At one dwelling the SIHF listed 12 'usual residents', five of whom had been at community B staying with a relative at the time of the enumeration there. They were listed on the SIHF of that relative's dwelling as visitors, and enumerated at community B.^[12]These people were still at community B when community A was enumerated, and so did not have SIPFs done at community A. A sixth 'usual resident' of the dwelling (person 10) had gone to community B to join the other five, and then left for a funeral ceremony elsewhere before the enumeration at community B. He was, nevertheless, enumerated at community B in absentia, because he was expected to return there to join his other 'visitor' relatives. However, he returned directly to community A and was enumerated there, as a 'usual resident'. He was not asked if he had been enumerated elsewhere, but even if he had been- 51 - asked he would not have known that he had already been counted at community B with the other five 'visitors' from community A.

Although this was an isolated case of double counting, it resulted from factors endemic to the context of the enumeration of the region. Other cases of the kind almost certainly occurred. The funeral mentioned above attracted many people from the region, who were then, of course, absent from their usual places of residence—or from the place where they were recognised as 'visitors'—for the duration. It would have been quite inappropriate to conduct the enumeration during the ceremony, and so this was not attempted: the community where it took place was enumerated after the ceremony was over. The enumeration at community B occurred while the ceremony was on, and anyone who was known to be at the ceremony, who would normally have been at community B (as a resident or as a visitor), was enumerated at there 'as if' they were there. Thus anyone who, like person 10, subsequently left the ceremony and went to an as yet unenumerated outstation instead of returning to community B was a potential candidate for double counting. There appears to be no perfect solution to this conundrum. Had the people attending the ceremony not been enumerated in absentia, those who returned to community B would not have been enumerated at all, because the count at community B had been completed before the ceremony ended.

Funeral ceremonies are, alas, all too common in the region. During the two months I spent at community A in 2001 there were two funerals in the community itself. Preparations for two other major funerals at nearby outstations were under way when I left. Funerals and other types of ceremony are a major cause of unpredictable short-term population mobility in the region: some people are almost continuously on the move from one ceremony to another, and some ceremonies attract large numbers of people for short periods from across the region. No dwelling-based enumeration strategy can ever be devised that will fully counteract the effects of this specific type of intra-regional mobility, because it would never be appropriate to conduct the census enumeration at a community during a ceremony. The practice of enumerating ceremonial attendees as if they were in their usual place of residence is a departure from the principle of enumeration in situ, but it is better than failing to enumerate them at all.

Person 1 from a particular dwelling in community A was at community A when the settlement B enumeration took place. However, she subsequently went to settlement B for a short visit, and was still there when community A was enumerated. She was therefore not enumerated either at community A or at settlement B. Note, however, that although no SIPF was done for her she did appear as person 1 on the SIHF for her dwelling at community A.

I suspect that this was not the only case of this kind in the region, despite the best efforts of the CFO and the CCs and enumerators. The residents of an entire dwelling at community A would have fallen through this same crack had they not serendipitously returned to- 52 - the community while the enumeration was in progress. Cases of double counting (other than ones of the 'ceremony effect' type detailed above), where someone happened to be present for the enumeration in two different places at two different times, might also have occurred, but this is less likely because 'visitors' were usually asked if they had already been enumerated elsewhere. However, at community A at least, 'usual residents' were not asked if they had been enumerated elsewhere as a visitor.

It would be possible, in theory, to assess the extent of under-enumeration and double counting by comparing the SIHFs for the whole region (i.e. settlement B and its associated outstations), looking for people who were listed as 'usual residents' but were absent, and then searching to see if they appear as enumerated 'visitors' at some other community. It would also be possible to check whether 'visitors' in one place had also been counted at the place where they were 'usual residents'. But this could not be done efficiently without the help of the local enumerators. They alone would be able to narrow down the range of possible alternative locations for such individuals: people's short-term movements are never random, but are constrained by considerations of kinship and attachment to place. People might also be under different names on different SIHFs, or their name might be spelled differently. Again, only someone with local knowledge would be able to make correct cross-identifications in such cases.

It is important to note, however, that this type of cross-checking is only possible if:

The count at community A approached this ideal, as did the count at settlement B and in the region generally.

A 'usual residence' count, where everyone was enumerated on a SIPF according to their usual place of residence, whether or not they were actually there at the time of the enumeration, would be an alternative, but less satisfactory, strategy. The quality of the data on the SIPFs would suffer, since many more SIPFs would be filled out in the absence of the individual concerned. Moreover, visitors who were actually present would not be enumerated, on the assumption that they were being enumerated at their usual residence, in absentia (see Sanders, this volume). Such a strategy is therefore also much more dependent on the robustness of the definition of 'usual resident'.

The worst-case scenario would be one where these two strategies were mixed within a single region. This would result in a high incidence of both double counting and under- counting, and it would be well nigh impossible to cross-check the data.

Whichever strategy is adopted, the SIHF is a crucial tool. Because it lists both absent usual residents and visitors, it provides a snapshot of the level of inter-community mobility, and it provides a means of checking for double counting and under-counting. However its usefulness is dependent upon two definitions: that of the 'dwelling' and that of 'usual resident'.

In small cohesive communities such as community A, the definition of 'usual resident' is unproblematic, although the local definition does not correspond to the ABS definition. For one thing, as we have seen, it is the community rather than the particular dwelling within it that people have in mind as the 'place' where they reside. The ABS definition of 'usual resident' is also based in the notion of calendar time, and this does not apply to the local definition, at least not in such an absolute sense.^[13] For example, some of the absent 'usual residents' of one dwelling at community A had been at community B more or less continuously for over a year, because one of their number was doing a year-long course at the local Health Centre. Nevertheless these people consider themselves, and are considered by others (including the enumerators at both A and B) to be 'usual residents' of community A and 'visitors' at community B, because community A is the locus of their most significant personal and spiritual attachments, and is the place they intend to live in the long term.

However, even within this region, and more so in other areas of Australia, the notion of 'usual resident' has been, and most probably continues to be, problematic. If the census is viewed strictly as a snapshot view, the data contained on the SIHFs accurately reflect the general situation: Indigenous people are highly mobile, and at any one time many will be 'visitors' at places other than their 'usual residence'. But it would be unwise to use the census data as the starting point for a more fine-grained analysis of attachment to place and patterns of mobility.

The case of the absent tent-dwellers from the yard of one dwelling highlights the other definitional problem alluded to above. These people were, for the moment, 'visitors' at community A, on an extended visit to their daughter, but happened to be absent (at community B) at the time of the enumeration at community A. Because their tent was a 'non-house' with no occupants who were usual residents of the community it was not designated as a vacant dwelling, and so it had no SIHF. Nor was it counted as part of dwelling that it was pitched next to. It is highly unlikely that these people were enumerated on SIPFs elsewhere. It is also possible that they did not appear on a SIHF anywhere. They have a permanent dwelling at another outstation, but that dwelling was currently vacant.

At the time of the enumeration, this happened to be the only tent in the community, but had the enumeration taken place in the following week, there would have been several more. Many 'usual residents' had returned by that time, and most houses in the community are too small to house their full complement of inhabitants comfortably. Until recently people got round this by sleeping on the veranda (if there was one), or out in the open. But tents have become very popular locally over the last two or three years, and are easily obtainable at affordable prices in the local town. They are now regularly used as extra bedrooms if a house becomes overcrowded—either temporarily (as when visitors arrive for ceremonies) or in the medium term, when the full complement of usual residents is present.

Functionally, these tents are extensions of a permanent dwelling, and their occupants are, more often than not, 'usual residents' of the dwelling and members of the dwelling's household. The difference is only in the eye of the non-Indigenous beholder, for whom a tent is a 'temporary dwelling' rather than an 'extra bedroom'.^[14]

Had I not been present, a significant number of 'person 1s' at the community would not have been enumerated on a SIPF. According to the CFO this did not happen anywhere else in the region, but it is worthy of discussion because it highlights a particularly salient point of difference between local Indigenous and mainstream social structure and cultural values. I did not notice the systematic omission immediately, because the first two dwellings we visited on the first day did not result in complete enumerations. At the third dwelling, however, I noticed that no SIPF was completed for person 1, in this case the oldest member of the community.

I asked the enumerators if they were going to do a SIPF for this man, thinking that they had just overlooked it because he was inside the house. To my surprise, they said that they did not think it was needed. I decided at this point that I would be doing everyone a favour if I temporarily abandoned my 'observer' status. (And besides I was intrigued: what could the reason be?) So I asked why. The answer was that it was not necessary for person 1 in each household to have a SIPF done, because their kinship relationship to everyone else in the household was already specified on the SIHF. Thus at this stage they were thinking of the SIHF as person 1's SIPF.

What led them to this conclusion? The reasons must have been powerful, for they produced what might be called a culturally induced 'blind spot': they overrode the principles which had been inculcated during training—principles which the enumerators had obviously otherwise taken on board. And indeed it took more than one conversation to convince them that SIPFs were needed for these individuals.

There were several factors that conspired to produce this 'mistake'. Question 4 on the SIPF asks for information that has already been recorded on the SIHF. So indeed do Q. 1, Q. 2, and Q. 3—but with an important difference. For local people, the most salient question on the entire SIPF is Q. 4, because kinship is the central organising principle of their society. The misapprehension that person 1 does not need a SIPF is reinforced by the design of Q. 4 (see Appendix C). It asks in big bold letters how the person is related to person 1. The little box for marking if the person is person 1 needed to have explicit instructions beside it, in lettering the same size as the main question.

For the enumerators, Q. 4 simply did not apply to person 1. This fact overrode all else, and their view was reinforced by the design of the SIPF. If Q. 4 had been placed further down the SIPF, it would not have acted as a brake to further questioning of person 1; placing Q. 4 in such a prominent position carried the implication that the information it elicited was as significant to the author of the form (i.e. the ABS) as it is to local people, which is to say that, by comparison, the other information elicited by the SIPF is relatively unimportant.

As will be evident from the general thrust of this paper, the advantages of employing local enumerators in Indigenous communities far outweigh the disadvantages, and no criticism of the enumerators is implied here. But Indigenous enumerators are being asked to do something very difficult, to deploy the skills and knowledge that they have acquired as members of their community, while simultaneously assuming the position of temporary 'outsiders'—agents of the state as it were. The designers of the SIPF need to be more aware of this delicate balance. Kinship is not as important in the 'mainstream' society as it is for Indigenous people. It is certainly not so important that it renders other information irrelevant. If questions on family relationships are to be asked, they should be in a less prominent place; or they should be radically rethought. This issue receives further discussion in the next section.