ale 
xture 
  
ifferent. 
the crop as possible. For obvious reasons, it was more possible to 
do so at the larger scale prints. Mean tone and standard deviation 
of the tonal measurements for each field (separately for each scale) 
were computed. The data so obtained is given in Table 1. 
Statistical Analysis 
The tonal and textural data obtained from the experiment were 
organised as univariate and multivariate sets and analysed 
statistically. The above data sets were analysed in a systematic 
manner to investigate the utility of textural variable separately 
and jointly with tone at each scale and in combination of scales 
to determine the relative utility of each textural and tonal vari- 
able for signature analysis. Correlation analyses were carried out 
to investigate the degree of independence of each variable in pro- 
viding additional discriminate information to specifically verify 
the role and utility of texture in addition to tone. Lastly cano- 
nical analysis was done to reduce the six variables into a lesser 
number of Canonical variables, and used for discriminant analysis 
and classification. 
The mean tonal values of each crop were assumed normally distri- 
buted. Histograms and normal plots of the mean tonal values 
generally showed normality except for departures attributable to 
the small sample size. 
Though the distribution of textural values, is a Chisquare distri- 
bution, it was assumed normal as theoretically, for degrees of 
freedom more than 30, the textural distribution could be considered 
approximately normal (Kandall 1947). Due to the robustness of the 
procedures used in the analysis of data, the inferences derived 
should be satisfactory based on the above assumption even when they 
are not fully met in practice (personal communication from Dr.Fowler) 
Standard deviation was selected in preference to variance as 
according to Kandall, it approaches normality faster than the 
variance as sample size increases. This measure also has a linear 
dimensional unit making it more compatable to tonal units than the 
variance. Histogram and normal plots of the standard deviation values 
did not indicate a serious departure from normality. 
For One Way or Two Way Anova and Discriminated Analysis, equality 
of variances - co-variances are assumed. During each stage of 
analysis, tests for equality of dispersion was carried out. In most 
cases, the assumption was met within the limits of significance. 
One Way Univariate Anova was carried out for each of the tonal 
and textural variables (at all the 3 scales) to test for pairwise 
group differences. These tests assisted in identifying the signi- 
ficance and the relative utility of each variable in discriminating 
the crop types pairwise, and independently at the same significance 
level. The results are summarized in Table 2. 
Discriminant analyses were carried out to determine the degree of 
success obtained in discrimination and classification of all five 
crop types by tonal variables, textural variables and combination 
of the two types of variables at one and more than one scale. Both 
linear and quadratic function scores were also determined to test 
if the quadratic function improved the scores. 
Stepwise discriminant analysis was also carried out to identify 
the more discriminating variables, for all the crop types simul- 
taneously. The effect on success scores by different combinations