Variability of the EEG reduces when we are attending to a stimulus.  Turns out however, that this change in variability during a stimulus dependent task is highly individual and has consequences for performance. 

Ever played a video game where you had to click something to score points and sometimes you just missed?  None of us are perfectly consistent even when doing the same task.  Why is that?

In earlier posts we have talked about intraperson variability in resting state EEG recordings.  Even an eyes closed state that does not have a changing visual stimulus environment is pretty variable no matter what aspect you look at.  (see related post Eyes Open, Eyes Closed and Variability in the EEG).  This is important because as a field we often ignore this variability in analysis, perhaps finding statistical significance when it’s not there simply because we haven’t factored in this variability.   On the other hand, this variability is not surprising, because even with our eyes closed, our thoughts wander and we are never thinking the exact same thing every time we close our eyes.

But what about when we are doing the same thing over and over again?

Neural Variability Decreases in Task Paradigms

Studies such as this one by Churchland et al using implanted arrays in cats and monkeys show that the presentation of stimuli result in decreased neural variability in various states and along various dimensions. For instance, the presentation of various stimuli resulted in decreases in membrane potentials and spiking of individual neurons and correlated spiking even though the mean spike rates did not change.  Surprisingly they report this phenomenon under various conditions including anesthesia suggesting that the impact of stimuli on these dynamics is not dependent on conscious attention.

Ilan Dinstein’s lab at Ben-Gurion University in Israel has studied this phenomenon extensively in human EEG.  Indeed, like phenomenology at the individual neuron level, variability in the amplitude of the EEG also decreases when presented with visual or auditory stimulus.  The figure below from Arazi et al, 2017 shows the EEG traces pre and post stimulus presentation for 28 subjects across 600 trials of presentation of the same stimulus. Overall across channels and subjects this decrease is an average of 30%.

While the phenomenon is reasonably consistent, the interesting aspect is the variability in this decrease or quenching (of variability) brought on by stimulus.  Interestingly while the average decrease in variability was about 30% post stimulus presentation, in some folks it decreased as much as 70% and in a couple of people it actually increased!  Further, in a different but similar study, this degree of quenching was virtually identical in each subject across trials conducted a year apart suggesting that the features of stimulus dependent variability are stable characteristics of an individual. This raises the very important question of what it means for behavior.

Variability in Stimulus-induced Quenching

Since there are many different references to variability, the reduction in variability is referred to as quenching.  What does the degree of quenching have to do with how a person perceives and responds to the world?

The first paradigm studied by Dinstein and others is the threshold of discrimination.  In this paradigm, also from Arazi et al, 2017, participants were asked to determine which of two successive presentations of the same visual pattern had higher contrast, a task that was repeated across many trials with different contrast pairs.  Indeed folks with greater quenching had better discrimination capability or ability to distinguish lower levels of contrast.   This is shown by the relationship between the quenching and the slope of the contrast discrimination relationship and the discrimination threshold (contrast where the subject discriminates correctly at least 80% of time).  A steeper slope which indicates better discrimination as at each lower contrast level was positively related to the degree of quenching with r values between 0.42 and 0.48, while the discrimination threshold was negatively correlated with the degree of quenching with r values ranging between 0.32 and 0.4.

In another study they looked at reaction time in a stimulus presentation task that required the subject to respond by clicking a button.  Faster reaction times were also related to greater quenching but with lower r values than the relationships above (here 0.15 to 0.2).  We also note that these studies took care to remove non-neural sources of variability such as eyeblink and muscle movements.

Finally, in yet another study they found that people diagnosed with ADHD had greater overall variability relative to control groups.  This was true not just for post stimulus periods but pre stimulus periods as well indicating that ADHD is associated with higher variability overall rather than the degree of quenching.

All this is strong indication that amplitude variability is a meaningful measure of performance on stimulus dependent tasks and perhaps attention.  However, the low correlation values also are an indication that while variability is important, the specific spatiotemporal aspect of variability that is directly related to the task is yet uncovered and may be hidden within the overall statistics of the phenomenon.

The Causes of Variability

There are many different contributors to variability. In the EEG some may be on account of uncontrolled external stimuli.  However, much is likely internal to the system – the thermodynamic noise at the molecular level as well as noise at the network level.  While thermodynamic noise of biophysical molecules is likely relatively similar from person to person, cellular level variability may be greater. One of the puzzles of the brain is how it reliably directs behavior at a system level when the failure rates of transmission at any individual synapse are extremely high.   It would be very interesting to know (but at this point difficult to study) what type of cellular features drive this individual variability – for instance do differences in overall synaptic transmission failure rates correlate?

The pros and cons of variability

Is more variability always bad? Not necessarily.  For a system to be flexible, intrinsic noise or variability is essential.  Without it the possible configurations of exploration would be limited and the system risks getting stuck in a sub-optimal paradigm that it has little chance of escaping.  Of course, beyond a certain point the system will largely produce random outcomes, a condition that is detrimental to the survival of an organism, let alone intelligent response.  In the repetitive task paradigms of these experiments consistency may be the measure of success. However, in the less controlled and more varied stimulus environment that we inhabit in our daily life, it is of benefit to society to have some people who have less neural variability and are therefore more consistent and perhaps diligent, and others who are more variable or unpredictable in their responses and therefore perhaps more creative.