Week Three: EEG and ERP

Limitations of EEG

- EEG is biased to signals generated in superficial layers of cerebral cortex on the gyri (ridges) directly bordering the skull - signals in the sulci are harder to detect than from the gyri, and may additionally be masked by the signals from the gyri - The meninges, cerebrospinal fluid (CSF) and skull 'smear' the EEG signal making it difficult to localise the source. This is known as the inverse problem: mathematically, if the sources are known, the resulting scalp configuration of signals can be reconstructed, but the reverse is not true - one given scalp configuration of signals can have multiple dipole solutions.

Analysing the signal in EEG data

- EEG signals measured from the scalp in relation to the reference electrode - the reference should be a neutral point (e.g. tip of the nose) but some people reference to the average of all scalp electrodes - EEG signals have a typical amplitude of 10uV to 100uV (SHOULD BE MEW SYMBOL) - these are tiny signals. This means they need to be amplified, typically by a factor of 1,000 to 10,000 - the signal is then typically digitalised, typical sample frequency is 256-1024Hz but can be >4000Hz - signal is band-pass filtered to remove the low (<0.5-1Hz) and high (typically >35-70Hz) frequencies because they cannot reflect brain activity - Signal is also notch-filtered (at 50Hz or 60Hz) to remove line noise which is also not brain activity

Analysing frequencies in the EEG

- when looking at frequency info, e.g. in sleep research, the raw signal might show systematic variations, and more of a specific frequency - the EEG is the sum of signals originating from many neural units - the important part is finding all the artefacts that are not brain signals. This can be a nightmare because there's so much going on so recording EEG is really hard

Two distinct advantages of ERPs over behavioural measures

1. An overt response reflecting the output of a large number of individual cognitive processes, variations in reaction time, and accuracy are difficult to attribute to variations in a specific cognitive process, but ERPs provide a continuous measure of processing between a stimulus and a response which makes it possible to determine which stages of processing are affected by a specific experimental manipulation. 2. An advantage of behavioural measures is that they can provide an online measure of the processing stimuli even when there is no behavioural response.

Disadvantages of ERPs compared to behavioural measures

1. the functional significance of an ERP component is virtually never as clear as the functional significance of a behavioural response. In most cases we don't know the specific biophysical events that underlie the production of a given ERP response or the consequences of those events for information processing. But when a computer records a button-press response we have a much clearer understanding of what the signal means. 2. ERPs are so small that is usually requires a large number of trials to measure them accurately. In most experiments, a reaction time difference can be observed with only abut twenty to thirty trials per subject in each condition, whereas ERP effects often require fifty, 100 or even 1000 trials per subject in each condition.

Inverse problem and forward problem

Aren't able to tell the locations and orientations of the diploes if given an observed voltage distribution, even though you can use equations to compute the distribution of voltage that would be observed if given the locations and orientations of a set of diploes (called the forward problem).

Artefacts in the EEG

Artefacts such as eyeblinks may contaminate the EEG and this problem can be addressed by identifying and removing trials with artefacts or subtracting an estimate of the artifactual activity from the EEG. The size and timing of ERP components is measured, and these measures are subjected to statistical analysis. - eye blinks ad movements have strong impacts on the EEG signal because the eye can be regarded as a dipole itself - signals originating from the eye will contaminate the signal of interest - and unfortunately will be much larger - these signals will be recorded by placing electrodes next to and under the eye to capture horizontal and vertical eye movement - eye related signals can then be removed by excluding contaminated trials or mathematical algorithms, such as independent component analysis (ICA)

Dipole

Dipole is a pair of positive and negative electrical charges separated by a small distance. Dipole from a single neuron is so small it wouldn't be possible to record it from a distant scalp electrode, but under certain conditions the dipoles from many neurons summate which makes it possible to record the resulting voltage at the scalp. For summated voltages to be recorded at the scalp they must occur at approx. the same time across thousands or millions of neurons and the dipoles from the individual neurons must be spatially aligned. The summation of the individual dipoles is complicated by the fact that the cortex is not flat but has folds, but fortunately physicists have demonstrated that the summation of many dipoles is essentially equivalent to a single dipole formed by averaging the orientations of the individual dipoles (called equivalent current dipole - ECD).

Duration of an action potential vs postynaptic potential

Duration of an action potential is usually 1ms, while duration of postsynaptic potentials last tens of hundreds of milliseconds and they are largely confined to the dendrites and cell body, occurring essentially instantaneously rather than travelling down the axon at a fixed rate.

Gehring et al 1993: what they were looking for and what they did

Investigated the ERN (error-related negativity), asking whether there is a cognitive mechanism for the detection of and compensation for errors. ERN is a negative deflection of up to 10 microvolts in amplitude observed at central electrodes 80-100ms after an erroneous response. Asked participants to emphasise accuracy or speed in a simple Flanker-task where participants had to respond to the central letter on the screen. Incongruent displays should lead to more errors and error detection should only matter in the accuracy condition.

Why use EEG

It has several advantages over other methods for investigating cognition, temporal resolution is great, but the spatial resolution is not so good

Mismatch negativity

Observed when subjects are exposed to a repetitive train of identical stimuli with occasional mismatching stimuli which elicits a negative-going wave that is largest at central midline scalp sites and typically peaks around 160 and 220ms.

Visual search: serial application vs parallel processing

Often the amount of time to detect increases as the number of objects in the stimulus array increases: some investigator say this reflects the serial application of attention to the individual objects in the array, other investigators argue it may be due to limitations in the processing capacity of parallel processing system that identifies multiple objects concurrently.

Using ERPs to investigate cognition

One reason for using ERPs to investigate cognition is that many components are very well studied, so finding a specific component that is modulated by the experimental task might shed light on what cognitive process is involved.

Gehring et al 1993: what they found

Overall, they found a clear ERN on incorrect trials in comparison to correct trials. They used muscle activity (measured using electromyogram) to determine response onsets. ERN was strongest when people emphasised accuracy and weakest for speed. But is the ERN indicative for compensating for errors? If that were true, one would expect that the ERN should also reflect the attempt to break the error. The ERNs from the entire experiment can be divided into quantities from small to extra-large. They then investigated how ERNs of different sizes were related to response parameters, which might in turn be related to correcting or avoiding errors. The greater the ERN, the lower the response force (trying to correct for error). The greater the ERN the higher the probability to get it right on the next trial (successful learning from errors). The greater the ERN the slower the response on the next trial (successful learning from errors).

Experiment without the artificial bias in the order

Possible that the probability manipulation caused participants to use a serial search strategy when they normally wouldn't so a third experiment was conducted without the means of artificial bias. Attention switches rapidly among non-target items even without artificial methods to bias the search order.

Neurophysiology of EEG

The EEG activity does not reflect action potentials but originates mostly from post-synaptic potentials - voltages that arise when neurotransmitters bind to receptors on the membrane of the post-synaptic cell. This causes ion channels to open or close, leading to graded changes in the potential across the membrane. This can be understood as a small 'dipole'. Signals from single cells are not strong enough to be recorded outside of the head, but if many neurons spatially align, then their summed potentials add up and create signals we can record. The pooled activity from groups of similarly orientated neurons mostly comes from large cortical pyramid cells.The functional unit is >10,000 simultaneously. The orientation of neurons determines the sign of the recorded potentials. Some orientations lead to signals that cannot be recorded (might cancel each other out)

What EEG is useful with and what the drawbacks are

The EEG has proven to be very useful in both scientific and clinical applications, but in raw form it is a very coarse measure of brain activity and it is very hard to use to assess the highly specific neural processes that are the focus of cognitive neuroscience. The drawback is that is represents a mixed conglomeration of hundreds of sources of activity making it hard to isolate individual neuro-cognitive processes. But it is possible to extract specific responses from the overall EEG through averaging technique - these specific responses are called event-related potentials because they are electrical potentials associated with specific events.

N2pc

The N2pc provides a continuous measure of the distribution of attention through its lateralised distribution - it is a negative-going voltage deflection typically observed 200-300ms after the onset of a visual search array and is largest over the areas of visual cortex in the hemisphere contralateral to the location of the attended object. If a visual search involves rapid, serial shifts of attention the N2pc component should shift rapidly between hemispheres as attention shifts rapidly between the hemifields.

Order of ERP components

The first major visual ERP component is usually called the C1 wave, which appears to be generated in area V1 (primary visual cortex). C1 wave is followed by a P1 wave which is largest at lateral occipital electrode sites and typically onsets 60-90ms post-stimulus with a peak between 100-130ms. P1 onset time is difficult to assess accurately due to overlap with C1 wave. Like C1 wave, P1 wave is sensitive to variations in stimulus parameter. N1 wave comes next, earliest subcomponent peaks 100-150ms post stimulus at anterior electrode sites, and there are two posterior N1 components that usually peak 150-200ms post stimulus. P2 comes next at anterior and central scalp sites.

Location and orientation of cortical generator source of ERP component

The location and orientation of the cortical generator source of an ERP component has a huge influence on the size of that component at a given scalp electrode site. Each individual has a unique pattern of cortical folding and the relationship between functional areas and specific locations on the gryrus or in a sulcus may also vary. Other factors that may influence the shape of the waveforms including drugs, age, psychopathology, even personality.

Woodman & Luck: visual search

asked participants to search for a target, a coloured square, which is open to the left. They were interested in whether people search all locations in parallel (all at the same time) or in serial fashion (one after the other). The idea was that the N2pc could help answering this question: only if people search in serial fashion should attention switch from one hemifield to the other, until the target is found. If the search is parallel, nothing should change.

Electromagnetic measures

excellent invasiveness, undefined/poor (ERPs) and undefined/better (ERMFs) spatial resolution, excellent temporal resolution, inexpensive ERPs and expensive ERMFs.

Hemodynamic measures

good (PET - issues with radiation) and excellent (fMRI) invasiveness, good spatial resolution, poor temporal resolution, and expensive cost.

Micro-electrode measures

poor invasiveness (requires inserting an electrode into the brain and therefore are limited to nonhuman species or in rare cases neurosurgery patients), excellent spatial resolution, excellent temporal resolution, and fairly expensive cost

Woodman & Luck: manipulating the target

Because it is impossible to determine which order objects are searched, a modified visual search paradigm was used so that subjects were biased to search the objects in a known order. Each search array contained four coloured squares with one in each quadrant along with 21 black distractor squares, the target (a square with a gap on the left side) was presented in 50% of trials. To bias the order squares were searched, target was presented in one prespecific colour in 75% of target-presented trials and in another prespecific colour in the remaining 25% - called C75 and C25. Subjects detected the target about 80ms faster when it was C75 with it being statistically significant. This result is consistent with both serial and parallel models of attention, but they make different predictions when considering the N2pc component. When C75 and C25 were in opposite hemifields, serial models would predict the N2pc component would appear first over the hemisphere contralateral to C75 and then shift to the hemisphere contralateral to C25 except if C75 was the target in which case attention would remain on that item. Then typical parallel models would predict N2pc component would not switch rapidly between hemispheres but instead would be consistently larger over the hemisphere contralateral to C75 because there would be a greater number of resources processing this item.e scanning the visual field.

Major advancements in EEG

But if subjects were required to press a button when they detected the large negative voltage observed at frontal electrode signalling the period that separated the warning signal and the target. The negative voltage was clearly not just a response, it appeared to reflect the subject's preparation for the upcoming target, which was an exciting new finding allowing researchers to start exploring cognitive ERP components. Next major advance was the discovery of the P3 component, they found that when subjects could not predict whether the next stimulus was auditory or visual, the stimulus elicited a large positive P3 component that peaked around 300ms post stimulus and the component got smaller when the modality of the stimulus was predictable. Over the next 15 years a lot of research on identifying various ERP component and methods for recording and analysing ERPs in cognitive components.

What is EEG? How it works and what it does

EEG is a method of detecting neural activity by placing electrodes on the scalp. These electrodes pick up small fluctuations of electrical signals, originating from activity of (mostly cortical) neurons. While the raw signals recorded are very noisy and might not look like much, they are systematically related to cognitive processes. This means we can use these signals to learn something about cognition when people perform tasks. EEG recorded at the scalp is non-invasive, but it is possible to record intra-cranial EEG by measuring activity directly at the exposed cortex. It is cheap and relatively easy to conduct.

Names of ERPs

ERPs are usually given labels like P1 and N1 that refer to their polarity and position within the waveform.

Recording EEG

Electrode cap used, amplifier, experimental stimulation (people need to do a task), EEG recording. To conduct an experiment you first have to attach some kind of electrodes to the subject's scalp to pick up the EEG. It must be filtered and amplified so it can be stored as a set of discrete voltage measurements on a computer.

More history of EEG

First unambiguous sensory ERP recordings from awake humans performed 1935-1936 by Pauline and Hallowell Davis. Modern era of research began in 1964 when Grey Walter and colleagues reported the first cognitive ERP component - in this study subjects were presented with a warning signal like a click followed 500 or 1000ms later by a target stimulus like a series of flashes. In the absence of a task, each of these stimuli elicited the sort of sensory ERP response one would expect for these stimuli.

Woodman & Luck: what they did and what they found

First, they tested what happened on trials where there was no target. When targets were absent, they found (C75) and (C25) were in opposite hemifields so the N2pc shifted. This is in line with the idea of a serial search (first the side of the more likely stimulus, then the side of the less likely stimulus). In the results, as predicted by serial search models, the N2pc was more negative over hemisphere contralateral to C75 from 200-300ms post-stimulus and then became more negative over the hemisphere contralateral to C25 from 300-450ms post stimulus. When C75 and C25 were both in the same hemifield the N2pc remained contralateral to that hemifield until the end of the recording epoch which is consistent with both serial and parallel models.

History of EEG: Hans Berger

Hans Berger (1873-1941) detected the first EEG signal in 1924 with electrodes attached to the scalp of a human (his wife) and reported the results in 1929. Berger initially studied medicine because he was convinced there is a 'psychic energy' which might allow for telepathy. He wanted to discover the objective activity in the brain and 'psychic phenomena' but he did not realise the basis and potential of his discovery at the time. Also the first to describe the alpha rhythm - when people close their eyes the electrical signal was not constant, but it varied with a characteristic frequency of 8-13 hz. Initially, he used two electrodes, one attached to the front of the head and one to the rear and recorded the potential (voltage) difference between them. Electrodes were silver wires placed under the scalp.

Trial with a single red item

The results could be explained by an intrinsically slower N2pc time course for the far item than for the near item, so trials with a single red item were included and the trials has the same time course whether it was near of far.

Studying cognition with ERPS Woodman & Luck 1999: using N2pc

This paper aims to use an electrophysiological marker of the moment-by-moment direction of attention (the N2pc component of the ERP waveform) to show that attention shifts rapidly among objects during visual search. Used the N2pc (second negativity, posterior contralateral), which is known to index attention: it is strongest over the posterior cortex contralateral to where the observer is attending.

Two main types of electrical activity associated with neurons

Two main types of electrical activity associated with neurons: action potentials and postsynaptic potentials. Action potentials are discrete voltage spikes that travel from the beginning of the axon at the cell body to the axon terminals where neurotransmitters are released. Postsynaptic potentials are the voltages that arise when the neurotransmitters bind to receptors on the membrane of the postsynaptic cell, causing the ion channels to open or close and leading to a graded change in the potential across the cell membrane. If an electrode is lowered into the intercellular space in a living brain, both types of potentials can be recorded.

Event related potentials and EEG

We want to know whether we can find brain activity that is reliably related to cognitive processes of interest. For this, single-trial EEG in our studies are probably not good and far too noisy. There is a lot of variance between sessions from the same participants, but also between participants. You need to average across all of them and compare the average in order to learn something. Positivity goes down and negativity goes up in the signals. There are different aspects of the ERP component of interest that can be analysed: - peak amplitude (used in 70% of studies) - looking at the first peak - area-under-the-curve (used in 20%) - peak-to-peak (used in 10%) There is no clear rule and results might differ between different measures (for example using peak amplitude you could find A is larger than B, find the opposite with area-under-the-curve, and find they are equal using peak to peak). Another option is to determine the onset of a component, but this can also be tricky to work out which rises first.

Dipole in a conductive medium like the brain

When a dipole is present in a conductive medium like the brain, current is conducted throughout that medium until it reaches the surface - called volume conduction. Voltage present at any point on the scalp will depend on the position and orientation of the generator dipole and on the resistance and shape of the various components of the head. Electricity doesn't just run directly between the two poles of the dipole in a conductive medium but spreads out through the conductor, so ERPs spread out as they travel throughout the brain and they tend to spread out laterally when they encounter the high resistance of the skull. These both blur the surface distribution of the voltage and an ERP generated in one part of the brain can lead to substantial voltages at distant parts of the scalp. There are algorithms that can reduce this blurring.

Recording many neurons simultaneously

When recording many neurons simultaneously, it is possible to measure either their summed postsynaptic potentials or their action potentials. Recordings of action potentials from large populations of neurons are called multi-unit recordings, and recordings of postsynaptic potentials from large groups of neurons are called local field potential recordings. In the majority of cases surface electrodes cant detect action potentials due to the timing of the action potentials and the physical arrangement of axons. When an action potential is generated current flows rapidly into and then out of the axon at one point along the axon and so on until the action potential reaches a terminal. If two neurons send their action potentials down axons that run parallel to each other and the action potentials occur at the same time (this rarely happens), then the voltages from two neurons will summate and the voltage recorded from a nearby electrode will be approx. twice as large as the voltage recorded from a single action potential. But if one neuron fires slightly after the other, then current given at a spatial location will be flowing into one axon at the same time is it flow out of the other axon, so they cancel each other out and produce a much smaller signal at the nearby electrode. Because this rarely happens, ERPs usually reflect postsynaptic potentials rather than action potentials.

What the results showed

When targets were absent and both stimuli were in the same hemisphere no shift in N2pc was seen. This is because participants' attention was drawn to the side of the more likely stimulus first then they scanned the less likely stimulus on the same side, so they never had to scan the other side. The results confirmed the predictions from the serial search hypothesis. They also tested what happens when one of the stimuli was actually the target. When the more likely stimulus was the target and stimuli were in opposite hemifields, no cross-over of the N2pc was observed. This is because no shift in attention between hemifields was required as the target was found immediately. When the less likely stimulus was the target and stimuli were in opposite hemifields, a cross-over of the N2pc was observed again. This is because attention was first directed to the hemifield of the more likely target but then a switch to the other side was required where the target was finally found. Again, all findings support the serial search hypothesis.