We're painting with a bit of a broad brush, but … even today, the implicit idea among theorists seems to be that synapses are essentially straightforward conduits used to transfer information between neurons. But, if that was really all there was to it, Nature could have used gap junctions instead of neurotransmitters, which would allow brains to be smaller, faster, and more energy efficient. Instead, the simple fact that the gray matter of brains is packed with presynaptic terminals is an indication that we really don’t understand the first principles underlying biological computation all that well at all.
So, a key question is: What can chemical synapses do that gap junctions cannot? Here’s something that has captured our attention: When excitatory chemical synapses are used in high frequency bursts - as they often are in vivo - the strength of the connections changes dramatically, and quickly. Above you can see examples of electrophysiolgical recordings from a variety of synapse types at times when the presynaptic axons are activated with short bursts of stimulation (second row of panels). Each downward deflection is a postsynaptic response to a single pulse of stimulation. For the Calyx of Held recording, you can see that the second response is smaller than the first and so on. This is termed short-term depression because - if left alone for a few seconds - the strength of the synaptic connection will return to the initial value.
Every chemical synapse seems to exhibit short-term changes, but the details vary dramatically. The second panel in the second row is an example of a synaptic connection between the cortex and thalamus. These synapses get stronger, and the phenomenon is termed short-term enhancement.
The third panel in the second row is from Schaffer collateral synapses in the CA1 region of the hippocampus. First one sees enhancement, and then depression. Each pulse of stimulation was strong enough to activate dozens of afferent axons, so the responses that are depicted were the average of many that were all activated at the same time. When one studies individual synapses, one at a time, each seems to do its own thing.
And differences between synapses is the rule, not the exception. For example, the single electrophysiological trace in the third row is the response of a calyx of Held synapse that exhibited enhancement instead of depression.
The electrophysiological responses above were all recorded in the presence of drugs that block NMDA receptors, which prevent the permanent or long-term changes in synaptic strength that are thought to underlie learning and memory. The remaining, temporary, changes have traditionally been termed short-term synaptic plasticity, which - to some people - suggests a short-term learning and memory mechanism. However, every type of synapse seems to exhibit short-term plasticity of one sort or another, including synapse types that aren't involved in storing memories. Because of this: some computational neuroscientists prefer the term frequency dynamics.
Whatever you call them, the short-term changes in synaptic strength are mostly caused by changes in the amount of neurotransmitter released from the presynaptic terminals. There are postsynaptic mechanisms that play an additional role at some synapse types, including at calyces of Held early during development. However, we are primarily focused on the presynaptic mechanisms, and, so, design our experiments so that the postsynaptic mechanisms don’t interfere.
We suspect that the mechanisms underlying synaptic dynamics function as elementary building blocks used by biological circuits to encode and process information. Because of this, we anticipate that we will ultimately need a comprehensive understanding of the nature of these building blocks before we will be able to truly understand how biological computation works.