FLUCTUATIONS IN A PARTRIDGE POPULATION 131 



The trouble with your particular method is that it may suggest that the 

 biggest mortality factor is necessarily the bigger cause of variation. A loga- 

 rithmic method of plotting, as Haldane first pointed out, is more reliable for 

 this kind of analysis. 



M. Graham: You said that the early diet of partridges depended on 

 insects, and this in turn will depend on temperature. Might not this one 

 factor be worth isolating for detailed study, and give the key to the whole 

 situation? 



T. H. Blank: We have no measure of the availability of insect food for 

 partridges. 



G. M. Dunnet: You have indicated a change in population between 

 September and March and March and September in terms of numbers. 

 Does this change proceed in a gradual manner or in a step-like system like 

 that in grouse, as illustrated by Jenkins and Watson ? 



T. H. Blank: Probably the curve is not stepped. Between September 

 and December the change is steady, apart from the effect of shooting. 

 Between December and March there are two effects, firstly a steady com- 

 ponent of mortality and secondly a stepped factor due to the break-up of 

 the coveys and the pairing of the birds — and a consequent movement off 

 the estate. This is a short phase, lasting about three weeks, and so introduces 

 a step into the curve. 



H. Klomp. You have shown a correlation between the survival of young 

 and the nature of the farm crops. Has this correlation been treated statistically? 

 If this could be done it should be possible to assess the importance of this 

 one factor out of your complex of variables. 



T. H. Blank: I think too many uncertainties are involved for this 

 treatment of the data, though I agree that it should be done. 



J. G. Skellam: An estimated correlation coefficient based on only 

 eleven pairs of values is not often found to be significantly different from zero 

 except when its value is rather high. In this particular case (where one of the 

 variables enters into consideration after its selection from among other 

 possible climatic variables), there is little to be gained by carrying out a 

 formal significance test. I would say that the authors are right in presenting 

 the basic information as they have done, thereby revealing a strong suggestion 

 of an interesting relationship. 



{Note by Dr Skellam) : Tliis discussion went on among various interested 

 people later. It became apparent : 



I. It is not correct to carry out a formal test a posteriori, rejecting all 



10 



