1 New dialects. Introduction
Long after the four Sula tribes settled the archipelago and their dialects diverged, several villages were established on Mangon by settlers from Sanana Island. In three of these villages; Capalulu, Wai U, and Orifola, a definable mixed dialect region developed beside its larger Mangon speaking neighbors. The new region’s early settlers spoke mutually intelligible Sanana language varieties (primarily of the Facei tribe), but generations of contact and intermarriage resulted in significant leveling and the development of a definable new dialect. The most apparent characteristic that sets apart the new settlement’s dialect from Sanana Island dialects is the presence of final epenthetic vowels. This epenthesis results in forms that are superficially similar to the neighboring Mangon-dialect’s vowel-final forms, but the new dialect’s word stems also maintain innovations that are unique to Sanana.
Dialect-forming changes were found to include phonological and lexical leveling: a phonological preference against word-final consonants was introduced from Mangon (the prestige dialect), and Mangon words that were phonologically simpler tended to get borrowed. In addition, some innovative lexical items were present, as were some interdialect forms that retain characteristics of both source dialects. Findings in this section expand the new dialect formation literature which is to date overwhelmingly represented by work on European languages. Also of note, this section represents the only published lexical data from the region in question, and it is one of a only a few descriptions of new dialect formation in an endangered language.
2 New dialects. Dialect Leveling
Multiple Sula dialects and sub-dialects are considered in this study. Over the past several generations, Facei tribe members from the island of Sanana have established three settlements on the neighboring island of Mangon: Capalulu, Wai U, and Orifola (see Figure 13). These settlements are roughly at the center of the island’s southern shore, just west of its southernmost point and in close proximity to the much more populous Mangon tribe’s ancestral homeland. The Facei newcomers still speak a sub-dialect of the Sanana variety of Sula, but it shares surface similarities to the Mangon variety of Sula. That is, generalizable innovations characterizing Sanana are still present in their speech, and importantly, while a number of unique Mangon lexemes have been borrowed, numerous forms are identical to those found on Sanana. A process of dialect leveling appears to be underway, resulting in a new dialect on the island of Mangon that is distinct from its Sanana progenitor and its Mangon neighbor. For the purposes of this study I refer to the region and dialect as Central Mangon Dialect (CMD).
This section analyzes data from basic vocabulary elicitation, and it addresses the questions: Do the new communities speak a Sula variety that is demonstrably different from other regions? and if so, what are the differences, and how did they arise? I argue that the dialect is indeed different and that it resulted from sociolinguistic processes and that a prestige scenario likely led to a regular practice of speech accommodation among the new community’s early settlers. As no previous studies have addressed Sula dialects, it is expected that this study will help our overall understanding of the language and to what degree the processes observed for European languages like English and Norwegian are generalizable to non-European, endangered language scenarios. By developing our understanding of new dialect formation in under-documented and endangered language communities, linguists will be in a better position to prioritize our research when facing populations with dwindling speaker bases.
As discussed earlier, the Sula Archipelago consists of three main islands: Mangon (aka Mangole, Mangoli), Taliabu (Taliabo), and Sanana (Sula) (see Figure 9). The town of Sanana on the island of Sanana is the main population center, but it is also the region with the most pronounced language attrition: all Sanana residents are native speakers of Malay, and it is uncommon to observe young people there communicating in the native tongue. However, language use remains vigorous in the less populous villages throughout the islands.
According to local knowledge, Sanana-based Facei tribe members began to establish new settlements roughly at the center of Mangon island’s southern shore beginning around 300 years ago. There are numerous additional Fagudu, Falahu, and Facei tribal settlements along the southern coasts of Mangon and Taliabo islands, but they are reported to be more recent and they have not been included in this study.
Figure 13. Historical tribal villages & new villages
The four Sula tribes are Fagudu, Falahu, Fecei, and Mangon. They traditionally inhabit the Sanana and Mangon islands of the Sula Archipelago (North Maluku, Indonesia). The Mangon tribe originally settled Mangon Island, and it traditionally speaks closely related, “Mangon” dialects. The Fagudu, Falahu, and Fecei tribes originally settled Sanana Island, and they traditionally speak closely related, “Sanana” dialects.
Figure 14. Facei tribe migration to Mangon
Facei tribe settlers first left the Facei village of Sanana village on Sanana Island to establish the CMD region approximately 300 years ago.
It is unclear how many dialects (or even separate languages) might be under the umbrella of what is loosely referred to as Sula; for example Lewis (2015) lists Sanana and Mangon as separate languages, however Collins (1982) considers them to be dialects of the same language. The main barrier to research has been lack of data from which to paint a more concise portrait of the language. No work to date has addressed Sula’s region-to-dialect mappings; the present study addresses this through targeted wordlist elicitation and analysis.
The sub-dialects of Sanana share definable characteristics, but they are yet to be thoroughly described. For the purpose of the comparisons in this section, “Sanana” forms are the forms received by Facei tribe members on the island of Sanana unless otherwise noted. Collins (1982: 83) infers that there might be a dialect division in the geographical region covered in this section. He remarks, “the status of Capalulu and Urifola [sic], also Sula dialects, is not clear.” My field research has revealed that CMD Sula is demonstrably different from both Mangon Sula and Sanana Sula, and the present research demonstrates that it genetically groups more closely with varieties of Sanana and has undergone dialect leveling due to contact with Mangon (see Figure 15).
Interviews conducted with Sula speakers in 2010 and 2014 indicated that there are not any additional primary dialect divisions in Sanana, though it is not out of the question that more could be found. As discussed in the previous section, the island’s dialect map is complex, and there are certainly a number of additional sub-dialects yet to be defined. One such example is the Sula variety spoken in the town of Malbufa. That town represents an opposite scenario to that described in this section: Malbufa is a Mangon tribe settlement on Sanana Island rather than a Sanana tribe settlement on Mangon Island. My short time spent in Malbufa left me with the impression that it too has developed a unique dialect. If that variety is indeed a newly formed dialect, a comparison between Malbufa and CMD could make a significant contribution to our understanding of dialect formation processes in general, particularly since nearly all putative cases of new dialect formation involve European languages (mostly English and Norwegian), and because the contact scenarios are nearly opposite, so a side-by-side comparison could help us differentiate universal tendencies in dialect-leveling from changes that are specific to a given contact scenario.
3 New dialects. Dialect forming
Figure 15. Genetic relationship of Sula’s dialects.
CMD Sula genetically groups within Sanana Sula, but it has undergone dialect leveling due to contact with Mangon Sula.
Dialect-forming processes include phonological leveling that disprefers word-final consonants, and lexical leveling including (a) a tendency for the new dialect to borrow unmarked words that are phonologically simpler, (b) a handful of innovative lexical items, and (c) interdialect forms that are intermediate to the source dialects. The most readily apparent change found that characterizes the speech of this region is the re-emergence of word-final high vowels, where they were previously lost in Sanana dialects (discussed in Chapter 2: dialects. part 1). That is, Sanana deleted word final high vowels following non-glottal consonants, but they remain following vowels and glottal consonants (e.g. tui ‘snake’, yau ‘far’, behi ‘to throw’, gahu ‘two’) (Table 27). That change and other levelings plus evidence of interdialect formation to be discussed, have set CMD apart from its progenitor.
Table 27. Loss and retention of final high vowels in Sanana dialects
i, u > Ø / [-cont, -glottal] _#
These data demonstrate the deletion of word-final high vowels following non-glottal consonants in Sanana dialects, and by comparison, help to demonstrate that CMD is a unique dialect that has undergone contact induced changes.
PSM Sanana English
*waki dabu > wak dab ‘to think’
*saŋa-petu > sanapet ‘thatch/roof’
*samamu > samam ‘to chew’
*nibu > nib ‘to sit’
*saku > sak ‘to pierce’
*afu-tuka > aftuka ‘ash’
*gasi > gas ‘salt’
*kabaresi > kabares ‘evil’
retention /h,ʔ_#
*sanohi > sanohi ‘family’
*gehi > gehi ‘to stand’
*kiʔi > kiʔi ‘3SG’
New dialects can form through leveling processes when speakers of one dialect settle in a region with an established, mutually intelligible dialect, and when a community that is comprised of speakers of mutually intelligible dialects is established in new territory—this being the New Town (immigrant) model (e.g. Kerswill 2002, Kerswill and Williams 2005). The case in this study describes the first scenario: Facei tribe members settled communities adjacent to established Mangon communities, and contact-induced changes took place. These changes included both phonological and lexical leveling and interdialect formation wherein characteristics from both source dialects were retained in the resulting forms (e.g. Britain and Trudgill 2005).
3.1 CMD background
Collins (1982) notes that most Sanana settlements on the island of Mangon are quite recent with the exception of those in the CMD dialect region (Figure 16). My own fieldwork consultants in the three CMD communities have self-reported to speaking the same dialect as one another, and this was corroborated by personal observation and observation from a Sanana-speaking assistant. As Collins suspected, the language variety of the CMD region is indeed different from both the Mangon variety and the Sanana varieties. Divergent features are both phonological and lexical, as will be demonstrated.
Figure 16. CMD—Mangon—Sanana community and dialect regions
This map shows the boarder between CMD dialect communities and Mangon dialect communities on the island of Mangon. The island of Sanana is home to Sanana communities (including the traditional home of Fagudu, Falahu, and Facei tribes).
CMD is a distinct Sanana dialect separated from other Sanana varieties. All of the generalizable innovations that define the Sanana dialect group are present in CMD (i.e. *ŋ>n; *d>r /V_V; *l>h/[V_V], [#_]; u>a/oσ_#)—including Sanana’s loss of final high vowels following non-glottal consonants (Table 27). This innovation can be observed in variant forms like gas ‘salt’ (< *gasi) and gatel ‘three’ (< *gatelu) as well as in compound words like napØfoa ‘hair’ (< *ŋapufou). That said, ongoing leveling in CMD, results in a dialect that is distinguishable from both sources.
The language change processes appear to be unconscious, as community members indicate a high degree of pride in their Facei heritage and many are under the impression that they speak a typical Sanana dialect, declaring either that they speak li Sanana ‘Sanana language’ (the Sanana variety) or li Facei ‘Facei tribe language’ (which is then clarified as being essentially the same as Fagudu and Falahu). A few linguistically astute consultants reported speaking li campur (mixed language), or li Orifola/Wai U/Capuli (the names of the village they reside in). The vast majority of people questioned self-reported speaking a dialect corresponding to their tribal affiliation. One crucial point to note is that no CMD speakers reported speaking li Mangon ‘Mangon dialect’, and during sessions, if asked whether particular forms were also acceptable, CMD consultants commonly respond that the forms were only acceptable in li Mangon. In essence, CMD residents know they’re not speaking Mangon, and they can perceive a difference between their dialect and what is spoken on Sanana island, yet due to tribal affiliation, they still identify as Sanana speakers.
3.2 Sula dialect contact
A symphony of phonological and social processes work in concert to form new dialects. Some of the social processes pertinent to the data at hand are founder population effects, swamping effects, and to a limited degree, even early stage creolization.
A newly-settled region’s founder population is theorized to limit the amount that subsequent waves of immigrants can affect dialectal change (Gordon et al. 2004). On the island of Mangon, the Mangon tribe is the original founder population and has been established in the area since long before recorded history. True to Gordon’s hypothesis, the Facei settlers who began arriving ~300 years ago do not appear to have significantly affected the Mangon dialect. While it does not appear that the introduction of Facei settlements resulted in significant changes to the Mangon dialect, the reverse did occur: the Facei immigrants Sanana dialect was largely modified by Mangon speakers.
Swamping effects, as discussed in Gordon et al. (2004), have to do with cases where the larger population’s variety overwhelms the minority’s variety. Thus, in the case of Mangon island, either (a) minority CMD forms would be pushed out in favor of majority Mangon forms or (b) Mangon forms could have been pushed out by if the Facei immigrant population had been enormous. Neither of those effects occurred, however: (a) although contact with the Mangon dialect resulted in significant changes, CMD retained many lexical items, and it has retained all of the innovations that define the Sanana dialects as distinct from Mangon. The second form of swamping (b) was not applicable, because the Facei settlers never approached the population numbers of the Mangon tribe.
The changes that took place in CMD, are not indicative of a creolization process, because the source language varieties were closely related. In a traditional view of creolization, a generation of speakers does not receive the previous generation’s languages to competence. Children instead receive a pidgin from multiple source strata and use that as the basis for a new language—the creole (e.g. Bickerton 1976, 1981). Although CMD underwent changes, the two sources were closely related, and the result does not approach new language formation; Mangon and Sanana dialects themselves are by many accounts mutually intelligible (at least among speakers who maintain a modest amount of contact), and CMD is still mutually intelligible with both. Additionally, intergenerational transmission was never interrupted, and modern CMD is a dialect that shares features with both its Mangon neighbor and also with Sanana dialects.
Dialect leveling occurs when there are two or more original varieties, and at least one of the marked forms is no longer transmitted the original way. This can be associated with highly mobile majority and minority communities settled in a small geographical range where minority forms are replaced (or modified) by their majority counterparts (Kerswill 2003). This scenario can explain the numerous Mangon borrowings present in CMD, and it fits the physical and social layout of the central southern coastal communities of Mangon Island. While the island of Mangon is large and mostly uninhabited, it is reported that the CMD communities have engaged in daily social interaction and trade with their Mangon-speaking neighbors since initial settlement. The communities straddling the CMD–Mangon boundary are less than three km apart (Orifola and Waitulia), and the length between the furthest CMD-speaking community, Capalulu, and the furthest Mangon-speaking community, Waitina, is less than 20 km—a journey that can be covered on foot within a day under most weather conditions.
3.3 Accommodation and change
Speech accommodation is a probable root of much of the leveling that took place in CMD. This phenomenon is known to happen when mutually intelligible dialects share cooperative communicative intent (Kerswill 2003), and that was the likely early contact scenario between CMD and Mangon. Mangon and Facei tribes consider themselves to be cousins within the Sula ethnic community, and the groups do not have a history of hostility or warfare. The island of Mangon is still mostly uninhabited today, and it is rich in resources, so competition was not necessary. Sula language consultants do indicate a level of prestige assigned to Mangon though, and it is likely that CMD speakers accommodate to Mangon speakers because of this and because Mangon is a much larger population group.
Perhaps in accommodating to Mangon speakers, CMD speakers developed a preference against word-final consonants to match Mangon’s near prohibition. This is in contrast to typical Sanana dialects where final vowels after non-glottal consonants were deleted. But since Mangon and Sanana each have unique lexical items, and since even the cognate forms often have differences beyond the word-final segment, CMD could not simply borrow the word-final vowel from Mangon. Instead a process of vowel epenthesis seems to be in place to fill the gap. This final consonant prohibition—likely rooted in early hypercorrection—was overgeneralized and applied to some final liquids and nasals, even where they are not prohibited in Mangon dialects.
Hypercorrection tends to coincide with a prestige imbalance between two varieties of a language, and it has also been demonstrated between a first and second language when an L2 speaker is aware of a partial mismatch between a phonological constraint or grammatical rule, but s/he over applies a correction to contexts where there is no mismatch (e.g. Eckman et al. 2013, Carey 2005). Corroborating evidence for a phonological dispreference is seen with CMD words whose primary semantic form is non-cognate to the Mangon counterpart: e.g. tilu ‘ear’ (where til is found in Sanana but talinga is found in Mangon). These examples show that they were not simply borrowing Mangon forms into their dialect but instead perceiving a gap and applying epenthesis to fill that gap—even when the epenthesis does not actually result in a match to the Mangon form.
4 New dialects. Data examination
This section examines elicitation data to answer the questions:
1. Do CMD speakers speak a Sula variety that is demonstrably different from others?
2. What are the differences, and how did they arise?
Epenthesis of final [i] and [u] on CMD forms was considered, because there is a near complete correspondence between word-final high vowels following non-glottal consonants in Mangon and final consonants in Sanana. CMD speech displays many instances of final [i] and [u], and that is unexpected considering that the innovations (to be described) genetically group the CMD dialect with Sanana dialects.
The paragogic high vowels given to words with final consonants appears to satisfy the previously postulated phonological preference against word-final consonants introduced from Mangon. In the analysis of CMD, presence or absence of final vowels was determined through target elicitation and sessions were recorded to ensure transcription accuracy. In the Mangon dialect, final [i] and [u] is determined by historical retention. In CMD, the presence of final [i] and [u] usually matches the final vowel in Mangon on cognate pairs, but there is no identifiable phonological condition determining how [i] or [u] is selected in CMD forms that do not share a Mangon cognate. This topic should be further explored in future studies.
4.1 Methods
The CMD data in this study were gathered during a two-week lexical documentation project conducted during the summer of 2014 and during a two week followup during summer 2015. Comparison data were collected on the islands of Sanana and Mangon during three three-month stays in Indonesia and several shorter duration trips. A total of 17 Sula-speaking sites were sampled on Sanana and Mangon islands for this project (Figure 17). Data were collected for the three dialects being compared at 11 Sanana sites, 3 Mangon sites, and 3 CMD sites.
Figure 17. Data collection sites on Sanana and Mangon islands
As with other studies carried out during my doctoral research, sites were chosen after consulting with the regional development office, the bureau of statistics, and community elders to find out, (1) which tribe settled each of the islands’ villages, and (2) which tribe’s dialect each village was reported to speak. A 230-word basic vocabulary list was elicited for each proposed tribal dialect. The list was adapted from the Austronesian Basic Vocabulary Database (Greenhill et al. 2008). Recordings were made on a Zoom H4n solid state recorder at 24bit/96kHz in WAV format. Recordings will be archived in Kaipuleohone, the University of Hawai’i Digital Ethnographic Archive. Each elicitation session had a minimum of two speakers (often 3–4) to help keep participants primed in Sula and to counteract interference that could arise due to Malay as the elicitation language. Sessions occurred onsite in the village whose dialect was being elicited to further maintain production of local word forms. The two primary speakers of each session were also born and raised in the village, but whenever possible, an additional speaker from another dialect was invited to sit in during word-list elicitation and interject where they could provide comparative observations that might otherwise go unnoticed. The CMD participants consisted of three women and three men ranging in age from early 40’s to mid-late 50’s.
After fieldwork data collection, CMD data were compared side-by-side in spreadsheet format to Sula data representing Falahu, Fagudu, and Facei tribe speakers on Sanana island and to Mangon tribe speakers on Mangon island. Forty-one words from the 230 item elicitation list were identified as containing an environment for word-final /i/ or /u/ epenthesis (Appendix F), and the tokens produced by each CMD participant were tallied both with and without the variant. Additionally, all of the items found to be non-Sanana vocabulary on the Mangon-dialect elicitation list (i.e. Mangon forms) were identified and compared to the CMD data to discover which had been borrowed by CMD and whether anything is generalizable about them. Speakers did not produce every word, nor did they each produce the same number of tokens, so there is insufficient data to determine which if any of the words might behave differently or what conditions might explain intra-speaker variation. Those topics should be explored followup studies.
5 New dialects. Results
Data comparison revealed CMD to be significantly different from the other speech varieties. The historical loss of word-final vowels in Sanana resulted in final consonants that are prohibited in Mangon, which allows only [+sonorant] consonants word finally (liquid and nasal), yet Mangon’s prohibition against word-final consonants has mostly transferred to CMD where it is in some cases even overgeneralized (resulting in epenthesis on some [+sonorant] final consonants), but the prohibition has not been borrowed completely, and forms with final consonants are sometimes produced without any identified condition triggering production or absence of the segments.
5.1 Phonological leveling
Forty-nine words were identified as containing a target environment for word-final /i/ or /u/ epenthesis. These include both (a) words that are cognate with Sanana forms ending in non-glottal consonants and (b) words corresponding to Mangon forms that end in CV[+high] (including liquid and nasals which are not always followed by vowels). Nine of the 49 forms were not produced with a final vowel. Seven of the nine end in liquid or nasal consonants, and two have a final /t/. A total of 183 tokens were present in session recordings and tallied by speaker. An additional 46 tokens were identified in which a target environment was only present in the first word of a compound, but these were excluded in the analysis because CMD retains Sanana’s deletion of high-vowels in that environment.
The average of all speakers’ final /i/ and /u/ production across lexical items was 75% of produced tokens representing words that are consonant final in the Sanana dialect. Percentages varied from speaker to speaker, ranging from 60% (Man 2) to 85% (Man 3). Figure 18 displays the frequency with which each speaker produced a token with a final vowel (either /i/ or /u/). Vocalic paragoge appears to be a change in progress, as there is intra-speaker variation without discernible conditioning: i.e. no other factors (e.g. word or phonological) were found to correlate with the variation. It appears the change is regular and unconditioned but still incomplete; however, followup studies should specifically test more factors.
Final epenthesis has progressed further with /i/ than /u/: all speakers but one produced a final /i/ a majority of the time, and none did so less than half the time. At least one final /i/ token was produced for all but three of the 22 forms where it was expected, and the remaining three forms all end in [+sonorant] consonants (Table 28)—which are not prohibited by Mangon dialect even though those particular Mangon cognates do happen to have final vowels. That is: the Mangon forms nonu boli ‘to lie down’, koli ‘skin’ and kafini ‘mosquito’ end in a final vowel, and the CMD counterpart was not produced with one; however, Mangon does not entirely forbid [+sonorant] consonants, so the preference against final consonants that was borrowed from Mangon might not phonologically prevent CMD from retaining the sonorant-final Sanana forms, bol, kol, and kafin. A followup study should examine whether final /i/ variants exist for these forms that were not produced during data collection. Appendix F also lists the percentage of times each form was produced with or without the final vowel.
Figure 18. Percentage of tokens produced with final /i/ & /u/ in CMD
Total number of tokens produced by each speaker indicated in parentheses.
Table 28. Words where final i would be expected in CMD.
Forms with final i
Sanana CMD Mangon English
ul uli mankawai ‘worm’
am ami gami ‘to squeeze’
duk duki duki ‘to come’
lif lifi lifi ‘to turn’
kag kagi kagi ‘to fear’
ek eki eki ‘neck’
mak maki maki ‘tongue’
banap banapi banapi ‘to shoot with a gun’
jub jubi jubi ‘to shoot with bow’
det deti deti ‘to cut/hack’
basel baseli batani ‘to plant’
pan pani pani ‘wing’
manip manipi manipi ‘thin’
dagat dagati dagati ‘narrow’
bal bali baali ‘shy’
bot boti boti ‘white’
mit miti miti ‘black’
gas gasi gasi ‘salt’
Forms without final i
Sanana CMD Mangon English
nona bol nona bol nonu boli ‘to lie down’
lan lan laŋi ‘sky’
kol kol koli ‘skin’
kafin kafin kafini ‘mosquito’
Final /u/ paragoge is also frequent in CMD. All but one speaker produced a final /u/ variant in the majority of instances, and the speaker who did not was underrepresented in the sample, producing only four tokens, so it is inconclusive whether his speech actually deviates from the norm. It is however worth noting that he produced a low percentage for /i/ as well.
Altogether, the final vowel variants were produced for thirteen of nineteen candidate forms where final /u/ was expected (Table 29). Of the six forms that were not produced with final /u/ variants, four end in a liquid or nasal, which as explained above, is not prohibited in the Mangon dialect. The remaining two forms, gapit ‘seven’ and sanapet ‘roof thatch’ were more problematic, but follow-up research showed a preference for vowel-final forms, gapitu and sanapetu in the village closest to Mangon speakers, and that preference fades moving westward away from Mangon speakers (reference data in Appendix I).
Table 29. Words with final /u/ target environments.
Final /u/ epenthesis attested in the data
Sanana CMD Mangon English
nib nibu nibu ‘to sit’
bamap baumapu bamapu ‘to cook’
fat fatu fatu ‘stone’
bag bagu bagu ‘thick’
til tilu talinga ‘ear’
nan nanu naŋu ‘bathe’
nap napu ŋapu ‘head’
as asu fako: ‘dog’
nib nibu nibu ‘sit’
sak saku saku ‘to pierce’
ak aku aku ‘1SG’
No final /u/ epenthesis attested
Sanana CMD Mangon English
samam samam samamu ‘to chew’
win win winu ‘to drink’
mantel mantel man(u)telu ‘egg’
gatel gatel gatelu ‘three’
gapit gapit gapitu ‘seven’
sanapet sanapet saŋapetu ‘roof thatch’
5.2 Lexical leveling
Although CMD has borrowed several lexical items from Mangon, most words remain from the Sanana dialect. Forty forms in the elicitation data were found to be unique to the Mangon dialect (i.e. not produced by speakers on Sanana island), and ten of these forms were also present in CMD, indicating borrowing. Kerswill and Williams (2000) propose that geographically and socially marked forms tend to be lost. However, Britain and Trudgill (2005) describe lexical simplification as a process that might help shed light on why some of the 40 were borrowed while others were not. That is, it is claimed that where there are competing dialects, simpler forms win out (e.g. Trudgill 1986:104, Britain and Trudgill 2005:184). In this situation, the minority CMD forms that were retained tend to be linguistically simpler than the majority Mangon counterparts. It is unclear to me precisely how Britain and Trudgill define the term simplification, but in the context of this study, it refers to forms containing fewer segments and/or lacking consonant clusters.
Tables 30 and 31 represent all forms from the elicitation list where Mangon speakers produced a variant that was not present on Sanana island. Some of these are borrowed into CMD while others are not. Campbell (p.c. 2015) proposed that ordinary dialect borrowing without regard to simplification might be able to account for the Mangon forms in CMD, and this should be further investigated in a followup study. However, given the available data, it would seem remarkable that such a strong pattern is present: only two of the non-borrowed forms are simpler than their Sanana counterparts. Conversely, of the ten Mangon forms that were borrowed, six are simpler than the minority counterpart (Table 30).
Table 30. Lexemes Borrowed from Mangon
Simplification principles favor borrowing of the majority word (Mangon) rather than retention of the minority word (Sanana):
Sanana CMD Mangon English
baked gena gena ‘to hear’
makariu bama bama ‘to split’
behifon kila kila ‘liver’
bisloi lika lika ‘to choose’
sagilaʔat sagila sagila ‘lightning’
tahun taun taun ‘year’
Equal complexity:
Sanana CMD Mangon English
gahu gama gama ‘to scratch’
Simplification would favor retaining the minority word (Sanana) rather than borrowing the majority word (Mangon):
Sanana CMD Mangon English
gada sakeu sakeu ‘to scratch an itch’
kaf manakem manakem ‘to hold’
sana daeti badaeti ‘branch’
Thirty Mangon forms were not borrowed. However in 28 instances, the Sanana forms that were retained are either simpler or of equal complexity to the Mangon counterparts. Followup research will probe deeper to determine to what degree the items (listed on Table 31) might indicate simple data gaps—that is, instances where both the Sanana and Mangon form exist in CMD but only the Sanana form was recorded.
Table 31. Lexemes Not Borrowed from Mangon
Forms where simplification would favor the minority form (Sanana):
Sanana CMD Mangon English
ihi ihi nisi ‘flesh’
suba suba patfo ‘rotten’
fua fua nceli ‘fruit’
kakon kakon maŋkuni ‘yellow’
aya aya fanini ‘big’
lal lal tuka ‘in’
neʔu neʔu bulela ‘below’
ota ota saka ‘hundred’
neʔi neʔe ncumi ‘nose’
ana-(nana) ana-(nana) gama-(nana) ‘child’
yai yai sanafa ‘road/path’
putar putar dagalili ‘to turn’
hoi hoi gifu ‘to open’
uka uka ŋasi ‘to bite’
as asu fako: ‘dog’
tena (kau) tena (kau) takau ‘belly’
til tilu taliŋa ‘ear’
ul uli mankawai ‘worm’
Forms where Sanana and Mangon forms are of equal complexity:
Sanana CMD Mangon English
hia hia gia ‘one’
safa safa fafa ‘rat’
buha buha boli ‘to blow’
apfei apfei bafei ‘smoke’
nana nana male ‘small’
yota yota togi ‘short’
do do di ‘and’
gareha gareha gadia ‘four’
gaya gaya giya ‘to eat’
basel baseli batani ‘to plant’
Forms where simplification would favor the majority form (Mangon):
Sanana CMD Mangon English
buakeu buakeu safe ‘to spit’
gahu gahu gai ‘to dig’
5.3 Other lexical leveling
In addition to borrowings from Mangon, there are a handful of forms in CMD that appear to be unique among Sula dialects in that they are not found in Sanana or Mangon. Some examples are CMD nuba ‘mouth’ for Sanana bayon and Mangon boni; CMD haku ‘to cook (rice)’ for which Sanana and Mangon did not report to have a special form; CMD buhi ‘night’ for Sanana bauhi and Mangon bedi; and CMD so ‘where’ for Sanana s(ah)oa and Mangon sibo. Some possible sources of these differences could be spontaneous changes, borrowing from unknown source languages, retentions that were subsequently lost in other dialects, and taboo word replacement. Collins (1989) identifies replacement of taboo vocabulary among Taliabo speakers on the neighboring island, and my consultants have reported that a similar system previously existed in Sula but has fallen out of use.
5.4 Other dialect-forming processes
Interdialect formation can cause forms to change in ways that are intermediate between the minority and majority parent forms (Britain and Trudgill 2005). CMD displays a number of such examples, such as those in Table 32. One possibility suggested by Campbell (p.c. 2015) is that CMD underwent its own normal sound changes that caused these words to appear halfway between Sanana and Mangon. This seems unlikely though, because the changes are not generalizable on phonological environment. For instance, if CMD were to have undergone a vowel paragoge process independently, we would expect a conditioned distribution between /i/ and /u/. Instead we see instances of both, for example: ami ‘to squeeze’ with a high front vowel following a nasal but nanu ‘to swim’ with a high back vowel following a nasal; uli ‘worm’ with /i/ following a liquid but also tilu ‘ear’ with /u/ following a liquid.
Table 32. Interdialect Forms
Words with Sanana bases and epenthetic final vowels matching Mangon’s preference against word-final consonants:
Sanana CMD Mangon English
am ami gami ‘to squeeze’
bal bali baali ‘shy’
basel baseli batani ‘to plant’
pan mpani/pani manpani ‘wing’
batut batutu dotu ‘to grow’
nan nanu naŋu ‘to swim’
nap napu ŋapu ‘head’
Glottal stop reduced in CMD, but vowel length (which is not a phonotactic violation in Mangon) is retained:
Sanana CMD Mangon English
maʔana maana mana ‘man’
Words whose Sanana bases differ markedly from their Mangon counterparts but take epenthetic final vowels in CMD to match Mangon’s preference against word-final consonants:
Sanana CMD Mangon English
as asu fako: ‘dog’
saik siku isuka ‘here’
til tilu talinga ‘ear’
ul uli mankawai ‘worm’
Interdialect forms are a result of sound change and not a mechanism of sound change. Some of the forms in Table 32 are also in the previous tables demonstrating paragogic epenthesis, if the epenthesis they underwent resulted in interdialect forms rather than a form that matches the Mangon counterpart. The forms found on both lists are additional evidence that words with final high vowels in CMD result from epenthesis rather than lexical borrowing (i.e. they result from dialect formation rather than dialect shift). These forms help make CMD recognizable as a distinct dialect.
Weak constraints in CMD—for instance constraints avoiding matching vowels or for lexical distinctiveness—could also be mechanisms responsible for forms like these, however there is no evidence for this in the data. These possibilities should be investigated in followup studies. Regardless of the mechanism(s) behind the changes, the result is 12 forms that are intermediate between the two input dialects, and these forms help to distinguish CMD as a dialect in its own right.
6 New dialects. Future work
Some of the findings presented in this section were not recognized at the time of the data collection and were not tested during the course of fieldwork. Moreover, the number of participants was small and not adequately balanced, and the data available is insufficient for making solid conclusions. Nevertheless, tendencies surfaced during analysis of field data corroborating that changes in the CDM region have caused it to become a recognizable dialect that is distinct from both Mangon and Sanana. The shortcomings will be addressed in future research.
7 New dialects. Conclusion
It is clear that CMD has undergone considerable language change in the generations since the community departed from Sanana to settle on Mangon, and ample evidence is presented indicating that CMD has developed into a distinct Sula dialect. However, the evidence is insufficient for making conclusive declarations about some of the specific mechanisms for these changes.
This research has found that (1) dialect leveling has occurred and may still be underway in CMD, (2) that dialect leveling was largely in the form of borrowing simpler lexical forms from Mangon and of borrowing Mangon’s phonological preference against word-final consonants, and (3) that the changes in CMD likely occurred due to a regular practice among early settlers of accommodating to Mangon speakers, who were more numerous and more established (both in terms of physical infrastructure and human resources/skilled labor).
The most apparent characteristic setting CMD apart from other Sanana dialects is its final epenthetic vowels. These final vowels resulted in forms that resemble their Mangon neighbors; however, Sanana-specific innovations are still visible within word stems, so it can be shown that these forms are not borrowings from Mangon but rather Sanana derived items that were subsequently modified because of final consonant disfavoring. Evidence for this is (a) that epenthesis is also applied to the Sanana forms (making them unique to CMD), and (b) that the process was overgeneralized to include some words with final [+sonorant] consonants (which are not prohibited in Mangon).
Forty forms were found in the elicitation data that were unique to the Mangon dialect. Ten of those forms appear to have been borrowed into CMD. Of the 30 for which borrowing was not found, all but two were of equal or greater complexity to the minority Sanana forms. In other words, of 40 available Mangon forms, only two of those not borrowed are simpler than the Sanana counterparts. The other 38 are either borrowed or simplification would not necessarily predict them to be. Additionally, a number of lexical items are found to be unique to CMD, further positioning it as separate from both its Mangon and Sanana cousins.
Lastly, this study also presents several interdialect forms that contain components from both of the source dialects. Any of the observations presented in this study might be speculated to result from numerous causes; however, when considered together, they are indicative of CMD as a distinct new Sula dialect that is likely still undergoing leveling processes.