This paper describes the diets of 201 larvae of Atlantic mackerel Scomber scombrus L. collected from Long Island Sound, New York (USA) during May-June 1982 and 1983. First-feeding larvae (3.5 mm in length) were phytophagous. The diets of larvae 4.5 mm were composed of the nauplii of Acartia hudsonica, Temora longicornis and Pseudocalanus sp. Larvae > 5 mm ate some copepodites of A. hudsonica and T. longicornis and smaller proportions of phytoplankton and copepod nauplii. Mackerel 3 6.5 mm were cannibalistic, eating larvae 3.5 to 4.5 mm in length. Weight of stomach contents averaged 1.8 % of an individual's body weight. We calculated that, in order to satisfy its daily energy requirement, a mackerel larva must consume 25 to 75 % of its body weight per day. Larvae fed selectively, taking a greater proportion of T. longicornis and Pseudocalanus sp. nauplii and a lesser proportion of A. hudsonica nauplii than expected by chance alone. Ivlev index values for T. longicomis nauplii were +0.66, for Pseudocalanussp., +0.25, and for A. hudsonica, -0.55. The Chesson index and Pearre's 'C' yielded the same result. We hypothesize that mackerel larvae select food primarily on the basis of prey visibility (both Pseudocalanus sp. and T. longicornis nauplii are more motile than A. hudsonica nauplii). Even though mackerel larvae consume large numbers of copepod nauplii per day and feed selectively on certain species, mackerel larvae are not sufficiently abundant in Long Island Sound to have an impact on copepod population dynamics.