A stocking, partial-poisoning and creel census experiment was carried out on a stratified lake in an effort to increase the crop of speckled trout,Salvelinus fontinalis. Except for a single year in which adult trout were planted, a negligible proportion of the introduced trout was recaptured. Application annually for five years of about one-half part per million of derris dust to the three-meter zone of the lake during summer stratification produced a heavy kill of coarse fish without harming the trout. Shiner,Notemigonus crysoleucas, perch,Perca flavescensand chub,Semotilus atromaculatuswere eliminated from the lake. Sucker,Catostomus commersoni, was drastically reduced. Killifish,Fundulus diaphanus, and eel,Anguilla bostoniensis, were not appreciably reduced, despite large annual kills. Smelt,Osmerus mordax, stickleback,Gasterosteus aculeatus, and trout,Salvelinus fontinalis, were killed in very small numbers during one or more poisoning years.Three independent estimates of the effect of the experiment upon the trout production of the lake are put forward: (a) The decrease in mass of the fish competing with trout for food, when calculated from the annual kills, amounts to 40 kg. The pre-poisoning mass of trout was 85 kg., calculated from a capture-recapture census. Thus if trout replaced its competitors quantitatively, the standing crop of trout would be increased to 150%. (b) The volume-time of water (i.e. percentage of the lake volume multiplied by the fraction of a year) which would be made available to trout by removing competitors from the part of the lake where summer temperature conditions would limit trout to between 50 and 100% of their full activity is calculated. If trout occupied this volume-time at the same density as they did the more favourable water, the standing crop of trout would be increased to 240%. (c) The total yield of the lake to anglers showed a steady increase following poisoning, to about 230% of the control value, or from less than one pound per acre to nearly two pounds per acre. These three estimates agree very well, considering the errors inherent in the method.