(2) Variability in the winds and background distortion in the stereo data 

 is sufficient to explain the difference in the two different sets of observed 

 values. 



(3) Sampling variations at the 1 percent significance level have 

 actually occurred. 



(4) The wave pole calibration is incorrect. 



(5) Weighted combinations of modifications of the above four hypotheses 

 taken 2, 3, or 4 at a time such as, for example (2-3-4). The wave pole cali- 

 bration is wrong by 30 percent at k = 15, sampling variation was at the 20 per- 

 cent level and the variability in the winds explains the rest of the differences. 



The first hypothesis can be checked by study of the original data as tabu- 

 lated. The fact that the histogram shown in figure 11.12 shows no effect of 

 distortion in area A at least suggests that most of this effect has been re- 

 moved. Also the uncorrected spectra come closer to agreeing with the theo- 

 retical Neumann spectrum than to agreeing with the theories of Roll and 

 Fischer [1956] and Darbyshire [1955]. The analysis has only served to re- 

 fine the results by what are in total rather small corrections, and the cor- 

 rections appear to be logically justifiable in all cases. If agreement with the 

 theoretical Neumann spectrum is not obtained, then the result would be that 

 there is no adequate theoretical wave spectrum in existence. 



Jtnthe light of these new results, the hypothesis of wind variability is 

 much less attractive than it was in Part 10. The wave pole and stereo obser- 

 vations were simultaneous in the sampling sense. The variation in the three 



213 



