Towards Utilising a Range of Neural Activations for Comprehending Representational Associations