When Language Maintenance Means Language Shift: Tibetan as an Heritage Language in Amdo Families in France

Abstract

This paper explores the initial steps of transgenerational language change in exile by comparing the varieties of “Common” Tibetan as spoken by parents born in Amdo, Tibet, and by their teenager children, born in Tibet or in South Asia, who arrived in France at an early age and who have spent all or most of their schooling in France. In these families, the parents speak a variety of Amdo Tibetan as their first language, which does not allow for inter-comprehension with “Common” Tibetan. They have acquired “Common” Tibetan during their stay in South Asia before they moved to France. The paper follows a descriptive approach to analyze the structural (dis)similarities between the parents’ and the children’s varieties of “Common” Tibetan. It also documents intra-generational variation (1) within the parents’ generation, where we can observe a variable extent of retention for Amdo Tibetic features, and (2) within the children’s generation, where variation is usually due more to the (often contact-induced) linguistic changes than to the retention of some of their parents’ linguistic features.

1. Introduction

The Tibetan diaspora in the world represents a population of circa 120,000 people. Originally resettled in South Asia after the Chinese invasion and the flight of the Dalai-Lama into exile in 1959, a worldwide diaspora has gradually emerged from the 1990s onward, with the departure of a number of Tibetans to certain Western countries (Robin, 2024). In France, the Tibetan population has grown rapidly over the last twenty years, from circa 100 people in 2000 to around 15,000 people now. Both in South Asia (India, Nepal, and Bhutan) and in the West (Northern America, Western European countries, and Australia), efforts have been and are still made to foster the transmission of the Tibetan language in the diaspora (Phuntsog, 2018; Robin & Simon, forthcoming in 2026).

The question of language maintenance outside Tibet and in the West raises the question of the language variety to be transmitted. “Tibetan” is, indeed, a cover term for more than 200 languages, not always mutually intelligible, directly derived from Old Tibetan (Tournadre & Suzuki, 2023, p. 57). In the face of such diversity, a variety of “Common” Tibetan has emerged as a lingua franca in South Asia, mainly based on Central Tibetic (Vokurková, 2022; Schmidt, 2022). The emergence of this lingua franca in exile is thus closely linked to the question of dialect contact (Trudgill, 1986) and wider language contact situations in the diaspora, but remains largely unexplored (Ward, 2015).

This paper explores the initial steps of transgenerational language change in exile by comparing the varieties of “Common” Tibetan as spoken by parents born in Amdo, Tibet, and by their teenager children, born in Tibet or in South Asia, who arrived in France at an early age and who have spent most or part of their schooling in France. In these families, the parents speak a variety of Amdo Tibetan as their first language, which does not allow for inter-comprehension with “Common” Tibetan. They have acquired “Common” Tibetan during their stay in South Asia before they moved to France. Their idiolectal variety of “Common” Tibetan is now the language they use on a daily basis in the nuclear family and with other members of the diaspora, whereas Amdo Tibetic is used in more limited contexts (e.g., with fellows from the same village or micro-area). Preliminary research has shown that, in such families, “while parents retain (to a highly variable extent) phonological, morphosyntactic, and/or lexical features of their native Tibetic language when they speak ‘Common’ Tibetan, only very few or no such features can be found in the variety of ‘Common’ Tibetan spoken by their children. Tibetan as a ‘heritage language’ is thus inherited less from the parents than from the wider community within which they grow up” (Simon, 2023). As we will see from the linguistic biographies below, the members of the interviewed families are or have been in regular contact with the Tibetan community around Paris, both with people from Amdo and from other regions.

Building on these preliminary findings, the paper follows a descriptive approach to document and analyze the structural (dis)similarities between the parents’ and the children’s varieties of “Common” Tibetan.

2. Languages, Methodology, and Corpus

2.1. Amdo and “Common” Tibetan

Amdo and “Common” Tibetan are two Tibetic languages, i.e., languages directly derived from Old Tibetan, the language of the Tibetan Empire (8th–9th century A.D.) (Tournadre & Suzuki, 2023, p. 44). Both languages belong to two different groups within the Tibetic languages (resp. the Northeastern and the Central group1). While the Amdo Tibetan varieties belong to the Northeastern section of Tibetic, “Common” Tibetan is a variety of Central Tibetan, i.e., the varieties spoken in the Lhasa region, some 2000 km away from Amdo, as shown in Figure 1 below. Despite being genetically related, and although the speakers share one and the same written language norm, Amdo and “Common” Tibetan are not mutually intelligible.

In this paper, “Common” Tibetan refers to the variety of Central Tibetan used as the vehicular language outside Tibet. This language emerged in the 1960s within the first communities of Tibetan refugees in South Asia, and more recently spread to North America, Europe, or Australia with the movement of “onward-migration” (Frilund, 2019) and the constitution of a worldwide Tibetan diaspora. Given its present worldwide spread and the various sociolinguistic background of its speakers, it is likely that structural linguistic and, at the very least, lexical differences exist or are developing in the different parts of the diaspora—hence, the quotation marks for “Common” Tibetan: it is the common language of members of the Tibetan diaspora, but, as this paper shows, structural dissimilarities exist, even within one family. Thus, we use “Common” Tibetan as a loose term, defined by its function as lingua franca in exile rather than by specific structural features. To date, Vokurková (2022) is the only attempt to provide a linguistic description of “Common” or “Diaspora” Tibetan in a contrastive perspective with Lhasa Tibetan.

This paper focuses on “Common” Tibetan as spoken by people with an Amdo Tibetic background to document contrastively the extent of adoption of Central Tibetic linguistic features in the first and the second generation of exiles.

2.2. Data Collection

The analysis is based on parallel mini-corpora recorded in 2024–2025, from parents and teenager (14–17 y. o.) children in four families living nearby Paris.

In Family 1, both parents are from Amdo. They arrived in South Asia as young adults. While Father-1 had completed a bilingual (Chinese–Tibetan) curriculum in Amdo, Mother-1 was illiterate upon arrival. Thus, she was enrolled in a programme of the Central Tibetan Administration to learn the basics of reading and writing, together with a bit of English. The people who follow this programme are brought together in a boarding school for a few months, and this stay thus constitutes the first daily contact between Tibetans recently arrived from different areas of Tibet and teachers born in exile or living there for a long time. Mother-1 and Father-1 had lived about one decade in India and Nepal. Their two children are born there, and they began their very first years of schooling in the network of Tibetan schools in South Asia. They moved to France when the two children were 6 and 9, respectively. At the time of the interview, the family had lived in France for ten years. Both children attend French school, and Child-1A, the elder, is also enrolled in a community school once a week.

Family 2 is a linguistically “mixed” family: the father has a Kham, i.e., Southeastern Tibetic, variety as his mother language, even though he also learned Amdo Tibetan from a young age. Both parents have completed a bilingual (Chinese–Tibetan) curriculum in Tibet. They have spent 5 and 7 years, respectively, in South Asia, before moving to France with Child-2A and Child-2B in 2013. Both children were of pre-school age at that time. After their arrival, they have lived some time relatively isolated from the wider Tibetan community because they have been granted accommodation in a fairly distant suburb. For a time, this location prevented them from enrolling their children in the Sunday community school. At that time, both parents being fully literate in Tibetan and particularly committed to the transmission of Tibetan, used to teach themselves written Tibetan and to read children literature with them. At the time of the interview, Child-2A and Child-2B attended a community school every week. Mother-2 is a stay-at-home mother and has deliberately decided not to put their youngest child (born in France) in daycare, to ensure a maximal exposure to the Tibetan language before entering the compulsory nursery school.

In Family 3, both parents are from Amdo, and Mother-3 was illiterate upon her arrival in exile, while Father-3 had completed a full curriculum in Amdo. They arrived in South Asia in the early 2000s, when they were in their twenties. Father-3 arrived in France some twelve years before the interview took place, but Mother-3 and Child-3 could come to France nine years later only, when Child-3 was 14. Thus, Father-3 has spent more than ten years, and Mother-3 more than twenty years, in South Asia. Child-3 has completed most of her schooling in a school run by the Central Tibetan Administration. Among our interviewees, she is the one who has been the most intensively exposed to the South Asian “Common” Tibetan variety, and the one who arrived most recently in France. During their stay in South Asia, both parents used to work in a “Common” Tibetan speaking environment. After the birth of Child-3, they also had the opportunity to visit their family in Tibet for two months. At the time of interview, Child-3 was enrolled in a French public school and attended a community school once a week: as there are no Tibetans in her neighbourhood, it is the main opportunity for her to meet young Tibetan speakers.

Family 4, like Family 2, is a linguistically “mixed” family where the father has a Kham Tibetic variety as his mother language. Mother-4 had left Tibet in her early twenties, and spent ten years in South Asia. She did not attend school in Tibet, and, like Mother-1, was enrolled in the Central Tibetan Administration programme for illiterate adults upon arrival. She settled in France ten years before the data collection, and Child-4 could join her in France four years later, at the age of 7. Child-4 thus completed her first year of schooling in the network of Tibetan schools in South Asia. At the time of the interview, Child-4 attended an online community school, run by a monastery based in South India, which offers a 1–2 h class once a week.

In these four families, a variety of “Common” Tibetan is the children’s first language and the language of daily communication with their parents (children speak both French and Tibetan with each other). The precise characterization of the variety spoken with the different interlocutors remains somehow fuzzy. Indeed, despite the significant structural differences between Amdo and “Common” Tibetan, neither the parents nor the children conceive of them as two distinct languages. Nevertheless, all parents clearly indicate that they do not speak to each other or their children in the same way as they would speak to members of their family back home: they tend to speak to them using a form of “Common” Tibetan more or less strongly influenced by an Amdo Tibetan substrate.

The data have been collected mainly through face-to-face interviews, and via Zoom with the members of Family 4. The interviews were conducted by speaking exclusively my own variety of “Common” Tibetan,2 in order to encourage them to replicate the way they speak to their children. The data include the following:

• A word list targeting specific phonological features and lexical items, repeated two or three times with each speaker (collected with help of a set of pictures to avoid interferences from Written Tibetan (WT), and because some speakers have limited reading skills in Tibetan);
• A narrative based on an excerpt of Shaun Tan’s graphic novel The Arrival (Tan, 2006, hereinafter referred to as “Arrival”): this parallel corpus is the source for grammatical analyses;
• Semi-structured, collective interviews about the family members’ linguistic biography and daily life have also been conducted in order to obtain a more precise overview of the linguistic profile of each family and provide the context to analyze the data. For Family 1 and 2, the linguistic biography had been recorded earlier, in 2022.

The corpus will be archived and made available as part of the Pangloss Collection, within the subcollection devoted to “Common” Tibetan.

Table 1. Data collected for the research.

Family	Family Member	Recorded Data and Duration
1	Mother-1 (Mo1)	Word list; “Arrival” (14 min. 11 s., with Fa1) Linguistic biography of the family (62 min. 8 s., with Fa1)
	Father-1 (Fa1)	Word list
	Child-1A (Ch1A)	Word list; “Arrival” (8 min. 1 s.)
	Child-1B (Ch1B)	Word list
2	Mother-2 (Mo2)	Word list; “Arrival” (7 min. 13 s.) Linguistic biography of the family (58 min. 20 s.)
	Child-2A (Ch2A) and Child-2B (Ch2B)	Word list; “Arrival” (8 min. 20 s.) Linguistic biography (50 min. 32 s.)
3	Father-3 (Fa3)	Word list; “Arrival” (21 min. 52 s.) Linguistic biography of the family (44 min. 16 s., with Mo3)
	Mother-3 (Mo3)	Word list
	Child-3 (Ch3)	Word list; “Arrival” (12 min. 27 s.)
4	Mother-4 (Mo4)	Word list; “Arrival” (26 min. 26 s.) Linguistic biography of the family (54 min. 26 s.)
	Child-4 (Ch4)	Word list; “Arrival” (14 min. 47 s.) Linguistic biography (21 min. 37 s.)

below indicates which members of the family were recorded and summarizes the data collected.

Based on this corpus, the next sections present a selection of linguistic features chosen to form a sample of different subdomains of the linguistic system, which differ significantly in form and/or function from Amdo and Central Tibetan, thus facilitating the contrastive approach adopted here. It should be noted that this paper is only a first step in the description: other linguistic features could have been analyzed and future research should extend the descriptive work to other features. These features are analyzed contrastively in the parents’ and children speech in order to document how “Common” Tibetan emerges in just one generation. Thus, this paper aims at illustrating how different domains of the grammar differ in being prone or resilient to dialect contact-induced linguistic changes, as well as documenting new dynamics of linguistic innovations and language inheritance, influenced by the new setting in the wider diasporic context, in this case, in France (Blackledge & Creese, 2008). By doing so, we also want to illustrate the extent of phonological and morphosyntactic variation among the first generation adopting “Common” Tibetan as their daily lingua franca, and, by contrast, the relative homogenisation attested in the children’s speech.

3. Phonological Features

As mentioned supra, the Tibetic languages are all derived from Old Tibetan, the language of the Tibetan Empire (7th–9th century A.D), and “Old Literary Tibetan orthography rather faithfully reflects the pronunciation of a historical Central Tibetan dialect at the time of the script invention.” (Bialek, 2022, p. 10). The orthography of written Tibetan has largely remained unchanged since the 10th century (Tournadre & Suzuki, 2023, pp. 194–196); thus, “Classical Tibetan is closely related to modern Tibetic languages and its orthography allows us to reconstruct many ancient forms and understand the evolution of these modern languages” (Tournadre & Suzuki, 2023, p. 137).

Several phonological features would deserve a thorough examination as part of the analysis of the development of Common Tibetan in Amdo families. Due to the limited scope of this paper, we will limit the analysis to one specific phonological feature, i.e., the distribution of word-initial aspirated stops.

3.1. Reflexes of Old Tibetan Initial Voiced Stops

Amdo and Central Tibetan varieties oppose the same three series of word-initial stops, which can be traced to Old Tibetan (Hill, 2007, p. 489): voiceless unaspirated, voiceless aspirated, and voiced initial stops. However, following the development of tones in the Central Tibetic varieties, their distribution differs in the two language groups. The reflexes under examination are highlighted with a grey background.

Table 2. Reflexes of WT initial stops in Central and Amdo Tibetic ^1,2.

Written Tibetan	WT Transliteration 3	Central Tibetic	Amdo Tibetic
ཀ་	<k>	/k/	/k/
ཁ་	<kh>	/kʰ/ (high tone)	/kʰ/
ག་	<g>	/kʰ/ (low tone)	/k/
ཏ་	<t>	/t/	/t/
ཐ་	<th>	/tʰ/ (high tone)	/tʰ/
ད་	<d>	/tʰ/ (low tone)	/t/
པ་	<p>	/p/	/p/
ཕ་	<ph>	/pʰ/ (high tone)	/pʰ/ (rare), or /h/ 4
བ་	<b>	/pʰ/ (low tone)	/p/ (rare), or /w/ 4
ཀ or པ་+ྲ	<k> or <p> + <r>	/ʈ/	/ʈ/ (rare) or other realizations 4
ཁ་ or ཕ་+ྲ	<kh> or <ph> + <r>	ʈʰ (high tone)	/ʈʰ/ (rare) or other realizations 4
ག་ or བ་+ྲ	<g> or <b> + <r>	ʈʰ (low tone)	/ʈ/ (rare) or other realizations 4

¹ Central and Amdo Tibetic also have voiced stops, but these are reflexes of initial consonant clusters and are not considered in this analysis. ² The WT labial stops are not always realized as stops in Amdo Tibetan: here, only the occurrences realized as labial stops were taken into account. ³ We use the “Wylie” transliteration system (Hill, 2012). ⁴ Not taken into account in this analysis.

summarizes the stops in Written Tibetan and their reflexes in Amdo and Central Tibetic.

Thus, as we can see from Table 2, the Written Tibetan voiced initial stops are realized as unvoiced, aspirated, low-tone stops in Central Tibetic, while in Amdo Tibetan they are realized as unvoiced unaspirated stops. The measurement and comparison of the Voice Onset Time (VOT) of these consonant types for each speaker can be used to know whether they have aligned their pronunciation with the Central Tibetic one, or whether they retain the unaspirated Amdo Tibetan pronunciation. For the labial and the retroflex series, Amdo Tibetan presents more than one regular reflex: only the pronounciations as labial stops and retroflexes (respectively) have been taken into account is this study.3

3.2. Transgenerational Contrast and Intra-Generational Variation

For each speaker, a selection of words starting with each type of WT stop described in Table 2 was extracted from the corpus,4 and the VOT was measured.5 The measurements were initially carried out separately in isolated words (from the word list) and in words included in natural speech (from the corpus “Arrival”). As no significant difference was found between the data from these two elicitation methods, the measurements are combined with the results summarized in

Table 3. Average VOT in the reflex of the three stop types in the parents’ generation.

Speaker	Type of Stop	Number of Items	Average VOT (in Rounded s.)	Average Deviation (in Rounded s.)
Mo1	<k, t, p, (retrofl.)>	14	0.03 [−ASP]	0.008
	<kh, th, (ph, retrofl.)>	24	0.07 [+ASP]	0.015
	<g, d, (b, retrofl.)>	31	0.03 [−ASP]	0.005
Fa1	<k, t, p, (retrofl.)>	23	0.02 [−ASP]	0.007
	<kh, th, (ph, retrofl.)>	25	0.08 [+ASP]	0.016
	<g, d, (b, retrofl.)>	25	0.02 [−ASP]	0.009
Mo2	<k, t, p, (retrofl.)>	17	0.03 [−ASP]	0.012
	<kh, th, (ph, retrofl.)>	24	0.06 [+ASP]	0.021
	<g, d, (b, retrofl.)>	45	0.03 [−ASP]	0.017
Fa3	<k, t, p, (retrofl.)>	49	0.025 [−ASP]	0.01
	<kh, th, (ph, retrofl.)>	37	0.07 [+ASP]	0.02
	<g, d, (b, retrofl.)>	52	0.03 [−ASP]	0.01
Mo3	<k, t, p, (retrofl.)>	29	0.02 [−ASP]	0.005
	<kh, th, (ph, retrofl.)>	10	0.07 [+ASP]	0.01
	<g, d, (b, retrofl.)>	26	0.03 [−ASP]	0.01
Mo4	<k, t, p, (retrofl.)>	21	0.02 [−ASP]	0.006
	<kh, th, (ph, retrofl.)>	24	0.07 [+ASP]	0.020
	<g, d, (b, retrofl.)>	34	0.02 [−ASP]	0.011

and

Table 4. Average VOT in the reflex of the three stop types in the children’s generation.

Speaker	Type of Stop	Number of Items	Average VOT (in Rounded s.)	Average Deviation (in Rounded s.)
Ch1A	<k, t, p, (retrofl.)>	25	0.02 [−ASP]	0.006
	<kh, th, (ph, retrofl.)>	29	0.06 [+ASP]	0.017
	<g, d, (b, retrofl.)>	20	0.07 [+ASP]	0.014
Ch1B	<k, t, p, (retrofl.)>	17	0.02 [−ASP]	0.007
	<kh, th, (ph, retrofl.)>	21	0.07 [+ASP]	0.016
	<g, d, (b, retrofl.)>	21	0.06 [+ASP]	0.011
Ch2A	<k, t, p, (retrofl.)>	25	0.03 [−ASP]	0.011
	<kh, th, (ph, retrofl.)>	29	0.09 [+ASP]	0.017
	<g, d, (b, retrofl.)>	29	0.04 [−ASP]	0.013
Ch2B	<k, t, p, (retrofl.)>	23	0.05 –[−ASP?] 1	0.026
	<kh, th, (ph, retrofl.)>	9	0.08 [+ASP]	0.012
	<g, d, (b, retrofl.)>	26	0.04 [−ASP]	0.01
Ch3	<k, t, p, (retrofl.)>	56	0.03 [−ASP]	0.001
	<kh, th, (ph, retrofl.)>	40	0.07 [+ASP]	0.012
	<g, d, (b, retrofl.)>	65	0.09 [+ASP]	0.022
Ch4	<k, t, p, (retrofl.)>	21	0.03 [−ASP]	0.011
	<kh, th, (ph, retrofl.)>	25	0.08 [+ASP]	0.02
	<g, d, (b, retrofl.)>	27	0.06 [+ASP]	0.02

¹ The data for Ch2B gives uncertain results due to a smaller corpus and a generally unclear pronunciation, probably due to the teenager’s brace.

:

The data show a clear contrast between the parents’ and the children’s generation for this phonological feature: while all the parents in our sample were consistent in not aspirating the reflexes of <g, d, (b, retrofl.)>, the children of three of the four families implied in the study presented an aspirated reflex of this consonant type. It can be noted that the change from an Amdo to a Central Tibetic type of pronunciation for this sound type is not correlated to the parents’ pronunciation, nor to the fact that both parents (Families 1 and 3) or only one (Families 2 and 4) parent is a native speaker of an Amdo Tibetan variety.

In his study of tone, voicing, and aspiration of stops and affricates by speakers of “Common” Tibetan in Kathmandu, Geissler (2021, p. 148) observes that the aspiration feature interacts with the two other above-mentioned features, and shows that the VOT length for this series of stops varies: some speakers consistently present an aspiration, others “only produce aspiration in historically high-tone words”, and a third category “only produce variable prevoicing and aspiration in the appropriate low-tone words”. The present paper does not address the question of tones or voicing. Thus, one may wonder whether combining these parameters of tone and voicing with VOT length might explain why the children in Family 2 do not produce an aspiration for this consonant series. In other words, is the lack of aspiration for these consonants in Family 2 a retention of an Amdo Tibetic feature or the alignment with another subtype of the “Common” Tibetan phonological system? More research is needed to answer this question with certainty, but my hypothesis is rather that this peculiarity can be explained by the linguistic biography of the family. First, Fa2 and Mo2 have spent relatively less time in South Asia than the other parents involved in this study, and, having a high level of education upon arrival, they also have spent less time in adult education centres run by the Central Tibetan Administration, where refugees from different areas of Tibet are mixed. Second, neither Ch2A nor Ch2B had started schooling in the Tibetan school system before arriving in France: they had not been in intensive contact with exile-born Tibetan children. In France, the family location was in relative isolation from other members of the Tibetan community (at least during certain periods of their childhood). Third, their mother has a high level of instruction and is a stay-at-home mother. Thus, she was very present during their early childhood and involved in storytelling and written language teaching; being more directly involved in her children’s formal learning, she might have influenced their pronunciation when correcting them in reading exercises. All these factors could explain the retention of an Amdo Tibetic phonological feature by these teenagers. Nevertheless, as the next sections will show, they do not seem to retain significantly more Amdo Tibetan morphosyntactic features than the other children.

Lexicon should also be mentioned as a factor influencing this phonological feature, this time in the parents’ generation. Words starting with <g, d, (b, retrofl.)> and copied from Central Tibetan are often pronounced as aspirated stops, like in Central Tibetic, even by parents. For example, while the word for “snow” is <khaba> /kʰawa/ in Amdo Tibetic and <gangs> /kʰaŋ˨/ in Central Tibetic, Mo3 pronounces /kʰaŋ/. Similarly, Fa3 copies the words <bu> /pʰu˨/ “son, boy”, <bu.mo> / pʰu˨.mo / “daughter, girl”, and <deb> /tʰep˨/ “book” from Central Tibetic (comp. to Amdo <zhi.lu> /ʃələ/, <zhi.mo> /ʃəmo/, and <dpecha> /xweʧʰa/, respectively), and constantly pronounces these words with an aspirated initial stop. Other speakers present a type of “partial” copy (Johanson, 1992) of the phonological form of such lexical items: for instance, Fa1 or Mo4 use the Central Tibetan word <deb> “book”, but with an unaspirated initial consonant /tep/.

Thus, this analysis of the distribution of aspirated vs. non-aspirated stops in 4 Tibetan families shows that most speakers settled in France tends to align with the Central Tibetic system from the children’s generation onward, even though some children retain the Amdo Tibetan pronunciation.

4. The Noun Phrase: Definiteness and Number

Several morphosyntactic features can be examined to assess the transgenerational linguistic changes in Amdo Tibetan families in France. In this section, we will focus on the noun phrase and examine the marking of indefiniteness, number categories, and spatial cases.

4.1. The Marking of Indefinite Noun Phrases

As shown in

Table 5. Indefinite markers in Amdo and Central Tibetic varieties.

	Central Tibetic	Amdo Tibetic
Indefinite marker	-ཅིག -ʧiˀ	-ཟིག -zək(a), -sək(a), -ʦək(a) 1

¹ The three allomorphs are in free variation; /a/ is added when the indefinite marker is followed by a numeral or a quantifier (Simon, 2016, p. 245).

, the indefinite marker has a distinct phonological form in Central and Amdo Tibetic varieties and is thus easily recognizable in the corpus.

The indefinite marker in Central and Amdo Tibetic varieties has the same, cross-linguistically classic etymology (Heine & Kuteva, 2002, pp. 219–222): it is grammaticalized from the number གཅིག་ <gcig> “one” (Bialek, forthcoming, p. 426).6 The pronunciation evolved differently, but its clitic position in the nominal phrase is identical in the two language groups: they occur at the end of the NP and can only be followed by the case marker and the focus marker (<yang> in “Common” Tibetan and <ra> in Amdo Tibetic). Their range of functions is slightly different: in Central Tibetic, ʧiˀ can also occur as an independent unit with the function of an indefinite pronoun, as a filler when the speaker hesitates, or as an attenuating particle meaning “once” or “a bit”. Such uses were not taken into account here.

The five children recorded for this study all exclusively use the morpheme -ʧiˀ to mark indefinite noun phrases, as illustrated for instance by examples in (1).7

(1)

a.

མི་-ཅིག་

འེའེ་

ས་ཆ་

Ch2B

mi-ʧiˀ

əəə

saʧa

person-indf

hes

place

ཧ་གོ་-མེད་-བ་-ཅིག་-ལ་

སླེབས་-བསྡད་-ཡོད་རེད་-བཱ།

hako-me-wa-ʧiˀ-la

lep-de-jore-wa

know-neg-nmlz-indf-dat/loc

arrive-aux:dur-perf-phat

“A man, er, has arrived to an unknown place.”

b.

ཁོ་-གིས་

རྫ་མ་

ཆེ་-ཅིག་-གི་

ནང་-ལ་

ཨ་ནི་

Ch3

kʰo-ki

ʣama

ʧʰe-ʧiˀ-ki

naŋ-la

ani

3sg-erg/instr

pottery

big-indf-gen

inside-dat/loc

then

སྐད་-ཅིག་

གོ་-ནས་

ཨ་ནི།

ke-ʧiˀ

kʰo-ne

ani

voice-indf

hear-conn

then

“He hears a noise in a big pottery, and…”

As we can see from

Table 6. Form of the indefinite morphemes found in the corpus “Arrival”.

	Central Tibetic	Amdo Tibetic	Double Marking
Mo1	11	25	-
Mo2	31	1	-
Fa3	60	30	3
Mo4	9	59	2
Ch1	33	-	-
Ch2A + Ch2B	16 + 27	-	-
Ch3	47	-	-
Ch4	46	-	-

while the children exclusively use the Common Tibetic morpheme to mark indefinite noun phrases, the parents’ variety of “Common” Tibetan can be divided into three groups for this morphosyntactic feature. Hence, Mo1 and Mo4 mostly use the Amdo Tibetic morphemes and the Amdo Tibetic morphemes are 2.3 times more frequent than their Central Tibetic counterpart for Mo1, and 6.15 times more frequent for Mo4. On the contrary, Mo2 uses almost exclusively the Central Tibetic form. Finally, Fa3’s system of indefinite marking mirrors Mo1’s: he uses the Central Tibetic morphemes twice more than the Amdo Tibetic ones.

On top of that, the corpus contains five occurrences of a double-marking of the indefinite, where the Amdo Tibetic morpheme is immediately followed by the Central Tibetic one as illustrated in (2) (in this example, the phonological form -ʧək instead of the expected -ʧiˀ corresponds to the regular reflex pronunciation of the Central Tibetic morpheme in the Amdo Tibetic phonology).

(2)	a.	འདི་-ན་	ད་	ཉལ་-ས་-ཟིག་-ཅིག་		འདུག་-གཱ	Fa3
		də-na	ta	ɲa-sa-zək-ʧək		du-gaː
		dem-loc	well	lay.down-nmlz-indf-indf		exist.sens-phat
		“Here, well, there is a bed (lit. place to lay down), right?”
	b.	རྒྱལ་ཁབ་	གཞན་དག་-ཟིག་ག་-ཅིག་-འ་		འགྲོ་-དགོས་-ཡོད་ས་རེད།		Mo4
		ʥɛkʰe	ʃenta-səka-ʧək-a		nɖo-go-osare
		country	other-indf-indf-dat		go-aux:must-exist.epist
		“[He] probably has to go to another country.”

4.2. Number: Plural and Dual Markers

The grammatical category of number (cf. Corbett, 2000) is grammaticalized in a more complex way in the Amdo Tibetic varieties compared to Central Tibetic, both in terms of subcategories and in terms of allomorphs, as illustrated in

Table 7. Markers of number categories in Central and Central Tibetic varieties.

	Central Tibetic	Amdo Tibetic
Definite plural	-ཚོ་ -ʦo	-ཚོ་ -zo, -so 1, -ཆབོ་ -ʧʰo, -ʧʰao 1,2
Dual	-གཉིས་ -ɲi, -གཉིས་ཀ་ -ɲika 1	-གཉིས་ -ɲi, -གཉིས་ཀ་ -ɲika 1
Collective	[not grammaticalized]	-ཅན་གོ ʧɛnko, -རིགས་འ་-rəka 1
Associative	[not grammaticalized]	-ཚང་ -ʦʰaŋ, -saŋ, -zaŋ 1

¹ Allomorphs in free variation. ² The use of -zo/-so vs. -ʧʰo/-ʧʰao is probably at least partly due to dialectal variation.

. (Skal bzang ‘gyur med & Skal bzang dbyang can, 2002; Simon & Noûs, 2021).

4.2.1. Plural Markers

Just like the indefinite marker, children in the sample almost only use the Central Tibetic definite plural form in the corpus “Arrival”, whether in pronouns or noun phrases, as illustrated in (3).

(3)

a.

ད་

ཁོ་-ཚོ་

ཡར་

ད་

གྲུ་གཟིངས་

འདི་-ལ་

བཏགས།

Ch1

tʰa

kʰo-tso

jaː

tʰa

ʈʰusiŋ

di-la

taˀ

disc

3-pl

up

well

boat

dem-dat/loc

tie

“Well, they, up there, well, [they] tie up the boat, and…”

b.

ཁོ་-གིས་

ཡིག་གཟུགས་

མཐོང་-འགྱོག་

དེ་-ཚོ་

མཐོང་ནས་

Ch1

kʰo-ki

jiksuˀ

tʰoŋ-ɟoˀ

te-tso

tʰoŋ-ne

3-erg/instr

writing.sign

see-nmlz

dem-pl

see-conn

“He has seen the writing signs [he] saw, and…”

Within the parents’ generation too, the Central Tibetic morpheme -ʦo is the one most commonly found in the corpus “Arrival”. A closer look, however, reveals a grouping of the speakers similar to that observed for the indefinite marker: Mo2 and Fa3 have switched completely or almost completely to the Central Tibetic morpheme, while the data for Mo1 and Mo4 are more varied, as shown in

Table 8. Plural morphemes found in the corpus “Arrival”.

	Central Tibetic	Amdo Tibetic
	-ʦo	-zo, -so	ʧʰo, -ʧʰao	-ʧɛnko	-rəka
Mo1	6	-	5	-	1
Mo2	17	-	-	-	-
Fa3	36	-	1	-	-
Mo4	5	1	6	-	-
Ch1	10	-	-	-	-
Ch2A + Ch2B	12 + 18	-	0 + 1	-	-
Ch3	34	-	-	-	-
Ch4	17	-	-	-	-

.

Examples in (4) and (5) illustrate this variability of the plural markers in Mo1 and Mo4, respectively.

(4)

a.

སེམས་ཅན་

དེ་-ཚོ་-ལ་

ལངས་-བསྡད་-འདུག་-གཱ།

Mo1

semʧen

te-ʦo-la

laŋ-de-du

animal

dem-pl-too

stand.up-aux:dur-perf.sens/infer-phat

“The animals too have stood up.”

b.

ཁུ་-ཚོ་-གིས་

ད་

ཁ་ལག

ཁ་ལག་

ཚང་མ་-ཟིག་

བཟས།

Mo1

kʰə-ʦo-ki

ta

kʰala

kʰala

ʦʰaŋma

ze

3-pl-erg/instr

well

food

food

all-indf

eat

“They, well, the food, [they] eat all the food, and…”

c.

ད་

ཁུ་-ཆབོས་

བྱ་ཤ་

བཟས་

ཆང་

འཐུང་།

Mo1

ta

kʰə-ʧʰe

ʃaxʰa

se

ʧʰaŋ

tʰuŋ

disc

3-pl.erg/instr

chicken

eat

beer

drink

“Well, they eat chicken, [they] drink beer, and…”

d.

ཚོད་མ་

འདི་-རིགས་འ་

ད་

འདི་

རེད།

Mo1

ʦʰoma

ndə-rəka

ta

ndə

re

vegetable

dem-coll

well

dem

equ.fact

ད།

འདིའ་

ཁྱེར་-འས་

ཡོང་།

ta

ndeː

ʨʰer-e

joŋ

well

dem.dat

take-conn

come

“All the vegetables, well, it’s that, well, [they] bring [them] here.”

(5)

a.

ཁུར་གེའི་

ལག་འཁྱེར་

དེ་-ཚོ་

ཕར་འ་

སྟོན་-གི་ཡོད་གི

Mo4

kʰərgi

lakʨʰer

te-ʦo

har-a

ton-kogə

3.gen

identity.papers

dem-pl

forth-dat

show-uncmp.sens

“[He] is showing his identity papers.”

b.

ད་

ཁུ་-ཚོ་-གི་

དེ་

ནང་-ནས་

ད་་་

འེའེ།

Mo4

ta

kʰə-ʦo-gə

te

naŋ-ne

ta

əəə

disc

3-pl-gen

dem

inside-abl/loc

well

hes

“Well, their… in [their] home, well, er [they prepare food].”

c.

དེ་ནས་

ཁུར་གེས་

དེ་-ཟོ་

ཚང་མ་

ལག་བ་

འཆང་-འས།

Mo4

tene

kʰərgi

te-zo

ʦʰaŋma

lakwa

ʧʰaŋ-e

then

3.erg/instr

dem-pl

all

hand

touch-conn

“Then, he touches them all.”

d.

ཁུ་-ཆོ་

ཚང་མ་

ཁྱིམ་ཚང་

ནང་-འ་

ཐོན་-བཏང།

Mo4

kʰə-ʧʰo

ʦʰaŋma

ʨʰimʦʰaŋ

naŋ-a

tʰøn-taŋ

3-pl

all

home

inside-dat

arrive-aux:cmp

“They all arrive home, and…”

Finally, Mo4 also has six occurrences of the associative plural category, i.e., a plural marker used to refer to one (main) entity together with the other entities frequently associated with this main entity (cf. Corbett, 2000, p. 101; Creissels, 2006, p. 123), and which has the distinctive feature of being able to be used with proper nouns. This specific plural category is marked with -ʦʰaŋ, -saŋ, or -zaŋ in Amdo Tibetic varieties. In our corpus, all the occurrences of this associative plural mark the third-person pronoun, as illustrated in (6).

(6)	a.	ཁུ་-ཚང་-གི་	ཁྱིམ་ཚང་		དེ་		ངོ་སྤྲོད་		བྱས།	Mo4
		kʰə-ʦʰaŋ-gə	ʨʰimʦʰaŋ		te		ŋorʈø		ʧe
		3-assoc-gen	home		dem		introduction		lightV
		“[He] shows [him] their home, and…”
	b.	ཁུ་-ཚང་-གི་	སེམས་ཅན་	དེ་-གཉིས་ཀ་		ཟ་མ་		འདྲ་བོ་	བྱིན་-འས།	Mo4
		kʰə-saŋ-gə	semʧɛn	te-ɲiga		sama		nɖawo	ʃin-e
		3-assoc-gen	animal	dem-du		food		similar	give-conn
		“[He] gives some kind of food to their animals.”

The lack of associative plural markers in other speakers is difficult to interpret with certainty. Given the relative scarcity of this plural category, a longer corpus could reveal the maintenance of this category of plural in other speakers. Nevertheless, it does not occur in the linguistic biographies either. Thus, a such marker being absent in Central Tibetic and in the languages in contact with Tibetans in exile the lack of associative plural marker in our corpuscould also correspond to the process of simplification and dialect levelling at work in the diasporic context, which is observed more clearly in the children’s generation.

4.2.2. The Grammatical Category of Dual

Another transgenerational difference observed in these families concerns the grammaticalization of dual. In both Central and Amdo Tibetic varieties, it is an obligatory marker, grammaticalized from the numeral <gnyis> “two”, which occupies the same morphosyntactic slot as the plural morphemes in the noun phrase (Simon & Noûs, 2021, pp. 60–61). The excerpt of the graphic novel used to elicit the corpus includes several situations featuring two characters and, as expected in the description of such situations, dual markers are frequent in the parents’ recordings: depending on the narrative length, they are found 13 to 23 times in these documents.

Such a frequency is not found in the data elicited from the children. One still finds a few occurrences of markers that seem to correspond to dual, as exemplified in (7). However, it can be observed that they are almost only limited to the marking of coordinated noun phrases, such as in (7)b. and c.

(7)

a.

ཁོ་-གཉིས་ཆ་

གྲོགས་པོ་

ཆགས་བཞག

Ch2A

kʰo-ɲiʧa

ʈokpo

ʧʰa-ʃaˀ

3-du

friend

become-perf.infer

“They (two) have become friends.”

b.

པཱ་ལགས་

འདི་-དང་

བུ་-གཉིས་-ཀྱིས་

ཨ་ནི་

འེའེ།

[…]

Ch3

paːla

di-taŋ

pʰu-ɲi-ki

ani

əəə

father.H

dem-and

son-du-erg/instr

then

hes

མགྲོན་པོ་

འདྲ་པོ་

སྐད་

བཏང་-ཡིན་ས་རེད།

ɖønpo

ɖapo

ke

taŋ-jinsare

guest

similar

voice

lightV-cmp.past.epist

“This father and the son then, er, [them] probably invited [him].”

c.

ད་

བསོད་ནམས་-དང་

སྐྱེས་དམན་-གཉིས་་་

ད་་་

གྲུ་གཟིངས་

Ch4

tʰa

sønam-taŋ

cemen-ɲi

tʰa

ʈʰusiŋ

disc

PN-and

woman-du

disc

boat

གནམ་-ལ་

ཕུར་-མཁན་-ཅིག་

ནང་-ལ་

བསྡད་-བསྡད་-ཡོད་རེད།

nam-la

pʰur-ken

naŋ-la

de-de-jore

sky-dat/loc

fly-nmlz

inside-dat/loc

stay-aux:dur-perf.fact

“Well, Sönam and the woman, well, [they] are sitting in a flying boat.”

The dual marker does not seem to be obligatory: in all children speakers, in such contexts, we find instances of the definite plural marker (like in (8)a. and b.), an alternation of dual and plural (in example (8)c.), or an alternation of dual and no number markers (like in (8)d.).

(8)

a.

ཁོང་-ཚོ་

འགྲོ་-ས་

ས་ཆ་

འདི་

སླེབས་-གི་རེད་-བཱ།

Ch1

kʰoŋ-ʦo

ɖo-sa

saʧa

di

lep-gire-wa

3.H-pl

go-nmlz

place

dem

arrive-fut.fact-phat

“[They] will arrive [to] the place where they (two) were going to.”

b.

ཁོ་-ཚོའི་

ནང་-ལ་་་

ཤོག་-ཡའི་་་་

ཟེར་-ནས་

Ch2B

kʰo-ʦøː

naŋ-la

ʃo-jeː

ser-ne

3-pl.gen

inside-dat/loc

come.imp-and

say-conn

“[They] also tell [him] to come to their (two) home, and…”

c.

ཁོ་-ཚོ་་་་

སེམས་ཅན་-གཉིས་ཀ་

ཨ་ནི་་་

གྲོགས་པོ་

འདྲ་པོ་

ཆགས་-འདུག

Ch3

kʰo-ʦo

semʧen-ɲika

ani

ʈokpo

ɖapo

ʧʰa-duˀ

3-pl

animal-du

then

friend

similar

become-perf.sens

“They… the two animals have become friends.”

d.

ཨ་ནི

པཱ་ལགས་དང་

བུ་

འདི་-གཉིས་

བསོད་ནམས་

རོགས་པ་

བྱས་ནས་་་།

[…]

Ch4

ani

paːla-taŋ

pʰu

di-ɲi

sønam

rokpa

ʧʰe-ne

then

father.H-and

son

dem-du

PN

help

lightV-conn

ཨ་ནི་

པཱ་ལགས་དང་

བུ་

འདི་

བསོད་ནམས་-ལ་

ཁ་ལག་

ཨ་ནི་

མ་འདྲ་བ་

ani

paːla-taŋ

pʰu

di

sønam-la

kʰala

ani

manɖawa

then

father.H-and

son

dem

PN-dat/loc

food

then

different

མང་པོ་

དཔེ་

བསྟོན་-རེད།

maŋpo

pe

tøn-re

many

very

show-cmp.past.fact

“Then, the father and the son help Sönam, and […] then, the father and the son show Sönam many different [kinds of] food.”

Finally, in three of the five children, we find a double marking combining the definite plural and the numeral “two” in distinct morphosyntactic slots, as illustrated in (9).

(9)

a.

ཁོང་-ཚོས་-གཉིས་ཀ་-ཡའི་

ཁག་ཁག་

འགྲོ་-གི་རེད་-དཱ།

Ch1

kʰoŋ-ʦø-ɲika-jeː

kʰaka

ɖo-gire-ta

3-pl.erg/instr-two-too

separate

go-fut-assert

“The two of them will go their way!”

b.

ཁོང་-ཚོ་-གཉིས་-གིས་

ཕར་-ཚུར་

སྐད་ཆ་

འདྲ་པོ་

བཤད་-ནས་་་་།

Ch3

kʰoŋ-ʦø-ɲi-ki

pʰaː-ʦʰuː

keʧa

ɖapo

ʃe-ne

3-pl-du-erg/instr

forth-back

speech

similar

tell-conn

“The two of them sort of discuss, and…”

c.

ཨ་ནི་

ཁོང་-ཚོ་-གཉིས།

སེམས་ཅན་-དང་

བསོད་ནམས་-གཉིས།

ཁོང་-ཚོ་-གཉིས་

Ch4

ani

kʰoŋ-ʦo-ɲi

semʧen-taŋ

sønam-ɲi

kʰoŋ-ʦo-ɲi

then

3-pl-du

animal-and

PN-du

3-pl-du

འེའེ།

ད།

བསྡད་-ས་-ཅིག་

ནང་-ལ་

བསྡད།

əəə

ta

de-sa-ʧik

naŋ-la

de

hes

well

stay-nmlz-indf

inside-dat/loc

stay

“Then, the two of them, the animal and Sönam, the two of them, er, stayed in a place.”

As we can see by comparing (9)a. and (9)b., the morphosyntactic structure of this construction is not fixed, with a variation in the position of the ergative case marker.8

As a conclusion to this section, we can observe that the grammatical domain of the plural clearly undergoes a simplification in the process of dialect levelling or the switch to “Common” Tibetan: the amount of distinct plural markers is reduced to a single one, and the corresponding subcategories of numbers merge into a simplified system. Such a simplification in the domain of plural markers in “Common” Tibetan, compared to the Lhasa variety, was already observed by Vokurková (2022, pp. 158–159). This process is documented here for speakers of Amdo Tibetic: we can see it starting from the first generation of exiles on, with the replacement of the varied allomorphs in Amdo Tibetic by a single morpheme. More than that, we can see that some speakers have already abandoned some peripheral number categories such as associative and collective plural categories. In the children’s generation, the simplification goes one step further, with the loss of a fully grammaticalized category of dual.

5. The Noun Phrase: Case Markers

The function and morphology of case markers is also a domain where we may contrast Amdo and Central Tibetic and identify individual as well as transgenerational variations. We will first examine the system of spatial cases and dative, which illustrates the transition from the Amdo to the “Common” Tibetan system in the parents’ generation. In a second paragraph, we will look at the ergative, instrumental, and comitative functions to highlight some ongoing linguistic changes in the children’s generation.

5.1. Spatial Cases

Amdo and “Common” Tibetan have a system of case markers in the form of enclitics. The system follows an ergative alignment (split-ergative in “Common” Tibetan) and shares the following case markers: absolutive, ergative/instrumental, absolutive, genitive, comitative, ablative/translocative, and dative/directive. On top of that, “Common” Tibetan has preserved a specific comparative case marker (replaced by a periphrastic construction in Amdo Tibetic languages). Finally, as we will see in this section, Amdo Tibetan has preserved a specific locative case marker, while this function is assumed by the dative/directive case in “Common” Tibetan. Like the category of number, the system of spatial cases (Creissels, 2009) is more complex in Amdo Tibetic than it is in Central Tibetic.

Table 9. Spatial case markers in Amdo and Central Tibetan.

	Central Tibetic	Amdo Tibetic
Ablative/Translocative	-ནས་-ne	-ནས་ -ni
Locative	-ལ་ -la, (rare: -ར་ -ː) 2	-ན་ -na, -ནས་ -ni 1
Dative/Directive	-ལ་ -la, (rare: -ར་ -ː) 2	-འ་ -a, , -ː 3

¹ The two allomorphs are conditioned by the stative/dynamic aspect of the verb (Sung & Rgyal, 2018, pp. 215–216). ² Phonologically conditioned allomorphs: -la after consonants; vowel lengthening in open syllables. ³ Phonologically conditioned allomorphs: -a after consonants; vowel lengthening, sometimes with a change in vowel quality, in open syllables.

shows that Amdo Tibetic varieties have morphemes to specifically mark locative, as opposed to directive/allative and dative (one of the two allomorphs is homophonous with the ablative marker). Moreover, “Common” Tibetan has a simplified system compared to other Central Tibetic varieties, such as Lhasa Tibetan: the vowel lengthening used in open syllables in plurisyllabic words in Lhasa Tibetan is rarely attested in “Common” Tibetan (Vokurková, 2022, pp. 155–158).

When we examine the proportion of dative–directive/allative markers of each form in the corpus, as shown in

Table 10. Number of occurrences of each type of dative–directive(locative) case marker in “Arrival”.

	Central Tibetic	Amdo Tibetic
Mo1	15	79
Mo2	29	43
Fa3	80	110
Mo4	13	173
Ch1	49	7
Ch2A + Ch2B	31 + 28	0 + 3
Ch3	67	0
Ch4	40	0

, we can observe the shift from the Amdo Tibetic morpheme, predominant in the parents’ variety, to Central Tibetic in the children’s variety: the most frequent morpheme for each speaker is highlighted in grey. Again, the parents can be divided into two subgroups. Mo1 and Mo4 have significantly more Amdo Tibetic forms in proportion: the Amdo Tibetic morpheme is 5.26 times more frequent than the Central Tibetic one in Mo1, and more than 13 times more frequent in Mo4, while it is only 1.48 and 1.38 times more frequent in Mo2 and Fa3, respectively.

A closer look at the morphemes used to mark the locative spatial relation can help to obtain a more precise picture of the shift from the Amdo to Central Tibetic spatial case system.

In the parents’ corpus, the locative marked by -na is found in Mo1, Fa3, and Mo4. Nevertheless, the linguistic change toward the functional simplification of spatial cases seems to be under way for these three speakers: to express locative, they not only use alternatively the Amdo Tibetic locative markers -na and -ni, as well as the Central Tibetic dative–directive–locative marker -la, but also the Amdo Tibetic dative–directive (and originally non-locative) marker -a, as illustrated in (10) to (12).

(10)

a.

ཨ་ནི་

ནང་-ན་

ཆི་ཟིག་

ཡོད་ནི་ནཱ་

བསམས་-འས་

ཁུ་-གིས་

Mo1

ani

naŋ-na

ʧʰəsək

jonənaː

sam-e

kʰə-ki

then

inside-loc

what

exist.wond

think-conn

3-erg/instr

རྣ་-གིས་

ཉན་-ཟུག

rna-ki

ɲen-zək

ear-erg/instr

listen-perf.infer

“Then, thinking ‘what is inside?’, he has listened with his ear.”

b.

འདི་-ལ་-ཡང་

ཟ་-རྒྱུ་-ཟིག་

འདུག

Mo1

ndə-la-jang

sa-ʥə-sək

du

dem-dat/loc-too

eat-nmlz-indf

exist.sens

“Here too, there is something to eat.”

c.

ད་

ཁོ་-གཉིས་

མེ་འཁོར་

ནང་-འ་

བསྡད་-བཏང་-ཟུག

Mo1

ta

kʰo-ɲi

meŋkʰor

naŋ-a

de-taŋ-sək

disc

3-du

train

inside-dat/loc

stay-aux:cmp-perf.infer

“Well, they sat on the train.”

(11)

a.

འདི་-ན་

ད་

ཉལ་ས་ཟིག་ཅིག་

འདུག་གཱ

Fa3

də-na

ta

ɲa-sa-zək-ʧək

du-gaː

dem-loc

disc

lay.down-nmlz-indf-indf

exist.sens-phat

“Here, well, there is a bed, right?”

b.

དེ་རིང་

དགོང་དག་

ག་བ་-ཟིག་-ནས་

བསྡད་-དགོས་-ན་

བསམས།

Fa3

teraŋ

goŋta

kawa-sək-ne

de-go-na

sam

today

evening

where-indf-abl/loc

stay-must-conn

think

“Where should I stay tonight, [he] thought, and…”

c.

ནང་-ལ་

རྐུབ་བཀྱག་-ཅིག་

འདུག

Fa3

naŋ-la

kəpʨa-ʧik

duˀ

inside-dat/loc

chair-indf

exist.sens

“Inside, there is a chair.”

d.

ཁང་པ་

ནང་-འ་

བསྡད་-ཀྱི་ཡིན་

Fa3

kʰaŋpa

naŋ-a

de-kijin

house

inside-dat/loc

stay-fut.ego

“I will stay in the room.”

(12)

a.

འབུ་

དི་

ཉལ་ཁྲི་

སྒང་-ན་

བསྡད་-ཡོད་གི

Mo4

nbu

tə

ɲaʈʰə

gaŋ-na

da-okə

bug

dem

bed

top-loc

stay-perf.sens

“The bug is on the bed.”

b.

ད་

འགྲོན་ཁང་

ནང་-འ་

བསྡད་-རྒྱུས་

ཟེར་-གི་ཡོད་-ནཱ།

Mo4

ta

nɖonkʰaŋ

naŋ-a

de-ʥi

se-ko-na

disc

hotel

inside-dat/loc

stay-fut.ego

say-uncmp-wond

“Well, does [he] say ‘I will stay in the hotel’, or…”

c.

ད་

འབྲི་དེབ་

སྒང་-འ་

ཡིན་-ནཱ་

Mo4

ta

nɖətep

gaŋ-a

jin-na

disc

notebook

top-dat/loc

equ-wond

དེའི་

སྒང་-ལ་

བྲིས་-བཏང་-འས།

teː

gaŋ-la

ʈi-taŋ-e

dem.gen

top-dat/loc

write-aux:cmp-conn

“Well, [he] has written [something] on that, is it a notebook?”

Thus, these speakers illustrate the first step of linguistic change in the domain of spatial cases. The functional simplification, i.e., the loss of a distinct grammatical category of locative, goes together with a morphophonological complexification due to the multiplication of allomorphs.

Mo2’s variety of “Common” Tibetan is one step further in the reconfiguration of spatial case markers: the Amdo Tibetic markers -na or -ni are not attested. As illustrated in examples (13), the locative noun-phrases in “Arrival” are all marked either with the Central Tibetic dative–directive–locative -la or with the Amdo Tibetic -a (which has gained a new function, compared to the original morpheme restricted to dative–directive/allative functions).

(13)

a.

རྫ་མ་

ནང་-ལ་-ཡང་

ཉ་

འདྲ་པོ་

དེ་འདྲས་-གི་

Mo2

ʣama

naŋ-la-jaŋ

ɲa

ɖapo

tenɖe-gi

pottery

inside-dat/loc-too

fish

similar

like.that-gen

སེམས་ཅན་-ཅིག་

བསྡད་ནས།

semʧen-ʧik

de-ne

animal-indf

stay-conn

“An animal, similar to a fish, like that, was in the pottery, and…”

b.

དེ་ནས་

ད་

ཁོ་-ཚོ་

གྲུའི་

ནང་-འ་

བསྡད་-ནས་

Mo2

tene

ta

kʰo-ʦo

ʈəː

naŋ-a

de-ne

then

disc

3-pl

boat.gen

inside-dat/loc

stay-conn

“Then, well, they stay in the boat, and…”

The children’s corpus represents the final stage of the linguistic reconfiguration of spatial relations. In fact, no example of the Amdo Tibetic locative marker -na is found in the children’s corpus. Some occurrences of -ne, with the proximal demonstrative pronoun འདི་-ནས་ /di-ne/ “here” found in Ch3 or the noun “after” བརྗེས་-ནས་ /ʤe-ne/ “later, after that” found in Ch1, can be interpreted as lexicalized units. Besides these specific cases, only two occurrences of -ne, reproduced in (14), are ambiguous between a locative, and ablative or translocative interpretation. But, except for these two occurrences, the locative is always marked with the Central Tibetic morpheme -la in the children’s corpus.

(14)	a.	ཆུ་ཁྲི་-ཅིག་	ནང་-ནས་		དེབ་-ཅིག་	མཐོང་།	Ch3
		ʧʰuʈi-ʧik	naŋ-ne		tʰep-ʧik	tʰoŋ
		desk-indf	inside-abl/loc		book-indf	see
		“[She] sees a book in / from inside the desk, and…”
	b.	train-གི་	འདི་	ནང་-ནས་	བྲོས་-སོང་-ངཱ།		Ch2A
		ʈʰweɪn-gi	di	naŋ-ne	ʈi-soŋ-aː
		train-gen	dem	inside-abl/loc	flee-cmp.past.sens-phat
		“[She] flew in / by this, the train, right?”

Thus, the shift from an Amdo Tibetic to a “Common” Tibetic spatial case system follows the steps summarized in

Table 11. Schematic transition from Amdo to “Common” Tibetic system of spatial cases.

Amdo			Stage 1			Stage 2			“Common”
dat-dir: -a loc: -na, -ni		dat-dir: -a (and -la) loc: -na, -ni, -a (and -la)			dat-dir-loc: -la (and, marginally -a)			dat-dir-loc: -la

, and the transgenerational shift occurs between Stage 1 and Stage 2.

5.2. Instrumental and Comitative

In the Tibetic languages, the associative case marker (Tournadre & Dorje, 2009, pp. 131–132; Tournadre & Suzuki, 2023) is a subtype of comitative, but with a more restricted function: it marks the role of a co-participant (co-agent of verbs such as “to talk/meet with sb.” or “to compete/fight with/against sb.” or co-patient of verbs such as “to exchange sth. with/against sth.”); the event must express some reciprocal meaning and the co-participant must have a similar degree of animacy and control over the event. The associative morpheme can also be used as a nominal coordinating conjunction.

Besides this case marker, we also find a comitative postposition with a broader function of accompaniment (Stolz et al., 2009), used to mark animated entities only. Contrary to most European languages including French (Stolz et al., 2013), the marking of instrumental and associative/comitative are strictly distinguished, as shown in

Table 12. Ergative, instrumental, and comitative markers in Amdo and Central Tibetic.

		“Common” Tibetic	Amdo Tibetic
Instrumental		-gi	-gə
Comitative	Associative Case	-taŋ (ɲamtu)	-la, -ra, -ʈa 1
	Postposition	ɲamtu, ɲampo	ɲamgə, ɲampe, ʧʰapka

¹ In several dialects, -la and -ra are free allomorphs, and -ʈa and -ra are phonologically conditioned allomorphs; -ʈa is not attested in the corpus recorded in France for this study.

.

The corpus contains two examples of the associative case marker, reproduced in (15). Thus, the associative case only occurs in two parents’ speech, and only in its Amdo Tibetic form.

(15)

a.

ད་

ཁུ་-གཉིས་-ར་

གྲོགས་པོ་

ཆགས་-འས་

ཁུ་-གཉིས་-གིས་

Mo1

ta

kʰə-ɲi-ra

ʈokpo

ʧʰak-e

kʰə-ɲi-ki

disc

2-du-com:ass

friend

become-conn

3-du-erg/instr

ཟ་-རྒྱུ་-ཟིག་འ་

མང་བ་-ཟིག་

ཉོས་-ལས།

sa-ʥə-səka

maŋa-sək

ɲe-le

eat-nmlz-indf

many-indf

buy-conn

“Well, after [he] became friend with them (two), they (two) buy a many things to eat, and…”

b.

སླེབས་ནས་

ད་

ད་ལྟ་-གི་

མི་

དི་-ར་

ཐུག་-བཞག

Fa3

lʰep-ne

ta

tanda-gə

mə

tə-ra

tʰək-ʃa

come-conn

well

now-gen

person

dem-com:ass

meet-perf.infer

“Once [he] arrived, [he] met this present person”

Given the semantic restrictions on the use of associative, it is a relatively rare morpheme and its absence in the corpus recorded cannot be interpreted.

More common is the use of the postposition. As we can see in examples in (16), both the parents and children use the postposition to mark co-participants, and the postposition mostly has the expected form in “Common” Tibetan (Vokurková, 2022, p. 172). In example (16)b., the form ɲampu, instead of ɲamtu, likely corresponds to a mix of the written Tibetan མཉམ་པོ་ ɲampo and the “Common” Tibetan form མཉམ་དུ་ ɲamtu.

(16)

a.

ཉ་

དི་

མཉམ་དུ་

ཕར་་་་

གྲུ་གཟིངས་-ཟིག་

ནང་-འ་

བསྡད།

Mo2

ɲa

tə

ɲamtu

pʰar

ʈəsaŋ-sək

naŋ-a

de

fish

dem

together

forth

boat-indf

inside-dat

stay

“He stays in the boat with the this fish, and…”

b.

འགྲོ་-དུས་སྒང་-འ་

ད་

ཁོ་རང་

མཉམ་པུ་-ཡང་

ཙི་ཙི་

Fa3

ɖo-tygaŋ-a

ta

kʰoraŋ

ɲampu-jaŋ

ʦiʦi

go-conn

disc

3

together-too

mouse

དི་

མཉམ་དུ་

ཡོང་-བསྡད་-ཀྱི་འདུག

tə

ɲamtu

joŋ-de-kiduˀ

dem

together

come-aux:dur-uncompl.sens

“As [he] walks, with him, too… [He] is coming with the mouse.”

c.

འབུ་

དི་

མཉམ་དུ་

ད་

དེའི་

ཕྱི་ལོགས་

དེ་འ་

ཡོང་།

Mo4

nbu

tə

ɲamtə

ta

teː

ʃilok

te-a

joŋ

bug

dem

together

disc

dem.gen

outside

dem-dat

come

“With this bug, well, [he] comes outside of that.”

d.

ཁོ་

པུ་གུ་

འདི་་་་

སེམས་ཅན་

མཉམ་དུ་

རྩེད་མོ་

བརྩེས་-བསྡད་-འདུག་- ཟེར།

Ch1

kʰo

pugu

di

semʧen

ɲamtu

ʦemo

ʦe-de-du-s

3

child

dem

animal

together

game

play-aux:dur-perf.sens-rep

“He, this child… [he] is playing with the animal.”

e.

ཁོ་

མཉམ་དུ་

སྐད་ཆ་

འདྲ་པོ་

བཤད།

Ch3

kʰo

ɲamtu

keʧa

ɖapo

ʃe

3

together

speech

similar

say

“[He] talks with him.”

Example (17) shows a change in the semantics of this preposition in one of the children’s varieties, here used to mark a noun referring to an inanimate entity, the book:

(17)

མོ་

དེབ་

འདི་

ཁྱེར་-ནས་

ཨ་ནི་

མཉམ་དུ་

མོ་་་་

འེའེ།

Ch3

mo

tʰep

di

cʰer-ne

ani

ɲamtu

mo

əəə

3f

book

dem

take-conn

then

together

3f

hes

མེ་འཁོར།

མེ་འཁོར་

ནང་-ལ་

འཛུལ་-ནས་

འདྲ་་་

བྲོས་

འདྲ་པོ་

ཡིན་-ས་རེད།

menkoː

menkoː

naŋ-la

ʣyː-ne

ɖa

ʈʰø

ɖapo

jinsare

train

train

inside-dat/loc

enter-conn

sim

flee

similar

compl.past-epist

“She took the book, and well, er, [she] got on the train, and probably ran off with [it].”

The ongoing linguistic change is also observed when we examine the marking of the instrumental function. First, in all parents’ and in some children’s corpora, we find the expected form: i.e., instruments are marked with the ergative–instrumental case, like in the examples in (18).

(18)

a.

ད་

ཁུ་གིས་

སྒོ་

གཅོག་གི་ཡོད་གི།

Mo1

ta

kʰə-ki

go

ʧok-kogə

disc

3-erg/instr

door

break-uncmp.sens

ལྕགས་-གིས་

ཁ་

ཕྱེས།

ʧak-ki

kʰa

ʃʰe

iron-erg/instr

mouth

open

“Here, she breaks the door. [She] opens with [a piece of] metal, and…”

b.

ཨ་ནི་

ཁོ་རང་

ཤོག་བུ་-ཅིག་-གིས་་་་།

ཨ་ནི།

ཝ་མོ་

འདི་

བཟོས་-ནས་

Mo2

ani

kʰoraŋ

ʃoku-ʧik-ki

ani

ʁwamo

di

zø-ne

then

3

paper-indf-erg/instr

then

fox

dem

make-conn

“Then, he, with a sheet, then, [he] makes this fox, and…”

c.

ཚང་མ་

སྐད་ཆ་-གིས་

འགོ་

འདྲ་པོ་

ཚོས།

Fa3

ʦʰaŋma

keʧa-ki

go

ɖapo

ʦʰe

all

speech-erg/instr

head

similar

ripen

“[They] understand everything with words, and…”

d.

ཐག་བ་-གིས་

བསྡམས་ནས་

Mo4

tʰokwa-ki

dɛm-ne

rope-erg/instr

tie.up-conn

“After [they] have tied up [the boat] with the rope.”

e.

སྒོ་

འདི་་་་་

ལྕགས་

འདི་-གིས་

ཕྱེས་-ནས།

Ch2A

go

di

ʧak

di-ki

ʧʰe-ne

door

dem

iron

dem-erg/instr

open-conn

“[She] opens this door with this [piece of] metal, and…”

f.

ཚལ་

ཚང་མ་-གིས་

ཨ་ནི་

འེའེ།

ཁ་ལག་

འདྲ་པོ་

བཟོས།

Ch3

ʦʰeː

ʦʰaŋma-ki

ani

əəə

kʰala

ɖapo

sø

vegetable

all-erg/instr

then

hes

food

similar

make

“[She] cooks a kind of food with all the vegetables.”

But two of the five children in our corpus regularly use the comitative postposition as an instrumental marker, as illustrated in (19).

(19)

a.

sabji

མཉམ་དུ་

འེའེ།

[…]

ཁ་ལག་

བཟོས་-རེད་-ཟེར།

Ch1

sabʤi

ɲamtu

əəə

kʰala

sø-re-s

vegetable

together

hes

food

make-acpXX.fact-rep

“With the vegetables, er, [they] made food.”

b.

ཉི་མ་-ཅིག་

ཁོ་

ལས་ཀ་

བྱས་དང་

བྱེད་ཡག་

མཉམ་དུ་

སྒོ་

བཅག

Ch1

ɲima-ʧik

kʰo

leka

ʧʰe-taŋ

ʧʰe-ja

ɲamtu

go

ʧaˀ

day-indf

3

work

lightV-and

lightV-nmlz

together

door

break

“One day, she broke the door with [something] to work again and again, and…”

c.

བསོད་ནམས་-གིས་

[…]

ཤོག་བུ་

མཉམ་དུ་

Ch4

sønam-ki

ʃoku

ɲamtu

PN-erg/instr

paper

together

ཞི་མི་

ཆུང་ཆུང་ཅིག་

བཟོས་རེད།

ʃimi

ʧʰuŋʧuŋ-ʧik

sø-re

cat

small-indf

make-cmp.past.fact

“Sönam […] made a small cat with paper.”

This use can be compared to the one illustrated in (17): in both cases, the postposition marks an inanimate entity. But it goes one step further, with a semantic change from accompaniment to instrument. It most probably corresponds to a copy (Johanson, 1992) of the semantic properties of the corresponding French preposition avec.

Thus, just like spatial cases, the distribution of instrumental and associative/comitative markers shows a grammatical reconfiguration both in the parents’ and in the children’s generation. The parents tend to adopt the “Common” Tibetan morphemes, while, at the same time, retaining their original morphemes to a variable extent. In the children’s generation, the Amdo Tibetic morphemes are largely absent, and further semantic or functional changes can be observed, which can be interpreted as contact-induced linguistic changes.

6. The Verb Phrase: TAME/E Categories

6.1. Verb Stem Distinctions

In the Tibetic languages, verbs have distinct stems, labelled “past”, “present”, “future”, and “imperative” stems in the traditional grammar, and used with different tense–aspectual and modal values. Typically, the “present” stem is used in theuncompleted9 aspect and before the connecting particle <dus> “when, while”, and the “past” stem is used in the perfect and completed past, and before the connecting particles indicating conditional and succession (cf. Zeisler, 2004, for a detailed description). Like Vokurková (2022, p. 153), we observe a reduced use of distinctive verb stems in the corpus. This simplification is more advanced in children’s speech, but it is also present in the parents’ data. Examples in (20) show that both Fa3 and Mo4 only use the stem <lta> /ta/ for the verb to “look” instead of alternating between /ta/ in uncompleted and /te/ (Central Tibetic pronunciation) or /^Fti/ (Amdo Tibetic pronunciation) in the completed past or perfect.

(20)	a.	ཁོ་	ས་བཀྲ་		དེ་-ལ་			ལྟ་-གི་འདུག		Fa3
		kʰo	sapʈʰa		te-la			ta-kiduˀ
		3	map		dem-dat/loc			look-uncmp.sens
		“He looks at the map.”
	b.	ད་	བཀྲ་ཤིས་-ཀྱིས་		ལྟ་-གི་ཡོད་གི					Mo4
		ta	ʈaʃi-gi		ta-kogə
		well	PN-erg		look-cmp.sens.dyn
		“Now, Tashi looks [at it].”
	c.	ཡག་པོ་	ཅིག་	ཉན་-ནས་			ལྟ་-བཞག			Fa3
		jakpo	ʧək	ɲen-ne			ta-ʃa
		good	indf	listen-conn			look-perf.infer
		“[He] has listened carefully and [he] has looked.”
	d.	ད་	དེ་གིས་			ལག་བ་		འཆང་-འས་	ལྟ་-ཟུག	Mo4
		ta	tegi			lakwa		ʧʰaŋ-e	ta-zək
		disc	like.that			hand		touch-conn	look-perf. infer
		“Well, [he] has touched, like that, and [he] has looked.”

For the verb “to look”, we can see that the loss of distinct verb stems is attested both in Fa3, a speaker whose variety of “Common” Tibetan shows very few retentions of morphosyntactic features of Amdo Tibetic, and also in Mo4, who, on the contrary, has adopted only a few Central Tibetic morphosyntactic features. Thus, for instance, in (20)b. and d., we can see that her TAME/E markers correspond to Amdo Tibetic ones (see below). The simplification of stem numbers observed for the verb “to look” is not observed for all the verbs in the corpus, though. A systematic examination of every verb stem in the corpus would probably reveal some differences in terms of the frequency of verb stem distinctions between these two types of speakers. Nevertheless, it can be noted that, even for speakers who show few adoptions of Central Tibetic features, the loss of verb stem distinctions is attested.

6.2. TAME/E Systems in Central and Amdo Tibetic

As we can see from

Table 13. Most frequent forms of the copulas in Central and Amdo Tibetic (affirmative).

		Central Tibetic	Amdo Tibetan
Equative copula	Ego-participative	jin	jən
	Ego-authoritative	-	jən.nə.re
	Factual	re	re
	Sensory inferential	re.ʃaˀ, re.duˀ	jən.taŋ.sək
Existential copula	Egophoric	jøˀ	jo
	Factual	jo.reˀ	jo.nə.re
	Sensory	duˀ	jokə, jokʰə
	Sensory inferential	-	jo.taŋ.sək

and

Table 14. Some TAME markers in Amdo and Central Tibetan (affirmative).

		Central Tibetan	Amdo Tibetan
Uncompleted	Egophoric [+dynamic]	V-gi.jøˀ	V-ko
	Egophoric [−dynamic]		V-nə.jən, -ni
	Factual [+dynamic]	V-gi.jo.reˀ	V-ko.nə.re
	Factual [−dynamic]		V-ni.re
	Sensory [+dynamic]	V-gi.duˀ, (V-giˀ) 1	V-ko.gə
	Sensory [−dynamic]		V-kə, V-kʰə
Completed past	Egophoric	V-jin, (V-pa.jin) 1	V-aux:cmp-ni, V-aux:cmp-nə.jən, V-aux:cmp-a 2
	Factual	V-re, (V-pa.re) 1	V-aux:cmp-nə.re
	Sensory	V-soŋ	V-aux:cmp-tʰa
Perfect	Egophoric	V-jø	V-aux:cmp-jo
	Factual	V-jo.re	V-aux:cmp-jo.nə.re
	Sensory	V-duˀ	V-aux:cmp-jo.kə, V-aux:cmp-jo.kʰə
	Sensory Inferential	V-ʃaˀ	V-aux:cmp-sək, V-aux:cmp-zək,

¹ Morphemes in brackets are rarely used in exile “Common” Tibetan, as opposed to Lhasa Tibetan (see Vokurková, 2022, pp. 163–165). ² For hypotheses regarding the functional difference between V-aux-ni, V-aux-nə.jən, and V-aux-a on the other hand, see Simon (2021, p. 312).

, most copulas and Tense–Aspect–Modality–Evidentiality/Epistemicity (TAME/E) markers differ in Central and Amdo Tibetic varieties, either due to a different etymology, a distinct pronunciation, or because the forms result from another morphosyntactic construction. In Amdo Tibetic varieties, some evidential and epistemic categories, non-existent in Central Tibetic, are grammaticalized with specific morphemes. For the sake of clarity and because it is not the purpose of this paper to present a contrastive analysis of Central and Amdo Tibetic copula and the TAME/E system, we only mention the affirmative forms and restrict the analysis to the most commonly found copulas and TAME/E markers in our corpus “Arrival”.

In the corpus “Arrival”, as expected in 3rd-person narratives based on pictures, the egophoric forms are rare, while sensory is the most common evidential category used by the speakers. It is also the evidential category for which the Central and Amdo Tibetic morphemes differ most in terms of morphophonology, thus making it easier to assign the morphemes in the corpus to one or the other of the two systems.

When we look at the relative proportion of Amdo vs. Central Tibetic TAME/E markers in the parents’ corpus, they can be placed on a continuum, summarized in Figure 2.

Thus, Mo4 almost only uses Amdo Tibetic markers (there are only three occurrences of the Central Tibetic form in her 26 min. speech). Mo1 presents a mixed system of TAME/E markers, with a similar amount of Amdo and Central Tibetic TAME/E markers. Example (21) illustrates this mixed system and shows how Mo1 alternates between the Amdo and Central Tibetic markers: in the first part of the utterance, she uses the Central Tibetic uncompleted sensory marker -gidu, and in the next sentence she uses the Amdo Tibetic inferential perfect construction -taŋ-zək. Thus, it results in the presence of doublets of morphemes for the different TAME/E categories, and it is not possible to establish whether these doublets share similar functions or whether the morphophonological differences are remotivated functionally. Fa3 mirrors Mo4’s system, with only two occurrences of Amdo Tibetic aspectual auxiliary verbs and one Amdo Tibetic TAME/E morpheme in an aside with his wife. Finally, Mo2 shows no trace of her native Amdo Tibetic variety in this domain of grammar.

(21)	ད་	འགྲོ་-གི་འདུག		དི་གི་	གཞུག་ནས།	ད་	ཨ་མ་	Mo1
	ta	ɖo-gidu		tə-gə	ʒək-ne	ta	ama
	disc	go-uncmp.sens[Central]		dem-gen	back-abl/loc	disc	mother
	ཙིག་འ་	ཐོན་-འས་	བུད་-བཏང་-ཟུག
	tsək-a	tʰon-e	wə-taŋ-zək
	indf-dat	arrive-conn	go-aux:cmp-perf.infer[Amdo]
	“Well, [they] go. After that, well, the woman has arrived somewhere and left.”

As expected, in the children’s generation, the systems of TAME/E markers closely correspond to the Central Tibetic forms. Some innovations can also be observed, and it is difficult to say whether they are transitory, idiolectal, or whether they will become widespread and characteristic of a Western or French variety of “Common” Tibetan.

First, in Ch1, we find several instances of a verb followed by the sequence /ja ciˀ duˀ/, where /ja/ is an irrealis nominalizer, /ciˀ/ is the indefinite article or an attenuative morpheme, and /duˀ/ is the sensory form of the existential copula. The construction is illustrated in (22) and seems to correspond to an uncompleted aspect.

(22)

a.

ད་

ཁོ་་་་

འེའེ།

ག་རེ་

དགོས་

ལབ་-ཡག་ཅིག་

འདུག

Ch1

tʰa

kʰo

əəə

kʰare

go

lap-ja-ʧik

duˀ

disc

3

hes

what

need

say-nmlz-indf

exist.sens

“Well, he, er, tells [the hotel owner] what [he] needs.”

b.

ཨ་ནི་

རྗེས་-ལ་

ཁོ་-ཚོ་

ནང་

འགྲོ་ཡག་ཅིག་

འདུག

Ch1

ani

ʤe-la

kʰo-ʦo

naŋ

ɖo-ja-ʧik

duˀ

then

after-dat/loc

3-pl

inside

go-nmlz-indf

exist.sens

“Then, later, they go home.”

This construction is absent from the rest of the corpus, and thus is probably idiolectal. Conversely, an innovative use pattern of the existential copula is found in all children’s data: ཡོད་བཱ། <jod-bA> /jø-wa/, the phatic form of the egophoric existential copula, is particularly frequent in the data. It occurs in contexts where the egophoric is not expected and apparently neutralized. As illustrated in (23), it is used with a discursive value and seems to be used to draw the hearer’s attention or understanding aligned with one’s own.

(23)

a.

མི་

འདི་

ཞེད་རེད།

དང་པོ་

འདི་ལ་

ལྟ་དུས།

Ch2A

mi

di

ʃe-re

taŋpo

di-la

ta-ty

person

dem

fear-cmp.past.fact

first

dem-dat/loc

look-conn

མ་འདྲ་བ་

འདྲ་པོ་

ག་རེ་

རེད།

བསམ་-རེད།

ཡོད་བཱ།

manɖawa

ɖapo

kare

re

sam-re

jø-wa

different

similar

what

equ.fact

think-cmp.past.fact

exist-phat

“This person was afraid, first, when [he] looked at it. ‘What is this kind of strange [thing]’, [he] thought, right ?”

b.

ཨ་ནི།

མོ་

ཡོད་བཱ།

འེའེ།

ཆུ་ཁྲི་-ཅིག་

ནང་-ནས་

དེབ་-ཅིག་

མཐོང་།

Ch3

ani

mo

jø-wa

əəə

ʧʰuʈi

naŋ-ne

tʰep-ʧik

tʰoŋ

then

3f

exist-phat

hes

desk-indf

inside-abl

book-indf

see

“Then, she (right?) er, sees a book in a desk, and…”

c.

ཨ་ནི་

རྗེས་མ་-ནས་

ཁོ་

ཁ་

སྐོམ་-ནས་

ཡོད་བཱ།

Ch1

ani

ʤema-ne

kʰo

kʰa

kom-ne

jø-wa

then

next-abl

3

mouth

dry-conn

exist-phat

ད་

ཅིག་

ཆུ་

འཐུང་-དགོས་-རེད།

tʰa

ʧik

ʧʰu

tʰuŋ-go-re

disc

indf

water

drink-aux:must-cmp.past.fact

“Then, next, he is thirsty, and, (right?), now, [he] needs something, water.”

As we will see in the next section, innovations in the morphosyntactic constructions or in their uses are also found in the expression of epistemicity.

6.3. Epistemicity

Epistemicity is here considered in a restricted sense, following Nuyts (2000, p. 21): “Epistemic modality is defined here as the linguistic expression of an evaluation of the chances that a certain hypothetical state of affairs under consideration (or some aspect of it) will occur, is occurring, or has occurred in a possible world which serves as the universe of interpretation for the evaluation process, and which, in the default case, is the real world”. In the Tibetic languages, it is mainly expressed though specific (morphologically derived) forms of the copulas and specific TAME/E markers on the verb phrase.

Vokurková (2022, pp. 166–168) summarizes the main differences in the epistemic forms found in Lhasa and “Common” Tibetan. The epistemic forms are not frequent enough in our corpus to allow a precise analysis. However, two observations can be made in this respect. First, the speakers in the adults’ or children’s generations have an inventory of epistemic forms limited to those constructed with sa.red (and, to a lesser extent, with gi.red), illustrated in (24). Such a restriction is congruent with Vokurková’s (2022, p. 179) observation that “in the diaspora most of the time only one type of epistemic ending, different from those of Lhasa Tibetan, is employed”.

(24)	a.	ཉ་		ཡིན་-ས་རེད་-གཱོ།					འདི།		Mo1
		ɲa		jin-sare-go					ndə
		fish		equ-epist-disc					dem
		“[It] must be a fish, this one.”
	b.	ཚོང་ཁང་-ར་			འདི་འདྲ་-ར་		མང་པོ་			མཐོང་-ཡིན་-ས་རེད།	Fa3
		ʦʰoŋkʰaŋ-ra			dinɖa-ra		maŋpo			tʰoŋ-jin-sare
		shop-and			like.this-and		many			see-cmp.past-epist
		“[He] has probably seen many shops and [things] like this.”
	c.	ད་	སོ་སོའི་			ནང་		འཚལ་-གི་ཡོད་-ཀྱི་རེད་-བཱ།			Ch1
		ta	sosøː			naŋ		ʦʰeː-gijø-kire-wa
		disc	oneself.gen			inside		look.for-uncmp-epist-phat
		“Well, he must be looking for his home, right?”
	d.	གྲོགས་པོ་	ཆགས་-ཡིན་-ས་རེད།								Ch2B
		ʈokpo	ʧʰa-jin-sare
		friend	become-cmp.past.epist
		“They probably became friends.”

In opposition to this general rule, we must mention the notable exception of Mo4, who, once again, retains several typical Amdo Tibetic epistemic forms, illustrated in (25)a. and b., together with several instances of “Common” Tibetan forms (e.g., in 25c.):

(25)	a.	མྱེ་		བཏང་-ས་-ཟིག་			ཡིན་-ན་ཐང་གི				Mo4
		mɲe		taŋ-sʰa-sək			jin-natʰaŋkə
		fire		lightV-nmlz-indf			equ.epist
		“[It] seems to be a place where you make fire.”
	b.	ད་		ཕལ་ཆེར་	ཚོད་མ་-ཟིག་			ད་		ཡིན་-ཁ་ཟིག་རེད།	Mo4
		ta		pʰaʧʰer	ʦʰoma-zək			ta		jin-kʰasəkre
		disc		probably	vegetable-indf			disc		equ-epist
		“Well, it’s probably a vegetable.”
	c.	ད་	འགྲོན་ཁང་-གི་			སྦྱིན་བདག་			ཡིན་-ས་རེད་-པཱ།		Mo4
		ta	nɖonkʰaŋ-gə			ʤindak			jin-sare-ba
		disc	hotel-gen			owner			equ-epist-phat
		“Well, this must be the owner of the hotel.”

But, at the same time, the corpus also comprises several instances of semi-grammaticalized constructions with, according to the context, a likely epistemic function. Such constructions are found in Fa3 and the children in Family 2, and attest the dynamics of this grammatical domain: Aikhenvald (2004, p. 296) had already noted that, for cognitive and social reasons, evidentiality is a grammatical domain particularly prone to develop and change in language contact situations, and the connex domain of epistemicity shows the same propensity of language change.

In Fa3, we can find a construction based on the adjective <’drapo> “like, similar”. The construction occurs in two forms: one with a nominalized clause followed by a main proposition with the adjective in a predicative position, illustrated in (26)a. and b. The full construction can be translated as “It is like […]”. In the second variant of the construction, illustrated in (26)c., the adjective predication is reduced to an epistemic ending <pa ’dra>, and the first clause is now the main clause. In (26)d., the construction can be analyzed as a fully grammaticalized completed past epistemic marker, or as a nominalization of the first clause, followed by the epistemic equative copula.

(26)

a.

འདི་

ཁྲོམ་རྭ་-ཟིག་

ཡིན་-པ་

འདྲ་པོ་

འདུག་-གཱ

Fa3

di

ʈʰomra-sək

jin-pa

ɖabo

du-gaː

dem

market-indf

equ-nmlz

similar

exist.sens-phat

“[It] is like this is a market, right?”

b.

ཙི་ཙི་

དི་-གིས་

ཁོ་འ་

ད་

ག་དུས་ཡིན་ན་

ཅིག་

Fa3

ʦiʦi

tə-ki

kʰoː

ta

katijina

ʧik

mouse

dem-erg/instr

3.dat

disc

whenever

hes

རོགས་པ་

འདྲ་པོ་

བྱེད་-གི་ཡོད་-བ་

འདྲ་པོ་

འདུག

rokpa

ɖapo

ʧe-kijø-wa

ɖapo

duˀ

help

similar

lightV-uncmp-nmlz

similar

exist.sens

“[It] is like this mouse kind of helps him all the time.”

c.

ད་

འདི་འདྲ་

ཡིན་-པ་འདྲ།

Fa3

ta

dinɖa

jinpaɖa

disc

like.this

equ-epist

“Well, it’s probably like that.”

d.

ཕལ་ཆེར་

ཉལ་-བ་

ཡིན་-པ་འདྲ།

Fa3

pʰaʧʰer

ɲɛ-wa

jin-paɖa

probably

lie.down-nmlz

equ-epist

“It seems to be that [he] probably slept.”

pʰaʧʰer

ɲɛ-wajin-paɖa

probably

lie.down-cmp.past-epist

“[He] probably slept.”

It is interesting to note that the grammaticalization of this construction has a parallel in Lhasa Tibetan (Vokurková, 2017, pp. 90–100), but the construction has not been maintained in “Common” Tibetan. Fa3 could have developed this epistemic marker in contact with Lhasa Tibetan speakers when he was still in Northern India. We can see here that he did not simply copy the grammaticalized structure, but replicated the grammaticalization process itself (Heine & Kuteva, 2005, pp. 79–122).

Similarly, in Ch2A and Ch2B, we can observe four variants of a construction with, apparently, an epistemic function, this time with the verb “to think”. These constructions are described in

Table 15. An incipient epistemic construction in Ch2A and Ch2B.

	Construction with an Epistemic Function
Syntactic construction	[finite clause] བསམ་ “think”-TAME/E
	[subordinated clause V-TAME/Eego] བསམ་ “think”-TAME/E
Morphological construction	clause mainV-བསམ་ “think”-TAME/E

, and ordered from the least to the most grammaticalized.

The least grammaticalized form, illustrated in (27)a., is merely a copy of the French construction with an epistemic function Je pense que [finite clause] “I think, [finite clause]”, but with the finite clause. In the second variant (27b.), the verb of the first clause has an obligatory egophoric TAME/E marker, a morphosyntactic restriction characteristic of an embedded clause in the Tibetic languages (Tournadre & Suzuki, 2023, p. 418). The last variant, illustrated in (27)c., is either a serial or an auxiliary verb construction,10 in which the verb “to think” is not the main verb anymore and has a grammaticalized function of marking epistemicity.

(27)	a.	ཨ་མ་	འདི་-གིས་			ཁ་ལག་		བཟོས་-རེད་	Ch2A
		ama	di-ki			kʰala		sø-re
		mother	dem-erg/instr			food		make-cmp.past.fact
		བསམ་གྱི་འདུག
		sam-giduˀ
		think-uncmp.sens
		“I think, the mother prepared food”
	b.	བག་ལེབ་-ཅིག་			བཙལ་གྱི་ཡོད་			བསམ་གྱི་འདུག	Ch2B
		paklep-ʧik			ʦeː-gijø			sam-giduˀ
		bread-indf			look.for-uncmp			think-uncomplsens
		“[I] think, [he] is looking for a bread.”
	c.	ཅིག་		བཙལ་-གྱི་			འགྲོ་བསམ་གྱི་འདུག		Ch2B
		ʧik		ʦeː-gi			ɖo-sam-giduˀ
		indf		look.for-conn			go-think-uncmp.sens
		“[He] presumably goes to look for something.”

6.4. Inner-Sensation Verbs and Endopathic Access

Finally, the category of inner-sensation verbs, comprising verbs such as “to be hungry/thirsty”, “to be afraid”, “to feel pain”, etc., must be mentioned to illustrate more subtle linguistic changes in the domain of TAME/E markers in the development of “Common” Tibetan in the diaspora. In the Tibetic languages, for such verbs, the sensory evidential markers can only be used in 1st-person utterances, but not in 2nd- and 3rd-person (Tournadre & Suzuki, 2023, pp. 349, 398). Thus, in Central Tibetic, an utterance such as “I’m hungry”, “I’m afraid”, “I can see sth.”, or “I have an headache” is normally constructed with an uncompleted sensory TAME/E marker, since only the speaker can perceive directly one’s own hunger, fear, vision, or headache. Conversely, in a third-person utterance, the speaker is merely able to perceive indirect signs of such feelings, and, accordingly, s/he must use an inferential TAME/E marker.

Examples in (28) below show that some speakers of both generations do use the expected sensory inferential TAME/E markers (in their Central (ex. 28a. and b.) or Amdo (ex 28c.) Tibetic version) with these verbs in 3rd-person utterances.

(28)	a.	ཨེ་འདྲ་			འདྲ་པོ་	བྱས་-ནས་		ཁོ་	ཞེད་སྣང་-ར་	ལངས་-བཞག	Fa3
		enɖa			ɖabo	ʧe-ne		kʰo	ʃenaŋ-ra	laŋ-ʃa
		like.that			similar	do-conn		3	fear-too	lightV-perf.infer
		“[It] happened like that and he was afraid!”
	b.	མི་			འདི་	ཞེད་-འདུག་-གཱ།					Ch3
		mi			di	ʃe-du-gaː
		person			dem	fear-perf.sens/infer-phat
		“This person was afraid, right?”
	c.	དེ་ནས་		བཀྲ་ཤིས་		ཁ་		ཐུག་གི་ཡོད་རེད་བཱ			Mo4
		tene		ʈaʃi		kʰa		tʰək-kora
		then		PN		face		meet-uncompl.fact.phat
		ད་	ཙིག་གེ་ཟིག་				སྐྲགས་-བཏང་-ཟུག
		ta	ʦəkesək				rʨak-taŋ-zək
		disc	a.bit				fear-aux:cmp-perf.infer
		“Then Tashi comes face to face [with the bug], right? There, he got a bit afraid.”

However, examples in (29) illustrate the fact that, in the same type of utterances, other speakers—again, both from the parents’ and the children’s generation—use direct sensory markers (in uncompleted or completed aspect).

(29)

a.

འདི་

ཞེད་-ཀྱི་འདུག

Mo1

ndə

ʃe-gidu

dem

fear-uncmp.sens

“This one is afraid.”

b.

ཨ་ནི་

རྗེས་ལ་

ཁོ་

འེའེ་

ཅིག

Ch1

ani

ʤe-la

kʰo

əəə

ʧik

then

after-dat/loc

3

hes

indf

གད་སྙིགས་

འདྲ་པོ་ཅིག་

མཐོང་-སོང་།

kʰeɲiˀ

ɖapo-ʧik

tʰoŋ-soŋ

garbage

similar-indf

see-cmp.past.sens

“Then, after, he saw some kind of garbage.”

c.

པཱ་ལགས་

འདི་

གནམ་གྲུ་

ནང་ལ་

ཅིག་

ཞེད/

ཞེད་-ཀྱི་འདུག

Ch2A

paːla

di

namʈu

naŋ-la

ʧik

ʃe/

ʃe-giduˀ

father.H

dem

plane

inside-dat/loc

indf

fear

fear-uncmp.sens

“This man, in the plain, er, [he] was af/ afraid.”

Thus, some speakers, adults as well as children, seem to have lost the grammatical category of endopathic, used to encode internal sensory access, while others retain this grammatical category. Arguably, the speakers who have lost the category of endopathic and do not distinguish internal and external access to information might have generalized the use of TAME/E markers on the basis of the syntactic person, rather than on pragmatic-semantic principles, and such a change could have been induced by contact with English and/or French.

More research on the use of evidential and epistemic markers in different sub-varieties of “Common” Tibetan is needed to understand the system and its evolution, especially in the contact of languages without grammaticalized evidential–epistemic systems. An analysis focused on the use of the different evidential and epistemic markers in discourse would, for instance, enable us to examine whether these categories retain their fundamentally semantic–pragmatic functions, or if they evolve into syntactic categories.

7. Elements for Discussion and Future Research

The aim of this paper was to analyze contrastively the parents’ and the children’s varieties of “Common” Tibetan for a series of phonological and grammatical features in order to document the initial steps of the adoption of the diasporic lingua franca, and the extent of intra- and transgenerational variation that goes with it.

While the phonological feature examined in this paper shows homogeneity within the parents’ generation, the morphosyntactic features show a significant variation: two speakers, Mo2 and Fa3, speak a variety of “Common” Tibetan much closer to the Central Tibetic languages than Mo1 and Mo4. Depending on the linguistic feature, this difference can be observed in absolute terms, with a complete switch from Amdo to Central Tibetic morphemes (e.g., for the indefinite or plural markers or the system of TAME/E markers), or in proportion (e.g., for the case markers or the use of distinct verb stems).

As mentioned in the introduction, all the adults recorded for this research have a comparable migration history, from Tibet to France through South Asia, with some slight differences in terms of stay duration in the different regions. But such intra-generational variation observed in our corpus might, in the parents’ generation, be explained by a different sociolinguistic background of the speakers: while Mo2 and Fa3 have been to school as children and have a high level of formal education in Tibetan, Mo1 and Mo4 have been taught to read and write as adults, after they arrived in exile. We can thus assume that a good knowledge of written Tibetan makes it easier to establish links between the different Tibetic languages. It allows a better linguistic consciousness and makes the switch from one to the other variety easier. At the same time, a high level of education is also related to a shorter stay in the linguistically mixed environment of the adult education centres run by the Central Tibetan Administration. This experience, although important in the refugee’s trajectory, seems to have significant consequences in the adoption of Central Tibetic linguistic features.

The corpus also indicates that children retain only a few characteristics of their parents’ original Amdo Tibetic variety. The transition from Amdo to Central Tibetic in exile is rather abrupt, between the first and the second generation. The structural differences observed between the parents’ and the children’s generation confirm the attitude of the parents, who claim to favour and encourage the transmission of “Common” Tibetan to their children, rather than their native variety (Simon, 2023). Such a shift is also documented in Chinese-speaking families settled in Canada (Ward et al., 2022), who, more or less consciously, choose not to transmit their original variety or varieties of Sinitic, but favour a standard variety associated with an international environment. The regular contacts between the members of the growing—but still relatively modest—Tibetan diaspora in France, especially during the weekly community classes, certainly further encourages this dynamics towards the adoption of “Common” Tibetan in the families involved in this study.

Besides the overall structural similarities observed in the children’s generation, some intra-generational differences are also observed. Certain differences, like the loss of a fully grammaticalized dual number category, the extension of comitative markers to mark instrumental functions, or the development of a new epistemic construction with the verb “to think”, can be attributed to language contact with French and/or English, while others, such as the grammaticalization of a new TAME/E construction by Ch1, seem to be language internal. Among the interviewees, Ch3 is the only representative of the children’s generation who has attended a Tibetan school in South Asia until high school: the other four children arrived in France after 0–3 years of primary school, and thus have spent most of their school career in France. This important difference in their linguistic socialization explains why Ch3 retained some morphosyntactic features lost by the other children in the sample, such as the specific TAME/E marking system for verbs referring to inner sensation.

Future research should expand on these initial analyses both by studying contrastively other phonological and morphosyntactic features, and by including Tibetan families with various dialectal backgrounds. This second line of research can however be conducted only with native speakers of Tibetic languages for which good documentation and descriptions are available. In the domain of contrastive Amdo/Central Tibetic phonology, of particular interest would be the study of the emergence of a tone contrast (absent in Amdo Tibetic varieties). In the domain of morphosyntax, a complete analysis of the case marking system and syntactic alignment would also shed light on changes from a full ergative alignment (in Amdo Tibetic varieties) to a split ergative alignment (in Central Tibetic), and other possible changes caused by the contact with languages without case marking. A detailed analysis of the incipient TAME/E categories observed in Section 6 would also enrich our understanding of the development of “Common” Tibetan in the West.

Finally, the data presented in this paper show that, in at least two of the Amdo Tibetan families included in this study (Families 1 and 4), the Tibetan language can be described as a dual language: in terms of linguistic structure, parents and children speak a significantly different variety of Tibetan, while having a passive knowledge of the other generation’s variety. Future research could try to precisely assess this passive knowledge of the second Tibetic language in each generation in order to highlight the points of continuity and discontinuity in the development of Tibetan as an heritage language in France. Nevertheless, from a sociolinguistic point of view, this generational language change is not conceived of as a language shift, but rather as a continuity and transmission of the heritage language.