I would guess that it is measured by two factors. First is the time it takes between a person's being confronted with a complex problem and that person's solving it. Second is the degree to which the person solves the problem. A person who quickly and excellently solves a problem would seem to have a high capacity of flow of information.

Currently, there are different methods and tools to examine the ability of the brain to process different types of information. These methods and tools will depend mainly on the type of information/system to be considered (i.e., anatomical, functional, biochemical) and on the experimental design to be carried (i.e., model). For example, graph theory and functional connectivity are concepts that help to know how the brain manages segregation and integration processes that we need to manage information. Other concepts such as global efficiency have also been developed to model the degree of capacity of a given system.

Article Efficient Computation of Functional Brain Networks: toward R...

Article Changes in Brain Network Efficiency and Working Memory Perfo...

I think this is a primarily anatomical question. If the question is how many on/off signals the brain can take in at the same time and derive meaning from then the existence of about a million axons in each optic nerve probably gives us the ball park figure. Thus the total number of input channels to a brain is about 10 million. Neurons tend to be able to fire about every 20 milliseconds so the input precession might be 500 million signals (including a lot of null signals).

There is no single central processor to take these signals so there is no suggestion that any one integrator unit can handle a million signals at once. However, individual neurons can have up to 50,000 input channels in the form of synapses in their dendritic tree. So we can expect an average brain neuron maybe to receive in the order of 2 million on or off signals per second. In practice the vast majority of synapses will be silent (off signals) in any one time frame.

There is also good reason to think that at all levels of brain processing, whether unconscious routine movement or conscious deliberation, involves the same pattern of signals being sent simultaneously to large numbers of neurons. Again, on average, neurons have about 10,000 output branches. So it seems reasonable to think that when we make a decision a pattern of about 10,000 signals indicating the 'data' for the decision to be made on is sent to at least 10,000 parallel cells and maybe more like 100 million (as achievable with a single 'broadcasting relay' step).

My suspicion is that beyond this basic anatomical information we really do not know much about 'bandwidth' because the investigators tools at larger scales cannot address this. But if one makes the simple assumption that bandwidth inasmuch as it is meaningful here has to relate to throughput through at least one single pathway then that has to relate to a single neuron because otherwise there is no integration at that point. So the 10-50,000 input channel figure for the average brain neuron may be the only real answer to the question (making about 2 million 'bits' per second).