FISHERY BULLETIN: VOL. 83, NO. 4 



tial bias in the estimates of annual dolphin kill. 



Directly testing for the existence of an observer 

 effect on dolphin kills would require comparison of 

 covert observations with observations by NMFS and 

 lATTC observers. Based on the large difference be- 

 tween the kill rate observed covertly by one crew 

 member and the kill rates recorded by NMFS 

 observers during other fishing trips made by the 

 same operator and vessel, Smith (1983) speculated 

 that a large observer effect existed. We investigated 

 the significance of the difference in kill rates 

 reported by Smith (1983) by grouping NMFS- 

 observed trips into sequences of trips with common 

 operator and vessel. A few of these sequences of 

 NMFS-observed trips revealed between-trip kill rate 

 differences as large or larger than in the sequence 

 that included the covert observations. 



The existence of an observer effect can be indirect- 

 ly tested without relying on data from covert 

 observers. Johnson and Bolstad (1973) established 

 the existence of an observer effect by comparing 

 measurements made by observers with various levels 

 of obtrusiveness to the human subjects whose 

 responses were being measured. They concluded that 

 the differences in the responses measured by 

 observers with different levels of obtrusiveness im- 

 plied that the observer's presence had affected the 

 subjects' behavior. They noted, however, that the 

 magnitude of an observer effect cannot be estimated 

 using this approach. 



Following this indirect approach, we tested for the 

 existence of an observer effect on the numbers of 

 dolphins killed by comparing dolphin kill data col- 

 lected by scientific observers who differed in their 

 purposes of data collection, and hence, in their levels 

 of obtrusiveness. 



DATA 



The scientific observers were placed aboard a ran- 

 dom sample of U.S. registered tuna purse seine 

 vessels (Lo et al. 1982). Assignment of an NMFS or 

 lATTC observer to vessels in the sample was also 

 made randomly, subject to the constraint that any 

 vessel sampled twice within a calendar year would 

 be accompanied by an NMFS observer on at most 

 one trip (Ikble 1). 



Information collected for each NMFS- or lATTC- 

 sampled fishing trip included departure date and 

 data pertaining to each set (such as set type, date, 

 and location), and for dolphin sets, the number of 

 dolphins killed. Data available to the authors from 

 NMFS-sampled trips included all of this information. 

 However, data available from lATTC-sampled trips 



did not include departure date, and set dates were 

 available only to the quarter of the year. 



While the data items collected by both types of 

 scientific observer have been similar over the years, 

 for NMFS observers the purposes of the data col- 

 lection changed after March 1981. The primary pur- 

 poses of data collection, as explained to each vessel 

 operator at a placement meeting held prior to depar- 

 ture, were as follows: 1) On NMFS-sampled trips 

 begun from 1978 through March 1981, data were col- 

 lected for estimating the annual kill of dolphins and 

 for monitoring compliance with dolphin-release 

 regulations; 2) on NMFS-sampled trips begun after 

 March 1981 through the end of 1982, the data were 

 still used for estimating dolphin kills but were no 

 longer used to monitor compliance with dolphin- 

 release regulations''; 3) on all lATTC-sampled trips, 

 since the inception of that sampling program in 1979, 

 data were collected for estimating total kill but were 

 never collected for monitoring compliance with 

 dolphin-release regulations. 



As described above, the data collected by both 

 NMFS and lATTC observers to estimate total 

 dolphin kill can be used by the NMFS to halt fishing 

 by U.S. vessels on specific dolphin populations for 

 the remainder of the year. The data collected before 

 March 1981 by the NMFS observers for monitoring 

 compliance with dolphin-release regulations, 

 however, can be used by the NMFS as evidence to 

 prosecute operators who failed to comply. Thus, the 

 operators are likely to be more conscious of the 

 presence of an observer who is collecting data both 

 for estimating dolphin kill and for monitoring com- 



■*The change in data collection purposes of NMFS observers after 

 March 1981 was prompted by a court order forbidding the NMFS 

 from using data collected by observers for monitoring compliance 

 with dolphin-release regulations. No NMFS observers were 

 placed on fishing trips begun from 1983 through part of 1984 

 because of a subsequent court order forbidding placement of NMFS 

 observers without a search warrant. 



Table 1. — Number of observed fishing trips which made 

 at least one dolphin set from 1978 through 1982, by 

 observer type and year. NMFS totals are subdivided ac- 

 cording to departure date of trips (previous year, Jan.-IVIar., 

 Apr-Dec.) and exclude trips in which fishing gear research 

 was conducted. 



522 



