(470) 
the reports of the international congress in Rome. There I indicated the same. 
I also agree with the fact that the application of the theory of least squares is 
only exact after the elimination of systematical errors. The value of the french 
method is that the instrumental errors are eliminated as well as possible. I 
admire that and also the improvement of the results that has been obtained. 
The difficult problem that remains concerns the so-called local systematicial 
errors or jumps. If it is true that by the method of Poivilliers these errors can 
be eliminated it is not necessary for Mr. Bonneval to include them in his paper. 
The result would be directly accessible to the application of the least squares. For 
me the question is still open if the elimination according to Poivilliers is indeed 
applicable. In this respect I want to refer to the experiments with the lottery- 
tickets published by Prof. Roelofs at the Congress in 1948. Until now these 
results are not definitely refuted. Prof. Roelofs explains the presence of local 
systematical errors mainly as an accumulation of accidental errors. In this case 
his compensation method is fully applicable. 
Mr. Bonneval: It is quite natural to try to eliminate the influence of 
systematical errors by the working method. But I don’t believe that in the 
actual state of technical development any of the practically used triangulation 
methods may claim to have obtained this result. Those developed by Mr. Poi- 
villiers, even if they give a contribution to the solution of the problem, have 
certainly not solved it completely. In photogrammetry as in every observation 
science every systematical error, even the smallest, is serious because it is syste- 
matical. It seems to me to be necessary to dispose of a procedure that permits 
to show, by an examination of the obtained results, the residual systematical 
errors of all kinds and that permits to correct the observations before the ap- 
plication of any compensation process, that is based on the exclusive existence 
of accidental errors. 
Mr. Reading: It is evident from all our experiences that there are three 
kinds of errors, the systematic which we can measure and correct for, the acci- 
dental repetitive errors which generally follow the Gaussian distribution, and 
larger excessive errors that occur at odd intervals. The systematic and acci- 
dental errors we can eliminate by computation and adjustment to terminal 
control; the excessive occasional errors give us trouble. In terrestrial trian- 
gulation we have quadrilaterals or other figures which give us the power 
by certain extra conditions such as excessive departures from mean, triangle 
closures, or side equations to reject the measurements with excessive errors due 
to excessive refraction or other unknown causes. Thus if a pointing with a 
theodolite differs by more than five seconds from the mean, if our triangles 
fail to close within one second in first order triangulation, we reobserve under 
different conditions. 
It is evident that in spatial air triangulation, we must find similar addi- 
tional conditions which will enable us to detect and safely reject the excessive 
errors. Whether the transverse parallax test disclosed by M. Poivilliers is suf- 
ficient we shall all be eager to test. Whether the procedures recommended by 
Mr. Bonneval or the elegant mathematical solutions of Prof. Roelofs are the 
most practical also require tests by others. 
As pointed out by Mr. Santoni there are available certain aids in his solar 
photography and in horizon photography as practiced in Finland, which give