Figure 1: (after Brandl 1961: figure 57) (after Baglin 1976:4)

I use 'figure' when speaking of the actual marks on the rock.

A 'group' has a number of images and their basis for membership may not depend upon their appearance. They could be from the same site and thus be location specific or from the one book and include images that are drawn from all over the world.

A 'genre' is an assemblage of similar images. The members of a genre look alike without any deep analysis. The following is a demonstration of the sense in which I use it.

I regard the following images as being in one genre:

Figure 2: (after Anati 1977: figure 6)

I show a relatively simple example but it is nonetheless complex when one needs to explain it. It is a process that is done quickly and with surprising accuracy by visual perception.

Aspects that unify this array into a genre are:

Figure 3.

The head-dress on 4, the hands shown on 1, 4, 6, 10, 14 and 15, the bits of leg that are not shown on 4, 6, and 10 do not strike the eye as being reasons for their exclusion. The right facing 12 and shaped body of 7 might be regarded as reasons for splitting this genre if there were more of them. The selection of images for this genre is simplified by five of the aspects being common to all images. It was selected for this purpose. It simplifies the clustering but is not essential. The essential element is that the images have enough in common to look alike. No aspect is essential and the possession of a unique aspect is not grounds for an image's exclusion.

Such analysis is not a conscious part of the selection of genres. Groups of figures could be split using the above type of analysis but it would be a long and tedious process. It is not necessary as the eye and the brain can do virtually the same thing at a subconscious level.

'Form of depiction' is used to signify the method employed to convey information through a figure. Examples of different forms of depiction could be painted figures, pecked figures, engraved figures, carved figures, imprint figures or stencilled figures.

'Landscape' is the sum total of the area in which the figure exists. It includes the immediate environment but is not restricted to it. Landscape includes the materials available for making figures on rock walls and it is in this sense that I will mostly be using it.

A 'model' is used in the sense of shaping an argument or a contention. A model's validity is always evidence dependent.

'Enhanced stick figures' are mostly thin, as are most stick figures, but there is frequently a depiction of muscle and other physical features. They could not be described as having a realistic shape but they appear to be heading in that direction.

Within the statistical element of this thesis the terms object/image should be regarded as interchangeable, as should attribute/variable/state.

The ground rules

An aim was to make no assumptions. This may be impossible, but the attempt can be made within a restricted field. We will see what can be done with what remains.

My sample consists of images of !people. The majority of the Bradshaw art is of !people. All archaeology—like most other disciplines—works on samples, because it must.

The restriction was a logical one. !people, that is pictures that look to the present observer like pictures of people, were likely to include pictures that actually were of people. There was no way of determining what any ancient picture meant to its makers or users in the absence of a gloss. Pictures of people are good to study. What can be more important to people than people? They are the stuff of human myths and narratives. Moreover there are cross-culturally plenty of such figures, whereas other sorts of pictures are culturally limited. There are no eland pictures in Australia nor kangaroo pictures in Europe.

I ignored a number of aspects of rock art that have been regarded as central by those who made the recordings I use. They may be central to its recording, but they are not central to its relation with other forms or genres.

The selection of aspects may cause controversy, for it was impossible to maintain complete consistency. An attempt was made to limit each image to only one person. Holding hands was not considered a close enough association to treat the two figures as a single image. Separate images were made. A child being carried papoose style was regarded as a part of the adult image. If a person carried an implement and such implement was regarded as part of that person's representation, then it was included. Making these decisions was harder than it would seem as various bits and pieces were ever closer or further away. A !person's clothing was always included but a bird apparently hovering overhead presented a more difficult problem. These aspects were included on a subjective basis.

Constant attempts were made to minimise interpretation. No attempt was made to interpret a scene—the subject of this study was the way the individual figure was depicted.

The age of a depiction was ignored. There are relatively few images for which a firm date exists. To have used only these would have restricted the selection of images to the point where the exercise would have been meaningless. The age of a figure was not used as a factor in this analysis.

Size was not regarded as a significant factor. To include it would have been to make a value judgement as to the meaning of size. Was bigger better or more important? Was the height of the head a more relevant factor? If these questions could be answered for all cultures for today would the same understandings have applied 10,000 years ago? There was no way of knowing so size was standardised.

Should the colour of a depiction be a factor? It would be an easy distinction to make if there were only three colours or three hundred colours and all were equally available at all sites. One would also need to assume that the colours had not changed over time. Such is not the case and on these grounds colour was not considered.

Allied to the previous factor was whether a distinction should be made between engravings, carvings, stencils, prints and paintings or any combination of these. Selection may have been a matter of choice by the artists, or it could have been imposed on the artists by their culture and so should be included. Possibly no choice was involved and the landscape was the deciding factor (a lack of sheltered sites could lead to engravings and carvings being favoured). No distinction was made on these grounds in this analysis.

Sexual differentiation was frequently contentious. I have taken the view that the only sure way of determining sex is by sexual anatomy. The images were classified as being sexed or unsexed. No distinction was made between female and male.

Finally and most importantly, the pictures were examined as they were imaged. There were no assumptions that because they are now contour drawings (a contour drawing is a drawing that is depicted only by an outline) that an infill colour had weathered away or had been missed by the recent copier. A gap in a depiction was not interpreted as anything other than a gap. The same system was applied to all aspects that were missing from images but could have been expected to have been shown. Some figures may have been processed in an incomplete form. There was no other way to do it.

There are some obvious exclusions that require explanation. Cape York in Queensland is an area that immediately comes to mind. The preservation characteristics of the rock in this area rule out the survival of much of the older art although some does exist and would have been included but for the lack of recordings. The older form is believed to be a contender for affinity with the Bradshaws.¹ This contention could not be tested but was supported by Judith Ryan (1993:14), curator of Aboriginal Art at the National Gallery of Victoria. She stated of the Bradshaws:

This archaic art form appears to be the westernmost variant of a style which exists in parts of Arnhem Land and Queensland.

Papua New Guinea was a more significant area that has been excluded. I was unable to obtain images of !people from this area. I noticed that Meredith Wilson (1994), in her research on Pacific rock art, struck the same problem.

The aspects that are appropriate to use in a world wide comparison of rock art are described in a later section of this paper that deals with the selection of attributes.

THE APPROACHES

A strict procedure was devised to seek genres that were similar enough to the Bradshaws for archaeologists to postulate a relationship of origin between them.

Order—Visual

1. A sample of rock art depictions of !people was obtained in an area that is roughly a surround of the Indian Ocean. Some photographs were not scannable successfully as the contrast was so ephemeral as to make a clear image impossible to extract.

2. Images within publications were scanned and distinguished into individual !people and the background removed so that they appeared as black images on white backgrounds.

3. The images were then arranged in geographical units—images that had been derived from the one area. The size of these areas was flexible. Such areas included a landscape of a size small enough to be regarded as a unit but large enough to allow for the geographical extent of trade, ceremony or other contacts. This ensured that there was the possibility of interaction and mutual cultural exchange that could lead to the development of similarities within specific art genres.

4. The group that included all images in a geographical unit was then split into what have been called genres. Each genre tended to contained a polythetic group of attributes that distinguished it from other genres within each geographical unit. The individual images in these genres were used as the basis of a visual comparison with the Bradshaws.

Order—Statistical

1. The next, independent, approach was a Correspondence Analysis on all of the images. A spreadsheet was constructed from the images ordered by genre. This made splitting the results into geographical units for statistical comparison simpler as each genre was a subgroup of a geographical unit.. At this stage no attributes had been selected, named or defined.

2. A number of characters were defined from familiarity with the images that had been acquired during the course of collection, scanning, cutting out, arranging into genres and seeking visual likenesses to the Bradshaws. Attributes were defined in the course of constructing the spreadsheet. At the completion of 117 genres, 79 attributes had been ascribed.

3. A Correspondence Analysis was carried out on the completed spreadsheet of data—79 attributes or states and 2230 objects or images. From the first two columns of the object scores, scattergrams were created of all images together, of the images and of the mean centroid of each geographical unit.

The task was to seek figures outside the Kimberley that look so much like the Bradshaws that they could be related to them. Other geographical areas displayed a variety of genres and were accordingly split into their appropriate groups for this exercise. The Kimberley too had a number of genres but this study was concerned only with the Bradshaws from this area. The Bradshaws do not lend themselves to clustering. They are a uniform progression from highly decorated to not decorated at all. They were therefore treated as one genre.

A single approach could be accused of being biased by an isolated incident or be a method that had been selected to show a desired result. To mitigate this possibility the approach used here had a troika of applications. The search for affinity used two independent methods. This was followed by considering whether the results were possible, even likely.

A description of the Bradshaws was given but space limitations prevents me displaying all but a few of the images here

Any conclusions suggested or reached had to satisfy all of the three approaches.

The first approach—visual.

A visual analysis showed that a number of factors were common to the Bradshaws and to other genres. All the groups were expected to be alike in some way—but some were more alike than others. Groups that did not possess commonality in this visual approach were not contenders as associated genres.

The second approach—statistical.

A statistical analysis was carried out and a result obtained that supported or refuted, and thus tested the previous analysis.

The third requirement—feasibility.

It had to be possible to construct a model to explain the existence of any commonality between images. Then and only was it be possible to make responsible suggestion that these genres are or have been in the past in some way connected.