996). Evidence for constraints on contrast: The dispersion theory of contrast. kontras. Chai-Shune K. Hsu (ed.) UCLA Working Papers in Phonology 1, 86-106. Flemming, Edward (2001). Contrast and perceptual distinctiveness. To appear in Bruce Hayes, Robert Kirchner, and Donca Steriade (eds.) The Phonetic Bases of Markedness . Cambridge University Press, Cambridge. Flemming, Edward (2002). Auditory Representations in Phonology . Routledge, New York. York. Fujimura, Osamu, M. Macchi, and LA Streeter (1978). Perception of stop consonants with conflicting transitional cues: A cross-linguistic study. Language and Speech 21, 337-346. Greenberg, Steven (1995). Auditory processing of speech. Norman J. Lass (ed.) Principles of experimental phonetics . Mosby, St. Louis. Gregores, Emma, and Jorge A. Suárez (1967). A Description of Colloquial Guaraní . Moton, The Hague. Hamilton, Philip (1996). Constraints and markedness in the phonotactics of Australian aboriginal language . Ph.D. Ph.D. dissertation, University of Toronto. Harris, Christopher M., and Daniel M. Wolpert (1998). Signal-dependent noise determines motor planning. Nature 394, 780-784. Herbert, Robert K. (1986). Language Universals, Markedness Theory and Natural Phonetic Processes . Mouton de Gruyter, Berlin. Hudson, Joyce, and Eirlys Richards (1969). The phonology of Walmatjari. Oceanic Linguistics 8, 171-189. Hume, Elizabeth (1998). The role of perceptibility in consonant/consonant metathesis. Proceedings of the West Coast Conference on Formal Linguistics 17, 293-307. Hume, Elizabeth, and Keith Johnson (2001). A model of the interplay of speech perception and phonology. Elizabeth Hume and Keith Johnson (eds.) The Role of Speech Perception in Phonology . Academic Press, New York. Itô, Junko (1989). A prosodic theory of epenthesis. Natural Language and Linguistic Theory 7, 217-259. Iverson, Gregory K., and Joseph C. Salmons (1996). Mixtec prenasalization as hypervoicing. International Journal of American Linguistics 62, 165-175. Jessen, Michael (1998). Phonetics and phonology of tense and lax obstruents in German. John Benjamins, Amsterdam.

Jun, Jongho (1995). Perceptual and Articulatory Factors in Place Assimilation: An Optimality-Theoretic Approach . PhD dissertation, UCLA. Kager, René (1998). Optimality Theory . Cambridge University Press, Cambridge. Kawasaki, Haruko (1982). An acoustical basis for universal constraints on sound sequences. Ph.D. dissertation, University of California, Berkeley. Kellogg, WN (1931). The time of judgment in psychometric measures. The American Journal of Psychology 43, 65-86.

Page 34 Page 3434 34 Kenstowicz, Michael (1996). Uniform exponence and base identity. J. Durand and B. Laks (eds.) Current Trends in Phonology . CNRS and University of Salford Publications, 363-393. Kewley-Port, D., Pisoni, DB, & Studdert-Kennedy, M. (1983). Perception of static and dynamic acoustic cues to place of articulation in initial stop consonants. J Acoust Soc America, 73(5) , 1779-1793. Kim, Kong-On, and Masayoshi Shibatani(1976). Syllabification phenomena in Korean. Language Research 12, 91-98. Kingston, John, and Randy L. Diehl (1994). Phonetic knowledge. Language 70, 419-454. Kirchner, Robert (1998). An Effort-Based Approach to Consonant Lenition . PhD PhD dissertation, UCLA. Kisseberth, Charles (1970). On the functional unity of phonological rules. Linguistic Inquiry 1, 291-306. Kondo, Yuko (1994). Targetless schwa: is that how we get the impression of stress timing in English? Proceedings of the Edinburgh Linguistics Department Conference '94 , 63-76. Koontz-Garboden, Andrew (2002). Voicing contrasts in medial stop clusters. Ms., Stanford. Stanford. Kuipers, Aert H. (1960). Phoneme and Morpheme in Kabardian (Eastern Adyghe) . Janua Linguarum, series minor, no. 8. 8. Mouton, The Hague. Ladefoged, Peter, and Ian Maddieson (1996). The Sounds of the World's Languages . Blackwell, Oxford. Blackwell, Oxford. Lass, Roger, and John M. Anderson (1975). Old English Phonology . Cambridge Cambridge University Press, Cambridge. University Press, Cambridge. Liljencrants, Johan, and Björn Lindblom (1972). Numerical simulation of vowel quality systems: The role of perceptual contrast. Language 48, 839-62. Lindblom, Björn (1983). Economy of speech gestures. Peter. Peter. F. MacNeilage (ed.) The Production of Speech . Springer-Verlag, New York, 217-246. Lindblom, Björn (1986). Phonetic universals in vowel systems. JJ Ohala and JJ Jaeger (eds.) Experimental Phonology . Academic Press. Academic Press. Lindblom, Björn, and Olle Engstrand (1989). In what sense is speech quantal? Journal of Phonetics 17, 107-121. Lisker, L., & Abramson, AS (1970). The voicing dimension: Some experiments in

comparative phonetics. In Proceedings of the 6th International Congress of Phonetic Sciences (pp. 563-567). Prague: Academia. Liu, Sharlene (1996). Landmark detection for distinctive feature-based speech recognition. Journal of the Acoustical Society of America 100, 3417-3430. Lombardi, Linda (1995). Laryngeal neutralization and syllable wellformedness. Natural Language and Linguistic Theory 13, 39-74. Lombardi, Linda (1999). Positional faithfulness and voicing assimilation in Optimality Theory. Natural Language and Linguistic Theory 17, 267-302. Lynch, John (1978). A Grammar of Lenakel . Pacific Linguistics, Series B, no. 55. 55. Asutralian National University, Canberra. Maddieson, Ian (1984). Patterns of Sounds . Cambridge University Press, Cambridge. Maiden, Martin (1995). Vowel systems. Martin Maiden and Mair Parry (eds.) The Dialects of Italy . Routledge, London, 7-14.

Page 35 Halaman 3535 35 McCarthy, John J. (2002). A Thematic Guide to Optimality Theory . Cambridge Cambridge University Press, Cambridge. University Press, Cambridge. McGregor, William (1990). A functional grammar of Gooniyandi . John Benjamins, Amsterdam. Amsterdam. Mithun, Marianne, and Hasan Basri (1986). The phonology of Selayarese. Oceanic Linguistics 25, 210-254. Nelson, WL (1983). Physical principles for economies of skilled movement. Biological Cybernetics 46, 135-147. Newman, Stanley (1944). The Yokuts Language of California . The Viking Fund Publications in Anthropology 2, The Viking Fund, New York. Newton, Brian (1972). The generative interpretation of dialect: A study of Modern Greek phonology . Cambridge University Press, Cambridge. Ní Chiosáin, Maire, and Jaye Padgett (2001). Markedness, segment realization, and locality in spreading. Linda Lombardi (ed.) Segmental Phonology in Optimality Theory , Cambridge University Press, Cambridge, 118-156. Ohala, John J. (1980). The application of phonological universals in speech pathology. Norman Lass (ed.) Speech and language: advances in basic research and practice, vol. vol. 3, 75-97. Academic Press, New York. Academic Press, New York. Ohala, John J. (1981). The listener as a source of sound change. Papers from the parasession on language and behavior , Chicago Linguistics Society, 178–203. Ohala, John J. (1985). Around Flat . Victoria Fromkin (ed.) Phonetic Linguistics: Essays in Honor of Peter Ladefoged . Academic Press, New York, 223-242. Ohala, John J. (1990). The phonetics and phonology of aspects of assimilation. John John Kingston and Mary E. Beckman (eds.) Papers in laboratory phonology I: Between the grammar and physics of speech . Cambridge University Press, Cambridge, Cambridge University Press, Cambridge, 258–275. Ohala, John J. (1993). The perceptual basis of some sound patterns. DA Connell and A. Arvaniti (eds.) Papers in Laboratory Phonology IV: Phonology and Phonetic

Evidence . Cambridge University Press, Cambridge, 87-94. Olive, Joseph P., Alice Greenwood, and John Coleman (1993). Acoustics of American English speech: a dynamic approach . Springer-Verlag, New York. Perkell, Joseph S. (1997). Articulatory processes. William J. Hardcastle and John Laver (eds.) The Handbook of Phonetic Sciences . Blackwell, Oxford, 333-370. Pisoni, David B., and Jeffrey Tash (1974). Reaction times to comparisons within and across phonetic categories. Perception and Psychophysics 15, 285-290. Plomp, Reinier (1975). Auditory analysis and timbre perception. Gunnar Fant and Michel Tatham (eds.) Auditory Analysis and Perception of Speech . Academic Press, New Academic Press, New York, 7-22. Podgorny, Peter, and WR Garner (1979). Reaction time as a measure of onter- and intraobject visual similarity: Letters of the alphabet. Perception and Psychophysics 26, 37-52. Prince, Alan, and Paul Smolensky. (1993) Optimality Theory: Constraint Interaction in Generative Grammar . To appear: Cambridge, MA: MIT Press. Puppel, Stanislaw, Jadwiga Nawrocka-Fisiak, and Halina Krasskowska (1977). A Handbook of Polish Pronunciation for English Learners . Panstwowe Wydawnictwo Naukowe, Warsaw.

Page 36 Halaman 3636 36 Redford, MA & Diehl, RL (1999). The relative perceptual distinctiveness of initial and final consonants in CVC syllables. Journal of the Acoustical Society of America , 106, 1555-1565 Robert-Ribes, Jordi, Jean-Luc Schwartz, Tahar Lallouache, and Pierre Escudier (1998). Complementarity and synergy in bimodal speech: Auditory, visual, and audio-visual identification of French oral vowels in noise. Journal of the Acoustical Society of America 103, 3677-3689. Shadle, Christine (1991). The effects of geometry on the source mechanisms of consonants. Journal of Phonetics 19, 409-424. Shepard, Roger N. (1972). Psychological representation of speech sounds. Edward David and Peter Denes (eds.) Human Communication: A Unified View . McGraw-Hill, New McGraw-Hill, New York, 67-113. Smeets, Rieks (1984). Studies in West Circassian Phonology and Morphology . Hakuchi Press, Leiden. Smith, Richard, and Connie Smith (1971). Southern Barasano phonemics. Linguistics 75, 80-85. Spajic@, Sinis&a, Peter Ladefoged, and Peri Bhaskararao (1994). The rhotics of Toda. UCLA Working Papers in Phonetics 87, 35-66. Steriade, Donca (1995). Neutralization and the expression of contrast. Ms, UCLA. Steriade, Donca (1997). Phonetics in phonology: the case of laryngeal neutralization. UCLA Working Papers in Linguistics 3, 25-146. Steriade, Donca (1999). Alternatives to syllable-based accounts of consonantal

phonotactics. fonotaktik. Osamu Fujimura, Brian D. Joseph, and Bohumil Palek (eds.) Proceedings of LP '98: Item Order in Language and Speech . Karolinum Press, Prague, 205-245. Steriade, Donca (2000). Paradigm uniformity and the phonetics-phonology boundary. Michael Broe and Janet Pierrehumbert (eds.) Papers in Laboratory Phonology 5: Acquisition and the Lexicon . Cambridge University Press, Cambridge, 313-334. Steriade, Donca (2001). Directional asymmetries in place assimilation. Elizabeth Hume and Keith Johnson (eds.) The Role of Speech Perception in Phonology . Academic Akademik Press, New York. Press, New York. Steriade, Donca (forthcoming). The phonology of perceptibility effects: the P-map and its consequences for constraint organization. To appear in Kristin Hanson and Sharon Inkelas (eds) The nature of the word: essays in honor of Paul Kiparsky. MIT Press, Cambridge. Cambridge. Stevens, Kenneth N. (1985). Evidence for the role of acoustic boundaries in the perception of speech sounds. Victoria A. Fromkin (ed.) Phonetic linguistics: essays on honor of Peter Ladefoged . Academic Press, New York, 243-255. Stevens, Kenneth N. (1998). Acoustic Phonetics . MIT Press, Cambridge. Stevens, Kenneth N., and Sheila E. Blumstein (1975). Quantal aspects of consonant production and perception: a study of retroflex consonants. Journal of Phonetics 3, 215-233. Stevens, Kenneth N., and Sheila E. Blumstein (1981). The search for invariant acoustic correlates of phonetic features. Peter D. Eimas and Joanne L. Miller (eds.) Perspectives on the study of speech . Lawrence Erlbaum, Hillsdale, 1-38. Stevens, Kenneth N., S. Jay Keyser, and Haruko Kawasaki (1986). Toward a phonetic and phonological theory of redundant features. Joseph S. Perkell and Dennis H. Klatt

Page 37 Page 3737 37 (eds) Invariance and Variability in Speech Processes . Lawrence Erlbaum, Hillsdale, 426-449. Trubetzkoy, Nikolai S. (1939). Grundzüge de Phonologie . Travaux du Cercle Linguistique de Prague 7. Tucker, AN (1962). The syllable in Luganda: a prosodic approach. Journal of African Languages 1, 122-166. Uno Y., M. Kawato and R. Suzuki (1989). Formation and control of optimal trajectories in human multijoint arm movement: Minimum torque-change model. Biological Cybernetics 61, 89-101. Van Bergem, Dick R. (1994). A model of coarticulatory effects on the schwa. Speech Communication 14, 143-162. Vennemann, Theo (1972). Zur Silbenstruktut der deutchen Standardsprache. Theo Theo Venneman (ed.) Silben, Segmente, Akzente . Niemeyer, Tübingen. Westbury, John, and Patricia A. Keating (1986). On the naturalness of stop consonant voicing. Journal of Linguistics 22, 145-66. Wright, Richard (1996). Consonant clusters and cue preservation in Tsou. PhD dissertation, UCLA.

Wright, Richard (2001). Perceptual cues in contrast maintenance. Elizabeth Hume and Keith Johnson (eds.) The Role of Speech Perception in Phonology . Academic Press, New York, 251-277.