This is depicted in the following figure:
validate -------> theory (model) experiment (analysis) adapt <----
Normally, this circle has to be passed through many many times as feedback from experiment is built into theory in order to recheck the theory to eventually obtain a powerful theory.
A theory is based on a model, mostly derived from first principles or very few assumptions as per Occam's razor or KISS principle. This shall be tested against the reality so that the theory can be either verified or falsified. An experiment is either built in such a way that it is suitable to check the theory, or alternatively it comes from the real nature and is a complex system (like the brain serving as the best example).
On the one hand, we have the theoretical approach to neuroscience mostly by physicsits or even mathematicians who want to incooporate their "first principles" or even "axioms" into a Stimulus–response model and thus try to model the brain as a mathematical function or mapping in the sense of a dynamical system that evolves with time by input configuration states. There is a paper "The Dynamics of Neural Populations Capture the Laws of the Mind" by Gregor Schöner in which he claims some hypotheses and cites Feynman for his conversation with his father about answering the "why" with just a more general physical reasoning.
On the other hand, there are people working with real data that comes from experiments. A lot of data-analysis is done there. This is mainly driven by biologist psychologists, and "data-scientists". They are also responsible for the setup on localizing where special neurons fire.
There are these persons I admire, popular as being nobel prize winner in 2014, O'Keefe and Moser. They managed to cover a broad range from real-world experiment to theory. Actually they localized these place and grid cells individually and detected them on new locations (pattern completion and separation).
In how far is neuroscience in general able to link theory and experiments, whereby with experiment I explicitly refer to those complex systems in real-world. And not the realization of a toy-model, which is built just in order to verify a theory. Actually in some part, this has to happen in order to isolate individual features of a model to verify, but still it is a difference of inventing an experiment to check a theory or to try to build a model to describe an experiment that is motivated from real-world behaviour. Can we acheive a linkage or are these two fields (theoretical and experimental) still separated from each other heavily?
What are other examples of broad-range covering from a model to describing a real-world complex system? (i.e. having a powerful theory from which directly many consequences can be derived, and the above mentioned circle has not to walked with a high frequency)
In this science, does it happen similar or same as in learning? Without initial knownledge, the human being adapts a model of its environment by simply being exposed (making experiments = experience) to it, i.e. passing through this above mentioned circle many times and obtaining much feedback to complement one's model of the world.