A fish was counted as "taken" by a seal when the knot that was tied around its neck was still left, if the knot was gone it was assumed it had been tied to loosely and the fish had just fell off. This could also be seen in some videos. 

For the dinner-bell effect test a total of ten fishes were taken, 3 from the Seal-safe pinger, 4 from the Banana pinger and 3 from the controls. This was only 2.7 % of the total occasions I hung out fish. But there where no significant difference between the three treatments suggesting that no dinner-bell effect was at play. 

For the fish preference test fish was taken on only four occasions. Unfortunately, this made it impossible to do any statistical test on it. The fish-pairs taken were herring-perch, herring-whitefish and perch-whitefish. On one occasion only one fish was taken, it was a herring and a perch was left.