by Martyn Harris
Co-authored with Tomi S. Melka.
Published in Journal of Quantitative Linguistics Volume 18, Issue 3, 2011
In Harris and Melka (2011a), we identified a fragment of text on the ‘‘Mamari’’ tablet (Cb) which echoes a short... more
In Harris and Melka (2011a), we identified a fragment of text on the ‘‘Mamari’’ tablet (Cb) which echoes a short section of a verbal list-like formula recorded by Englert (1948) concerning the kills made by warriors during battle. We highlighted the possible
correlation between ‘‘frigate bird’’ /600/ and ‘‘fish’’ /700/ glyphs and their presence in delimiter lists (Barthel, 1958; Horley, 2007) through an analysis based on the whole RR corpus, and a reduced corpus of texts, which we propose relate to the same literary genre; the ika and timo genres (see Routledge, 1919; Fischer, 1997). We presented the ethnographic data (Harris & Melka, 2011a), a Key Word In Context concordance, and explained the methods adopted for a further statistical analysis of these glyphs as a concordance is not normally used to demonstrate significance, without some statistical measure to support it.
In this paper, we present the final results, and demonstrate that there is some evidence to assume that these two glyphs /600/ and /700/ are highly correlated in other fragments of text across the RR corpus, as either separate constituents /600/ and /700/, with one or more intervening glyphs, or as part of a compound form /605/, or /606/, requiring the ‘‘hand’’ glyph /006/ to appear before the ‘‘fish’’ glyph /700/.
Furthermore, we propose that these glyphs suggest the presence of ika lists – a record of warriors killed in battle – and timo – vengeance chants imbued with the power to bring death upon the named victim. What these inscription types share are the presence of proper names, which may be indexed by glyphs /430/ and /530/ previously identified by Davletshin (2002), as possible ‘‘title’’ glyphs.
Although this study highlights many observations made by previous research (Pozdniakov, 1996; Guy, 1998; Davletshin, 2002; Horley, 2007) through palaeographic methods and frequency analysis, this paper attempts to take this further by adopting multivariate methods from lexicographic and latent semantic statistical methods. These methods are useful for reducing noise created by an inadequate transliteration scheme (Barthel, 1958), and issues associated with frequency analyses over a corpus of texts of variable length. The results in this paper also show these methods correspond to observations made by previous analyses based on palaeographic methods.
Download (.pdf) 
