The Electrical Nature of Conscious Memory Formation and Retrieval

When you memorize something, the brain creates a nerve-impulse code to create a representation of the information represented in brain, and this code can get stored in memory. Upon retrieval, the code is replayed, and thus what the code represents becomes consciously available again as a simulation. At least that’s the theory. Until now, the evidence for this explanation has been derived mostly from rodents. But now rather direct evidence is available from humans.

In one new study, human subjects created memory associations between word pairs, while experimenters simultaneously recorded single-neuron impulses and their associated field potentials from an implanted microelectrode array in the medial temporal cortex, which is known to participate in memory formation. The EEG was also recorded from subdural electrodes implanted over the temporal cortex immediately above the microelectrode array. This allowed simultaneous observation of the local nerve impulse discharges, their associated local field potentials, and the EEG during memory formation and retrieval after a brief distraction period.

Recordings revealed the well-known relationship that EEG signals often have superimposed low-voltage high-frequency waves, which are called ripples. As expected, the ripples appeared at the same time of the impulse discharges from the microelectrodes, indicating that the impulses actually cause the small field potential changes of ripples.

In the top signal, we have the sum of a fast and slow oscillations, where the power of fast oscillation's envelope changes with the phase of the slower oscillation. The bottom signal shows only the filtered fast oscillation and the variation in its power. As it is obvious from comparison of two signals, the fast rhythm's power is always maximum at a certain coupled phase of slower oscillation (From Samiee et al.).

In the experiment, impulse burst clusters occurred throughout the presentation of word pairs while subjects were encoding the pairs. Trial-specific spike sequences observed during encoding were replayed during correct recall. As expected, ripples during recall appeared at the same time as the impulse sequences.

Not mentioned by the authors is that their findings have implications for neural correlates of consciousness. After all, forming the word-pair associations was a conscious operation. In the field of consciousness research, neural correlates are clearly evident in the EEG in that the frequency of voltage shifts predictably as the brain progresses from large slow waves during anesthesia or sleep to increasingly faster and smaller waves during alert arousal.  Relatively high frequencies (40-200 waves per second) appear more prominently when the brain is working on difficult tasks. Moreover, hard tasks are associated with more phase-locking of the EEG oscillations at different locations of the cortex.

Conscious perceptions seem to involve short- and long-range oscillations in the vertically oriented network columns in the cortex. Each column contains a local network that processes input locally in oscillatory activity that is gated at certain frequencies by inhibitory neurons in the circuit.  

At the same time, local oscillations from large pyramidal cell firings spread to distant columns both within and between the cortical hemispheres. The frequencies of this long-range activity may be slower because of the longer impulse conduction and synaptic delays. Collectively, local and distant networks interact and may likely be the basis for consciousness. The electrographic correlate is that of fast frequencies from local processing being nested within more globally generated slow frequencies. The timing phase relationships would clearly influence how much integration of local and distant processing occurs and the likelihood that the processing could be consciously perceived.

Many experiments have shown that selective attention is needed for conscious perception. Such attention activates local processing (and ripples in the local field potential). Bear in mind, however, that the ripples are not the source of processing but rather an associated manifestation of the processing that is actually occurring via the impulse timing in the local circuitry.

Two basic kinds of coupling can be seen in brainwave activity: 1) the phase of the slower frequency modulates the faster frequency, and (2) the phase coupling between two overlapping frequencies occurs when one frequency is a harmonic multiple of the other.

Conscious processing seems to be crucially dependent on the cross-frequency coherence of neural activity that can be seen at the local circuit level in multiple local sites of neocortex, hippocampus, and basal ganglia. There are different varieties of cross-frequency coupling (phase-phase, amplitude-amplitude, and phase-amplitude coupling), each of which may reflect distinctive processing. Such coherence differs across brain areas in a task-relevant manner, and changes quickly in response to sensory, motor, and cognitive events, and correlates with performance in learning tasks. Moreover, cross-frequency coherence increases with level of task demand. For example, continuous EEG recordings obtained during an arithmetic task, rest and breath focus revealed that cross-frequency alpha and theta peak-frequency coherence significantly higher when cognitive demands increased (Rodriguez-Larios and Alaerts, (2019). What is likely to remain enigmatic is how such cross-frequency coupling yields a conscious perception.

The most significant neural correlation of consciousness may prove to be time locking of nested oscillation of different frequencies whose underlying impulse patterns carry different aspects of information. The time locking of nested high- and low-frequency activity likely increases information throughput in the local circuits participating in selective attention, occludes noisy disruption from other inputs, and improves the signal-to-noise ratio of neural activity that is processing the target of attention. Parsimonious as this view might be, it still does not fully explain how a conscious percept emerges.

Sources:

Rodriguez_Larios, Julio and Alaerts, Kaat (2019). Tracking transient changes in the neural frequency architecture: harmonic relations between theta and alpha peaks facilitate cognitive performance. J. Neurosci. 7 August, 39 (32) 6291-6298; DOI: https://doi.org/10.1523/JNEUROSCI.2919-18.2019

Samiee, Sohelila et al. (2019) Phase-amplitude coupling. Nov. https://neuroimage.usc.edu/brainstorm/Tutorials/TutPac

Vaz, Alex P. et al. (2020). Replay of cortical spiking sequences during human memory retrieval. Science. 367,1131-1134.

Where We Stand in Understanding Consciousness

Many scientists, even physical scientists, assert that the Holy Grail of science is to understand human consciousness. This human state is even hard to define, but is characterized by a state in which we know what we believe, know, and imagine, know what we decide and plan, and feel what we feel. That explains nothing.

The problem in understanding is not only that the mechanisms must surely be complicated, but also that we don’t have good non-invasive experimental tools. There are only two useful tools, a metabolic proxy of neural electrical activity (functional fMRI) and scalp monitoring of electrical activity (the electroencephalogram {EEG), or its magnetic field counterpart. Among the problems with fMRI are that it is only an indirect measure of the actual signaling within the brain that generates thought and feeling and enables consciousness. Its time resolution is about one-second or more, whereas signaling in the brain occurs on a millisecond scale. Although the EEG monitors activity on the appropriate time scale, it has very poor spatial resolution, inasmuch as voltage fields over various regions of cortex overlap, because the voltage extends in progressively diminished amplitude throughout the conductive medium of brain from its source of generation to other source generators. Although the EEG does monitor the appropriate target (electrical activity), that activity is an envelope of the algebraically summed signals from heterogeneous neuronal ensembles, which are nerve impulses and their associated postsynaptic potentials nearest the sensing electrodes.

                                             By Davidboyashi - Own work, CC BY-SA 4.0

Nonetheless, we do know many useful things about brain function that are surely involved in conscious functioning. Neuroscientists have discovered much of this in lower animals from invasive procedures that are not permissible in humans. In summary, we can list the following brain functions that are relevant to consciousness:

• The brain is a network of richly inter-connected networks.
• Functions are modular. Different networks have different and shifting primary functions, and some may be selectively recruited when their function is needed.
•  Some networks can perform multiple functions, depending on which other networks have recruited them into action.
• Some aspects of functional connectivity of different networks differ in unconscious and conscious states.
• Wakefulness and consciousness are not the same. Wakefulness is necessary but not sufficient for consciousness.
•  A great deal has been learned about the neural mechanisms causing wakefulness but that has not helped much in understanding consciousness.
• The messaging signals of brain are nerve impulses and their neurotransmitter postsynaptic effects.
• The summed voltages of the messaging have electrostatic effects that alter the excitability of the neurons within the voltage field.
• The frequency of bursts of impulses and their EEG envelope impose important effects on gating and throughput of information as it propagates and is modified throughout the global workspace of networks.
• There are multiple neural correlates of consciousness, but we have not identified with certainty which ones are necessary and sufficient for consciousness.

Oscillatory electrical activity is thought to have a key role in selective routing of information in the brain. Oscillations seem to modulate excitability, depending on phase relationships of linked neuronal ensembles. Two prominent hypotheses have been advanced as crucial for consciousness, and they are not mutually exclusive:

• Phase-locked activity in two or more ensembles (coherence)
• Inhibitory gating that directs pathways for propagation within networks.

The key to discovering mechanisms of consciousness is to identify all the neural correlates and then winnow the list to those that are both necessary and sufficient for consciousness. Sometimes, important discoveries occur when you study the opposite of what you want to study. This principle is manifest in studies on brain function during various states of unconsciousness (like anesthesia, coma, or non-dream sleep). A recent review of research compared the neural correlates of unconsciousness with those of consciousness. The evaluation showed disrupted connectivity in the brain and greater modularity during unconscious state, which inhibited the efficient integration of information required during consciousness. Additionally, the review made the key point that the neural correlates of consciousness that matter are the ones that occur in consciousness but not in unconscious states. Of particular relevance are the correlates related to functional connectivity among networks, because multiple lines of evidence reveal that this connectivity degrades during unconscious states and returns when consciousness resumes.

In rodents, multi-array recordings in visual cortex indicate that connectivity patterns are the same during anesthesia as in wakefulness. Perhaps this indicates that rodents do not have the needed network architecture to enable consciousness. They can be awake but not conscious. Being awake is clearly necessary for consciousness, but not sufficient. In addition (if you don’t believe me, see the classical U-tube basketball-game video on inattentional blindness). At any given instant, we are only consciously aware of the specific cognitive targets to which we attend.

 Statistical co-variation of activity in linked networks is a measure of functional connectivity. The activity in linked networks may randomly jitter or be in phase or locked at certain time lags. Operationally, the connectivity may enable one group of neurons to mediate or modulate activity in another for past, present, or future operations. The temporal dynamics of these processes differ depending on the state of consciousness.

A very popular view on consciousness among neuroscientists these days is that higher-order thinking, especially conscious thinking, is mediated by extracellular voltage fields that oscillate in the range of 12 to 60  or more waves per second. Changes in oscillatory frequency and coherent coupling of the oscillations among various pools of neurons are thought to reflect the nature and intensity of thought.

The issue arises as to how these voltages, commonly called field potentials, can influence the underlying nerve impulse activity that causes the oscillation in the first place. The messages of thought are carried in patterns of nerve impulses flowing in neural networks. Field potentials are not signaling, at least  not directly. They may well indirectly influence messaging by electrostatically biasing networks to be more or less able to generate and propagate nerve impulse traffic.

Neuroscientists attach much importance to the temporal dynamics of EEG voltage frequencies. For example, at one time neuroscientists believed that 40/sec synchrony was critical to consciousness, but later studies revealed that this synchrony can be maintained and even enhanced during anesthesia. Later, investigators thought they had found a crucial role for higher frequency gamma synchrony, but that too is now called into question. This gamma synchrony can be present or even enhanced during unconsciousness. However, the spatial extent of synchrony may be the meaningful correlate of consciousness. Widespread synchrony breaks down during unconsciousness, while more localized synchrony remains intact or even enhanced.

Numerous studies show a breakdown of functional connectivity during various states of unconsciousness. For example, fMRIs reveal cortico-cortical and thalamocortical disconnections during sleep, general anesthesia, and pathological states. EEG analysis shows similar connectivity breakdowns. Additionally, the repertoire of possible connectivity configurations that can be accessed diminishes during unconscious states and is restored as consciousness resumes. This obviously limits the robustness of information processing that can occur in unconsciousness. Conscious selective attentiveness likely requires a different repertoire of connectivity than inattentive consciousness.

Neuroscientists are also discovering the importance not only of multi-area coherence at a given frequency band, but also that the phase synchrony to two different frequencies can also modulate network communication. Cross-frequency coupling of the alpha and beta oscillations with higher frequency gamma oscillations can amplify, inhibit, or gate the flow of nerve impulses throughout circuitry.

Future advancements will surely include more emphasis on monitoring functional connectivity as the brain shifts into and out of various states of consciousness and unconsciousness. I think, however, that we will not make definitive progress in consciousness research until we make progress in one area of theory and another of tactical methodology.

The theory deficiency lies in models of neural networks. Computer models of man-made networks yield interesting results, but they are probably  not relevant. Brains do not work with the same principles that computers do. Moreover, brain networks have intrinsic plasticity that cannot yet be duplicated by computers.

The method deficiency is that we have no non-invasive way to monitor the actually signaling in even a significant fraction of all the neurons in all the networks. Moreover, even if we had a way to monitor individual neurons noninvasively, it would likely be necessary to selectively monitor neurons in defined circuits. Ultimately, we may confirm that some things are just not knowable. Surely, however, we can learn more than we do now.

Sources:

Bonnefond, Mathilde et al. (2017). Communication between brain areas base on nested oscillators. eNeuro. 10 March. 4(2) ENEURO.0153-16.2017. doi: https://doi.org/101523/ENEURO.0153-16.2017.

Mashour, George A., and Hudetz, Anthony G. (2018). Neural correlates of unconsciousness in large-scale brain networks. Trends in Neurosciences. 41(3), 150-160.