Methodology
Chapter 4: Theoretical Framework
5.6. Methods of motif and scene analysis
Two statistical methods of analysis were applied to the motif and scene data of this research. These methods were employed to compliment the formal archaeological analysis (discussed above); which provided counts of certain attributes of the motifs and scenes, such as human figure form or material culture types. This section details the multivariate analysis methods used to investigate the data produced from this formal analysis and investigate trends and groups within this data.
In studies of rock art, quantitative research and analysis is often employed in response to the concerns associated with directly interpreting rock art and the perceived understanding of objectivity associated with quantitative methods (Conkey 2001:280; Ross 2003:98). Conkey argued that quantitative studies have been useful for empirically challenging long held notions in rock art and archaeological discourses, e.g., hunter gatherers mostly depicted animals they ate (Conkey 2001:280). However, as the data collected and generated by researchers is influenced by their decisions, methodology and research agenda one must as be critical of quantitative results as much as the interpretation of a rock art scene (Ross 2008:99). This research employed two multivariate methods of analysis which have proved useful for comparing large amounts of data with numerous variables or attributes (see McDonald 2008; Ross 2003:101-103; Taçon 1989; Taçon et al. 1996; Travers 2015; Wilson 1998,2004). Multivariate analysis has also been used to illustrate the relationships (trends) between recorded variables in rock art data (Franklin 2004:33). The trends observed within Dynamic Figure art were used to answer aspects of each of the research questions.
5.6.1 Metric motif analysis
The metric motif data consists of the measurements recorded for each individual motif from the first section of the motif recording form. This metric data was analysed using GenStat (v18.1.0.17005) software which performed multiple regression analysis. Multiple regression analysis is a ‘method of describing the relationship’ between multiple metric variables (Payne 2015:3,17). In relation to this study, I have used this analysis to examine if a meaningful relationship exists between the various measurements of Dynamic Figure human figure motifs. Or, do a motif’s arms increase proportionally to its body and is this relationship relatively consistent across the
I foreshadowed that the metric motif analysis would not be particularly insightful to isolate Dynamic Figure types which could be developed into a Dynamic Figure chronology in an early study (see Johnston et al. 2017). This is because Dynamic Figure artists specifically used size as part of their narrative constructions within scenes, e.g. some motifs are bigger than others in one scene (see Johnston et al. 2017). This precludes motifs grouped be size being a meaningful type, despite Chaloupka’s (1993:106) use of size in his Dynamic Figure chronology (see Section 3.5; Johnston et al. 2017).
I determined that an insightful investigation was to examine the relationship between the proportional size of motifs and compare this to the proportional size of the headdresses they wore. Determining if headdresses were not painted proportionally larger under the same relationship to increasing body or arm length supports the contention that artists intended to paint specific headdress types on certain motifs. In short, I examined if big motifs have the biggest headdresses or is headdress size more influenced by the intended ritual messages of the artists. Other material culture objects were not recorded frequently enough to examine in similar manner.
5.6.2 Motif and scene analysis
The final analysis technique employed was correspondence analysis (CA) also using the GenStat (v18.1.0.17005) software. Correspondence analysis applies a chi-squared test to measure the distance between variables or, in this thesis, how often variables occur together (Wilson 2004:176; see also Harding and Payne 2015; Ross 2003:101-103). When multiple variables are tested this technique is called multiple correspondence analysis (MCA). After the software applies a chi-squared algorithm it displays points on XY graph were the proximity between points represents the relationship between attributes. Closely plotted attributes on the graph indicates they were more often recorded in the same motif or scene, distance between attributes indicates the opposite. MCA or CA analysis was applied to the scene and motif data, where a meaningful relationship could be determined. For example, testing whether a scene depicting sex and female motifs group together would be an unmeaningful plot, as females made up a small percentage of the overall sample and for a scene to be identified as depicting sex it needed a female motif. A more useful test would be if a certain therianthrope type more often had a specific weapon type, spears or boomerangs, as this trend was observed by Taçon and Chippindale (2001a). The value of this CA is that it could show that even though macropod headed therianthropes are more likely to have boomerangs, all
therianthropes are more likely to have boomerangs; therefore, demonstrating that although these variables group together this result is not overly significant. This type of analysis replicated studies of rock art from other regions of Australia (McDonald 2008; Ross 2003; Taçon et al. 1996; Travers 2015).
Correspondence analysis was employed to investigate two primary questions: 1. Do relationships exist between attributes of Dynamic Figure motifs?
2. Do clusters and groups exist within the data? Specifically, are there distinct types of Dynamic Figure human figures represented in the Jabiluka data?
The correspondence analysis provided insights into patterns within Jabiluka Dynamic Figure art and was valuable for examining some of the ritual practice indicators such as formalism and style.