The act of reaching for an object is a fundamental yet complex skill for a robotic agent, requiring a high degree of visuomotor control and coordination. In consideration of dynamic environments, a robot capable of autonomously adapting to novel situations is desired. In this paper, a developmental robotics approach is used to autonomously learn visuomotor coordination on the NICO (Neuro-Inspired COmpanion) platform, for the task of object reaching. The robot interacts with its environment and learns associations between motor commands and temporally correlated sensory perceptions based on Hebbian learning. Multiple Grow-When-Required (GWR) networks are used to learn increasingly more complex motoric behaviors, by first learning how to direct the gaze towards a visual stimulus, followed by learning motor control of the arm, and finally learning how to reach for an object using eye-hand coordination. We demonstrate that the model is able to deal with an unforeseen mechanical change in the NICO's body, showing the adaptability of the proposed approach. In evaluations of our approach, we show that the humanoid robot NICO is able to reach objects with a 76% success rate.