The patch clamp technique has been used to study channels in a membrane inside a cell. A single muscle fiber is skinned in relaxing saline (high K+, low Ca2+ with EGTA and ATP), leaving the native sarcoplasmic reticulum (SR) membrane exposed for patching. Fibers are dissected from the second antenna remotor muscles of the American lobster, Homarus americanus. Transmission and scanning electron microscopy confirm the large volume fraction of SR (approximately 70%) and absence of sarcolemma in this unusual skinned preparation. The resting potential of the SR was measured after the resistance of the patch of membrane was broken down. It is near 0 mV (-0.4 +/- 0.6 mV). The average input resistance of the SR is 842 +/- 295 M omega. Some 25% of patches contain a K+-selective channel with a mean open time of seconds and the channel displays at least two conducting states. The open probability is weakly voltage dependent, large at zero and positive potentials (cytoplasm minus SR lumen), and decreasing at negative potentials. The maximal conductance of this channel is 200 +/- 1 pS and the substate conductance is 170 +/- 3 pS in symmetrical 480 mM K+ solution. The current-voltage relation of the open channel is linear over a range of +/- 100 mV. The selectivity is similar to the SR K+ channel of vertebrates: PK/PNa is 3.77 +/- 0.03, determined from reversal potential measurements, whereas gamma K/gamma Na is 3.28 +/- 0.06, determined from open-channel conductance measurements in symmetrical 480 mM solutions. Voltage-dependent block in the lobster SR K+ channel is similar to, but distinct from, that reported for the vertebrate channels. It occurs asymmetrically when hexamethonium is added to both sides of the membrane. The block is more effective from the cytoplasmic side of the channel.