Introduction
Paul Kent Andersen notes that in Classical Tibetan certain types of zero realisation operate under syntactic limitations: there are “boundaries” that block zero anaphora and zero cataphora.Footnote 1 Among other conclusions, Andersen notes that zero anaphora is blocked by a perfective verbal noun marked with the associative case -dang, which he describes as creating a “boundary” entailing a switch in subject reference as well as allowing for a “new spatio-temporal setting”.Footnote 2 While he uses the term “boundary” loosely to refer to a number of affixes, including terminative and associative converbs, his research points to a grammatical function not articulated in traditional Tibetan grammars: the structures distinguishing co-reference from switch-reference.
By framing his discussion in terms of zero anaphora, Andersen does not explicitly recognise that Tibetan shows a system of switch-reference. Zadoks appears to have been the first researcher to take this step, contrasting V-pa-dang “discontinuity” and V-nas “continuity”.Footnote 3 Felix Haller builds on Zadoks’ laconic and unpublished remarks in a full-length paper “Switch-Reference in Tibetan”.Footnote 4 Confirming Zadoks’ findings, on the basis of his analysis of Shigatse-area colloquial Tibetan and the language of the Mi la ras pa'i rnam thar by Gtsang smyon he ru ka rus pa'i rgyan can (1452-1507), Haller settles on a “rule of thumb” that “-ne [-nas] is employed if the verb of the preceding clause is a control verb and two actants are co-referential and occupy the highest possible position in the hierarchy, respectively. -pata. [-pa-dang] is used in all other cases”.Footnote 5 Implicit in this conclusion are several other useful observations, namely that both structures typically convey anteriority, and that there is a “hierarchy” of modes of involvement in an event, at the top of which is the agent of a verb of control.Footnote 6 More importantly, while in governing co-reference -nas is highly restricted to joined clauses in which the actor functions in the role of agent in both, switch-reference serves as a broader catch-all category for a number of syntactical relationships, which may relate two people or entities joined in a number of other ways that are not necessarily as agents of control verbs. Having said this, as Haller points out, attuning ourselves to the nuance of distinction between the subordinators -nas and -pa-dang may help us to better interpret certain complex passages.
With the hope then of demonstrating the usefulness of discerning the switch-reference behaviour of -dang and other related converbs from the co-referential behaviour of the elative converb -nas, this article will analyse a sūtra translated in the early Tibetan translation period known as the Ye shes rgyas pa'i mdo (YG).Footnote 7 While relatively unknown, this lengthy text is useful as a locus for studying switch-reference largely because of its content. Within its presentation of the ten powers (stobs bcu) of the Buddha, it also contains dozens of past life stories of the Buddha prior to his awakening, each with its implicit causal rationale as to why he possesses one of the many subdivisions of his ten powers. These tales provide—in the numerous interactions and conversations they encompass—a rich mine of examples of the workings of co- and switch-reference.
Co- and switch-reference in the YG
In attempting to assess the distinct functions of different subordination markers, it may be most helpful to notice how they work in proximity in the same syntactical units. One story in the YG is about the Buddha's past incarnation as a ship's captain who journeys to find distant treasures. After acquiring them:
(1) de 'gro ba shin tu grub nas phyir log ste, rgya mtsho las de rgal nas rgya mtsho la nye reg byas pa dang, nor bu rin po che lag pa la btags pa rgya mtsho chen por lhung ngo Footnote 8
He, having thus accomplished the purpose of his journey (-nas), began his return home (-ste). Yet, while crossing the ocean (-nas) he took a bath in the sea (-pa dang), [and] a precious jewel tied to his hand fell into the sea.
The connections among the first three clauses behave as Haller describes, with the voyager (here represented by the pronoun de) acting as agent for the control verbs grub, log, and rgal, and then bathing (nye reg byas). The verbs grub (accomplish) and rgal (cross) are marked with the converb -nas that shows co-reference. The verb log (return) is marked with the converb -ste. So far this suffix has not been the subject of research in terms of co- or switch-reference, but in this case it coincides with co-reference. The third and fourth clauses are then joined by the subordinator -pa dang because although nye reg byas is a control verb and could potentially link via -nas to another verb (i.e. if it were followed by a verb meaning “drop”), in this case the change to a different subject (nor bu) necessitates -pa dang to make clear the change of reference.
In another story in the YG a man is wandering through the woods when a sudden storm brews and a bear takes him into its den:
(2) de nas zhag bdun 'das pa dang sprin nag po de med de de'i gnas nas phyung ste gtong ngo Footnote 9
After seven days passed and (-pa dang) the dark clouds cleared (-de), they emerged from the den and (-nas) [the bear] sent the man off.
Semantically, the seven days (zhag bdun) are the subject of the initial clause, the dark clouds (sprin nag po) govern the action of the middle clause (“cleared”—med), while the bear, represented by the pronoun de, is the agent of the main verb in the final, “sent” (gtong). Thus there are two changes of reference. The first occurs with -pa dang, a switch-reference that seems to have the added effect of solidifying the temporal sequence. The second and third clauses are joined by the semi-final converb de, which here appears to allow for switch-reference. A comparable yet subtler example appears in a story of the Buddha's life as a Brahmin, in which the latter relates the following:
(3) yab dang yum bdag de ring mtshan mo'i tho rangs gnyid sangs pa dang, bar snang las kun 'byung gi chos gang yin pa de thams cad ni, 'gog pa'i chos can no, zhes sgra thos te Footnote 10
“Father, mother, this morning I awoke at the break of dawn (sangs pa dang) and heard a voice in the atmosphere (antarikṣa) saying, ‘It is the nature of everything that originates to come to an end.’”
Although in the English translation “awoke” and “heard” have the same subject, in fact this passage again conforms to the established pattern. Here -pa dang is chosen because the criteria Haller mentions for co-referential use of -nas are not met: sangs is not a verb of control, nor is the Brahmin at the top of the hierarchy of involvement in the action—it is, literally, sleep (gnyid) that is broken (sangs)—and thus switch-reference occurs.
A sentence from the tale of the ship's captain which immediately follows example (1) in the YG seemingly presents more direct counter-evidence for -pa dang as a marker of switch-reference, while further confirming the use of -nas for co-reference
(4) des nor bu rin po che de lhung ba tsam nyid kyis chu phyags thogs te, « nor bu rin po che'i phyir de ring rgya mtsho ‘di skam par bya'o, » snyam nas, des chu phyags thogs te rgya mtsho chen po bskam par brtsams pa dang, des chu phyags bzung ma thag tu, lha bye ba khrag khrig stong phrag brgya dag grogs byed par grub par gyur te Footnote 11
He, as soon as the jewel had fallen, picked up a cloth and thinking “On account of the precious jewel, I shall today dry up the sea!” (snyam nas) he lifted the cloth and endeavoured to dry up the ocean (brtsams pa dang). After he had grabbed the cloth immediately (ma thag tu) one hundred quintillion gods came to his assistance.
Here -nas shows that it is the captain who has the thought (snyam) and also picks up (thogs) the cloth, but the syntactic relationship of the final three clauses is more complex. The captain is the agent of both “set out” (brtsams) and “grabbed” (bzung), and a -pa dang between these two verbs appears to contradict the pattern of co-reference established so far. The explanation is that des chu phyags bzung ma thag tu functions adverbially to the clause with the verb grogs byed par grub par gyur te. In other words, -pa dang does not coordinate brtsams and bzung, but rather it coordinates brtsams and grogs byed par grub par gyur.
We see now that co-referential -nas is governed by strict rules that necessitate a clear sense of an agent remaining in control over separate clauses, whereas -pa dang always accompanies a switch in the reference. Tibetan allows many other ways to connect clauses, not all of which can be discussed here. We have seen evidence that the semi-final converb (-te, -ste, -de) is neutral with respect to switch-reference: in example (1) -ste coincides with co-reference but in example (2) -de coincides with switch-reference. Another convention that Hill has noticed in Old Tibetan, which has seeped into Classical Tibetan, is the usage of the -pa las to convey a sense of surprise.Footnote 12 In the YG the -pa las appears only rarely, compared to the extreme frequency of -pa dang, but it seems to reflect this surprise meaning, while also coinciding with switch-reference. This function is seen in the following passage, in which a man on a journey is suddenly beset by a storm:
(5) maudgal gyi bu sngon byung ba 'das pa'i dus 'khor ba dpag tu med pa de'i pha rol gyi yang ches pha rol na, bram ze bha ra dwa dza dang rus gcig pa zhes bya ba zhig shing gzhan zhig gi drung du song ste, 'dug pa las de'i tshe dus ma yin pa'i sprin chen po zhig byung nas de nyid du char bab bo Footnote 13
Long ago in limitless cyclic existence, Maudgalyāyana, in the distant past there was a Brahmin named Descendant of Bharadvāja. Once he was travelling to visit someone. He was resting along the way (-pa las) when a great untimely storm cloud appeared (-nas) and it began to rain.
Here there is a distinct contrast between the surprise implied by the sudden rain, as opposed to the matter-of-course nature of clouds forming and then (-nas) rain falling. One can note that although -nas is here used for co-reference, if one understands that the cloud rains, bab is not a control verb. Thus, Haller's specification that -nas is associated with co-reference specifically with control verbs, requires some further elaboration.
Switch-reference in conversation
By far the most active domain for the functioning of reference relations lies in passages of dialogue, which occur frequently in the YG. It is at these junctures that the subordinators -nas and -pa dang help the reader (more than anywhere else) to keep track of whom among the participants in a conversation—which at times may be numerous—is speaking.Footnote 14 As an example:
(6) « bcom ldan 'das legs so » zhes gsol nas, byang chub sems dpa' 'od 'phags bcom ldan 'das kyi ltar nyan pa dang, bcom ldan 'das kyis 'di skad ces bka' stsal to Footnote 15
“Excellent, Blessed One,” replied the bodhisattva Noble Light, (-nas) and as he the listened keenly (-pa dang), the Blessed One explained…
Here, the co-referential -nas tells us the bodhisattva is speaking, and then still listening, while the switch-reference occurring with -pa dang indicates the change in dialogue to the Buddha's reply. Here and elsewhere in the YG these subordinators serve a crucial role as consistent reference cues in dialogue, aided by the slightly less consistent placement of the quotative clitic zhes, and honorific and humilific verb forms. On certain occasions -nas and -pa dang are vital, as for instance when zero-marked vocatives appear in conversation to render the reference relations even murkier:
(7) rgyal rigs sa ga'i bshes gnyen shing sā la chen po lta bus « bcom ldan 'das legs so » zhes gsol nas bcom ldan 'das kyi ltar nyan pa dang, bcom ldan 'das kyis de la 'di skad ces bka' stsal to Footnote 16
Viśākhamitra of the warrior caste, who was like a great sal tree, said, “Of course, O Blessed One!” (-nas) and he listened attentively as the Blessed One had directed. (-pa dang) The Blessed One continued…
These relations of course become much more complex in stories in the YG in which three or more characters interact. Although the co-reference signalled by -nas may be partial,Footnote 17 in the following passage three people are together co-referentially indexed by -nas. Note as well how the switch-reference of surprise (-pa las) shifts the focus of the narrative abruptly to the demon, while the ensuing speech is clearly attributed to him, largely due to the presence of -nas:
(8) maudgal gyi bu de nas rgyal po tshangs pas byin btsun mo dam pa dang, bu thu bo gzhon nu dang lhan cig tu bkad sar song nas 'khod pa las, de nas sha za gzugs mchog tu mi sdug pa zhig 'ongs nas, rgyal po de 'dug pa'i bkad sa'i bar snang la skyil mo krung bcas nas 'di skad ces smras so Footnote 18
Maudgalyāyana, King Brahmadatta went with her majesty the queen and the crown prince to the dining hall. (-nas) Once there, (-pa las) a hideous piśāca demon arrived. It sat cross-legged in mid-air in the middle of the dining hall where the king was sitting, (-nas) and spoke to him…
The special context of verse narrative
As a final note, it is worth mentioning that it is distinctly useful to pay attention to these dynamics when reading narrative conveyed in verse, as frequently occurs in the YG. Such narrative passages can be lengthy at times. In general, they are governed by the same principles, and yet of course, for the sake of meter, zero anaphora becomes more common, coupled with the frequent elision of case markers and quotative clitics. As a simple yet representative example I present selections from the YG of a meeting between the Buddha and a group of nāgas; the latter seek the former's counsel, who in turn predicts their future awakening:
(9a) ston pa'i gsung gis klu de mgu gyur nas,
sangs rgyas la ni snying nas skyabs song zhing,
byang chub gzhol la yid ni bskyed pa dang,
'jig rten mkhas pas byang chub de lung bstan Footnote 19
Overjoyed by the Teacher's words, (-nas)
The nāga took refuge in the Buddha sincerely,
Formed the resolve in his heart for awakening, (-pa dang)
And the Knower of the World prophesied his awakening.
(9b) rnam par snang byed dge ba tshol ba'i klus,
'jig rten mgon de la ni zhus pa dang,
nyon mongs kun sel yan lag brgyad ldan zhing,
snyan cing 'jebs pa'i sgras ni lung bstan pa Footnote 20
Thus Vairocana, a nāga in pursuit of virtue,
Asked the Guardian of the World, (-pa dang)
Who prophesied (lung bstan pa), with sweet and pleasant words
With eight qualities to remove all afflictions.
(9c) klu yi rgyal po bsgyings pas rnam par 'dren,
'jig rten mkhas pa de yis zhus pa dang,
yid du 'ong ba'i tshigs bcad sna tshogs kyis,
klu yi rgyal po yid ni dga' bar gyur Footnote 21
Thus did the nāga king Haughty One ask
The Guide and Knower of the World. (-pa dang)
The numerous beautiful verses
Then delighted the nāga king.
Each of these examples (9a-c) exhibits a certain twist of grammar that is no doubt muddled further by the translator's attempts to render Sanskrit structures as faithfully as possible. In the midst of these, the reference relationship functions of -nas and -pa dang remain constant and offer us our best chance of properly interpreting the interchanges. In (a) for instance we notice that co-reference necessitates the use of -nas because the nāga is understood to be the primary agent, even though his joy is stimulated by the Buddha's words. All of the ensuing actions therefore belong to him, until the switch-reference that occurs with bskyed pa dang¸ which is required by the Buddha's subsequent reply. In (b) the switch-reference makes clear that the flowery descriptions that follow it describe the sounds of the Buddha's prophecy. Had it been elided as well in the interests of meter, we might wonder if the third line of the verse in particular is to be construed with the question or the reply. In (c) the switch-reference serves the important role of not simply marking where one person's speech ends and that of the next begins, but rather conveys important syntactic information regarding the reference of the adjoining clauses. Moreover, in combination with the humilific verb zhus, the switch-reference aids us in understanding that the convoluted syntax of the antecedent clause holds one referent or, in other words, that des in the second line of the verse refers back to klu yi rgyal po bsgyings pas and not the Buddha, ‘jig rten mkhas pa.
Conclusion
The humble aim of this article has been to apply the research of Andersen, Zadoks, and Haller into the reference-relation governing rules of subordination to a sūtra that provides many examples of their functioning, and thereby extend the range of these rules’ applicability. We have seen that, while reference relations do not constitute a topic in traditional Tibetan grammars, converbs and declined verbal nouns such as -nas, -pa dang, and -pa las encode a set of syntactical rules that, at times, reinforce other aspects of subordination and, at other times, render such other factors unnecessary. In short, -nas indicates co-reference, whereas -pa dang and -pa las indicate switch-reference, the latter with an added meaning of surprise. The weak evidence marshalled here also indicates that the converb -te, -ste, -de is neutral with respect to switch-reference. These grammatical constructions may often hold the key to the interpretation of many complex passages, especially in the contexts of conversation or when numerous agents are interacting with one another. Perhaps most significantly then, for those in the situation of having to decipher Tibetan translations of Sanskrit texts, it has been shown that being properly attuned to the nuance of reference relations offers a promise of guiding readers to a more accurate understanding of the texts in front of them, in verse passages more than anywhere else.
