55 

 the basis of such effects (following section). 



Surveys that are based on stratified random samples depend greatly 

 upon the control of biases. While the actual types of sampling gear 

 vary and are subject to certain difficulties, these difficulties are well 

 understood and adjustments are made. For example, plankton samples depend 

 utterly upon the mesh size of the sampling net: a large mesh allows the 

 small organisms to escape, thereby skewing the results toward selection 

 of larger individuals than actually exists. However, such biases have 

 long been recognized, and, through use of similar gear over time, are 

 generally discounted. 



The problem lies not with the taking of samples but with their 

 analysis. There presently exists no automatic method of determining the 

 species selection of plankton tows. Individuals must be counted by hand, 

 a laborious and slow process that consumes time that might otherwise be 

 used more productively. 



A similar situation exists with regard the aging fish samples. While 

 the measurement of fish size and weight is relatively straightforward, 

 determination of age through analysis of scales or otoliths (ear bones)is still 

 a slow process that is characterized by some as an "art." There is 

 a great deal of interest in designing a system that can estimate the age of 

 fish rapidly and automatically. The present system is costly, requires 

 highly-skilled analysts, and is subject to error. Yet age determination is 

 critical, for upon such information the abundance of various year-classes 

 is estimated. 



Other forms of sampling (for example, temperature soundings) have 

 been refined and are highly accurate. However, as in all such methods. 



