The Umpithamu language

Umpithamu is a language that is spoken on the east coast of Cape York Peninsula in northeast Australia. It belongs to the Paman subgroup of the Pama-Nyungan family, as established by Hale 1964. It is one of the four languages associated with a group of people currently known as Lamalama (a post-classical group type that has been called a “language-named tribe”; see Sutton 2003:72–73), together with the languages Morrobalama (also known as Umbuygamu), Mba Rumbathama (also known as Lamalama), and Rimanggudinhma (see Rigsby 1992 for more information on the languages of this group). Like the other languages associated with the Lamalama people, Umpithamu is moribund, with two remaining full speakers, a handful of people who can use Umpithamu with varying degrees of fluency but whose main language is one of the other Lamalama languages, and a number of younger people who can understand Umpithamu but do not themselves speak it. In spite of the small size of the speech community, the language and the stories told in Umpithamu are still very much part of everyday life. For the last speakers, it is the basic language of interaction when talking to each other, and they also frequently speak Umpithamu to other members of the community who can understand the language. With older speakers of other Lamalama languages, such conversations are often bilingual in Aboriginal languages, with the Umpithamu speakers addressing the interlocutors in Umpithamu and the interlocutors replying in Morrobalama or Mba Rumbathama. With younger people who can understand Umpithamu, the conversations are typically bilingual with English, with the Umpithamu speakers addressing the interlocutors in Umpithamu, and the interlocutors replying in English.

There has been a large amount of historical and anthropological work about the past and current lives of the Lamalama people (see, e.g., Hafner 1999, Jolly 1997, Rigsby 1999, Rigsby & Chase 1998, Thomson 1934), but there is little published linguistic work on the languages associated with the Lamalama (see Rigsby 1992, 1997 on Umpithamu). The analysis in this article is based on the first author's linguistic work with Mrs. Florrie Bassani since 2003, and on recordings made by Bruce Rigsby with Ms. Joan Liddy in 1972 and 1974. The analysis of the narrative studied here is checked against a larger reference corpus consisting of 30 narrative texts (27 recorded by the first author, and 3 recorded by Bruce Rigsby). The corpus is annotated for aspects of participant marking and case marking, and amounts to approximately 6,000 words or 1,700 clauses. Any quantitative information on aspects of grammatical structure in this article is based on this reference corpus.

The historical context

In 2003, the first author recorded a text from Mrs. Florrie Bassani, in which she describes the removal of her family from their home country at the mouth of the Stewart River on the east coast of Cape York Peninsula (Port Stewart, Yintyingka in Umpithamu), their subsequent deportation to an Aboriginal reserve 400 kilometers further north (near Bamaga at the tip of Cape York Peninsula), the attempts of various family members to visit their home country or to receive visits from other family members who remained near Port Stewart, and their final return to Coen, a town about 70 kilometers west from their home country.

The events recounted by Mrs. Bassani are documented in various historical sources, and they are by no means uncommon in the history of contact between Aboriginal people and white settlers and officials in Australia. For extensive documentation of the historical circumstances, the reader is referred to the official documents excerpted in McIntyre-Tamwoy (2000), and to the historical surveys in Rigsby & Williams (1991) and Genever (1997). In this section, we will merely summarize the events on the basis of these sources, to enable the reader to better understand the social and historical background of the narrative.

The people currently known as Lamalama are a post-contact group of Aboriginal people, formed in the 1920s around the mouth of the Stewart River, from those who survived dispossession and introduced diseases in the lower Princess Charlotte Bay area, on the east coast of Cape York Peninsula. Their traditional way of life was studied at the end of the 1920s by the anthropologist Donald Thomson (e.g. Thomson 1934), and they continued to live on their home country around Port Stewart until their removal in 1961. The removal was instigated by the owners of a nearby cattle station, who considered the presence of a group of Aboriginal people a nuisance and in the late 1950s started lobbying for their removal with the Department of Native Affairs. After some discussion with the local Protector of Aborigines from the nearby town of Coen, who initially defended the Lamalama against the accusations, the Department of Native Affairs decided to give in to the owners of the cattle station, and to remove the Lamalama people from Port Stewart to an Aboriginal reserve near Bamaga. In 1961, white officials used the excuse of a medical examination to convince a group of twenty-three Lamalama people to board a boat, and then took them 400 kilometers north to Bamaga at the tip of Cape York Peninsula. A number of other Lamalama people, who were working on cattle stations at the time, were not removed. In spite of continuous requests to return to their home country, the deported Lamalama people were kept on the reserve, and for a long time family members who were not deported were refused permission to visit their deported relatives in Bamaga. Several elderly people, including the father of Mrs. Florrie Bassani, passed away near Bamaga without seeing their home country again. In the 1970s, the remaining people in Bamaga started moving back south, and settled in the town of Coen. They soon started going back to Port Stewart, and now they have three outstations in the area, with some people staying there permanently and others coming in from Coen for shorter periods of time to fish, hunt, and spend time on their home country (see Rigsby & Williams 1991 for an account of the Lamalama people's re-occupation of their home country).

The narrative

In this section, we will present the narrative as it was recorded from Mrs. Florrie Bassani in 2003. The version reproduced here is slightly edited, in the sense that we removed some English lines that translate Umpithamu lines and were obviously meant to provide clarification for the sake of the recorder, whose proficiency in Umpithamu was limited at the time. We realize, of course, that issues like switching between languages, and the choice which lines to translate and how, are not trivial and constitute a topic of research in their own right, especially for questions like the interactive dimension of narrative performance. They are not immediately relevant to the questions of narrative structure investigated here, however, so we have chosen to leave out these translations for the sake of brevity. We have only removed clauses that repeat information already provided in Umpithamu.¹

The narrative is represented in Table 1. It is subdivided into numbered lines (1–97), and the lines are grouped into episodes, defined further in the following section. The lines are presented in the form of a table, with the line number in the first column, the Umpithamu clause in the second column, a glossed version in the third column, and a free translation in the fourth column. Each episode is identified by a character (A–M) and preceded by a brief summary.

The narrative.

Participant marking in Umpithamu

In Umpithamu, as in many other languages of the Pama-Nyungan family, clause participants are marked primarily with pronominal means, and secondarily with lexical means. The example clauses in (1), (2), (3), and (4) below represent the most typical ways to mark participants in Umpithamu. Participants can be marked only with pronouns, as in (1), where the nominative pronoun ilu refers to the subject role and the accusative pronoun ungku refers to the object role; or only with lexical items, as in (2), where nayngka ‘child’ is the only element referring to the direct object role; or with a combination of a lexical item and a cross-referencing pronoun, as in (3) and (4), where the subject angkutha athuna ‘my grandfather’ and the direct object anharra yawul ‘big crocodile’ are cross-referenced with a nominative pronoun iluwa and an accusative pronoun ungku, respectively.

Within this system, lexical marking of participants is the least typical option overall, as shown by the figures in Table 3. Of all verbal clauses in our reference corpus, about two-thirds have no lexically marked participants at all and rely exclusively on pronominal marking. Only one-third have lexically marked participants, hardly ever more than one per clause. For clauses that use lexical marking, there is a further degree of markedness, in that subjects are less often marked lexically than objects, representing slightly less than one-third of all lexically marked NPs.³

Frequency of pronominal and lexical marking of participants.

What these figures show, therefore, is that in Umpithamu discourse, the most typical way to mark participants is to use only pronouns, and the least typical way is to use lexical marking for subjects, with lexical marking of objects in between. This is somewhat reminiscent of patterns of preferred argument structure (e.g., Du Bois 1987), but what stands out here is the generalized discrepancy between subjects and objects,⁴

Within the system of pronominal marking, Umpithamu has a further contrast between two positions in the clause: Pronouns most typically occur in their standard position as enclitics to the verb, as in (1)–(4) above, but they can also occur in a less frequent position at the start of the clause, as in (5) below. This marked position signals focal status for the referent in question; in such cases, the referent is central in the local development of the discourse, for instance in establishing a relation of contrast or an explanatory relation between two subsequent clauses. In (5), for instance, the fronting of the pronoun is motivated by the contrast between the fact that the referent of the 3rd person singular nominative pronoun ilu- in (5b) claims ignorance about the whereabouts of a fugitive (5a), and the fact that this same person is the one who told the fugitive to go into hiding (5b).

Within this grammatical opposition, the clause-initial position is marked not only in qualitative terms, because of its association with a marked information-structural option, but it is also clearly marked in terms of its frequency in natural text. As shown in Table 4, out of the 1,409 clauses that contain pronominal forms (both with verbal and nonverbal predicates), only 99, or 7%, have a clause-initial pronoun.

Frequency of pronoun positions.

Participant marking and episode structure

The literature on episode structure has shown that systems of participant marking can play an important role in the delimitation of episodes, in the sense that switches in the constellations of participants can be one of the factors that trigger thematic breaks between episodes (see Hinds 1977 for descriptive evidence, and Tomlin 1987 and Ji 2002 for experimental evidence). From a linguistic perspective, moreover, Fox 1987 has shown that in contexts allowing choice between various types of participant tracking devices, the morphosyntactically marked types tend to be associated with the start of episodes, and thus come to function as formal markers of episode boundaries. If we apply these two principles to the system of participant marking in Umpithamu, they make the following predictions. For non-speech act participants, there is a choice between the default system of pronominal 3rd person marking and the marked system of lexical marking, which predicts that lexical marking will be used at the start of episodes, and pronominal marking within episodes. For speech act participants, there is only pronominal marking, but in this system there is a choice between the default postverbal position and the marked clause-initial position, which predicts that the clause-initial position will be used at the start of episodes, and the postverbal position within episodes. If we look at the data, both in our narrative and in our reference corpus, these predictions are only partially borne out, because lexical marking is also found within episodes, and clause-initial pronominal marking is even found exclusively within episodes. On the basis of a qualitative analysis of this unexpected distribution of marked forms, we will argue for a refined version of the Principle of Morphosyntactic Markedness, which not only looks at formal markedness oppositions in participant tracking as such, but also takes into account their specific functions. In addition, we will argue that McGregor's Expected Actor Principle, originally formulated for episode structure in Gooniyandi (McGregor 1992, 1998), provides a good explanatory framework for the overall distribution of participant marking devices found in Umpithamu narratives, including the unexpected distribution of clause-initial pronouns.

If we look at the start of episodes in the narrative, these generally confirm the prediction about the distinction between lexical and pronominal marking found in the literature. For non-speech act participants, we find lexical subjects in the first line of the episode, which are different from the final subject in the preceding episode, while for speech act participants we find pronominal subjects that are different from the final subject in the preceding episode. Thus, episodes B, D, E, G, H, K, L and M each start with a line that contains a lexical subject. In episodes B, D, E, H, L and M, the rest of the clause is in the same line as the lexical subject. In the case of G and K, the lexical item is separated from the clause for which it forms the subject by an aside about the person(s) involved – that is, an explanation about the makeup of the family in line 51 in G, and an indication of the location of the person involved in line 86 in K. In any case, the distribution of lexical subjects in the narrative strongly suggests a link with episode structure, in that lexical subjects tend to occur in the first clauses of episodes, and thus seem to mark the start of a new episode. The first clause in episodes F and I (and the English one in J) all use a pronominal subject referring to a speech act participant that is different from the one used at the end of the preceding episode. Thus, episode F starts with a 1st person plural subject, switching from the 3rd person singular subject at the end of episode E, and episode I starts with a 1st person plural subject, switching from the 3rd person singular subject at the end of episode H (not including the quoted line in 68). In this sense, the prediction about the distribution of lexical versus pronominal marking is borne out: When dealing with non-speech act participants, we tend to find lexical subjects at the starts of episodes (as in B, D, E, G, H, K, L and M).⁵

There are also some predictions that are contradicted by the data, however. On the one hand, we also find lexically marked subjects that occur within episodes,⁶

The existence of such counterexamples to the Principle of Morphosyntactic Markedness suggests that markedness oppositions in systems of participant tracking cannot be taken at face value when we are dealing with their distribution in narrative. Specifically, we argue that the marked forms have different functions that are compatible with their use inside the structure of episodes. Take, for instance, what happens between lines 23 and 26 in episode C, reproduced in (6a–d) below.

In this excerpt, there is a switch to a new lexical subject nhunha ama ‘other people’ in (6b), and a switch to a focal 1st person plural subject antyampa in (6d), all of this within the episode. Such participant switches are similar to the episode-delimiting ones in that they all signal distinct units of action within the narrative – as mentioned in the introduction to this section, this was precisely one of the instances for which there was some doubt about the delineation of the episode. In this case, the switch in (6b) introduces the movement and return of another set of deported people (in an administrative procedure before settling them at the reserve), and the switch in (6d) introduces the movement and return of the Lamalama people. What makes the switches discussed here qualitatively different from undoubted episode-delimiting switches, however, is that the units of action they introduce are thematically linked to each other in the function of the larger episode in which they occur. In this case, for instance, the switches introduce the successive moves of the Lamalama people and a group of other deportees, which together form an episode defined by the general goal of resettling people in the Aboriginal reserve. This is also why we chose to group the whole set of clauses in 21–30 as one single episode consisting of distinct subunits, rather than as distinct episodes. The same applies to an episode like E, where waypala ‘whitefella’ in 38 and omoro athuna ‘my father’ in 41 are episode-internal lexical subjects. Again, these uses delimit closely linked units of action (request, refusal, and effect) that together form a larger unit that we chose as our basic episode.

There is linguistic evidence for the relevance of this analysis, in that pronominal equivalents of these situations use a different position for episode-internal and episode-delimiting participant switches – the focal rather than the neutral position. As already mentioned, the clause-initial focus position is associated with participants that are prominent in the local development of discourse, typically in establishing an interclausal relation between two subsequent clauses, such as the relation of contrast between (6c) and (6d). It is such interclausal relations that build the thematic links between subsequent units of action to cluster them into an episode. The same can be observed in the episode-internal pronoun switch in line 4 in episode B, repeated in (7a–c) below. The switch from the policeman in (7a–b) to the focal 1st person singular in (7c) introduces a distinct unit of action. The policeman, representing the white officials' goal of removing people, is contrasted with the Lamalama people, whose goal is to stay on their home country. Again, however, these two units of action are thematically linked in the larger build-up of the episode, in dramatizing a conflict between the contrasting goals of the white officials and the Aboriginal people. And again, there is a focal, clause-initial pronoun to mark the contrast between the two units.

To sum up, we can say that episode boundaries are associated with switches in subject participants, formally realized either by lexical marking of the subject in the case of non-speech act participants, or by pronominal marking of the subject in the case of speech act participants. More generally, switches in subject participants signal the start of a distinct unit of action, which means that they can also occur within episodes. In such cases, however, the units of action separated by the participant switch are thematically related, for instance in a relation of contrast, and cluster together as related subunits of one single episode. In the case of pronominal marking, these episode-internal switches can also be formally distinguished from episode-introducing switches because they use the clause-initial focus position for pronouns instead of the standard postverbal position.

From a broader perspective, the apparent exceptions to the Principle of Morphosyntactic Markedness examined here also suggest some refinements to the principle. Markedness oppositions in the system of participant tracking cannot be taken at face value, because we also need to take into account their function before we can make predictions about their distribution in narrative texts. Unlike the opposition between lexical and pronominal marking for subjects, the distinction between focal and non-focal pronouns has a specific function to establish local semantic links between units of action, which explains why it is associated with the middle of episodes rather than their boundaries. If we look at our analysis in terms of the literature available on episode structure in Australian Aboriginal narratives, our observations can most easily be integrated into the Expected Actor Principle formulated by McGregor 1992, 1998 for episode structure in Gooniyandi. According to this principle, “[t]he episode protagonist is – once it has been established – the expected (and unmarked) Actor of each main narrative clause; any other Actor is unexpected” (McGregor 1998:516). This principle explains why episode structure hinges on subject marking. On the one hand, subjects are the only possible locus for Actor roles in Umpithamu,⁷

Case marking in Umpithamu

In Umpithamu, the grammatical roles of participants are marked with suffixes, in an ergative-absolutive system for lexical items and a nominative-accusative system for pronouns. Thus, lexical transitive subjects can receive the ergative marker –mpal,⁸

While the split between an ergative system for lexical items and an accusative system for pronouns is typical for many Australian languages, Umpithamu has an additional option for case marking that is much rarer, and that is exploited with rhetorical effect in the narrative studied here. When referring to human participants undergoing actions, Umpithamu allows a choice between the standard accusative pronoun, as with inangku in (10) – and similar structures in (1) and (3) above – and a genitive pronoun, as with ulangana in (11) (a genitive pronoun precludes presence of a nominative pronoun in the same structure; see note 5). This type of pronoun is labeled “genitive” because its primary function is to mark possession with nouns, as in the structure in (12).

The semantic effect of the use of genitive pronouns for human direct objects is to mark an experience of extreme lack of control on the part of the person affected by the action (McGregor 1999, who describes a comparable construction in Nyulnyulan languages, argues for an analysis in terms of actions “befalling” people rather than people “being affected” by them). There is evidence for this semantic effect from at least three sources: (i) the type of objects that typically receive genitive marking, (ii) the type of subjects that typically trigger genitive marking for human direct objects, and (iii) the rhetorical exploitation of the accusative/genitive contrast in discourse. The first factors will be discussed in this section, and the last one will be discussed in a following one, using the narrative studied here to demonstrate rhetorical uses of the genitive/accusative distinction. There is additional evidence for the analysis with which we will not deal in this context, most importantly the grammatical similarity of the genitive construction to experiencer constructions describing physiological and psychological states or events.

In the majority of cases, the distinction between genitive and accusative is a matter of choice for the narrator and relates to the perspective the narrator wishes to impose on the events. There is a small number of constructions, however, that have obligatory genitive marking, and therefore provide an interesting perspective on the semantics of the genitive pronoun. The first factor that can make the genitive obligatory relates to the type of human direct object used in the construction. If the human direct object is inherently without control, the genitive will be used. Some typical examples are a dead body being handled for burial, as in (11) above, or a baby being handled for care, as in (13) below. In both cases, we are dealing with human referents undergoing an action, but unlike the typical human referent, these are inherently without any control over the action, and accordingly also receive genitive marking.

Another factor that can make the genitive obligatory relates to the type of subject used in the construction. If the subject refers to an inanimate entity affecting a human direct object, the genitive is obligatory for the direct object. A typical example is in (14) below, with the inanimate entity yuma ‘fire’ acting on the human direct object, which receives obligatory genitive marking. Moreover, if the subject refers to a lower animate entity or a supernatural being, genitive marking is typical – though not obligatory – for the human direct object, as in (15) and (16) below, with ewen ‘mosquitoes’ and wooperri ‘Stories’ (mythological beings) representing the lower animate and supernatural categories, respectively.

Taken together, the types of subjects and objects that can trigger obligatory genitive marking provide good evidence for the semantics of lack of control associated with the genitive. The evidence from objects is most direct: Human direct objects inherently without control, illustrated in (11) and (13), always receive genitive marking. The evidence from subjects is more indirect, but points the same way. What is special about inanimates and lower animates, illustrated in (14) and (15), is that they are unexpected as agents of actions (compare Wierzbicka 1980). Unlike higher animates, inanimates and lower animates are not typically monitored as potential agents, and are therefore associated with actions that take the human object by surprise and are thus outside his or her control. Supernatural beings, by contrast, illustrated in (16), are expected agents, and are therefore typically monitored as potential agents, but what distinguishes them from normal animate agents is that their actions are inescapable and therefore inherently uncontrollable. In this sense, the use of genitive marking in constructions with inanimate, lower animate, or supernatural subjects indirectly confirms the semantics of lack of control associated with the use of genitive marking. For all other subjects – that is, the higher animate range, which is the default type of subject – use of genitive or accusative is a matter of choice, and therefore reveals the perspective the narrator wishes to impose on a particular event.

Genitive pronouns and white-Aboriginal confrontation

A third piece of evidence for the semantic effect of genitive marking is its rhetorical exploitation in actual discourse. The narrative studied here is particularly interesting in this respect because it uses the genitive marker more frequently than normal texts do, and also exploits it strategically to highlight the dehumanizing effect of the white officials' actions on the Aborigines.

If we look at the quantitative profile of genitive marking in our reference corpus of 30 texts, genitives for direct objects are an extremely rare feature overall, as shown in Table 5. Within the reference corpus, genitives constitute a little less than 10% of all pronominally marked direct objects.

Frequency of accusative and genitive pronouns for direct objects.

From this perspective, the use of the genitive marker is unusually high in this particular text. Out of the seven pronominally marked direct objects found in the whole text, four are genitive-marked. This is not only more than the number of accusative pronouns in the same text, but it is also one-fifth of the total number of uses of genitive markers in the whole reference corpus, which suggests that the profile of this text deviates strongly from the normal profile.

If we look at the distribution of genitive pronouns within the text, moreover, it is obvious that the narrator exploits the distinction between accusative and genitive objects for rhetorical effect. All genitive pronouns are used to refer to the Aboriginal people in constructions where they undergo actions by white officials or their associates. This is the case in lines 1, 9, 21 and 23, repeated below as (17)–(19).

All three instances are from the complication episodes describing the deportation to Bamaga. Given the semantics of lack of control associated with the genitive pronoun, the use of genitives to refer to Aboriginal people undergoing deportation can be regarded as a rhetorical choice by the narrator. Instead of construing the actions with accusative pronouns – that is, with normal involvement for the human direct object – the narrator chooses to represent the actions as involving less than normal control. By doing this, she represents the actions of taking Aboriginal people away in (14), depositing them in a reserve in (16), and the ruse of medically examining them in (15) as actions that take place completely outside any control of the human beings involved in them.

In this sense, the text does not just provide an account of what happened at the deportation, and what happened subsequently to get the family back to their home country; its use of case marking also contains a subtle way to highlight the actions of the white officials, and to construe the confrontation between white officials and Aboriginal people as one involving less then normal control for the Aboriginal people involved.