31 
reduce once more all the obtained raw data. There is good 
reason to expect that the mean error of the triangulation 
will be slightly reduced, bringing the computed value in 
closer agreement with the theoretical value, and that the 
distribution of the residuals will follow more closely the 
theoretical distribution law. 
Resume 
In its present state of development satellite triangula 
tion is capable of determining a single station to satellite 
direction, relative to the Right Ascension-Declination 
system, with an accuracy of one part in 500,000, or +01'*+. 
With sufficient redundancy in the observations statistical 
improvement by a factor of two can be justified, so that 
the attainable accuracy of the final results, in terms of 
the computed station coordinates, will therefore be one part 
per million of the average station-satellite distance. 
Stated in more general terms, the results already obtained 
suggest that it is possible to maintain high performance 
standards in all phases of a photogrammetric program even 
though data acquisition covers a period of years and encom 
passes the whole earth and is therefore influenced by a 
significant range of environmental conditions, both geo 
graphical and seasonal. Specifically, the results indicate 
that large-scale data reduction activities can be carried out 
over such extended periods in compliance with pre-established 
performance tolerances and meet the demands for calibrating 
evaluation equipment, including the control of its physical 
environment, and furthermore cope with the consequences of 
the turnover in manpower and unavoidable changes in organi 
zational elements of the program supervision. 
With respect to specific photogrammetrie operations or 
components the results show a significant difference in 
performance between the members of a group of comparator 
operators, suggesting, at least in theory, that the use of 
automatic measuring equipment could effect an improvement 
by normalizing the precision of the measuring operation. On 
the other hand, the results prove that each operator per 
forms reliably at his own average performance level, essen 
tially free of systematic errors. The histogram of Figure 10 
displays the measuring errors of a group of 34 operators, 
involving approximately 14,000 hours of comparator operation 
and close to 1.3 million settings on about 900 plates. The 
histogram shows convincingly that the work of the group as a 
whole can be considered as the work of a fictitious single 
operator, performing almost ideally according to the principle