Biological evidence shows that there are neural networks specialized for recognition of signals and patterns acting as associative memories. This work present the quaternionic quantum neural networks using just one quaternion quantum autonom neuron. The Quaternion Quantum Neural Network (QQNN) is a feed-forward neural network, based in the so called qubit neuron model, represented and handled in the algebra of the quaternions. We take advantage of the property of the quaternions to make rotations with a simple product, so we can at the same time and with the same objects (quaternions) represent and operate the quantum bits. Since the QQNN generalizes the complex valued quantum neural networks, it is more appropriate to handle more difficult classification problems. According the results of the experiments using bench mark examples, we show that the QQNN have a good performance.