The brain is often thought to be organized in a way that it optimizes its information processing and storage capabilities, supported by studies indicating that neuronal networks self-organize to a critical state. This is characterized by the emergence of scale-free statistics in neuronal avalanches close to those predicted by mean-field theory, suggesting that neuronal avalanches propagate in a tree-like network. Here, we use optical imaging techniques to study neuronal avalanches in different preparation of neuronal cultures and different dynamical regimes. We present an experimental case study where the avalanche statistics is significantly different from mean-field, suggesting that feed-back loops play a leading role in propagation of neuronal avalanches. Another phenomenon is that cortical and hippocampal neurons alternate between high and low activity regimes (up and down states) characterized by distinct firing rates. In contrast with previous theoretical works, our experimental findings suggest both states exhibit scale-free behavior with different intrinsic time scales.