36 
Characters (Table 1) were selected on 
the basis of an examination of the variation 
observed among taxa and the critical 
assessment of past studies. The character-state 
data for all OTUs were entered into a data 
Austrobaileya 8(1): 35-46 (2009) 
matrix originally prepared as a spreadsheet in 
Microsoft Excel 7.0. Original data collection- 
sheets were cross-checked with data in the 
spreadsheet to guard against errors. 
Table 1. List of characters used in phenetic analyses. 
Numbers refer to those in Appendix 1. 
1 
Culm length (mm) (measured to the base of the primary inflorescence bract) 
2 
Culm width (mm) (measured at the mid-length of culm) 
3 
Culm cross section: (1) trigonous, (2) triquetrous 
4 
Involucral bracts - septate nodules presence: (1) absent, (2) present 
5 
Primary inflorescence bract length (mm) 
6 
Primary inflorescence bract width (mm) (at the widest part of the bract) 
7 
Base of primary inflorescence bract - teeth: (1) absent, (2) present 
8 
Inflorescence structure: (1) simple, (2) compound, (3) decompound 
9 
Spikelet width (mm) (measured at the mid-length of the spikelet) 
10 
Rachilla wing: (1) absent, (2) present 
11 
Glume length (mm) 
12 
Glume mucro length (mm): (1) <0.3, (2) 0.3-0.7 
13 
Style length (mm) 
14 
Nut length (mm) 
15 
Nut width (mm) (measured at the widest part of the nut) 
Data was analysed using a number of 
numerical methods. For the phenetic analysis, 
the Gower distance coefficient (which includes 
range standardization of data) was applied 
to all data matrices as it handles mixed data 
(Crisp & Weston 1993). The unweighted 
pair-group method with arithmetic mean 
(UPGMA, with (3 = -0.25; Belbin 1993) was 
used. 
Ordination was performed using semi- 
strong-hybrid multidimensional scaling (SSH) 
in 2-dimensions with 200 random starts on 
non weighted character states to minimise 
stress values. Ordinations were assessed by 
examining stress values and correlations 
of character states with ordination vectors 
(Belbin 1993). 
Correlations between character states and 
ordination vectors were performed to assess 
which character states were contributing to the 
pattern of ordination. Although all character 
states contribute to the overall ordination 
pattern, correlations above 0.7 are considered 
diagnostic of the taxa involved (Crisp 1991). 
