The analysis Is as follows ; 



where * Indicates an P value for which P is less than 0.05, and indicates an F 

 value for which P is less than 0.01, in all cases using the second order inter- 

 action as the error term. 



There are no significant differences among the mean responses to the 

 different foods which were fed. However, as shov/n in the following values of the 

 grand means, there is a general decrease from Week 1 (Fi) to Week 3 (F3) : 38.4, 

 36.4, 33.2. In view of the results which follow, this trend, if real, is more 

 likely related to time than to food. Of necessity, time (weeks) and food are con- 

 founded in the design. 



There are highly significant differences in the mean responses of the 

 fish between successive days of the v/eek. as shown by the grand means for succes- 

 sive days, there is a progressive decrease in response from Di (Tuesday) to D3 

 (Thursday); 42.0, 36.0, 27.3. The decrease is consistent within weeks, as is ap- 

 parent from the lack of any significant interaction involving days in the analysis 

 of variance table, or as may also be seen from the data of table 1. It may be 

 shovjn that the regression of response on days is highly significant (mean square 

 for regression is 1290.667**) and accounts for most of the variation between days. 

 A study of tidal heights, direction of tidal current, and weather at the time of 

 each test revealed nc apparent relationship to account for the regression. It is 

 most likely due to progressively increasing fatigue or a dulling of response with 

 successive testing. It may reflect a learning process — the fish can be deceived 

 Into a feeding reaction by .the extract but become less responsive when they find 

 that no food is avallableL/. Although the fish were allowed to rest on Saturdays 

 and Sundays, the regression may have been carried over from one week to the next, 

 particularly if it involved a progressive dulling of response or a learning pro- 

 cess, thus accounting in part at least for the decrease in response between weeks 

 noted in the preceding paragraph. 



There are significant differences in the responses to the various ex- 

 tracts. The grand means for squid, skipjack llesh, shrimp, and blend extracts 

 are respectively as follows: 29.3, 42.4, 31.0, 37.7. According to these results 

 the skipjack flesh extract gave the greatest response and the squid extract the 

 smallest response, with the blend slightly greater than the mean of the other 

 three. Although heterogeneity is indicated, it may not reflect differences In 

 response between the substances per se , but rather an interaction between the ex- 

 tracts and the food fed (no significant differences between the grand means for 

 ex'tracts are indicated if the FxE interaction is used as the error term). 



There is a significant first order interaction between extracts and 

 foods (or weeks). This indicates that the fish responded differently to the four 

 extracts when they were fed on different foods (or, in different weeks). The 

 nature of the differential response may be seen from the following summary, which 

 gives the mean response according to both food and extract: 



1/ This has been suspected to occur between successive tests within days on 

 other occasions. 



