268 



POPULATIONS 



is essentially normal or, as is frequently 

 true, aggregated or "contagious" (p. 365), 

 and must test liis method and his data by 

 appropriate statistical techniques." 



Before sampling, the investigator should 

 ask himself questions sometliing hke these: 



1. Is this a case (i.e., his study) in 

 which sampling can be used at all? Can the 

 population size be approximated by samples 

 with an error that the investigator is will- 

 ing to exclude as negligible or unimportant? 



2. Can the magnitude of this error be de- 

 termined objectively? Then, can the deci- 

 sion be reached as to whether the method 

 needs to be refined or, as is sometimes true, 

 to be coarsened? The determination of the 

 magnitude of the error involves statistics; 

 what to do thereafter largely involves good 

 judgment. Defection in either aspect preju- 

 dices the entire study. 



3. Is enough known of the general 

 ecology and distribution of the species to 

 determine how the samples should be taken 

 both in space and in time? 



4. Is it technically feasible to take the 

 required samples both from the point of 

 view of the method and the labor involved? 

 Patently, a sampling method that is too 

 laborious defeats its own purpose. 



We stress these obvious points because, 

 in our opinion, the ecologist too frequently 

 derives a datum from inadequate sampling. 

 We appreciate that it is often diflBcult, if 

 not impossible, to live up to these rubrics. 

 But we do think that population ecology 

 will improve when more serious attention 

 is given to these matters. The problem is 

 particularly vexatious for the student of 

 complex natural populations. It is relatively 

 easy for a protozoologist working with a 

 liter volume of paramecia in a dense culture 

 to stir his population thoroughly, quickly 

 draw oflf a cubic centimeter of fluid, count 

 the protozoa in this sample, collect, say, 

 thirty such samples, and then compute with 

 considerable accuracy the number of para- 

 mecia in the whole volume and per cubic 

 centimeter. But it is harder for a student 

 of forest-floor Collembola to determine 

 the number, density, and time trend of 

 that population for the entire locality. In 

 each case the problem is one of sampling; 

 but the first requires only a simple treat- 

 ment with a modicum of judgment, while 

 the second demands much knowledge 



• An unusually cogent discussion of sampling 

 appears in Simpson and Roe ( 1939, Chap. IX). 



about the form, its habitat, its distribution, 

 and various statistical methods before it 

 can be assured that an adequate sample is 

 obtained. Later in this section, when dis- 

 cussing Contagious Distributions (p. 365), 

 and in the chapters on Communities, we 

 shall return to this subject. 



5. The method of marking. Marking, a 

 technique of much promise, but containing 

 many pitfalls for the unwary, is being more 

 and more adopted. It has at least two var- 

 iants: (a) In the first, animals such as small 

 mammals are individually marked and 

 turned loose in an area that is extensively 

 and systematically supphed with suitable 

 five-traps. Each time an unmarked animal 

 is caught it is marked. If a marked animal 

 is caught, this fact is recorded. By main- 

 taining this routine for an appropriate time 

 interval the investigator learns much about 

 the density of the species in the study area. 

 Also, he is able to plot territories or "home- 

 ranges" for those forms that have them. 

 There are objections to this technique: one, 

 the animals may become "trap-shy" or "trap 

 addicts"; another, animals may move in or 

 out of the area. These bias the sample some- 

 what, especially in the case of certain spe- 

 cies. On the whole, the data thus collected 

 can be relatively trustworthy, (b) The 

 second variation is adaptable to more types 

 of populations. A known number of marked 

 animals is turned loose in the original area 

 (or volume). Since these animals sup- 

 posedly redistribute themselves in their orig- 

 inal population as they were before they 

 were withdrawn, prior knowledge about the 

 pattern of their distribution is not required. 

 Then, after an appropriate interval of time, 

 which must be based on the judgment of 

 the investigator, a sample of the population 

 is taken and the proportion of marked to 

 unmarked forms is computed. 



The second variant, sometimes referred 

 to as the "Lincoln Index," may be clarified 

 by an actual case. Green and Evans (1940) 

 worked with snowshoe hares in the Lake 

 Alexander area of Minnesota. They set five- 

 traps for the hares. Those caught were 

 marked before they were released. This was 

 called the "precensus period." In a certain 

 instance they banded 948 hares. At a later 

 trapping, "the census period," they caught 

 167 marked and 254 unmarked rabbits. The 

 following simple proportion then was set 

 up: 



