We have applied solutions with varying osmotic pressures symmetrically to the inside and outside of perfused, TTX-treated, giant axons. The potassium conductance G decreased with increasing osmotic stress, but there was no effect on either the shape or the position of the voltage-current curve. One must distinguish three possible actions of the osmotic agent: osmotic stress, channel blocking, and lowered solution conductivity. To do so, we compared results obtained working with pairs of internal and external solutions of either (a) equal osmotic stress, (b) equal conductivity, or (c) the same blocking agent. There was the same change in G irrespective of the type of stressing species (sorbitol or sucrose); this provides some evidence against a blocking mechanism. The conductivity of the external solution had a small effect on K currents; internal solution conductivity had none. A change in series resistance of the Schwann cell layer could account for the small effect of external solution conductivity. The primary cause of G depression appears, then, to be the applied osmotic stress. Using this result, we have developed models in which the channel has a transition between closed states under voltage control but osmotically insensitive and a closed/open step that is voltage-independent but osmotically sensitive. We have assumed that the conductance of this open state does not change with osmotic stress. In this way, we estimate that an additional 1,350 +/- 200 A3 or 40-50 molecules of solute-inaccessible water appear to associate with the average delayed rectifier potassium channel of the squid axon when it opens.