OBSIDIAN 51 
selected for freshness, thus avoiding possible chemical changes (ie. clay formation) 
inherent in weathering processes. The analyses were carried out by the method of 
Norrish & Hutton (1969), using a Fhillips PW 1410 x-ray spectrometer. Major 
elements (those generally occurring in abundances of >0.1 weight percent (wt %)) 
were analysed using glass discs consisting of sample powder and lanthanum flux. 
Those elements of lower abundance (reported as parts per million, or ppm) were 
analysed on boric acid backed discs of pressed powder. Analysis was run under the 
detection limits and instrumental settings as outlined by Parker (1983). 
Results 
The analytical results are presented in Table 2. The elemental weight percent 
values are also shown in Table 2, and can be used for direct comparison with the values 
presented by Duerden et al (1987, Table 1; see extract in Table 3) which were obtained 
by the PIXE/PIGME method. The relative percent errors between the control 
samples of differing aliquot weights were found to be within those listed in Parker 
(1983). 
Comparison of the elemental weight percent values of the control analyses of this 
study with the corresponding values of Duerdxen et a/ (1987) shows a general lack of 
similarity, except for Ca, Sr, Zr, Y and Mn, and, Ti andSi (for Mayor Island only), 
with Rb from Great Barrier also similar to these workers. Unless gross rock chemical 
inhomogeneity (strangely affecting some elements and not others) can be used to 
explain this discrepancy, it would appear that the differences are due to differences in 
methodology. PIXE/PIGME is more sensitive to sample surface effects than XRF. 
Sample 679/26 from Awana stands out, especially with respect to LOI and H,0- 
which total 7.02 weight percent. This may be explained by weathering processes, as 
volcanic glass (obsidian) is particularly vulnerable to hydration reactions. Thus this is 
the only sample which cannot be considered fresh — generally denoted by low LOI 
and H.0- values — and is thus not considered in the following attempts at sourcing. 
Sources 
The following sources are suggested by comparing the chemical data (known and 
unknown sources) presented in Table 2, and backed up by the corresponding values of 
Duerden et al (1987): 
(a) green obsidian pieces (#’s 54, 105, and 444) are sourced to the Mayor I ‘green’ 
source (control samples 1514/2 and 504); 
(b) grey pieces (#’s 49, 14, 60, and 59) are sourced to Great Barrier | (samples 680/ 
4 and 2935). 
As an additional quantitative comparison of this data, clustering analysis, 
utilising Euclidean distances, was employed. This technique supported the above 
conclusions with respect to source. The means and standard deviations of the two 
clusters are listed in Table 4, and provide a numeric summation of the raw data. 
Possible shortcomings include the small sample size, and it should be noted that the 
