Evidence for these things

Neurons are a great place to start modelling AI because your model can be tied to real biological neural behaviour.
Evidence for fLIF neurons:
1. It seems pretty clear that neurons fire.
2. Integration is also quite clear.
3. Leak is a little less clear, but there seems to be widespread evidence.
4. Fatigue also has some evidence but is less clear.
5. The nice things with all of this is that you can take a neuron out of a body, send spikes to it, and measure it.
6. Clearly, fLIF is a simplification, but it should be evaluated on the quality of fit to neural data and the efficiency of simulation.
Evidence for CAs:
1. Abeles on firing patterns in behaving monkeys is good evidence. Eckhorn et al 1988.
2. Gelbard-Sagiv et al 2008
3. High firing (fMRI) in different brain areas for words (Pulvermuller 1999).
4. This is a lot harder to find really solid evidence for because you can't see lots of behaving neurons.
Evidence for Hebbian Learning:
1. This is also nice because you only need two neurons to test it.
2. However, the effects may be quite long term.
3. LTP: Marr 1969; Hubel and Wiesel 1965; Bliss and Lomo 1973 etc.
4. STP: Hempel et al 2000; Buonomano 1999.
There is also an emergent level of evidence.
That is, systems based on fLIF neurons, Hebbian learning and CAs are meant to behave like people do.
So, psychological behaviour is another bit of evidence that should be used.