The 65 solutions published in the fault-plane project of the Dominion Observatory have been based on 2,476 observations of P and 722 observations of PKP. Of these observations 18.3 per cent have been inconsistent with the published solutions, but a small number of stations have contributed a high percentage of these inconsistencies. Applying a criterion of rejection to the data reduces the percentage of inconsistencies to 14 per cent. Considering that most stations have tried to coöperate as fully as possible in the programme and have reported observations even when the arrivals were recorded only as weakly emergent phases, this percentage seems satisfactorily small.
Because the circles represent the boundaries between zones of compression and zones of dilatation, it might be expected that a higher percentage of inconsistencies would occur close to the circles. The reverse has been found; the observations close to the circular boundaries are slightly more accurate than those remote from them.
The inconsistent observations show no significant variation with epicentral distance except that the percentage of inconsistencies is high at the shorter distances. This is the effect of the crustal layers, and had been anticipated.
There have been 282 observations of PP and 135 observations of pP. These observations were examined to determine whether they could be used to determine the value of Poisson's ratio under the continents and under the oceans. It was found that the percentage of inconsistencies was too high to allow a final conclusion.
There were 17 observations of PcP. These were examined following a method suggested by Båth in an attempt to define the density ratio at the boundary of the core. The attempt was a failure because of the inconsistency of the data.
The failure of the reflected phases suggested that the data from them might be random. To test this, 23 new solutions were carried out based only on P and PKP data. When the solutions were complete, observations for the reflected phases were plotted on the diagrams. It was found that for all reflected phases the inconsistencies approached 50 per cent. It is clear that reflected phases, at least when reported by questionnaires, are not sufficiently accurate to be useful in fault-plane studies.