The “Edge-of-Chaos”: Brain Activity Underlying Consciousness

Post by Lani Cupo

The takeaway

Modeling the electrical brain activity underlying stages of consciousness reveals that the conscious brain exhibits activity poised at the “edge-of-chaos”, a critical point between stability and chaos. The loss of consciousness, such as with anaesthesia, corresponds with a transition away from the critical point, while psychedelics induce a state closer to the critical point.

What's the science?

As scientists seek to understand consciousness, they investigate patterns of electrical brain activity during various stages of consciousness, such as during generalized seizures, under anaesthesia, and after exposure to lysergic acid diethylamide (LSD). In doing so, they can examine the transition of brain activity from mathematically stable to chaotic dynamics. While previous research suggests that the conscious brain’s electrical activity exists at a critical point at the boundary of stability and chaos, it is yet unknown what phases (from a mathematical perspective) exist on either side of the critical point of wakefulness. This week in PNAS, Toker and colleagues sought to provide empirical evidence for the cortical dynamics (patterns of electrical brain activity) that underlie stages approaching consciousness and how these patterns relate to information richness.

How did they do it?

The researchers used a previously published model of low-frequency electrical activity in the brain reflecting cortical oscillations that allow for tuning of parameters associated with neuronal inhibition and excitation. By setting the parameters of the model based on the literature, the authors could simulate data for different brain states, including waking consciousness, generalized seizure, and anaesthesia, which have been previously validated in the model with acquired data. In the data they simulate, the authors assessed chaotic dynamics with the largest Lyapunov exponent, where a positive exponent corresponds with chaos and a negative exponent corresponds with periodicity (ordered pattern). The authors applied a modified 1-0 chaos test to this value where a value of 1 is assigned to chaotic systems and 0 to periodic systems. They also assessed the richness of information in the model with a measure known as the Lempel-Ziv complexity, which gives an estimate of the amount of non redundant information in a signal—in this case brain activity. The authors could then relate measures of chaos to measures of information richness, comparing between models of various states of consciousness. They also performed a “parameter sweep” where they simulated data with diverse parameters not corresponding with a specific brain state to explore the relationship between chaos and richness across the models. In addition to the simulated data for waking consciousness, generalized seizure, and anaesthesia, they also examined previously published data from two macaques and five humans during wakefulness, two macaques and three humans under anaesthesia, three humans experiencing generalized seizures, and 16 people after exposure to either saline or LSD.

What did they find?

The authors hypothesized that if they plotted the measure of information richness on the y-axis and a transition from periodicity to chaos on the x-axis of a plot, they would observe an inverse-U shaped curve, indicating that the richest information was correlated with the critical point at the edge-of-chaos, and information was lost as the system became more periodic or more chaotic. In line with their hypotheses, the authors found an inverse-U relationship between chaos and richness, with 0 on the x-axis representing the critical point of edge-of-chaos. Visualizing the simulated models for wakefulness, seizure, and anaesthesia reveals that wakefulness falls to the right of 0, towards the side representing chaos. This finding supports an old hypothesis that at a large scale, the brain’s electrodynamic system is at least weakly chaotic. Anaesthesia fell farther towards instability, but seizure fell on the periodic side of the graph, as predicted. Data from participants exposed to LSD suggests the psychedelic increases the information richness of the system and stabilizes it, moving the patterns of activity closer to the critical point when compared to consciousness.

What's the impact?

This study found a relationship between the chaotic dynamics of brain activity and the complexity of information in models representing the brain across different states of consciousness, suggesting wakefulness occupies a critical point between chaos and periodicity. The findings provide information to better understand states of consciousness, both in healthy brains (such as during sleep) and disorders related to consciousness.  

Access the original scientific publication here.

Selective Attention Modulates Activity in the Auditory Nerve

Post by Lina Teichmann

The takeaway

Studying the effects of selective attention on subcortical structures is usually not feasible in humans, however, testing cochlear implant (CI) users offers a unique opportunity to examine how top-down effects modulate activity in the auditory nerve. Using a cross-modal attention task, the current study shows that activity in the auditory nerve in humans is modulated by attention.

What's the science?

Attention relies on selecting relevant features from our environment while ignoring irrelevant features. For example, when listening to someone speak, we are able to focus our attention on the words they are saying while ignoring irrelevant background sounds. Direct evidence from animal studies suggests that these attentional mechanisms modulate auditory nerve action potentials. Studying similar effects in humans is usually difficult, as direct recordings from the auditory nerve are generally not feasible. However, this week in the Journal of Neuroscience, Gehmacher, Reisinger and colleagues present data from CI users, showing that auditory nerve activity in humans is modulated by attention.

How did they do it?

A group of CI users completed a cross-modal attention task while recordings were taken from a coil that temporarily replaced their CI. In every trial, participants saw a cue on a computer screen (either an eye or ear) to indicate whether to attend to an auditory or visual stimulus. Then an audiovisual stimulus was presented. The auditory stimulus was a tone delivered directly to the CI coil. The visual stimulus was a circle with black and white stripes oriented vertically. In some trials, oddball auditory (slightly different tone) and visual (slightly tilted version of visual stimulus) stimuli were presented, and participants were asked to press a button when they detected an oddball in the cued domain. 

What did they find?

Using a frequency analysis, the results showed that cochlear activity was modulated by selective attention. In the theta frequency range (5-8Hz), a higher power was associated with attending to the auditory domain. Relating these results to a concurrently recorded electroencephalography (EEG) dataset from one participant, the authors showed that the auditory nerve, as opposed to a source located elsewhere in the brain, was the most likely origin of the signal. Lastly, the authors showed that classification algorithms trained on single-trial activity recorded from the CI could distinguish whether the participant was attending to the visual or auditory stimulus. Together, these results support the hypothesis that auditory nerve activity is modulated by attention in humans.

What's the impact?

Previous work has shown that the neural signal is modulated by attention at the cortical level. However, evidence for attentional modulation in subcortical structures such as the cochlea was scarce, partially because direct recordings in humans are usually not feasible. The current study addressed this gap in the literature by studying auditory nerve activity directly in CI users. The results highlight that auditory nerve activity is modulated by attention in humans, providing new insights into the interplay between top-down and bottom-up effects in hearing.

Access the original scientific publication here.

High Cognitive Load is Associated with Increased Associative Interference

Post by Shireen Parimoo

The takeaway

Associative interference occurs when our prior knowledge interferes with our memory for newly learned associative information. This effect is enhanced when processing resources are reduced under high cognitive load. 

What's the science?

Our brain functions by linking related items in memory - otherwise known as associative memory. Sometimes, our prior knowledge or associations can hinder our ability to learn new associations, an effect known as associative interference. How our brain reacts to associative interference when cognitive resources are low, is not clear. This week in Scientific Reports, Baror and Bar conducted a series of associative memory tests with varying levels of cognitive load and memory demands to investigate the impact of reduced processing capacity on associative interference.

How did they do it?

The authors conducted several memory experiments that assessed associative interference under different levels of cognitive load using explicit (Exp. 1-3) and implicit (Exp. 4) memory paradigms. In Exp. 1a and 1b, participants first intentionally learned word pairs (learning phase) that consisted of semantically related (e.g., Salt-Pepper) or unrelated words (e.g., Salt-Mouse). They then completed a cued recognition test in which a cue word (e.g., Salt) was followed by a target and distractor. There were three conditions based on the semantic relatedness of the cue-distractor pair: (i) related target, unrelated distractor (Pepper/Tree), (ii) unrelated target, unrelated distractor (Mouse/Tree), and (iii) unrelated target, related distractor (Mouse/Cheese). During the test phase, participants also performed a working memory task ranging from low to high cognitive load. High cognitive load was hypothesized to reduce the processing resources available for memory encoding. Exp. 1b included an additional block of learning and test phase trials that were expected to further reduce processing resources over time.

In Exp. 2 and 3, rather than relying on prior knowledge, the authors assessed associative interference from incidentally learned associations. In Exp. 2, individual words were sequentially presented in different colored fonts and participants were instructed to associate consecutive words that appeared in the same color (cue-target pair; intentional learning). These word pairs appeared four times throughout the learning phase and were always preceded by the same word in a different color (pre-cue word), forming an incidental association with the cue word. In the cued recognition test, the distractor was either the pre-cue word or an unrelated word. As before, participants completed a working memory task with low and high cognitive load during the test phase. Exp. 3 was similar, except participants learned associations between pairs of pictures (intentional learning) that were always preceded by the same pre-cue picture (incidental learning).

Lastly, the authors used a contextual priming task to assess associative interference under implicit memory conditions (Exp. 4). They used prime-target pairs that were unrelated or were weakly, moderately, or strongly related to each other. Participants provided object/non-object judgments for the targets while concurrently performing the digit span task under low and high cognitive load. Reaction times to target object recognition were examined as a function of cognitive load and prime-target relatedness.

What did they find?

Target recognition was generally higher under low cognitive load than high cognitive load (Exp. 1 and 2). However, the effect of cognitive load was only present when the distractor was related to the cue word. Thus, reduced processing capacity under high load led to interference from previously learned associations between the cue and distractor. Similarly, reduced processing resources from completing an additional block of the experiment (Exp. 1b) only affected memory when the distractor was related to the cue, but not when the distractor was unrelated to the cue. Together, these results indicate that reducing processing resources by increasing cognitive load and time on task independently contribute to associative interference during recognition.

A similar pattern of results emerged when cue-distractor associations were incidentally learned during the learning phase (Exp. 2 and 3). There was no load effect on target memory with unrelated distractors, but memory accuracy was reduced under high load when the distractor was incidentally associated with the cue. High cognitive load, therefore, interfered with associative retrieval and generalized to both words and pictures. Finally, participants were faster to identify objects that were related to the prime than those that were unrelated (Exp. 4). Interestingly, object recognition for strongly related targets was fastest under low load but slowest under high load, suggesting that reduced cognitive processing capacity also delays the perceptual processing of strongly related information.

What's the impact?

The results of this study provide evidence in favor of the idea that decreasing the available processing resources increases associative interference in memory. These findings are important for informing social and educational domains, where increased stress or too much cognitive load might result in biasing towards previously learned, and potentially misleading information.

Access the original scientific publication here.