Pinging the brain to reveal hidden working memory states

Maintaining information for short periods of time in working memory, without its existence in the outer world, is crucial for everyday life, allowing us to move beyond simple, reflexive actions, and towards complex, goal-directed behaviours. It has been the consensus that the continuous activity of specific neurons are responsible to keep these information “online” until they are no longer required. However, this classic theory has been questioned more recently. Working memories that are not actively rehearsed seem to be maintained in an “activity-silent” network, eliciting no measurable neural activity, suggesting that it is the short-term changes in the neural wiring patterns that is responsible for their maintenance. These memories are thus hidden from conventional measuring techniques making it difficult to research them.
This thesis proposes an approach to reveal hidden working memories that is analogues to active sonar: Hidden structures can be inferred from the echo of a “ping”. Similarly, by pushing a wave of activity through the silent neural network via external stimulation (for example a white flash), the resulting recording patterns expose the previously hidden memories held in said network. This approach is demonstrated in a series of experiments where both visual and auditory working memories are revealed. It is also used to reconstruct specific working memories with high-fidelity after different maintenance periods, showing that the maintenance of even a single piece of information is by no means perfect, as it tends to randomly and gradually transform within 1 to 2 seconds (for example purple becomes blue).