why we learn

More from Sikström¹, neuropsychology, and cognitive theory…

To make the network usable after repetitive learning of items, the increase and decrease in synaptic strength should balance so the sum of changes over all synapses in the network is zero. To accomplish this equilibrium, the sum of the expected increase and decrease of synaptic strength must be zero (p. 253)

and,

To keep the expected synaptic strength at zero, changes that occur with a low probability must be associated with a large magnitude of synaptic strength. Conversely, changes that occur with a high probability must be associated with a low magnitude of synaptic strength. This implies that a postsynaptic cell that is frequently inactive should make large increases in the synaptic strength on those few occasions when it is active and small decreases when it is inactive. In contrast, a postsynaptic cell that is frequently active should make small increases in the synaptic strength when it is active and large decreases when it is inactive.

At the most basic level, the author is saying that it is impossible to continuously increase synaptic strength for all connections. Instead, the sum of all increases (potentiation) and decreases (depression) must, over time, reach equilibrium. I wonder what size of displacement is possible, i.e., do the processes occur at such a high rate that equilibrium is constant, or are there more pronounced shifts (during peak activity) and subsequent readjustments? I’m still reading the article, so I’m uncertain as to whether or not this is addressed.

1. Sikström, S. (2006). The Isolation, Primacy, and Recency Effects Predicted by an Adaptive LTP/LTD Threshold in Postsynaptic Cells. Cognitive Science 30(2), 243-245.

An amazing piece of logic from an article I’m reading¹…

The primacy effect is accounted for because the adaptive LTP–LTD threshold is low during encoding of the first few items; this lack of proactive interference leads to synaptic strengthening during encoding (LTP) and better performance compared to items in the middle of the list. The adaptive threshold is high following encoding of the last list item, and the recency effect occurs because no retroactive interference or LTD has yet occurred on the last item or items at immediate testing. The isolation effect in the middle of the list occurs because the adaptive threshold is low for the isolate, and this lack of proactive interference increases the synaptic strength (LTP). Furthermore, the presynaptic cells tend to be inactivated following the isolated item, causing a blocking of LTD, or a lack of retroactive interference, which also leads to an isolation effect at the first serial position. Synaptic boundaries cause weaker retroactive interference on an isolated item presented in a congruent list than on a corresponding item in a heterogeneous list, yielding a better performance for the isolate compared to a heterogeneous control list. Synaptic boundaries also account for power-function forgetting curves due to retroactive interference (p. 251).

The author is examining the primacy affect, the recency effect, and the isolation effect. When presented with a list of items, individuals are able to recall several items from the beginning (primacy) and the end (recency). They are also able to recall items from within the list that are contextually different (isolation). Synapse refers to the small gap between cell within which impulses traverse. The acronyms LTP and LTD stand for “long-term potentiation”, an increase in the strength of impulses, and “long-term depression”, a decrease in this strength. Proactive interference occurs prior to the encoding of information, retroactive after.

1. Sikström, S. (2006). The Isolation, Primacy, and Recency Effects Predicted by an Adaptive LTP/LTD Threshold in Postsynaptic Cells. Cognitive Science 30(2), 243-245.