84 PHYSIOLOGICAL REGULATIONS 



more constant rates of water exchange than less constant ones ; that 

 is a question for investigation only after some criterion of "better" 

 to which even the dog will agree has been discovered. 



In brief, the rates of exchange that occur under standard condi- 

 tions for water balance vary by measured coefficients, rates through 

 two paths fluctuating simultaneously by diverse amounts. Avail- 

 able data do not indicate that adjustments are more regular or ac- 

 curate by gain than by loss, or vice versa, nor that control is exerted 

 upon intakes more or less vigilantly than upon outputs. Both are 

 coordinate in the maintenance of water content. Only in the 

 average of successive periods of time does total gain equal total 

 loss. 



§ 24. Variations in rates at diverse loads 



Recognition that contents and rates each vary while the dog is in 

 a state of water balance allows the introduction of variabilities into 

 the equilibration diagram. First of all, the point Wo on the abscis- 

 sae (fig. 47, C) is actually not a point but a zone. Since periods of 

 1.0 hour were used in the measurements of rate that go into it, this 

 zone has a width of o or ± 0.10 per cent of the body weight. It may 

 be supposed as a crude guess that the variability of any body weight 

 that enters into the estimates of water load is at least this large. 

 Second, the ordinate rate of turnover (Rq) is a zone about ± 16 per 

 cent of 0.25 %/hr. or ± 0.04%/hr. in height. Rate and load together 

 give a rectangle of random variations 0.08 %/hr. X 0.20%, that usu- 

 ally includes 2/3 X 2/3 = 4/9 of the observations. Third, each 

 series of rates for which averages have been plotted, such as figure 

 2, has a variation peculiar to itself. Every point in every graph 

 might therefore be rendered as a rectangle or ellipse. These 

 rectangles or ellipses might also be employed, with modifiers for 

 numbers of observations, as measures of error and of significance. 



One question that may be decided is whether the variation of 

 rates at several water loads is proportional to the mean rate or 

 whether it is independent of the mean. The former is approxi- 

 mately the case, and for this reason coefficients of variation are 

 much more useful than standard deviations in characterizing it. 

 Next, its magnitude may be found. Among 7 tests on two individu- 

 als (B, fig. 2), in the first 1.0 hour after water was put into the 

 stomach, the rate of total loss varied by C.V. ±: 24, of urinary loss 

 varied by C.V. ± 27, or identical values. Further, selecting 24 tests 



