Most excitatory synapses in the brain occur on dendritic spines. Stronger synapses contain more AMPA receptors. Stronger synapses also tend to occur on larger spines. Spine size and AMPA receptor content have been linked, but never quite like this. Park et al. showed that all the things you expect to induce increases in synaptic strength/AMPAR content (calcium, NMDAR activation, theta burst stimulation) also cause localization of recycling endosomes to spines and fusion of those endosomes with spine membrane. This increases the spine surface area. Endosomes are little balls of membrane with lots of proteins stuck in them (sort of like synaptic vesicles). At least sometimes they carry AMPA receptors, so now they have a dual purpose. When they fuse with the membrane they make the spine larger and insert AMPARs at the same time.

You can ignore all the above and just take a look at these spines dancing.



The hotter points indicate that the inside of an endosome has opened up and probably fused with the cell membrane. When the white dot appears a form of long-term potentiation is being induced. I don't understand why the hotspots aren't right on the membrane instead of appearing just inside, but I can live with it.