1.1 Where to begin?
When I took on the task of documenting Sula, naturally one of the basic questions that came to mind was: how many language varieties are spoken by the Sula people? I set out boldly to answer that question. And I boldly failed.
The original research plan was predicated upon a local-folk-wisdom-derived hypothesis that seemed logical at face value but which was flatly disproven: i.e. there are four dialects of Sula, one for each tribe. This notion is commonly accepted throughout Sula and it may well have been true in the distant past, but it is demonstrably false today. Today there are many more dialects marked by sometimes subtle and other times abrupt isoglosses.
When the dust began to settle, and I had emotional distance from my failure, a far more interesting and complex situation was revealed than anyone I consulted with had anticipated, but it is a situation that I expect is common throughout the region (if not island communities around the entire world). It is, however, a situation that I unfortunately lacked sufficient time and resources to tackle head on.
Although this dissertation is unable to provide a detailed dialect map, my fieldwork did collect the linguistic data necessary to populate such a map, and that data is archived and available for future endeavors. My original research plan fell short in anticipating the scrambled picture that would surface after analyzing my data. The data was carefully collected around the islands, and while I should have anticipated potential problems and should have worked to analyze it during the data collection process, I was victim to hubris and believed the research plan was failsafe and I would be better dedicating my limited time exclusively to gathering dialect data from sites far and wide. So I utilized nearly all of my time collecting as much as possible, and I saved the majority of the analysis for when I returned home. Home is where it became evident that I had collected more than enough data, and it was useful and well organized, but I had neglected key pieces of necessary non-linguistic information for describing the nonlinear structure of Sula’s complicated dialect continuum: oral histories for the target communities studied. As will be discussed further, a detailed dialect map of Sula is an eminently attainable goal, and I am confident that a single, well organized research project could satisfactorily fill in the remaining blanks and unscramble the puzzle.
Although a granular level dialect map of Sula remains elusive, great strides were made regarding Sula dialects at the macro level, and this chapter demystifies the most important parts of my original question: How many language varieties are spoken by the Sula people?
Broadly speaking, there are two main dialect groups and at least two additional newer dialect groups formed by contact, mixing, and leveling—one of which represents a Sanana population settlement region on Mangon Island that was heavily influenced by Mangon speakers, and the other represents a Mangon population settlement on Sanana that was heavily influenced by Sanana speakers. The new dialects are discussed in Chapter 2: dialects. Part 2.
1.2 Geographical and social factors
As explained in the introduction, Sula is one among a number of under-documented Austronesian and Papuan languages of the Maluku Utara province of Eastern Indonesia. The language is spoken in the Sula Archipelago, which includes the islands of Sanana, Taliabo, and Mangon. The archipelago has been frequently listed among the least studied regions in Indonesia (Collins 1981, 1982); Most of the work published on the language is built on wordlists collected by early explorers and missionaries (e.g., Wallace 1869, Holle c. 1900 via Stokhof 1980, Fortgens 1921), and the precise number of Sula dialects remains unknown, but field interviews indicate that, as Collins (1981, 1982) suggested, there is only one primary dialect division, and this has also been borne out by lexical comparison (list of Mangon, Sanana, and Proto forms in Appendix E). There are many additional subtle but definable dialect divisions spread throughout Sula, but it appears clear that the dialect split described in this chapter (henceforth referred to as Mangon, and Sanana) represents the oldest and most dramatic divergence that is still represented by modern Sula varieties.
The choice to use the names, Mangon and Sanana, is intended to limit ambiguity in this study. Mangon is a Sula endonym for the Mangon tribe and it is also the name of one of the oldest villages on the island and of the island itself. The island is also referred to both in speech and writing by the terms Mangole and Mangoli. Although these terms seemed at first to be exonyms, it is likely that they represent two variations of an early endonym (l later shifted to n and the final vowel was dropped in some dialects and then that pronunciation, Mangon, spread). As the forms, Mangole and Mangoli were already recorded on numerous official records, those terms did not drop from usage entirely, and all three forms are now commonly used. It is unclear if the name Sanana is an exonym or endonym. It refers to the primary administrative town in the region as well as the island it is situated on. That island is also known by the name Sula/Sua. However “Sula/Sua” is too ambiguous a term to use in this chapter when addressing the dialect, because it is also the name of the entire island group, the ethnicity of the entire speaker population, and the language as a whole. So in this chapter, and without any intended social or political implications, the following definitions apply:
• Sula/Sua: the general name for the language, ethnic group, and island archipelago.
• Mangon: the northeasternmost island; the dialect and tribe that settled this island.
• Sanana: the southernmost island of the Sula archipelago; the city located on that island; the general dialect of the Falahu, Fagudu, Facei tribes that settled there.
While not modern by objective measures, Sanana city does have access to basic services: there are at least two banks, several local shops where food staples can be purchased, schools for all age ranges, numerous mosques, around a half dozen one-room restaurants, administrative agency offices, a governor’s mansion, a police force, a large open-air market, and a primitive hospital staffed by competent medical professionals who are stationed by the Indonesian central government. The city lacks recognizable business franchises and recreational destinations like movie theaters or malls. There is a small city park next to the pier that is in disarray, and several vacant land expanses converted into makeshift soccer fields. The island’s sole ‘tourist attraction’ seems to be ruins of a Dutch era fort where, fittingly, the regional tourism office is located. The tourism office is a bare room with a desk, a guestbook, and a handful of government employees.
During my work in Sula, I stayed for extended periods of time with a number of families in Sanana and in rural satellite communities alike. I observed that the city boasts limited urban infrastructure: including paved roads, and in many neighborhoods, intermittent electric power and cellular coverage with limited data access. But as the most populous area in the region, it also exhibits the most conspicuous language attrition; observed public interactions are usually in Malay, and it is rare to hear youth in particular interacting in Sula even in more private domains. In rural communities, on the other hand, the use of Sula language is more vigorous; however the populations are smaller, and children still frequently speak to each other in Malay. Elderly language consultants often remark that children ‘no longer speak the language well’.
The Sula ethnic group consists of four tribes—Falahu, Fagudu, Facei, and Mangon. According to local lore, the first three originally settled on the island of Sanana, while the fourth, Mangon, settled on Mangon island. The physical separation of the islands is the likely the main reason for the Sanana–Mangon dialect division. Since original settlement, the Mangon tribe has settled two additional areas on Sanana Island: a neighborhood in the region that has become the greater urban area of Sanana (this neighborhood is also confusingly named Mangon), and a village named Malbufa, on the northern part of the island’s west shore. The Malbufa dialect of Sula is significantly different from other Sanana dialect(s); however, the Mangon neighborhood in Sanana seems to have adopted a dialect that is of a typical Sanana variety. Sanana tribes have also settled numerous villages throughout the islands, the oldest of the villages external to Sanana island may be the Facei tribe settlements of Capalulu, Wai U, and Orifola on the island of Mangon, where a contact dialect appears to have developed. Additional Falahu and Fagudu villages were also established on Taliabo Island and along the southern coast of Mangon.
Figure 9. Sula archipelago
The Mangon tribe settled the northeastern island, and the Falahu, Fagudu, and Facei tribes settled the southern island
Figure 10. Sula Isogloss Map
An interactive version of this map is available at the following web address: http://www.arcgis.com/apps/View/index.html?appid=e776186a151c4e1c94ce03bbfece6dd
1.2.1 Sula isoglosses
As stated previously, I cannot provide a highly granular dialect map and explain all of the variations and innovations unique to each sula community. I can however provide a broad overview that identifies where dialects and languages (in the case of Bajo) are located. From Figure 1.2.1:
Southwest Mangon Falahu (1): This sparsely populated region is settled by Falahu tribe members. It is a mix of primary settlement communities from pioneers who left what is now Falahu village in Sanana city and of newer, second and third generation settlements. There also may be settlements or areas with significant influx of people from the Southeast Mangon Falahu region. The language here is of the Sanana dialect, but interrelationships between varieties spoken here are not known.
Central Mangon Dialect (CMD) (2): this region consists of three villages, and it was settled long ago by Facei Tribe settlers coming from Facei village, Sanana. In this region, a new dialect has emerged via contact to the more populous Mangon tribe members who speak Mangon dialect. In this region an east–west contact dialect continuum is present where the easternmost village is closest to the Mangon tribe and has taken on more features of Mangon dialect than the neighboring villages to the west.
Mangon Island Mangon (3): This region is the home settlement of the Mangon tribe, and it is where the Standard Mangon dialect is spoken.
Southern Mangon Fagudu (4): Villages here were settled by Fagudu tribe pioneers from Sanana. The area speaks a Sanana dialect of Sula that has also begun taking on several features of Mangon. Further study is be required to determine whether Mangon words and features have spread widely enough for the language to be considered another example of a contact dialect.
Southeast Mangon Falahu (5): This sparsely populated region is settled by Falahu tribe members. It is a mix of primary settlement communities from pioneers who left what is now Falahu village in Sanana city and of newer, second and third generation settlements. There also may be settlements or areas with significant influx of people from the Southwest Mangon Falahu region. The language here is of the Sanana dialect, but interrelationships between varieties spoken here are not known.
Northern Fagudu Sanana (6): Fagudu is the primary tribe in this region. Sula spoken here is of the Sanana dialect type and it is quite similar to the language as spoken in Sanana city. Individual villages are said to have some unique characteristics, but it is unclear if any are different enough to be dialects in their own right.
Western Sanana Mangon (7): this region centers around the large town of Malbufa and its satellite neighborhoods. It was settled long ago by pioneers from Mangon island (Waitulia, Mangon, and Waitina villages). A new contact dialect has emerged in Malbufa via contact to the more populous Sanana tribe members nearby. This dialect was formed in a near opposite scenario to that of the Central Mangon Dialect on Mangon island. Side-by-side, the two dialects could help shed light on new dialect formation by revealing points of convergence and divergence.
West Fagudu Sanana (8): Fagudu is the primary settler tribe in this region. Sula spoken here is of the Sanana type, but it is recognizably different from the language as spoken in Sanana city. The language has some unique local words and sound innovations. As a Sanana dialect, it shares the innovations that differentiate Sanana dialects from Mangon.
West Facei Sanana (9): Facei is the primary settler tribe in this region. Sula spoken here is of the Sanana type, but it is recognizably different from the language spoken in Sanana city. The language has a number of local words and unique innovations, but as a Sanana dialect, it shares the innovations that differentiate Sanana dialects from Mangon.
Southwest Fagudu Sanana (10): Fagudu is the primary settler tribe in this region. Sula spoken here is of the Sanana type though significantly different from the language spoken in Sanana city. The language has numerous local words and unique innovations, but as a Sanana dialect, it shares the innovations that differentiate Sanana dialects from Mangon. It is a mix of primary settlement communities from pioneers who left what is now Fagudu village in Sanana city and of newer, second and third generation settlements. There also may be settlements or areas with significant influx of people from the West Fagudu Sanana region. The number of definable dialects and interrelationships between varieties is not known.
South Facei Sanana (11): Facei is the primary settler tribe in this region. Sula spoken here is of the Sanana type though it is recognizably different from the language spoken in Sanana city. The variety spoken in Fatkouyun is perhaps the most divergent among Sanana varieties. There is said to be a significant number of unique lexical items and, by observation, there seem to have either been some vowel substitutions or a modification of the phonetic space assigned to certain vowels. Unfortunately I could only spend an evening in the village and did not have a chance to further explore the topic. Several targeted recordings were made in the village and archived, so the topic can be further explored by a future researcher. The language does appear to share all of the innovations that differentiate Sanana dialects from Mangon.
Southern Sanana Mixed Tribes Region (12): This area encompasses Manaf village and its surrounding areas. Residents report the the settlement to have arisen as a crossroad between the Facei villages to the south, Falahu villages to the north, and Fagudu villages to the west, which are connected by the only navigable route across the island interior. Given my limited interaction in the village, I did not notice remarkable dialectic differences from the variety as spoken to the north, and it is unclear if the area has a distinct language variety, if it is similar to Bega village to the north, or if it is a hodgepodge of families with different language backgrounds, and I only happened to communicate with individuals who have a Southern Falahu Sanana linguistic heritage.
Southern Falahu Sanana (13): Falahu is the primary settler tribe in this region. Sula spoken here is of the Sanana type and it begins to deviate from the language spoken in Sanana city. The deviation is not strictly linear, indicating some leapfrogging, some counterclockwise settlements, and some additional, later intermediate settlements from Falahu village in Sanana in addition to the expected clockwise settlement expansion. Additionally, lateral borrowing further obfuscate the direction of migration in this region, so while a non-geographically linear dialect continuum likely does exist here, the varieties are quite similar to begin with, and the variations are clouded by lateral borrowings in this highly mobile area. As such, defining the linguistic relationship between dialects here will require a dedicated study that leverages comparative linguistics, historical documents, and oral histories.
Southern Falahu and Fagudu Blend Sanana (14): This is a boundary region between Falahu and Fagudu areas. Sula spoken here is of the Sanana type and it deviates somewhat from the language spoken in Sanana city. The region has some identifiable characteristics, but it shares the innovations that differentiate Sanana dialects from Mangon.
Southern Fagudu Sanana (15): Fagudu is the primary settler tribe here. Sula spoken in this region is quite similar to the standard Sanana variety in Sanana city to the north. Individual villages have unique characteristics, but it is unclear if any of the communities here could be said to have their own dialects.
Standard Sanana Dialect (16): Sanana city formed over the historical home territories of the Fagudu, Falahu, and Facei tribes. The language spoken here is the most generic Sanana variety. That is, Sula as spoken here is seems most difficult for other Sanana speakers to place geographically, as it does not contain innovations that people might pick up on to identify a local variety. That is: all of what is present in Sanana city is also present in a number of other communities, but unique features to those other communities are not found in Sanana city.
Bajo Language (17): this region is a very old Bajo (Sama–Bajau) settlement that is built approximately a third over the coast and two thirds over water on stilt houses connected via a network of boardwalks. Additionally, some percentage of the extended community is aquatic nomads who live aboard ships and migrate between this and other Sama–Bajau communities. The basic vocabulary of this community’s language variety is listed in this dissertation appendix, and recordings are archived along with other dissertation materials. Also archived are several targeted conversation recordings. Unfortunately investigating the community’s language was outside of the purview of my research, but the residents here are warm and welcoming, and I highly encourage a future researcher to live among this community for an extended period.
1.3 Overview
The two main dialect groups have similar but not identical phoneme inventories. Sanana dialects have a phonemic glottal stop and voiceless glottal fricative that are both absent (or rare) in Mangon dialects, while Mangon dialects retain a voiced velar nasal that has been lost in Sanana (ŋ>n). Both dialects have five vowels (i, u, e, o, a), but Sanana dialects also retain several vowel clusters that have been reduced in Mangon (ei, oa, and ao). Field data provided the present day phoneme inventories, and the Proto–Sanana–Mangon (PSM) inventory was reconstructed using the Comparative Method. While the phoneme inventories are quite similar, not every phoneme corresponds perfectly to its counterpart in the other dialect group. Table 10 lists phoneme correspondences between the dialect groups alongside PSM, the progenitor.
Table 10. Sanana–Mangon phoneme correspondences
Consonants:Vowels:
Sanana Mangon PSM Sanana Mangon PSM
p – p *p i – i *i
b – b *b Ø – i *i
t – t *t (word final)
d – d *d u – u *u
r – d *d a – u *u
(intervocalic) (if preceding syllable contains o)
k – k *k Ø – u *u
g – g *g (word final if preceding syllable
ʔ – Ø *ʔ does not contain /o/)
m – m *m e – e *e
n – n *n o – o *o
n – ŋ *ŋ a – a *a
r – r *r ei – e *ei
f – f *f oa – o *ou
s – s *s ao – o *ao
h – Ø *h
y – y *y
h – l *l
(subset 1)
l – l *l
(subset 2)
c – c *c
j – j *j
w – w *w
While this chapter describes the macro-level dialect division in Sula, the topic of dialect differentiation was complicated at the micro-level by an interplay between three factors: time, tribe, and terrain. If asked how many dialects of Sula there are, Sula residents answer without hesitation, “four: Fagudu, Falahu, Facei, and Mangon,” (the four tribes, recited in various orders). That response is so universal that when I designed a geographical study of dialects, I predicated my research plan upon an assumption that while these dialect divisions were no longer evident in Sanana city, they were likely still present in rural settlements. I still believe that is likely the case, but once I began analyzing data, it became clear in short order that the picture was not so simple.
The region’s only city, Sanana, is the ancestral home to three of the four tribes, and while there are widely accepted geographical delineations separating each tribe’s range, no such isoglosses can be drawn. In fact, I begrudgingly came to the conclusion that although there are a handful of deliberate identity-marking word variants used by tribal members, there is truly only one extant dialect that covers all of Sanana city. It does, however, appear that the tribes might have indeed had discernible dialects in the distant past, and that those dialects may yet be reconstructable by comparing modern Sula as spoken in outlying settlements. This is where the factors of time and terrain enter.
Beginning at some point in the distant past, community settlements branched out to distant (primarily) coastal locations around the islands, and many of those settlements then spread to form segments of dialect chains along the coast. But this did not happen in a geographically linear pattern. This fact makes logical sense in the context of an island where groups prefer to settle along the coast. Costal conditions vary in Sula from swampy, to sandy, to steep cliffs, to large fertile plains, to dense jungle, to arid. The quality and reliability of rivers available to support settlements also varied, and because of all of these varying conditions it seems that it was frequently the case that the most preferable place to establish a new settlement was not necessarily an origin community’s directly neighboring river valley. New settlements were often established numerous rivers down, where superior conditions were present, and subsequently, the space in between was backfilled with new settlements from both directions, and also in a nonlinear manner. This way of settling new villages results in a situation where a grandchild or great-grandchild settlement can be geographically situated beside its grandparent settlement with the intermediate generation of settlements geographically leapfrogging over both on either side.
This geographic incongruity means that while describable dialect continua do exist in Sula, they are nonlinear and nigh impossible to arrange into a linear descent model without historical context for reference. Figure 11 illustrates Sula’s settlement tendency and why historical context is vital to help decode it.
Figure 11. Sula village settlement pattern
Illustration of geographically non-linear settlement patterns around the Sula Islands. A, B, and C parallel the original settlements of Sanana’s three tribes, and the sub versions represent generations of communities descended from the original tribes.
Figure 11 shows an example of a dialect continuum that is geographically influenced but not geographically determined. As complex as the figure looks, the actual situation on the ground is far more complex. This is because settlement periods happened at various time depths and in both directions on the islands, and because numerous interspersed settlements were established both from the four original tribes and simultaneously by descendants of the new settlement communities—in both directions—and the language varieties were also influenced by lateral transfer from neighboring communities of non-corresponding settlement generations. It would therefore be a daunting if not impossible task to describe Sula’s dialect continua using the Comparative Method alone.
Setting foot on the lower half of the island in Figure 11 with no historical background, one would likely expect to find a dialect continuum that descends from left to right or from right to left, but in Sula’s case, a comparison of dialect features would not bear that out. That is, one might expect the Comparative Method to reveal the following settlement pattern (or its clockwise counterpart): A1.2.1 > A1.2 > A1 > A1.1 > A1.1.1 >A1.1.1.1 > A1.1.2.1.1 > A1.1.2.1 > A1.1.2 > A1.1.2.2
Instead, however, the data would seem un-interpretable, because random noise and lateral features overwhelm the signal. The situation would not necessarily be a lost cause though. In Sula, as with most Austronesian communities, there are elders in every community who take on the sacred responsibility of accurately preserving oral histories, and I believe that gathering these histories and applying a comparative historical framework to their details should in most cases provide what is necessary to re-approach each community’s language data, correct it for lateral transfer and noise, and then construct the internal structures of Sula’s dialect continua and reveal the islands’ migration history along with it. In the case of Figure 12, the structure would turn out to be:
Figure 12. Rlationships between dialects in hypothetical set
A
|
A1
__________|________
| |
A1.1 A1.2
_________| |
| | A1.2.1
A1.1.1 A1.1.2
| |_________
A1.1.1.1 | |
A1.1.2.1 A1.1.2.2
|
A1.1.2.1.1
As explained, a fair amount of fieldwork time and effort went into collecting data at numerous locations around the islands, and rather than revealing the structure of Sula’s dialect map, subsequent research revealed that the Comparative Method in a vacuum is probably insufficient to unwind the complexity of the situation on the ground. Subsequent inquiries with elders like Ismael Duila, of Waibau, gleaned an intricate pattern of nonlinear settlement trends on the Archipelago and also revealed the degree of detailed knowledge that elders tend to possess about their communities’ histories and the histories of their neighbors. A revisiting of the communities to gather oral histories is needed to reveal Sula’s detailed dialect map, but the broad divisions are unmistakable, and using the Comparative Method, this chapter can accurately and adequately describe Sula’s modern dialect groups, and use them to identify the rules for reconstructing Sula’s proto ancestor and demonstrate how it transformed into the varieties that are spoken today.
1.4 Previous research
As discussed in chapter one, few publications analyze the Sula language directly, and as a result of this lack of descriptive data, comparative linguistic work has been limited (e.g., Blust 1981, Collins 1983). Fortgens (1921) does an impressive job describing Soboyo on the neighboring island of Taliabo having only scant data, and Devin (1989) and Grimes (1992) quite impressively describe the main indigenous language of Buru island immediately to the south. Based on the material available at the time, Blust (1981) and Collins (1981) argued for an Austronesian subgroup of Buru–Sula–Taliabo under Proto–West–Central Maluku.
Blust (1981) also made use of the Soboyo data in Fortgens 1921 to show that PAN *S, which has disappeared in all other known languages of eastern Indonesia, is reflected consistently in Soboyo as h. The paper also argues for the inclusion of Soboyo in a subgroup alongside Sula and Buru. Collins (1989) picked up the analysis of Taliabo and explained that his field research showed the island to contain a single native language spreading over a long dialect continuum. The languages of the B-S-T subgroup were next taken up by Blust (1993), who disputed a claim in Esser 1938 proposing a subgroup containing Taliabo–Sula–Bacan. Blust demonstrates that Esser’s proposal is not based on linguistic reasoning.
At a higher level up the family tree, Mark Donohue and Charles Grimes (2008) challenge the Central Eastern subgrouping favored by Blust, and Blust’s 2009 follow-up defends the grouping. No new data source was cited either as basis for the challenge or the follow-up.
There have been subsequent challenges to Collins’s and Blust’s early subgrouping proposals, but to my knowledge none of them has included any new primary data, and Collins’s work remains the most complete historical evaluation. In a 1982 publication, Collins (1982:82) was first to suggest that there is one primary dialect division in Sula and that it separates Mangon and Sanana. My research corroborates this hunch. However, Collins’s analysis is mostly limited to consonants, and the data he uses is mainly from the Mangon dialect of Sula. Collins writes (1981:35) that our knowledge about the languages of Sula is in an elementary state, and this limits our ability to analyze the reflexes of PAN vowels.
Collins (1981:41) also notes that the Sanana dialect has “undergone a number of obfuscating innovations.” By comparing Sanana and Mangon dialects, the present chapter identifies these obfuscating innovations, and makes sense of Sula dialects’ vowel correspondences. The research in this chapter augments Collins’ findings and provides a more complete dataset for Austronesianists to use in refining higher-level subgroupings. Prior to these findings, Sula could appear to group more closely at times with Buru and at other times with Taliabo, depending on the Sula dialect that words were pulled from. The findings in this chapter lift that ambiguity by establishing a higher level Proto–Sula that should instead be used for comparative work with neighboring languages—a task I am eager to take on once I or another scholar conducts more thorough fieldwork on Taliabo and uses the island’s numerous present dialects to similarly reconstruct Proto Taliabo.
This chapter helps to develop the academic literature in an area of Austronesian linguistics where there remains significant debate and little data available, and it provides new, more-comprehensive primary data that should be used to help settle ongoing disputes that have thus far been based on weak foundations.
1.5 Methods
Data used in this study was gathered in North Maluku during three three-month long trips between 2010 and 2015. Language consultants were chosen as described below to help ensure that data would be reliable for comparative work. Elicitations were conducted at seventeen sites on Mangon and Sanana, where the majority of Sula speakers reside. Before choosing sites, I consulted with the regional development office, the bureau of statistics, and community elders. I inquired about the settlement history for each village to learn which tribe settled the area and which of the (then hypothesized) four dialects it was reported to speak. A minimum of ten sessions were conducted for each targeted dialect. Sessions included a series of three videos that were watched by speakers, two at a time, within three age ranges as well as elicitation of a 230-word basic vocabulary list adapted from Greenhill et al.’s (2008) Austronesian Basic Vocabulary Database (see Table 11). Typically, various other data collection and documentation sessions were also conducted. These ranged from ethnographic interviews to performance recordings; demonstrations of traditional medicine and farming/hunting/fishing methods; and topical vocabulary elicitation.
Table 11. Elicitation sessions by tribe
Sessions were conducted to determine whether there is tribe-based variation. To control for geographic variation, research did not cross village boundaries. In addition to the four Sula tribes, the same elicitation materials were used with Sanana’s Bajo community, and also the Facei tribe community settled on Mangon’s southern coast (CMD).
Dialect Video 1 Video 2 Video 3 Swadesh
18-29|30-49|50+18-29|30-49|50+18-29|30-49|50+Mixed age
Falahu ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Fagudu ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Facei ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Mangon ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ CMD ✓ ✓ ✓ ✓
Bajo ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
At least two speakers participated in all vocabulary elicitation sessions. This helped to counter code switching effects, because all participants are also fluent Malay speakers, and Malay was the language of elicitation. Sessions had three or more participants in situations when there was a group dynamic that facilitated thoughtful responses. All sessions—vocabulary and video alike—were conducted in the target geographical area. For example, an elicitation session for Facei could not be conducted in a Fagudu village, even if the consultants were native Facei speakers, and the two main participants for each targeted elicitation session were required to have been born and raised in the particular village where the session was being conducted (i.e. they could not come from another village even if that village reported to be of the same tribe or the same dialect). During vocabulary elicitation sessions, dialect-external participants were invited to join as tertiary observer-consultants when available. Their task was to listen and identify when unexpected or interesting dialect differences were encountered. These participants did not tend to contribute a large quantity of material; however, their occasional observations were thought provoking.
Three videos were created and used to elicit conversation controlled for topic. Videos were produced locally to (a) keep discussion from centering on any foreign elements rather than scenes and actions being depicted and (b) contextually prime participants to speak Sula rather than Malay, which is the language that typically accompanies media content depicting the outside world. The videos include a movie with no audio track that shows a hungry young man earning money to buy a bread roll (Video 1) and two compilation videos consisting of various scenes filmed around the island (Video 2 and Video 3). Each video was presented in the same manner to two participants at a time. Videos were filmed and presented in HD 1080 on an iPad with Retina display.
Nine sessions were conducted for each target dialect. During each session, a video was watched and responded to twice by two participants who were: (a) born and raised in the community where the session took place, and (b) within a given age range (either 18–29, 30–49, 50+). It became clear that broad age ranges were needed, because many participants were unsure of their chronological age. To correct for age uncertainty, a reported age was recorded alongside an age estimate made by the researcher. A small number of elicitation sessions were excluded from the study and repeated with different participants. This happened when either a reported or observed age was outside the target range. The study included men and women; however it was impractical to balance for gender due to population availability.
A video was watched by only one participant during the first viewing. The viewer described what was seen to the second participant, who would ask questions to clarify and gather additional information. Instructions were provided primarily in the Sula language. After the first viewing of video 1, the non-viewer recounted what was remembered from the story. The video was then played a second time when participants watched it together and discussed it freely. Videos were shown in this way to help collect conversational language data from which vocabulary was later extracted and added to Swadesh list terms for analysis with the Comparative Method.
1.6 Comparisons
Table 12. Phoneme inventory
The Proto–Sanana–Mangon phoneme inventory consists of 24 phonemes—nineteen consonants and five vowels
ConsonantsVowels
p b t d k g ʔ i u
c ([t͡ʃ])j ([d͡ʒ])
m n ŋ e o
r
fsha
l
w y ([j])
1.6.1 Clusters and diphthongs in Proto–Sanana–Mangon
Many Mangon forms seem to have prenasalized word-initial onsets, but it is not clear whether they are Prenasalized consonants or consonant clusters. They do however appear to reflect a fossil morpheme N-. It is uncertain whether they arose in Mangon after the dialect split (likely from the possessive marker –in), or if they reflect an earlier affix and were reduced in Sanana. If the morpheme existed in PSM, the following clusters would have been present word-initially: Ny, Nb, Nl, Nc, Np
Various vowel sequences are also present, but their phonemic status is uncertain. These segments are: ia, ui, ua, ei, eu, oi, ou, ai, au, ae, ao
1.6.2 Changes in Proto–Sula, prior to PSM
During fieldwork, an unexpected vowel correspondence kept conspicuously popping up between Sula and Malay, the language of elicitation. This marked an interesting finding in reconstructing the Proto–Sanana–Mangon dialect: the process revealed two early synchronic snapshots of the language instead of one. That is, whereas the structure of PSM was reconstructable, it necessitated that at least one important change must have occurred at a level higher than PSM but lower than Buru-Sula-Taliabo, if what we know of the sibling languages is complete enough. I refer to this intermediate stage as Proto–Sula. The change in question is: PAN *uCi,u > oCi,u. Here, *u became o when followed in the next syllable by a high vowel (*u>o/i,uσ_#).
Malay cognates containing uCu such as kutu ‘louse’, bulu ‘fur’, and sepuluh ‘ten’, corresponded to oCu in Proto–Sula (oCu in Mangon; oCa in Sanana). This correspondence appeared to reflect a change from PAN *u to Proto–Sula *o in syllables preceding *u. Comparison to other Austronesian language data from the Austronesian Comparative Dictionary (Blust and Trussel, on-going), the Austronesian Basic Vocabulary Database (Greenhill et al. 2008), and personal communication with Robert Blust, revealed this correspondence to be a change to PAN *u in the penult when the ultimate vowel is [+high]. That is, *u became o in the penultimate syllable only where the final vowel was *i or *u. This is corroborated by the fact that uCV[+high] sequences do not violate the phonotactics of either dialect, and also by reconstructed PSM forms like *kuli ‘right’, *duki ‘come’, *jubi ‘shoot’. The modern forms for these words are, kuli (M), kul (S); duki (M), duk (S); and jubi (M), jub (S), but if the change had happened lower than the level of P–S–M, these forms would instead be, koli (M), kol (S); doki (M), dok (S); and jobi (M), job (S). The *u > o change is demonstrated in the following table, and many more examples are found in Appendix E and in the Chapter 4 vocabulary list.
Table 13. *u > o examples
The following table lists a few typical examples of the u > o change.
*uCi
PMP PSM Mangon Sanana English
*kulit *koli koli kol ’skin’
*ma-putiq *boti boti bot ‘white’
*duRi *loi loi hoi ‘thorn’/‘bone’
*uCu
PMP PSM Mangon Sanana English
*bubu *fofu fofu fofa ‘bamboo fish/eel trap’
*buku *foku foku foka ‘joint, e.g. finger’
*bulu *fou foː foa ‘hair, feathers’
Comparing words like ma-tua ‘old’ (people) from *tuqa, uha ‘shrimp, lobster’ from *qudaŋ, uma ‘house’ from *Rumaq, uya ‘rain’ from *quzan, and fua ‘fruit’ from *buaq, we see that *u changed only before high vowels. Furthermore, we can see that the change did not affect other vowels in the same environment by comparing forms like ga-pitu ‘seven’ from *pitu, nihi ‘tooth’ from *ŋisi, nui ‘coconut’ from *niuR (< met.), timu ‘cucumber’ from*qatimun, and winu ‘to drink’ from *inum.
This sound change poses a problem to the Neogrammarian hypothesis that sound change can be conditioned only by phonetic factors (e.g. Osthoff and Brugmann 1878, Hock 1991). Mechanistic change to phonetic factors is postulated to be an unconscious result of speakers slightly missing their targets when attempting to recreate a mental representation of a sound. These repeated mistakes over time are thought to cause modifications to the underlying mental representation of the sound (Blust ### FIND CITATION); however, there is no identifiable biological reason why repeatedly missing a *u target in the environment of / _Ci,u would move a speaker’s prototypical mental representation of the segment closer to o. As it stands, the data at hand are most in keeping with the Neolinguist assertion that change is possible in the absence of a generalizable underlying phonetic motivation (Bartoli 1925). Blust (2018) further discusses Sula’s PAN *u > o change in an expanded discussion on the theoretical implications of sound changes that are not phonetically conditioned and why they are odd.
1.6.3 Sound changes from PSM to Sanana
To help describe the structure of PSM, the subsequent sections illustrate changes that occurred as the language split and evolved into modern Sanana and Mangon dialects. Let us first chart the path of evolution from PSM to the present-day Sanana dialect group.
1.6.3.1 Consonants
*d>r /V_V. Intervocalic *d became a trill in Sanana (usually produced as a flap)
Sometimes historical sound changes stand on their own without affecting other phonological processes or sound changes, but often they can only be understood diachronically as stages in a sequence of processes. The *d>r /V_V sound change is one such instance; we can see that it must have occurred before a final-vowel reduction event that took place in Sanana, otherwise some of Sanana’s modern words would have been different. Take for example tar ‘horn’. If Sanana’s final vowel deletion had not followed the intervocalic *d>r change, the target environment would not have been present, and rather than the modern word having followed the path *tadu > taru > tar, the process of events would have truncated with the final vowel deletion and there would have been no opportunity for *d to change to r. That is, we would expect to see a reflex of tad instead of tar.
Table 14. *d>r /V_V
PSM Sanana English
*badaganabaragana‘to dream’
*gad(i,e)ha gareha ‘four’
*padomu paroma ‘knee’
*tadu tar ‘horn’
1.6.3.1.1 *l>h/[V_V], [#_]
There are many instances of l in Sanana basic vocabulary; however PSM *l became h in intervocalic and onset positions in the words of the following list. This change is apparently sporadic rather than regular; a regular sound change applies across the board. A sporadic sound change affects only a subset of candidate words in a language, and there is no identifiable regularity determining where a change of this sort will occur (e.g. Campbell 2004:17).
Table 15. *l>h/[V_V], [#_]
PSM Sanana English
*geli gehi ‘to stand’
*lai hai ‘earth/soil’
*lama hama ‘eye’
*lani -han ‘near’
*loi hoi ‘bone’
*(N-)losa hosa ‘leaf’
Collins (1981, 1982) speculates that PAN *d/D became h in Sanana and l on Mangon. He provides the following evidence:
PAN *(dD)uRi ‘thorn’ > loi
PAN *dakep ‘embrace’ > hakṾkotṾ
PAN *DuSa ‘two’ > guu
PAN *ke(dD)eŋ ‘stand’ > keli
PCM *dama ‘eye’ > lama
I cannot account for all of Collins’ reconstructions; however the forms themselves are mostly consistent with my data for the Mangon dialect. One difference concerns the form, hakṾkotṾ ‘embrace’, which he identifies as having a Fagudu origin; his use of Ṿ represents “an unspecified devoiced vowel” (1981:42). During my fieldwork, I heard closures and geminates rather than devoiced vowels and recorded both hak and hakkot ‘to hug’ from a Fagudu tribe member in Waibau village. As for the word itself, I am unconvinced that it is in fact a reflex of PAN *dakep: first, neither Collins nor myself collected an l onset variant for Mangon, and furthermore Collins does not account for *p > t /_# in Sula nor *e > o. In fact, other higher level forms listed by Collins containing *e do not become o in Sula. Take for instance, PCM **seget > segi ‘high tide’ and PAN *ke(dD)eŋ > keli ‘to stand (Collins 1981:32).
The form, guu from PAN *DuSa is also problematic: Collins (1981:42) analyzes PAN *d and *D merging to **ḍ in B–S–T and then **ḍ subsequently becoming l in Sula but h in “Falahu and Fagudu dialects.” he further states that the loss of l from ḍ is regular in Sula but that there is an unexpected loss of final -a and that there is an unexpected appearance of u for a in the numeral prefix ga-. Presumably the **ḍ > l that Collins describes occurred at the Proto–Sula stage, as I am able to reconstruct the PSM form *gahu ‘two’. In the Mangon dialect, PSM *h is deleted, resulting in *gau and **au is then reduced to u (and lengthened for syllabic weight). Thus:
PSM *gahu >
gahu (Sanana)
gaØu > gØØu > guu (Mangon)
1.6.3.1.2 *l – l
This is not a sound change but rather, a peculiar and notable example of what appears to be the halting of a sound change across a particular stratum of words. I hypothesize that as a bilingual population, Sula speakers unconsciously interrupted (or perhaps even reversed) a sound change in this part of the lexicon because of the words’ similarity to cognates in Malay—the speakers’ other ‘first language’. In other words: PSM *l > h was a regular change, but bilingualism in Malay either blocked it on some words or put it back.
Whatever the cause, the /l/ phoneme was apparently retained in a number of words. It is not clear that all additional forms I collected with an l–h correspondence reflect PAN *d/D; however, the forms with an l–l correspondence do appear to reflect PSM *l.
One explanation for why some instances of *l did not become h, could be that l was retained in intervocalic and onset position in early loans and forms that are very similar to Malay cognates or false cognates. These forms might have been ‘protected’ from the spread of *l > h. That is, most Sula speakers have been bilingual in Malay for many generations, and they would thus have frequently produced l in these cognate (and false cognate) forms while conversing in Malay. This fact might logically have halted the spread of *l > h in these particular Sula forms. The subset of vocabulary in question is demonstrated in the following examples.
Table 16. *l – l strong examples
PSM Sanana English
*bahali bahal (cf. Malay malu) ‘shy, ashamed’
*bet pila bet pila (cf. Malay bila) ‘when?’
*galima galima (cf. Malay lima) ‘five’
*kalo kalo (cf. Malay kalau) ‘if’
*koli kol (cf. Malay kulit) ‘skin’
*la la (cf. Malay layang) ‘to float, fly’
*laka laka (cf. regional Malay laka) ‘to walk’
*laŋi lan (cf. Malay langit) ‘sky’
*lawa lawa (cf. Malay labah-labah) ‘spider’
*lika lika (cf. Malay milih) ‘to choose’
*lima lima (cf. Malay lima) ‘hand’
*lua lua (cf. Malay luah) ‘to vomit’
*(manu)telu (man)tel (cf. Malay telur) ‘egg’
*tilu til (cf. Malay telinga) ‘ear’
The similar-word hypothesis is not fully satisfactory though, because the following forms are not transparently similar to Malay. In these particular words, perhaps a more logical explanation is a usage frequency hypothesis; however, as Lyle Campbell points out, usage frequency has not been adequately shown to be a mechanism for halting the spread of sound change (p.c. 2020).
Table 17. *l – l counter examples
The following words have no obvious, common Malay counterpart with l.
PSM Sanana English
*nonu boli nona bol (cf. Malay goler?) ‘to lie down’
*sagila(ʔat) sagilaʔat (cf. Malay halilintar?) ‘lightning’
*dalena dalena (cf. Malay lebar?) ‘wide’
*balela balela (cf. Malay ?) ‘to laugh’/‘smile’er
*gatelu gatel (cf. Malay ?) ‘three’
*kila kila (cf. Malay ?) ‘liver’
*lepa lepa (cf. Malay ?) ‘above’
*lifi lif (cf. Malay ?) ‘to turn’
The inability to reconcile the *l–l counter examples underlines the fact that there are multiple sources for Sula l: one that descends from PAN *D/d and at least one that is more mysterious. Unfortunately my data do not reveal more about this source, and I can only speculate that the words entered the language subsequent to *l>h/[V_V], [#_] and then spread to both dialects.
1.6.3.1.3 *ŋ>n
Table 18. *ŋ>n
PSM *ŋ became n in Sanana without exception.
PSM Sanana English
*baifoŋi baifon ‘to hide’
*daufoŋi daʔufon ‘to hide’
*laŋi lan ‘sky’
*maŋa mana ‘sharp’
*maŋa paumana pau‘beat, pound’
*meŋa mena ‘to cry’
*(N-)caŋa sana ‘branch’
*naŋu nan ‘to swim’
*ŋa na ‘name’
*ŋapu nap ‘head’
*ŋihi nihi ‘tooth’
*saŋa-petu sanapet ‘thatch/roof’
*yaŋa yana ‘to see’
1.6.3.2 Vowels
1.6.3.2.1 *u>a/oσ_#
Lowering of *u to a where the preceding syllable contains a mid back vowel is another particularly interesting sound change which has no apparent phonetic motivation, similar to the lowering of PAN *u before high vowels described in the previous section. In an oσu environment, the tongue raises between the first and second vowel. A change from PSM *oσu > oσa requires the tongue to lower between the first and second vowel, so this change not only lacks an obvious phonetic motivation, it appears to behave in opposite manner to phonetic expectation and provide even more sound corroboration the Neolinguists assertion that change is possible in the absence of a generalizable underlying phonetic motivation. As mentioned above, see Blust (2018) for an expanded discussion on the theoretical implications of sound changes that are not phonetically conditioned.
During dissertation review, Lyle Campbell offered the following hypothetical phonetic motivation: the *u from *oCu could have lowered to o (**oCo) being influenced by the lowness of the vowel in the preceding syllable, and afterward the final o might have lowered even further as a form of weakening in an unstressed syllable (p.c. 2020). Whether or not that was the mechanism for this change, it is certainly counterintuitive at face value, and it is remarkable that a parallel to this this chain of events (*uCu > **oCu > oCa) did not take place in any of the Austronesian Language Family’s other 1,200+ languages, where the *u segments diverged with the final *u lowering owing to it being in a less prominent syllable.
This change is also another example of a change within an ordered sequence of changes: it must have occurred prior to a final high vowel deletion that will be discussed. If it had not, forms like ‘to suck’, ‘to burn’, and ‘to hit’ would have ended up as bos, don, and dot instead of bosa, dona, and dota, because the final vowel deletion would have prevented the change occurring in those items.
Table 19. *u>a/oσ_#
Lowering of *u to a where the preceding syllable contains a mid back vowel (full table in Appendix E)
PSM Sanana English
*bagou bagoa ‘cold’
*bosu bosa ‘to suck’
*donu dona ‘to burn’
*dotu dota ‘to hit’
1.6.3.2.2 *i,u>Ø
*i,u>Ø / [-glottal] _#
Sanana deleted word final high vowels following non-glottal consonants. Presence of forms like tapa ‘left’ and laka ‘walk’ demonstrate that this change occurred only to high vowels (i and u). The vowels remain in words where they follow other vowels and glottal consonants (e.g., tui ‘snake’, yau ‘far’, behi ‘to throw’, gahu ‘two’). They may also remain following affricates; however there are too few examples to state this conclusively (e.g., gaji ‘grease’, which could be a loan).
Table 20. *i,u>Ø
Word final high vowel deletion following non-glottal consonants. (Full table in Appendix E.)
PSM Sanana English
*(t,d)ufi duf ‘to stab’
*aku ak ‘1SG’
*api ap ‘fire’
*asu as ‘dog’
*bagu bag ‘thick’
*bahali bahal ‘shy’
*baifoŋi baifon ‘to hide’
*bamapu bamap ‘to cook’
1.6.3.3 Anomalies
In addition to what is described above, there are also a handful of changes that cannot be generalized or adequately accounted for. Pending further data collection, these instances have to be written off as evidence of the inconvenient fact that language is a messy, organic, organism with occasional peculiarities that are hard to characterize, when their explanations have been obfuscated by time.
1.6.3.3.1. Proto *c became s in ‘branch’
PSM *(N-)caŋa >
sana ‘branch’ (Sanana)
ncaŋa ‘branch’ (Mangon)
*c is so infrequent that this sound is likely a loan phoneme. Interestingly, the Malay form cabang ‘branch’ also contains the affricate; however these forms are not transparently similar to one another. Lyle Campbell notes that this may only be an apparent difference and that the forms might be the same in reality, because transition between n and s in various other languages frequently causes an “excrescent” [t] in between to ease articulation of the segments. Some examples he provides are the English words prince [prIn(t)s], dance [daen(t)s], and incense [InsEn(t)s]. Thus it may be sufficient to state that (at least for some speakers), /n-s/ is realized as [nts], and these forms do not represent a difference between the two dialects. More investigation is needed to determine whether other analogous forms exist in the language.
1.6.3.3.2. -y- epenthesis occurred in ‘mouth’
PSM *baØoni > **baØon >
bayon ‘mouth’
boni ‘Mangon’
Epenthesis here is a natural change, and there are several examples of it in Sula as an optional pronunciation (e.g. giya how some Mangon speakers transcribe gia ‘one’), but in this Sanana form it seems to always be present, so it should be listed among the anomalies.
1.6.3.3.3. -l- epenthesis occurred in ‘to steal’.
PSM *biØnaka >
bilnaka ‘to steal’ (Sanana)
binaka ‘to steal’ (Mangon)
This l could not have been present in PSM, or it would still exist in the Mangon dialect. That said, there is also another Sanana variant of this word with a geminate n (binnaka). These three forms reflect a mysterious history and they are difficult to reconcile.
1.6.3.3.4. The numeral prefix ga- was dropped in the morpheme for ‘one’ and all ordinal derivatives.
While less problematic in that it does not involve sound change, this occurrence is nevertheless interesting, because the prefix is retained in the lexical item, gahia which means ‘alone’. In the Mangon form, the h was also deleted and reduced from gaia to gØia. No analogous items have been identified. Examples:
PSM *fatu-ga-hia >
**fat-ga-hia > fat-Ø-hia ‘one unit’ (Sanana)
**fatu-ga-ia > fat-gØ-ia ‘one unit’ (Mangon)
PSM *ga-hia >
Ø-hia ‘one’ (Sanana)
**ga-Øia > gØ-ia ‘one’ (Mangon)
PSM *ca-ga-hia >
ca-Ø-hia ‘one thousand’ (Sanana)
ca-ga-ia > ca-gØ-ia ‘one thousand’ (Mangon)
1.6.3.3.5. *g deletion in two words.
PSM *gami >
**gam > Øam ‘to squeeze’ (Sanana)
gami ‘to squeeze’ (Mangon)
PSM *gifu >
**gif > Øif ‘to open’ / ‘to uncover’ (Sanana)
gifu ‘to open’ / ‘to uncover’ (Mangon)
Proto *g was deleted in two Sanana forms. Problematically, it does exist in onset position elsewhere in the dialects.
1.6.3.3.6. *mf to /f/ reduction
PSM *mamVfai >
maØfai ‘to swell’ (Sanana)
mamfai ‘to swell’ (Mangon)
If *mf is a consonant cluster, it was reduced to f in ‘to swell’. It may however be a prenasal segment realized as [m] due to the previous syllable [m], in which case it was simplified in Sanana along with the other likely prenasal segments. No analogous forms are present.
1.6.3.3.7. Final ŋ deletion
PSM *rekiŋ (L) >
**rekiØ > rekØØ ‘to count’ (Sanana)
rekiŋ ‘to count’ (Mangon)
Final ŋ was deleted in the Dutch loan meaning ‘to count’. This either occurred prior to final vowel deletion, or Sanana and Mangon each independently borrowed this from Dutch, Sanana in the form of reki and deleted the final i, and as reking in Mangon. While the second scenario is possible, there did not seem to be a lot of direct interaction with Dutch people in Sula during the colonial period. It is more likely that Dutch loans entered by way of intermediary languages: Malay, the regional lingua franca, and Ternate, the administrative tongue.
1.6.3.3.8. PSM *s > h in ‘flesh’
PSM *(N-)isi >
ihi ‘meat/flesh’ (Sanana)
nisi ‘meat/flesh’ (Mangon)
PSM *s likely became h in the Sanana form for ‘flesh’ as it would be improbable for *h to have become s in Mangon.
1.6.3.3.9. Proto *y > ʔ
PSM *fayata >
faʔata ‘heavy’ (Sanana)
fayata ‘heavy’ (Mangon)
Proto *y became a glottal stop in the form meaning ‘heavy’. The reverse is not possible, because the Mangon form would have deleted the segment if the proto form had contained a glottal stop.
1.6.4 Sound changes from PSM to Mangon
This section illustrates changes that occurred as the proto language split and evolved into the modern Mangon dialect group.
1.6.4.1 PSM to Mangon. Consonants
1.6.4.1.1 PSM to Mangon. Consonants: *h>Ø
With few exceptions, PSM *h was lost in Mangon in all positions.
Table 21. PSM to Mangon: *h>Ø
The following forms demonstrate *h deletion.
PSM Mangon English
*baha baː ‘to buy’
*bahali bali ‘shy’
*behi beː ‘to throw’
*gad(i,e)ha gadia ‘four’
*gahu guː ‘two’
*gatahua gatua ‘eight’
*han an ‘who?’
*kahiku kaiku ‘grass’
*mahi mai ‘sea’
*nahu nau ‘long’
*ŋihi ŋiː ‘tooth’
*pougahu pogu(ː) ‘twenty’
*sahafa fafa ‘rat’
*samohu samo ‘needle’
*tahaga taga ‘lake’
*tahun taun ‘year’
Exceptions: A few counterexamples have been recorded in which forms are produced optionally with h. All are high frequency words, and all also have variants without h. These forms could either be back borrowings from Sanana, examples of *h retention, or indication that the change has not been fully adopted by all speech communities.
Table 22. PSM to Mangon: *h>Ø exceptions
The following forms are produced optionally with h.
PSM Mangon English
*dahi dahi ‘true’
*deha deha ‘be located’
*hapa hapa ‘what’
1.6.4.1.2 PSM to Mangon. Consonants: *ʔ > Ø
Surface glottal stops are produced in Mangon (e.g., [faʔ.ko] ‘dog’, [saʔ.ka.fi] ‘to hold, as a baby’); however, these do not appear to be phonemic, as no minimal pairs have been found, and hypothetical pronunciations with and without surface glottal stops have been judged equally correct. The following forms demonstrate *ʔ deletion.
Table 23. PSM to Mangon: *ʔ > Ø
Instances of *ʔ deletion
PSM Mangon English
*daʔufoŋi daufoŋi ‘to hide’
*fa-ʔoki faoki ‘forest’
*kiʔi kiː 3SG
*maʔana mana ‘man’
1.6.4.1.3 PSM to Mangon. Consonants: /N-/ prefix
Many Mangon forms have prenasalized word-initial onsets. It is common in Sula to frame nouns in isolation using the morpheme in, and this might be related. It is unclear whether the nasal segment in Mangon’s nasal-initial clusters reflect a fossilized morpheme that was lost in Sanana, or whether they arose in Mangon subsequent to the dialect split—possibly derived from the in morpheme. If the segments in Mangon do derive from in, it is not clear by what process. The syllable Maximal Onset Principle often explains similar phenomena, but it would be inadequate here, as Mangon’s apparent prenasal applies also to forms that already have onset consonants.
Whatever the origin, the result is a series of prenasalized consonants or nasal-initial clusters whose phonemic status is debatable, as tends to be the case with prenasals (e.g. Herbert 1975). An odd if not eerie coincidence, however, is that nasal clusters (or prenasals) existed in an earlier stage in the language family. It is tempting to speculate that the prenasalized consonants in Mangon could reflect previously lost segments, but they do not seem to be reconstructable, and furthermore, that would require convergent loss of those segments in sibling languages and also the sister Sula dialects. The following forms demonstrate word initial prenasalized clusters present in Mangon but not in Sanana. Many additional examples can be found in the Chapter 4 vocabulary list.
Table 24. PSM to Mangon: /N-/ prefix
Words with prenasalized word-initial onsets.
PSM Mangon English
*(N-)caŋa ncaŋa ‘branch’
*(N-)losa nlosa ‘leaf’
*(N-)yai nyai ‘leg/foot’
*(N-)boyu mboyu ‘tail’
There are several exceptional forms as well: manpani ‘wing’ contains a prenasalized cluster internally. The form ‘wing’ is also produced as, npani. The form manpani is a clear contraction of the compound word, manu-(n)pani ‘bird wing’. What’s not certain is whether the np onset in npani is original or if the u was contracted and the n coda was rebracketed to the onset of pani. The form fantui ‘star’, which also contains an internal prenasalized consonant, belongs to a class of environment nouns that all include a non-analyzable morpheme, fa-/pa-. Evidence that the nasal segment in these clusters likely derives from a separate morpheme is however seen in the fact that it also occurs on otherwise vocalic syllables such as nisi ‘meat’.
1.6.4.2 PSM to Mangon. Vowels
Many vowel clusters are reduced in Mangon; however, it is difficult to find a single phonetic condition to account for all of the changes. In instances of *V1V1>V1, what happens is the two like vowels are reduced to a single long vowel. The result is phonetically identical to a double vowel in the same syllable, because these reductions land on stressed syllables and stress in Sula is largely realized by vowel length, but Sula does not permit multiple vowels in the same syllable, so in regions where this reduction is far along, the resulting long vowel is not an instance of a double vowel; Sula’s stress pattern already explains the vowel lengthening. It must be noted that there are regions where this change is somewhat ambiguous, and speakers do occasionally accept the word reanalyzed with the long vowel split across syllables (two vowels), but where this change is far along, speakers strictly permit only a long vowel within a single syllable.
With *ei>e and *ou>o, the second vowel is deleted, and it is a higher vowel. But in *ao>o, the first vowel is deleted, and it is a lower, less back vowel.
1.6.4.2.1 PSM to Mangon. Vowels: *V1V1>V1
Sequences of like vowels are reduced to a single vowel with length determined by a minimal word requirement.
Table 25. PSM to Mangon: *V1V1>V1
Examples of double vowel reduction.
PSM Mangon English
*baha>baa> baː to buy’
*bahali>baali> baːli ‘shy’
*ki’i>kii> kiː ‘3SG’
*maʔana>maana> maːna ‘man’
*ŋihi>ŋii> ŋiː ‘tooth’
*sahafa>saafa> faːfa ‘rat’
*tahaga>taaga> taːga ‘lake’
1.6.4.2.2 PSM to Mangon. Vowels: *ei>e, *ou>o
Sequences of *ei and *ou were reduced to the vocoid with the most prominence, e and o respectively, and this reduction can leave a mark on the word’s stress pattern. Where a CVCV word typically has a stressed first syllable in Sula, the reduction can result in a bimoraic (heavy) ultimate syllable when the ultimate syllable was previously a stressed penultimate syllable (e.g. [apˈfe.i] > [apˈfeː]). These stress marks are indicated in parentheses, because the stress pattern in these words is inconsistent among speakers—perhaps an indication that it is shifting to the more expected penultimate location for CVCV words. In this reduction, the vowel that is retained is both first in sequence and lower than the vowel that is deleted.
Table 26. PSM to Mangon: *ei>e, *ou>o
Examples of *ei to e reduction.
PSM Mangon English
*apfei apfe(ː) ‘smoke’ (fire)
*bafei bafe(ː) ‘smoke, fog’ (general)
*bagou bagoː ‘cold’
*behi>bei> beː ‘to throw’
*fou foː ‘hair, feather’
*ganei gane(ː) ‘six’
*pou poː ‘blood’
*pou poː ‘ten’
*pougahu poguː ‘twenty’
*pougalima pogalima ‘fifty’
*samohu>samou> samo ‘needle’
1.6.4.3 Non-generalizable
1.6.4.3.1 PSM to Mangon. Non-generalizable: *a lost from *au sequence
In forms containing the number ‘two’ only, *a from the sequence *au was lost, and the remaining u was lengthened to meet syllabic weight requirements.
PSM *gahu>gaØu>guː ‘two’
PSM *gatahua>gataØua>gatua ‘eight’ (from ‘minus two’)
PSM *pougahu>pougaØu>poguː ‘twenty’
1.6.4.3.2 PSM to Mangon. Non-generalizable: balfoŋi
If derived from PSM *baifoŋi, this peculiar word undergoes both -l- epenthesis and a vowel cluster reduction from *ai to /a/. This is not mirrored elsewhere in the lexicon and cannot be explained. It might make more sense to consider the Sanana form baifon and the Mangon form balfoŋi as similar but separate forms with the same meaning. If they are in fact both descended from the same PSM form, however, a possible explanation for balfoŋi could be that it is a compound of the active auxiliary bal and the abbreviated (or possibly back-formed) variant, fongi. If it is indeed directly derived from PSM *baifoŋi, the derivation followed the following path: *baifoŋi>baØfoŋi>balfoŋi ‘to hide’
1.6.4.3.3 PSM to Mangon. Non-generalizable: *s > f
In the form fafa ‘rat’, *s became /f/ . It is not likely that the proto form was *safafa, because sahafa, saafa, and safa all occur in Sanana, where intervocalic h deletion appears to have begun but is still optional and limited mostly to rapid, non-deliberate speech. Example: PSM *sahafa>saØafa>saØØfa>fafa ‘rat’
1.6.4.3.4 PSM to Mangon. Non-generalizable: *ao>o
Two PSM words are present with *ao sequences that were reduced to o. That is, the vowel that was retained is both higher and it is the second in the sequence, whereas in *ou sequences above, the opposite occurs, u was deleted: the retained vowel is lower and it is the first in the sequence. Because of the conflicting phonological motivations, these similar reduction processes seemingly must be taken as independent instances of change. Examples: PSM *baoni>boni ‘mouth’; PSM *saotu>sotu ‘dry’
1.6.4.4 PSM to Mangon: PSM in summation
In Proto–Sula (before the Sanana–Mangon split), PAN *u was lowered in words where the following syllable contained a high vowel.
In the Sanana dialect:
*d became an alveolar trill.
- *l became h in intervocalic and onset position in native vocabulary, but it appears that a subset of the lexicon with surface similarity to Malay was shielded from this change, and there also appears to be an unidentified source for some other instances of l in the language.
- *ŋ became n in Sanana in all environments.
- peculiarly, u was lowered to a where the preceding syllable contains a mid back vowel—a change that seems at face value to display the opposite of phonetic motivation. This is an order dependent scenario where the lowering must have proceeded the following in E.
- word-final high vowels were deleted following non-glottal consonants but not following vowels and glottals.
In the Mangon dialect:
- *h was lost in Mangon in all positions (with a few exceptional lexemes).
*ʔ was deleted.
- Initial nasals, which might reflect an early explicative (possibly genitive) marker, were fused to many nouns. These often resulted in either prenasalized segments or nasal-initial consonant clusters (in a dialect that otherwise avoids clusters).
- Sequences of like vowels were reduced to long vowels.
*ei was reduced to e.
- *ou was reduced to o.
- The study of the Sula language is still very much in its infancy; however, this section has built upon the groundbreaking work begun by Robert Blust and James T. Collins nearly four decades ago, and it helps to narrow the academic literature gap. It provides a more complete picture of the inner structure of Sula—especially with regard to the confusing vowel correspondences, and it provides data from which hypotheses about higher branches of the Austronesian language family can be evaluated.
Sula is indeed a single language with a clear primary division between two dialect groups. Speakers across this dialect divide can at times strain for mutual intelligibility, but the patterns of divergence are mostly regular and speakers are able to quickly adapt to the differences. This section has provided an account of the dialect differences that describes correspondences and the sound changes responsible for Sula’s synchronic variation. It can also be read as an algorithmic workflow that can compare a word from both Sula and Mangon and generate the corresponding Proto–Sanana–Mangon form in nearly all cases where cognates are present. The data provided will aid in refining Sula’s position within the Austronesian family as well as helping to understand Sula’s place in its local subfamily of languages.