In recent years, the field of cognitive neuroscience has experienced a paradigm shift. Moving from controlled laboratory settings to more naturalistic, real-world environments. This transition is driven by a growing interest in human cognition in unrestricted contexts.
The leap from a controlled to a real-world environment was made possible by significant advancements in tech. Innovations such as the miniaturization of sensors and the development of portable EEG devices, have played a big role.
In this study, scientists used one such device – mbt’s Smartfones, which are a powerful sensor used for unobtrusive recordings. Smartfones allowed them to collect EEG during a setup modulating mental workload levels.
Technological breakthroughs have not only enhanced the ecological validity of cognitive research but also expanded its applicability in diverse fields such as safety, healthcare, and human-machine interaction.
Manipulating Mental Workload using multi-tasking
Researchers in the study decided to use multi-tasking to achieve different mental workload levels. They did this by varying the number of simultaneous tasks the participant had to perform at once. They recruited 11 participants for the analysis.
The first level consisted of 1 task – Auditory oddball task.
For the second and third level, scientist added tasks from the MATB (Multi-Attribute Task Battery) software. MATB is a simulation of tasks usually performed by the cockpit crew during flights.
Single-Task: includes only the auditory oddball task. Dual-Task: adds a MATB task to the auditory one. This task requires you to track the target with your aim. Multi-Task: Including 2 more MATB tasks. One requires you to track the green and red lights + bar levels below the light indicators. The second one is a resource management task, where you track the level of liquid in the tanks. In total, this level has 4 tasks active at the same time.Auditory Oddball task – participant performance
The oddball paradigm is used as a secondary task to measure mental workload. Participants hear 2 different tones in a random fashion, where one of the tones is a target one. Target tones occur at a ratio of 1:4 compared to non-target (distractor) tones. They had to count the target tone only, during all 3 levels of difficulty.
So, how did the participants perform?
The ability to count target tones was affected by how complex the primary task was (MATB). More counting errors occurred as the primary task became harder. The fewest errors happened in the Single-task, while the most errors occurred in the Multi-task.
Performance on the oddball task can show the level of resources used for the primary task, but it’s not a practical marker in most settings. Imagine doing your job, but having to count high-pitched tones to estimate your workload all the time. It’s not very convenient over a long period of time.
That’s where EEG measures can come in handy.
Task difficulty effects on P300 amplitude for Target vs. Distractor sound
First step was to look into ERPs (Event Related Potentials) appearance on different electrode sites. The ERP component in question is P300.
Smartfones have electrodes in the earcups and 3 central electrodes in the headset’s rim. You can find the average P300 response on all electrodes below:
A grand average ERP waveform (N = 11 participants) for each channel + different difficulty task levels (single-, dual-, muti-). We can see that P300 response is mostly visible on central electrodes, so for this reason further analysis is done using data on these sites.From the previous image, we can see that P300 amplitude seems to decrease as the demands of the primary task increase. Let’s take a more detailed look at this behavior.
(A) Grand average ERP waveforms recorded at Cz electrode to target and distractor sound, during different task levels. To reduce the number of factors to the analysis, they also calculated a Difference waveform between target vs distractor (B). We can also see a spectrogram representing frequency changes around the stimulus time stamp. There’s decreased activity in delta and theta bands as task level increases. (C) Lastly, we look at the spatial distribution of P300 amplitude over electrode sites. This plot confirms that central electrodes best show changes in EEG to all task levels.Looking at Difference waveform (target vs distractor ERP), we can see participants distinguished less and less between the two, as task load got bigger.
Can classifiers successfully distinguish Target vs Distractor ERPs?
The classifier was trained on ERP responses to target + distractor stimuli for each workload condition. The mean achieved accuracy was around 61% across all task load conditions, but only single-task accuracy was above chance level.
These results highlight the inconsistency in classification of ERPs elicited by different auditory stimuli.
How about using ERPs to classify workload levels?
This time, the classifier was trained only on ERP responses to target stimuli. The classification is binary, between pairs of workload levels: single-dual, single-multi, dual-multi task.
Researchers were able to consistently classify different workload levels with mean accuracies from 80-87%. You can see the classifier’s performance for each participant in the table below:
Signal-to-Noise ratio (SNR) analysis over time
Researches in the study also wanted to confirm that changes in P300 amplitude were not caused by increased participant motion in high task loads.
They chose to observe SNR changes over time to make sure artifact bursts didn’t affect the results. To be more specific, they looked into the RMS amplitude of an averaged signal over all electrodes.
We can see RMS changes over trials and participants in this plot. These SNR changes don’t reveal improvement nor degradation of the EEG signal, which confirms artifacts didn’t affect P300 changes across task conditions.
So, the take-home message is that unobtrusive EEG sensors such as Smartfones can capture neural signals related to auditory processing, as well as mental workload changes. If you’re interested in more detail about the study findings, you can read the original paper here – https://www.sciencedirect.com/science/article/abs/pii/S0166432823005454
Curious if this setup is right for your study?
Our team works with researchers across neuroscience, sports science and clinical applications - we'll help you figure out if mobile EEG fits your study design.
