The results of the network analysis for all houses
The above discussion of the network graphs suggests several possible conclusions. First, the network analysis for all houses suggests that the marble trends occurred both within specific (see Laurence 1994, 34-35, for definitions of neighbours and neighbourhoods in Roman cities, including Pompeii) and across the city. However, this datum is not entirely unexpected because social contact between the inhabitants of Pompeii is likely to have taken place at both local and city-wide levels. The absence of strong links between many directly neighbouring houses shows that the distance between houses did not always affect trends in marble use. On the other hand, we do see strong links based on geography (at least within individual Regiones). The prominent similarity between pavements within atria with impluvia shown in our network analysis of five houses from Regio I, all located along the same street, highlights the seeming importance of geographical proximity in influencing the spread of marble trends. Therefore, geographical proximity may have been a key factor in the social
- 1.06.07
- 1.07.03
- 1.08.08
- 1.09.01
Figure 5.14 The graph shows atrium impluvia network with nodes arranged
according to their geographic position. Link width and colour according to Brainerd-Robinson similarity. The darker the shade of grey and the greater width, the higher the similarity value. Links representing higher similarity are drawn on top of links with lower similarity. Node size is scaled to house size. Node colour represents the period: light grey represents Third Style and dark grey represents Fourth Style.
distribution of marble tastes. Our Regio I network also shows trends operating at specific social levels - the other four houses were all of similar size and their pavements had similar marble volumes as seen in the case of house at 1.07.03, which is the smallest house within the network with the smallest impluvium pavement.
There also emerges a strong correspondence between house size and the types of marbles employed in pavements. For instance, houses VII.16.17 and 1.07.01 are both large houses, a factor which seems to suggest, and thus confirm, a top-down influence. At the same time, houses 1.07.01 and 1.09.01 are strongly linked to house 1.13.09. It is possible that these links signify a local trend in Regio I with house 1.13.09 drawing trends from the other two houses (Figure 5.16). The strength of the links between houses VII.16.17, 1.07.01 and 1.08.08 indicates that these houses were important interaction points in the transmission of marble trends within our network (Figure 5.15, Figure 5.16). The important factor for house I.O8.O8 seems to be the total volume of marble on display within the premises, while the overall size is perhaps the most important factor for houses VII.16.17 and 1.07.01.14
Figure 5.15 The graph shows the network for all houses with attribute mappings to visual variables. Link width and colour according to Brainerd-Robinson similarity. Only the links with the highest and lowest similarity value are shown. Links representing higher similarity are drawn on top of links with lower similarity. Node size is scaled to house size. Node colour represents period: light grey represents Third Style and dark grey represents Fourth Style.
Therefore, if the total volume of the marble (surface area) constitutes an important factor, meaning that it would have had a higher visual impact on the viewers and visitors of the premises, this must have been significant in transmitting the marble trends. Indeed, house 1.08.08 was also commercially active, with a bar counter selling drink and food, thus more accessible to the public than other houses - an aspect that makes us wonder whether it would have been crucial in aiding the spread of marble. Similarly, public rooms, like triclinia and tablina within elite houses, but also of other business premises, such as fulleries, which included marble decoration, further supports the idea of marble spreading through visibility.
Lastly, several houses present strong links, however most of the Fourth Style houses only have weak links to those of the previous period (Figure 5.16). The lack of very strong links between houses across different periods suggests that marble trends changed significantly between periods. This datum might be due to the adoption of new marble types in the Fourth
Figure 5.16 Network layout by Stress Minimization on Brainerd-Robinson distance for all houses. Link width and colour according to Brainerd-Robinson similarity. Only the links with the highest and lowest similarity value are shown. Node size is scaled to house size. Node colour represents period: light grey represents Third Style and dark grey represents Fourth Style.
Style - a hypothesis that is supported by archaeological evidence from Pompeii where recent studies have shown that newly imported materials made their debut on a large scale in the Julio-Claudian period (Fant et al. 2013; cf. Fant 2007, 340-343). Nevertheless, the examples that do share strong links between periods suggest that houses from these two periods were influential in spreading marble trends. The impetus for these changes may well have come from Rome; however, we should not discount the idea that house owners of Pompeii having seen and interacted with earlier pavements sought to surpass existing elite levels by having a clear break with previous trends. Unfortunately, the introduction of new varieties of marbles is something for which our network analysis was unable to account. Interestingly, the network graph layout by Stress Minimization on Brainerd-Robinson distance points towards links between houses 1.08.08, VII.l6.17 and 1.07.01 and suggests that these houses were important interaction points in the transmission of marble trends within our network. It is possible that these houses with late Third Style and early Fourth Style pavements record the transition of marble tastes between these periods. Perhaps, the owners were trying to keep ahead of previous elite standards - a hypothesis that we aim to test as we continue to examine and record pavements from Pompeii and Herculaneum.
Summing up, if we examine the houses that act as ‘links’, a number of factors seem to be important: (1) the total volume of marble; (2) the size of the house; (3) the location of the marble within the premises; and (4) geography. Particularly strong links between date, size and geography were found when looking at a specific pavement type in atria with impluvia in Regio I. The examination of a single pavement type emphasises the importance of geographical influence, especially between neighbouring houses, all located on the same street. This final observation is worth exploring with a larger and more complete dataset.
