Consonantal Debuccalization and Deletion in Minangkabau

Indonesian and Minangkabau: Speakers and Location

Indonesian (locally known as Bahasa Indonesia) and Minangkabau (locally known as Baso Minangkabau) belong to the Malayo-Polynesian branch of the Austronesian language family. Although linguistically related to each other to a great extent in terms of phonemic inventory, morphology, syntax, and vocabulary, the two languages are not completely mutually intelligible (Crouch, 2009). Indonesian is the national and official language of the Republic of Indonesia. It is spoken predominantly across Indonesia by approximately 200 million speakers; it is a first language (L1) for 43 million speakers and a second language for 157 million speakers (Indonesia’s Census Bureau, 2010). Compared to Indonesian, Minangkabau has a smaller population of 6.5 million speakers (Antoni et al., 2019; Halimiyah et al., 2014; Wiratsih et al., 2014). It is the principal standard regional language in West Sumatra where it is spoken as L1, and is regarded as a symbol of nationality and culture, and is used as a tool for intergroup communication (Amri et al., 2013; Ayub, 1993; Erwandi, 2018; Medan et al., 1986; Mulyadi, 2017; Rahmadani et al., 2013; Restifiza & Rosa, 2012; Sumidita et al., 2013; Wulandari & Nugraheni, 2019; Zurmita et al., 2013). Both Indonesian and Minangkabau have different dialects in different provinces.

Consonantal Inventories

As shown in Tables 1 and 2, Indonesian (Adelaar, 1992; Adisasmito-Smith, 2004; Batais, 2013; Lapoliwa, 1981; Soderberg & Olson, 2008) and Minangkabau (Adelaar, 1992; Almos, 2012; Erwandi, 2018; Medan et al., 1986; Zurmita et al., 2013) share 19 consonants: three bilabials /p, b, m/, six alveolars /t, d, s, n, r, l/, two post-alveolars /ʧ, ʤ/, two palatals /ɲ, j/, four velars /k, g, ŋ, w/, and two glottals /ʔ, h/. However, compared to Minangkabau, Indonesian has four additional consonants, namely /f, z, ʃ, x/, which are limited to foreign loanwords (as a result, they are placed in parentheses in Table 1).

OPEN IN VIEWER

Table 1.

The Consonantal Inventory of Indonesian.

	Bilabial		Labio-dental		Alveolar		Post-alveolar		Palatal	Velar		Glottal
Stop	p	b			t	d				k	g	ʔ
Fricative			(f)		s	(z)	(ʃ)			(x)		h
Affricate							ʧ	ʤ
Nasal		m					n		ɲ		ŋ
Trill							r
Central approximant									j		w
Lateral approximant							l

OPEN IN VIEWER

Table 2.

The Consonantal Inventory of Minangkabau.

	Bilabial		Labio-dental		Alveolar		Post-alveolar		Palatal	Velar		Glottal
Stop	p	b			t	d				k	g	ʔ
Fricative					s							h
Affricate							ʧ	ʤ
Nasal		m				n			ɲ		ŋ
Trill						r
Central approximant									j		w
Lateral approximant						l

Phonotactics

The basic syllable structure in both Indonesian and Minangkabau is CVC (Batais, 2013; Wiratsih et al., 2014; Zaharani, 2005; Zurmita et al., 2013). In both languages, only singleton onsets and codas are allowed, and onsets and codas are not mandatory. With respect to phonotactic constraints, according to Macdonald (1976), Lapoliwa (1981), and Batais (2013), in Indonesian words, all consonants except the glottal /ʔ/ can occupy the syllable onset position, but only /p, t, k, ʔ, s, h, r, l, y, m, n, ŋ/ are permitted in coda positions. On the other hand, in Minangkabau, only /m, n, ŋ/ are allowed to occupy coda positions, while /ʔ, h/ can only appear in word-final coda positions (Medan et al., 1986; Wiratsih et al., 2014). In addition, /ʔ, h/ cannot occupy onset positions.

Language Contact

Indonesian has been in close contact with many major regional languages in Indonesia, including Minangkabau. This language contact is inevitable, as Indonesian is Indonesia’s official and national language, the medium of instruction, and the language that is increasingly used for inter-group communication across the archipelago (Nababan, 1985). In addition, for many speakers of regional languages, the use of Indonesian by Indonesians generally denotes social prestige and modernity, and broadens their access to the job market in metropolitan Jakarta as well as in regional cities. As a result, superstrate Indonesian has linguistically impacted substrate regional languages (Sneddon, 2003), affecting phonological, morphological, syntactic, and lexical changes. For example, on the lexical level in Minangkabau, Indonesian has contributed a large number of vocabulary items to domains such as religion (e.g., agamo“religion,”puaso“fasting,” and sarugo“heaven”), medicine (e.g., ilimu“science,”kulik“skin,” and panyakik“disease”), geography (e.g., gampo“earthquake,”lauik“sea,” and gunuang“mountain”), body parts (e.g., kaki“feet,”mato“eye,” and rambuik“hair”), education (e.g., guru“teacher,”belajar“study,” and kuliah“lecture”), and technology (e.g., komputer“computer,”masin“machine,” and ubuangan“connection”). This is not an exhaustive list and demonstrates the contributory impact of Indonesian on Minangkabau.

Theoretical Approaches to Loanword Adaptation

Two major opposing approaches have emerged to address the question of whether phonology or phonetics is involved in loanword adaptation. The purely phonetic approach claims that all loanword adaptation processes always take place in phonetics/perception. This claim is motivated independently by psycholinguistic evidence that non-native phonological structures, such as segments, suprasegments, and syllables, are distorted (i.e., misperceived) by L1 listeners during perception. Along the lines of this phonetic claim, Peperkamp and Dupoux (2003), Peperkamp (2005), and Peperkamp et al. (2008), make arguments which will now be highlighted. First, they argue that the borrowing speakers cannot access the phonological system of the lending language. Second, the incoming input is not purely segmental; instead, it is the acoustic (or surface phonetic) representation (with syllable-sized chunks) of the source language. Third, during speech perception, the phonetic input form as a whole, rather than individual segments, is mapped onto the closest available well-formed native counterpart, as determined by phonetic distance between the source and target categories, with as minimal phonetic adjustments as possible. Therefore, in the view of these authors, loanword adaptation processes are on-line “minimal transformations” (Peperkamp, 2005) of a phonetic nature.

Fourth, in addition to being driven by minimal phonetic distance, the loanword adaptation processes are sensitive to the fine phonetic details of foreign speech, but not to its phonological properties. To illustrate this, Peperkamp (2005) and Peperkamp et al. (2008) examine the asymmetry in the online Japanese adaptations of French and English loanwords ending in [n]. They show that word-final [n] in English loanwords is directly mapped on a final moraic nasal (e.g., Japanese [peN] for English “pen”), whereas, in adapting the French loanwords, word-final [n] is realized with a following epenthetic vowel (e.g., French [dwan] → Japanese [duan:ɯ]). As noted by the authors, the vowel epenthesis in the French loanwords is not phonotactically necessary since words ending in a moraic nasal consonant are permitted in native Japanese phonology (e.g., [teN] “point”). To solve this puzzle, Peperkamp (2005) and Peperkamp et al. (2008) explain that, based on experimental evidence, the above asymmetry is due to the phonetic differences of word-final [n] in English and French which, as a result, have largely influenced the way they are perceived by native speakers of Japanese. This is particularly notable, because in French, unlike English, word-final [n] has a strong vocalic release; therefore, Japanese listeners perceive a vowel after [n]. Subsequently, the “perceived” vowel is mapped onto the phonetically closest native vowel [ɯ].

By contrast, the purely phonological model proposes that the loanword adaptation is done by bilingual speakers (i.e., main adapters) and occurs within the realm of production (i.e., phonology). This means that the speakers can access the phonological structure of the source language; consequently, the source phonological representation, not the phonetic output, with its L2 distinctive feature combinations is mapped directly onto a native representation (L1 input). This runs counter to the purely phonetic position which argues that the loanword adaptation takes place in phonetics/perception, that borrowing speakers have no knowledge of the borrowing language, and that the input to perception is merely a superficial acoustic signal. Among the prominent works couched in the purely phonological framework are Paradis and LaCharité (1997, 2002, 2005) and Paradis and Tremblay (2009).

Paradis and LaCharité (1997) propose the Theory of Constraints and Repair Strategies (TCRS) to account for the adaptations of segmental and syllabic malformations in a corpus of 545 French loanwords in Fula. In line with TCRS, Paradis and LaCharité stress that loanword adaptation is constraint-based and is triggered when an illicit source form violates the L1 constraints. Next, universal repair strategies, which either insert or delete content or structure, are initiated in a predictable fashion so that the malformed form meets the violated constraint. Paradis and LaCharité (2005), along the same lines as Paradis and LaCharité (1997), moreover, maintain that loanword adaptation is largely a phonological process, not a case of misperception of the L2 phonetic sound by monolingual speakers. They further advance the notion of phonemic approximation and test its predictions, as well as the predictions by the phonetic approximation approach, against the 12 corpora of English and French loanwords. For example, Paradis and LaCharité (2005) show that /b/ in English words borrowed in French is preserved as [b] despite the fact that English /b/ is phonetically (acoustically) closer to French /p/. The same feature combination of voiced stop /b/ exists in both English and French; accordingly, English /b/ is phonologically identified with French /b/. On the other hand, if this combination does not exist in L1, Paradis and LaCharité (2005) suggest, the foreign sound is usually adapted (provided that it is not imported) by the deletion of problematic features.

Another piece of evidence against the phonetic approximation comes from Paradis and LaCharité (2002). They consider three cases that include the adaptation of English voiced stops in Spanish, adaptation of English rhotic in Japanese, and a case of phonetic metathesis. For example, with respect to the Spanish adaptation of the voiced stops in English loanwords, it is predicted, from the view of phonetic approximation, that monolingual Spanish speakers will map the English voiced stops, namely /b, d, g/, on their Spanish voiceless counterparts, based on the fact that the VOT (Voice Onset Time) values of the English voiced onset stops and Spanish voiceless stops are more or less similar. This prediction does not appear to be true. Paradis and LaCharité test this prediction against their corpus data of 1,514 English loanwords (containing 563 voiced onset stops) in Mexican Spanish and observe that the English voiced stops /b, d, g/ are integrated as /b, d, g/ (and then realized as [b, d, g]) in Mexican Spanish, based on their phonemic status in the phonology of the borrowing language, not their phonetic properties, as advocated by the purely phonetic approach. Therefore, employing the phonetic closeness would give the wrong adaptation.

Paradis and Tremblay (2009) also strongly favor the phonological stance in their attempt to explore the adaptation of voiced and voiceless stops of English loanwords in Mandarin Chinese. The phonetic view predicts that the borrowing speakers of Mandarin Chinese will pay close attention to the tiny phonetic details of English stop aspiration and consequently equate the English aspirated stops and unaspirated stops with Chinese aspirated voiceless stops and unaspirated voiceless stops respectively. Contrary to the phonetic prediction, the study results show that English aspirated and unaspirated voiceless stops are both invariably treated as aspirated stops (despite the fact that Mandarin Chinese enables its speakers to differentiate between aspirated and unaspirated stops), whereas English voiced stops are systematically adapted as voiceless unaspirated stops in Mandarin Chinese.

This study tests the predictions made by both approaches against the analyzed data. It provides evidence from Minangkabau loanword adaptation in support of the phonological approach and strongly argues for the irrelevance of the phonetic approach.

Participants

Ten adult native speakers of Minangkabau participated in the study. They were five male and five female speakers. Their ages ranged from 25 to 41 years. They speak Minangkabau as a first language. In addition, all 10 participants were born and raised, and presently reside, in the city of Padang, West Sumatra. Padang was selected in order to rule out any dialectal differences among the participants (thus ensuring they all speak standard Minangkabau).

Data

The total number of source words used in this study is 52. They were drawn from Rusmali et al.’s (1985) Minangkabau-Indonesian Dictionary and Burhanuddin and Zabadi’s (2013) Indonesian-Minangkabau Dictionary. The selected source words must meet the following criteria: They must be native Indonesian lexical items; consequently, foreign loanwords (including Arabic, English, Portuguese, and Dutch words) were excluded. They must vary in their syntactic categories (e.g., verbs, nouns, adjectives, etc.) and syllable lengths (monosyllabic, bisyllabic, etc.). Most importantly, they must be well-established lexical items. To verify this, before the start of the study, the list of selected words was presented in writing to three adult native speakers of Minangkabau living in Padang. They confirmed that all 52 words were quite familiar to them (see Appendix 1 for the list of words).

Data Collection and Transcription

The data in this study was collected through word elicitation. Each word of the 52 Indonesian loanwords was elicited only once from each participant. After completing the questionnaire (see Appendix 2), the 10 participants were individually face-to-face interviewed and recorded by a native speaker of Minangkabau. The portable digital recorder used for this purpose was a Marantz PMD660 model with an internal stereo microphone. All of the recordings occurred over a period of 10 days.

Each interview started by the participant stating his or her name. Next, the interviewer elicited from each participant the list of 52 words through one of the following tasks: fill-in-the- blanks, short definitions matching the meaning of the target loanword, and object-naming. Only the pronunciation of each target loanword was digitally recorded. If, after many attempts, the participant failed to guess and hence pronounce the word, the interviewer would pronounce it and ask the participant whether he or she was familiar with the word. If it was known to the participant, he or she was next asked to pronounce it. Loanwords thus pronounced were recorded and designated as (*** Yes) so to differentiate them from the words which were spontaneously pronounced (designated as Yes). However, if the word was not familiar to the participant, the participant was not further asked to pronounce it. In this case, the word was labeled as (No). None of the elicited words were unfamiliar to all participants.

Next after recording, both types of familiar elicited loanwords were transcribed in International Phonetic Alphabet (IPA) symbols. It is worth noting that only on a few occasions did I resort to two native speakers of Minangkabau to help me discern a handful of indistinct sounds in the elicited data (owing to some recording complications).

Per each speaker, the final draft of the transcription sheet of the elicited list of loanwords, after being transcribed, has seven columns. A sample transcription sheet is provided in Appendix 3. The first leftmost is No. column which gives a number to each elicited word. The second and third columns, from the left, list the standard Minangkabau spellings of each elicited loanword according to the two used dictionaries: Rusmali et al. (1985) and Burhanuddin and Zabadi (2013) respectively. The fourth column indicates the participants’ familiarity with each elicited word. The fifth column displays the phonetic transcription of the participant’s pronunciation of each elicited loanword. The sixth column displays the phonemic representations of the elicited words in the source language (i.e., Indonesian). The source phonemic representations of the Indonesian words were taken from Kamus Besar Bahasa (Pusat Bahasa, Departemen Pendidikan Nasional, 2008) and were later reviewed by a native speaker of Indonesian. Finally, the seventh column contains the English gloss of each elicited loanword in Minangkabau.

Model

The model used in the study to examine and account for the phonological processes of consonantal debuccalization and deletion employed in Minangkabau is Optimality theory (OT) (McCarthy & Prince, 1993; Prince & Smolensky, 1993). Being a constraint-based framework, the basic premise of OT is that any language consists of a set of hierarchically violable constraints. These constraints are universal, but their ranking differs from one language to another (i.e., language-specific).

In OT, there are two types of constraints: markedness and faithfulness. The markedness constraints ensure structural well-formedness of the output forms, penalizing structures that are hard to produce. For example, Coda requires that syllables end with a coda, while NoOnset militates against any syllables starting with an onset. Faithfulness constraints, on the other hand, guarantees identity between the input and output forms; therefore, all linguistic features in the input must be preserved in the output. For example, Max prohibits any deletion of segments, while Dep prevents any insertion of segments. Both markedness and faithfulness constraints can conflict with each other to determine the best candidate known as the optimal output.

The optimal (winning) output is selected out of an infinite set of candidates which are generated by the function Gen (Generator). Next, based on the language-specific constraint ranking, the function Eval (Evaluation) evaluates the candidates generated by Gen and selects the candidate that minimally violates the constraint ranking. In Table 3, there are three candidates generated from the input by Gen. These candidates are evaluated by Eval against three constraints ranked as follows: Constraint 1 is ranked higher than Constraint 2 which dominates Constraint 3 (C1>> C2>> C3). Each of the three candidates incur one violation. However, Candidate A and Candidate B are excluded because they violate higher constraints (i.e., Constraint 1 and Constraint 2, respectively), hence marked with an asterisk and an exclamation mark (*!). Candidate C, on the other hand, is selected as the optimal output because it violates the lower-ranked constraint (Constraint 3) with the minimal punishment (thus, marked with an asterisk *). In sum, lower ranked constraints can be minimally violated in order to satisfy higher ranked constraints whose violations are most serious and fatal.

OPEN IN VIEWER

Table 3.

Explanation of OT: Candidates and Constraints.

Input	Constraint 1	Constraint 2	Constraint 3
Candidate A	*!
Candidate B		*!
Candidate C			*

Debuccalization of the Word-Final Obstruents /p, t, k, s/

When occurring word-finally, the obstruent consonants /p, t, k, s/ in Indonesian words that are incorporated into Minangkabau are replaced with /ʔ/ and /h/ respectively, as detailed in Table 4

OPEN IN VIEWER

Table 4.

Examples of Minangkabau Debuccalization of the Word-Final Obstruents /p, t, k, s/.

Indonesian word	Minangkabau word	Gloss
/asap/	/asoʔ/	“smoke”
/tutup/	/tutuәʔ/	“close”
/sakit/	/sakiʔ/	“sick”
/berat/	/bareʔ/	“heavy”
/otak/	/utaʔ/	“brain”
/pedas/	/padih/	“spicy”
/lapis/	/lapih/	“layer”

To account for the debuccalization of the above word-final obstruents in Minangkabau, this study proposes the following OT constraints:

OPEN IN VIEWER

Align-Right (Align-R)	The right edge of the Grammatical Word coincides with the right edge of the Prosodic Word (McCarthy & Prince, 1993).
Align- Left (Align-L; Non-glottalObstruents, σ)	Non-glottal obstruent consonants must be left-aligned with a syllable (adapted from Itô & Mester, 1994).
Ident (Place)	An input segment and its output correspondent must be identical in their primary place of articulation (McCarthy & Prince, 1995).
Ident (Manner)	An input segment and its output correspondent must be identical in their manners of articulation (McCarthy & Prince, 1995).
Dep-IO (V)	Vowels in the output must have corresponding vowels in the input (McCarthy & Prince, 1995).
Max-IO (C)	Consonants in the input must have corresponding consonants in the output (McCarthy & Prince, 1995).

See the Minangkabau adaptation of Indonesian /asap/ into [asoʔ], /sakit/ into [sakiʔ], and /otak/ into [utaʔ] in Table 5.

For the word-final voiceless stops /p, t, k/ to be mapped onto the voiceless glottal stop [ʔ], the Align-L (Non-glottalObstruents, σ), Dep-IO (V), Max-IO (C), Ident (Manner), and Align-R constraints must be undominated. Any violation of these five constraints is fatal to the candidates. As shown in Table 5, the winning outputs are the candidates in Row (a). By replacing the three word-final obstruents with the voiceless glottal [ʔ], the candidates in Row (a) satisfy the five highest ranked constraints at the expense of violating the lowest ranked Ident (Place) since the only mismatch between /p, t, k/ and /ʔ/ is the place of articulation. Because it is violable in actual outputs in Minangkabau, this mismatch is considered a small violation.

OPEN IN VIEWER

Table 5.

OT Analysis of Debuccalization of Indonesian Word-Final Voiceless Stops in Minangkabau.

	/asap/	/sakit/	/otak/	Align-L (Non-glottalObstruents, σ)	Dep-IO (V)	Max-IO (C)	Ident (Manner)	Align-R	Ident (Place)
a.	[asoʔ]	[sakiʔ]	[utaʔ]						*
b.	[asop]	[sakit]	[utak]	*!
c.	[aso]	[saki]	[uta]			*!		*
d.	[asopi]	[sakiti]	[utaki]		*!			*
e.	[asoh]	[sakih]	[utah]				*!		*
f.	[ason]	[sakiŋ]	[utam]				*!		*

On the other hand, the candidates in Row (b) are excluded because they violate the Align-L (Non-glottalObstruents, σ) constraint, which requires that the word-final non-glottal obstruents occupy the left edge (i.e., onset) of the word’s final syllable. The obstruents in the candidates in Row (b) are in the coda position. Furthermore, the candidates in Row (c) lose, owing to their disagreement with Max-IO (C). The candidates in Row (d) are also ruled out due to their violation of Dep-IO (V). Preserving word-final obstruent consonants via vowel epenthesis and subsequent re-syllabification as onsets yields incorrect outputs, namely the candidates in Row (d). Note that, as a result of violating either Max-IO (C) or Dep-IO (V), the right alignment of the candidates in Rows (c) and (d) is affected (hence the violation of the Align-R condition). Finally, the candidates in Rows (e) and (f) are eliminated on the basis that they do not satisfy the highest ranking condition of Ident (Manner). In these sets of candidates, the three word-final input stop consonants are realized as output consonants (i.e., fricative [h] and nasal [m, n, ŋ]) with different manners of articulation.

Now, consider the Minangkabau adaptation of the Indonesian word-final voiceless fricative /s/. As illustrated in Table 4, it is adapted into the phonotactically permissible word-final [h]. For this repair to occur, the constraints Align-L (Non-glottalObstruents, σ), Dep-IO (V), Max-IO (C), Ident (Manner), and Align-R must be equally non-violable and dominate Ident (Place), as shown in Table 6.

OPEN IN VIEWER

Table 6.

OT Analysis of Debuccalization of the Indonesian Word-Final Voiceless Fricative /s/ in Minangkabau.

	/pedas/	Align-L (Non-glottalObstruents, σ)	Dep-IO (V)	Max-IO (C)	Ident (Manner)	Align-R	Ident (Place)
a.	[padi]			*!		*
b.	[padisi]		*!			*
c.	[padas]	*!
d.	[padiʔ]				*!		*
e.	[padin]				*!
f.	[padih]						*

As a consequence of their violation of Dep-IO (V) and Max-IO (C), the candidates (a) and (b), respectively, are excluded. Furthermore, owing to vowel epenthesis and consonant deletion to repair the word-final /s/, the right edges of candidates (a) and (b) are not well-aligned with the right edges of the inputs, thereby incurring a serious violation of the undominated Align-R condition. Candidate (c)ends with the voiceless fricative /s/; therefore, it is eliminated for disobeying the highly ranked Align-L (Non-glottalObstruents, σ) condition. Moreover, in candidates (d) and (e), the manners of articulation for the word-final consonants [ʔ] and [n], respectively, are different from that of the word-final /s/ in the input. This repair strategy is prohibited in Minangkabau; therefore, candidates (d) and (e) lose for violating the undominated Ident (Manner) constraint. In addition to Ident (Manner), candidate (d) disagrees with the lowest ranked constraint of Ident (Place). Finally, candidate (f) wins because, unlike the other candidates, it satisfies the five high ranking constraints due to sacrificing the lowest ranked Ident (Place) condition.

Deletion of the Word-Final Liquids /r, l/

As seen in Table 7, the word-final liquid consonants /r, l/, unlike word-final obstruent consonants, are adapted to Minangkabau phonotactics through consonant deletion.

OPEN IN VIEWER

Table 7.

Examples of the Deletion of the Word-Final Liquids /r, l/ in Minangkabau.

Indonesian word	Minangkabau word	Gloss
/mahal/	/maha/	“expensive”
/gatal/	/gata/	“itchy”
/tebal/	/taba/	“thick”
/tukar/	/tuka/	“exchange”
/lembar/	/lemba/	“sheet”
/kasar/	/kasa/	“rude”

In addition to Dep-IO (V), Ident (Manner), Ident (Place), Align-R, and Max-IO (C), this paper proposes the following constraint to explain the word-final deletion of Indonesian liquids in Minangkabau:

OPEN IN VIEWER

Align-L (Liquids, σσ)

Liquid consonants must be left-aligned with a syllable (adapted from Itô & Mester, 1994).

The preference for the deletion of word-final liquid consonants in Minangkabau is illustrated in Table 8. This deletion occurs by demoting Align-R and Max-IO (C) while keeping non-violable Align-L (Liquids, σ), Dep-IO (V), Ident (Manner), and Ident (Place) at the top of the constraint hierarchy. In light of this constraint ranking, the candidates in Row (a) are the winning outputs because they incur non-fatal violations of the lower ranked Align-R and Max-IO (C) conditions in that the word-final liquids are deleted and thereby cause harm to the right alignment of the word; however, they satisfy the higher ranked Align-L (Liquids), Dep-IO (V), Ident (Manner), and Ident (Place) conditions. The remaining candidates in Rows (b) through (e), by contrast, are eliminated based on their violation of the fatal high ranking four constraints. For example, the candidates in Row (b) lose because they do not comply with Align-L (Liquids, σ) condition as they end with word-final liquids. In addition, the candidates in Row (c) are excluded as a consequence of fatally violating the Dep-IO (V) condition. Finally, the candidates in Rows (d) and (e) are incorrect outputs because they disagree with Ident (Manner) and Ident (Place). That is to say, substituting an output consonant (with or without the same place or manner of articulation) for word-final input liquid consonants is a fatal violation in Minangkabau.

OPEN IN VIEWER

Table 8.

OT Analysis of the Deletion of Indonesian Word-Final Liquid Consonants in Minangkabau.

	/ga.tal/	/tu.kar/	Align-L (Liquids)	Dep-IO (V)	Ident (Manner)	Ident (Place)	Align-R	Max-IO (C)
a.	[ga.ta]	[tu.ka]					*	*
b.	[ga.tal]	[tu.kar]	*!
c.	[ga.ta.li]	[tu.ka.ri]		*!			*
d.	[ga.taʔ]	[tu.kah]			*!	*
e.	[ga.tan]	[tu.kan]			*!

The OT analyses in this section confirm the major role of phonology in loanword adaptation (Paradis & LaCharité, 1997, 2002, 2005; Paradis & Tremblay, 2009), as explained here. First, the observed Minangkabau repairs of word-final Indonesian obstruents and liquid consonants are governed by Minangkabau phonotactics. Therefore, the OT constraints of Align-L (Non-glottalObstruents, σ) and Align-L (liquids, σ), which ensure that the obstruents /p, t, k, s/ and the liquids /r, l/ do not appear in the word-final coda position, are always undominated. To fulfill the restriction against word-final obstruent consonants in Minangkabau, the repair via consonantal debuccalization is triggered, resulting in the inputs /p, t, k/ and /s/ weakening and becoming the glottals /ʔ/ and /h/, respectively. Moreover, in line with the Minangkabau phonotactic constraint, the Indonesian word-final liquid consonants /r, l/ are repaired via consonant deletion in Minangkabau. As a result, such consonant deletion, unlike debuccalization, has an effect on the right alignment of the output (hence, the demotion of the Align-R and Max-IO (C) conditions).

Second, the analyses confirm that the word-final voiceless obstruents are debuccalized into glottal consonants in Minangkabau corroborates Lombardi’s (1999, 2001) phonological position that, in many languages, some Place and Laryngeal features are restricted in the coda position (i.e., Coda Condition). As a result, to meet this phonotactic condition, a process known as neutralization is automatically triggered. For example, in German, syllables cannot end in voiced consonants; therefore, laryngeal neutralization via devoicing takes places, yielding devoiced consonants. Likewise, subject to the same coda condition, syllable-final obstruents with different Place features in languages such as Slave, Burmese, and Kelantan Malay are disallowed; hence, they are neutralized to consonants with an unmarked Place. The latter case of Place neutralization is evidenced in the data analyzed in this paper in which Minangkabau prohibits any Place distinction among the obstruents /p, t, k, s/ word-finally. As a consequence, the bilabial, alveolar, and velar Place distinctions among these coda obstruents are lost in Minangkabau. In other words, these obstruents are automatically neutralized and, thus, appear as the glottals /ʔ/ and /h/ with an unmarked Place (therefore, the Ident (Place) condition is always demoted).

The third piece of evidence drawn out is that adaptation through debuccalization exhibits the preference of Minangkabau phonology to maintain the same syllable count of the input word, preserve input-output manner identity, and keep the right alignment of the output word intact. As a result, neither vowel epenthesis nor consonant deletion is employed to adapt the Indonesian obstruents to Minangkabau constraints (hence, the OT constraints Dep-IO (V), Max-IO (C), Ident (Manner), and Align-R are the highest ranked conditions). In a similar way to the obstruent consonants, the liquid consonants /r, l/ cannot be adapted through vowel epenthesis, or be mapped onto consonants with the same or different place or manner of articulation. Therefore, in this case, Align-L (Liquids, σ), Dep-IO (V), Ident (Manner), and Ident (Place) are always non-violable.

Finally, the purely phonetic approach (Peperkamp, 2005; Peperkamp et al., 2008; Peperkamp & Dupoux, 2003) cannot account for the deletion of post-vocalic liquids /r, l/. Per this approach, these two liquids have strong perceptual/acoustic cues in post-vocalic coda contexts (Silverman, 1992; Yip, 1993, 2006); accordingly, they should be perceived and realized in the adapted output. Contrary to the phonetic prediction, the two liquids in the current study are deleted, by virtue of the restriction against liquids in coda positions and the preference for bisyllabic minimality in Minangkabau phonology. Moreover, while the purely phonetic approach predicts the retention of voiceless obstruents /p, t, k, s/ on the basis of their high acoustic audibility and perceptual detectability in post-vocalic positions (Shinohara, 2006), it does not explain why /p, t, k/ are matched with [ʔ] and /s/ with [h]. In order for the phonetic-only approach to account for the observed adaptations, the process of phonetic adaptation must be driven by phonological constraints. In other words, the phonetic assimilation must interact with other aspects of the native language phonology; in this case, it will be called phonology-based phonetic assimilation. The purely phonetic approach cannot be solely responsible for loanword adaptation based on phonetic details only.

In sum, the findings drawn from this study are important for the following reasons. Firstly, they contribute to the literature on loanword phonology by reaffirming the findings of previous phonology-only research that loanword adaptation is a phonological process, and disprove the claim that it is a phonetically/perceptually motivated operation. Secondly, the findings can add to the Austronesian linguists’, as well as laypeople’s, theoretical understanding of the differences in terms of sounds and syllable structures between Standard Minangkabau and Standard Indonesian. Thirdly and finally, it was hoped that the study findings would fill some research gaps in the under-researched areas of Minangkabau phonology and Minangkabau loanword phonology by investigating (using the OT model) the phonological processes of consonantal debuccalization and deletion that are both triggered to assimilate Indonesian loanwords into native Minangkabau phonology, and this has been achieved.