How Big is Your Working Memory?

Evanescent working memory may be defined by gamma brain waves, whose number is limited by the capacity of theta waves containing them.

Human working memory is sadly minuscule. We can keep only a handful of things in immediate mind at a time, like a new telephone number. How luxurious it is, in comparison, to program a computer with its gigabytes of ram, which can be consulted instantaneously! Humans have lots of intermediate and long-term memory, which are accessible quite rapidly. But working memory is a special class, happening (as far as we know) without any neural cell alterations, rather purely on the electro-chemical level. A recent paper pursued the theory that working memory is mechanistically constituted by the encoding of gamma electrochemical rhythms within the theta rhythm cycle, a bit like AM radio carries sound amplitudes encoded in its carrier wave.

This theory (more generally reviewed here) would imply that gamma waves individually mark different bits of content, which is somewhat difficult to understand, really. Neural oscillations have come to be seen as entraining selected networks across the brain, allowing attentive synchronization and binding of content from various anatomical regions. The waves do not carry the content, rather the anatomy does. But each network actuated by separate peaks of the gamma oscillation could be different, thus "carrying" different information, though the wave is simply a timing and separation device. This mechanism of enclosing a set of distinct gamma patterns (typically running at roughly 40 Hz) within a theta wave (typically running at a much slower 5 Hz, but ranging from 3 to 8 Hz) is already understood in the case of place cell firing/encoding in the hippocampus, so it is not a stretch to think, as many seem to, that it is also responsible for working memory in many different subsytems of the brain. In that place cell system, the encoding is not only differentiated by gamma cycle, but time-compressed, such that a physical traversal of a space that takes a second may be encoded by gamma peaks only tens of milliseconds apart. So there is true encoding of information going on here.

Experimental protocol. Subjects where asked to memorize a pattern of colored dots for roughly a second. They were randomly directed to memorize the left side set (experiment) or the right side set (control). The number of dots that they successfully recalled as staying the same or changing was the measured outcome.

A recent paper (review) sought to support this hypothesis by using transcranial AC current stimulation (TACS) to entrain the theta rhythm in the visual cortex to faster or slower pace than normal, and asking subjects to memorize visual features. TACS is a very interesting technique, different from the transcranial magnetic stimulation you may have heard of before. These AC currents are specifically designed to alter neural oscillations, not general activity. The authors found that slowing down the theta rhythm allowed for a slightly increased memory capacity, consistent with the theory that the slower theta wave could fit in more gamma waves. Conversely, speeding up the theta rhythm significantly cut the subject's memory capacity. Amazing! Control experiments that sent the TACS current over a more superficial path from the left visual field, or, even better, which asked for memorization in the right visual field rather the left one that was being stimulated, showed no significant effect.

Theory of the experiment. If the theta rhythm is slowed down (left), more gamma waves, and thus distinct working memory engrams, could be enclosed within each theta wave, increasing effective memory capacity.

The authors mention that the difference in induced theta rates (between 4 and 7 Hz), would have theoretically have allowed two more or less gamma cycles to be enclosed, thus two more items memorized in working memory at 4 Hz than at 7 Hz. The effect size was very small, (about 0.8 items more or less were memorizable), but the experimental intervention was rather diffuse and blunt as well. This kind of work helps gives specific shape to our models of what oscillations do- how they can organize information transfer and binding within the brain, while not themselves really carrying anything in their waves/waveforms.

Data. Memory retention (in terms of items remembered, vs average) is plotted on the right vs the induced theta current. See the paper for controls.

• Yes, to a job guarantee. And a detractor, making as though GDP growth were the point.
• Science? Not broken.
• Bernie on foreign policy.
• The meritocracy that devours itself. By not sharing the wealth.
• Millennials are depressed, and no wonder.
• Mainstream, supply-side economics remains wrong. We need to go back to the Keynesian model.
• The ECB saves the day, but Europe's reckoning will come nevertheless.
• More on Europe's enormous imbalances.
• Down syndrome, or no Down syndrome?
• When California's primary is too much of a good thing.
• We are seeing obstruction of justice in public, on a daily basis.