So anyway I recently made up my mind that I want to go to grad school for neuroscience (currently doing linguistics) and I am learning just as much as I can

I had read that working and short term memory may be implemented in terms of recurrence that lasts just as long as these particular memories. It sounded feasible and cool when I first read it but The Computational Brain suggests strong limitations to this approach, noting that accurate preservation of a signal only through recurrence in a biological neural network, even over a short-time period, would be very difficult

The book suggested, however, that there could be recurrence among small populations of neurons (whisper down the alley with redundancy I take it), and that short-term and working memory might also make use of short-term synaptic weight modification

Does anyone have a clear account of our current understanding of how working and short-term memory are implemented at a neural level? If so please relate it here