The ability of some neural networks to produce multiple motor patterns required during different behaviors is a well-documented phenomenon. We describe here a dramatic transition from coordinated inhibition between two functionally antagonistic groups of motoneurons to their co-activation in the feeding neural network of the predatory mollusc Clione limacina. To seize its prey, Clione uses specialized oral appendages, called buccal cones, which are controlled by two groups of motoneurons: Cr-A (cerebral A) neurons controlling buccal cone protraction and Cr-B (cerebral B) neurons controlling buccal cone retraction. When Cr-A neurons are active, Cr-B neurons usually receive strong inhibitory inputs that terminate their firing, which leads to the full protraction and elongation of the buccal cones. We have found, however, that the Cr-A and Cr-B motoneurons sometimes burst simultaneously without any traces of inhibition in the Cr-B motoneurons. This transformation of the neural network activity from inhibitory interactions to co-activation presumably occurs during the late "extraction" period of the feeding behavior when buccal cones become partially retracted and rhythmically active. The transition from the inhibitory interaction to co-activation is controlled by the activity of a single pair of interneurons (Cr-Aint interneurons), which are electrically coupled to the Cr-A neurons and monosynaptically inhibit Cr-B neurons. Normally, the Cr-Aint interneurons are active along with Cr-A motoneurons and inhibit Cr-B motoneurons. During a period of co-activation, however, these interneurons do not produce spikes, thus allowing Cr-A motoneuron activation without inhibition of the Cr-B motoneurons.