3Let’s start with a simple experiment: variant of the oddball task 80%20%XXOOXXMarkerAlso called continuuous performance taskLuck, 2005
4Simple experiment 80% 20% X X O O X X Also called continuuous performance taskEEG from one electrode site midline over parietal lobesLuck, 2005
5Simple experiment ERP Components: P = Positive N = Negative P1 = P100 etc.Also called continuuous performance taskERP components have labels like P1 and N1 refer to polarity and position in the waveform. These labels are NOT linked to the nature of the underlying brain activity!Luck, 2005
6ERP Components – P1 P1: Luck, 2005 Sensory peak, elicited by visual stimuli no matter which task is usedstrongly influenced by stimulus parameters (luminance)Onset: ms, peak: msLatency varies with contrastEarly portion may come from middle occipital gyrus, late portion from fusiform gyrusAlso called continuuous performance taskLuck, 2005
7ERP Components – P3 group Depends on which task is performedno clear consensus about what neural or cognitive process the P3 wave reflects. (‘context-updating’)P3 amplitude gets larger as target probability gets smaller.Also local probability matters, because the P3 wave elicited by a target becomes larger when it has been preceded by more and more nontargets.know a great deal about the effectsof various manipulations on P3 amplitude and latency, but there isMoreover, it is the probability of thetask-defined stimulus class that matters, not the probability of thephysical stimulus. For example, if subjects are asked to press a buttonwhen detecting male names embedded in a sequence containingmale and female names, with each individual name occurringonly once, the amplitude of the P3 wave will depend on the relativeproportions of male and female names in the sequence (seeKutas, McCarthy, & Donchin, 1977). Similarly, if the target is theletter E, occurring on 10 percent of trials, and the nontargets areselected at random from the other letters of the alphabet, the targetwill elicit a very large P3 wave even though the target letter isapproximately four times more probable than any individual nontargetletter (see Vogel, Luck, & Shapiro, 1998).e.g. here much larger for the rare O then for the frequent X stimuliLuck, 2005
8ERP Components –Mismatch Negativity (MMN) observed when subjects are exposed to a repetitive train of identical stimuli with occasional mismatching stimulinegative-going wave that is largest at central midline scalp sites and typically peaks between 160 and 220 ms.Several other components are sensitive to mismatches if they are task-relevant, but the MMN is observed even if subjects are not using the stimulus stream for a taskthought to reflect a fairly automatic process that compares incoming stimuli to a sensory memory trace of preceding stimuli.However, the MMN can be eliminated for stimuli presented in one ear if the subjects focus attention very strongly on a competing sequence of stimuli in the other.
9ERP Components –Error-related negativity (ERN) Can be elicited byBeing aware of an errornegative feedback following an incorrect response- observing someone else makingan incorrect responseMost investigators believe that the ERN reflects the activity of a system that either monitors responses or is sensitive to conflict between intended and actual responses.Source could be ACC.
11Advantage over pure behavior 1. Provide a continuous measure of processing between stimulus and responseResearch Question: Are slowed responses (RTs) due to slower perceptual processes or slower response processes?Latency of the P3 wave becomes longer when perceptual processes are delayed, no increase in latency in the Stroop Task ERPs might proof useful for determining which stage of processing is influenced by a taskLuck, 2005
12Advantage over pure behavior 2. Can provide an online measure of the processing of stimuli even when there is no behavioral responseAttended versus ignored stimuliLanguage comprehension can assess processing of a word embedded in a sentence at the time the word is processedLuck, 2005
14Disadvantage over behavior 1. Very small signal, requires a large number of trials to measure them accurately. (50, 100 or even 1000 trials)2. Functional significance: we don’t know the specific biophysical events that underlie the production of a given ERPLuck, 2005
15What are ERPs? Evoked Model EVOKED MODEL: The fact that ERPs consist of a sequence of componentsseems to imply a sequential activation process. As anexample, the processing of a visual stimulus can becharacterized by three early components, the C1 with alatency around 80ms which is generated in the striate cortex(area 17), the P1 with a latency around 110ms which isgenerated in the extrastriate cortex (areas 18 and 19) and theN1 with a latency around 160ms which probably is generatedin the temporal cortex (cf. Clark et al., 1995 ; Di Russo et al.,2002 ; Allison et al., 2002 ). These findings invite theinterpretation that a visual stimulus is first processed in theprimary visual cortex and then in secondary areas before itcan be identified as an object, a process that is related withareas in the temporal cortex. Thus, the timing for thegeneration of ERP components appears as sequential neuralprocesses that results solely from anatomical properties.
16What is an ERP?Two main types of electrical activity in the brain: action potentials and postsynaptic potentialsAction potentials are discrete voltage spikes that travel from the beginning of the axon to the cell body to the axon terminals where neurotransmitters are releasedPostsynaptic potentials are voltages that arise when the neurotransmitters bind to receptors to open or close and leading to a graded change in the potential across the cell membraneIf an electrode is lowered into the intercellular space in a living brain both types of potentials can be recorded. Its fairly easy to isolate the action potentials arising from a single neuron by inserting a microelectrode into the brain but it is virtually impossible to completely isolate a single neuron’s postsynaptic potential in an in vivo extracellular recording. Single unit recordings thus measure action potentials rather than postsynaptic potentials.When recording many neurons simultaneously it is possible to measure either tehor summed postsynaptic potential 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 potentialsLuck, 2005
17?Which of the two, action potentials or postsynaptic potentials, do you think we see reflected in EEG?
18What is the ERP?Mostly surface electrodes can not detect action potentials due to the timing and physical arrangement of axons. Will most likely cancel each other outWhen an action potential is generated,current flows rapidly into and then out of the axon at onepoint along the axon, and then this same inflow and outflow occurat the next point along the axon, and so on until the action potentialreaches a terminal. If two neurons send their action potentialsdown axons that run parallel to each other, and the action potentialsoccur at exactly the same time, then the voltages from the twoneurons will summate and the voltage recorded from a nearbyelectrode will be approximately twice as large as the voltagerecorded from a single action potential. However, if one neuronfires slightly after the other, then current at a given spatial locationwill be flowing into one axon at the same time that it is flowing outof the other axon, so they cancel each other and produce a muchsmaller signal at the nearby electrode.Luck, 2005
19What is an ERP? - Postsynaptic Potential If an excitatory neurotransmitter is released at the apical dentrides:current will flow from extracellular space into the cell, yielding a net negativity on the outside in the region of the dentrides.Current will also flow out of the cell body and basal dentrides yielding a net positivity in this areaAction potential duration: 1msCortical pyramidal cellLuck, 2005
20Best guess of a biophysical event that gives rise to a scalp ERP This creates a tiny dipoledur: mslocation: largely dentrides & cell bodydelay: occur instantaneously and do not travel down the axonsize: Can summate under certain conditions rather then cancel each other out and then be recorded at great distance (scalp)Luck, 2005
21Best guess of a biophysical event that gives rise to a scalp ERP The dipole of a single neuron is tinyBut under certain conditions the dipoles from many neurons will summate 1. occur approximately at the same time across 1000 – neurons2. be spatially aligned3. receive all excitatory or all inhibitory inputThis is most likely in cortical pyramidal cells, which are aligned perpendicular to the surface of the cortexLuck, 2005
22?Purkinje cells in the cerebellar cortex are beautifully aligned with each other and oriented perpendicular to the cortical surface. Can we measure them in EEG?
23Assume we have many aligned dipoles now.. …then the signal still needs to reach the scalpThe brain is conductive material (volume conduction).The voltage on the surface will thus depend onThe position and orientation of the generator dipole (equivalent current dipole)The resistance and shape of the various components of the head (brain, skull, scalp, eye holes)Luck, 2005
24Assume the signal has reached the skull… …what does it look like?Electricity spreads out through the conductor blurTends to follow the path of least resistanceThe skull has high resistanceTravels laterally when reaching the skullMore blur *BUT nicely electricity travels nearly as fast a light. So the signal is instantaneous!Luck, 2005* There are clever algorithms that calculate the ‘skull’-blur and reduce it.
25How do I find out where the signal comes from? Forward problem: If you knew the location and orientation of the dipoles and the conductance of the volume, the you could compute the distribution of voltage.Inverse problem: ‘ill-posed’. An infinite number or different dipole configurations can produce any given voltage. use specific constraintsNo perfect tool out there yet talks on source reconstructionLuck, 2005
26Sum: ERP components in the Evoked Model tx = xth point in timei = trial indexk = total number of single trials- reflect neural activity in rather localized brain regions that are involved in the processing of a stimulus and/or task.- reflect a sequential processindependent of the background activity- ERP components do not interact with prestimulus EEGCentral assumption is that the noise component apporximates zero or many trialsKlimesch et al., 2007; Luck, 2005
27Critique ERP Components in the Evoked Model (e1) EEG oscillations do not serve a specific function Negative evidence(e2) No correlation/ interaction between preand poststimulus EEG Negative evidence(e3) ERP components and (power of) ongoing EEG are additive Preliminary negative evidence(e4) ERP components do not interact with prestimulus EEG Negative evidence(e5) ERP latencies/ interpeak latencies and evoked power are not associated with frequencies of dominant EEG oscillations Negative evidence(e6) Dipole source analysis yields meaningful results Preliminary positive evidence(e7) ERP components are generated along a pathway of localized neural activation Preliminary positive evidenceKlimesch et al., 2007
28Oscillations in EEG Klimesch et al., 2007 Brain oscillation theory offers an alternative explanationfor the generation of ERP components that is completelydifferent from that of the evoked model. The interpretationof the event-related EEG response is a logical consequencefrom results obtained for the ongoing EEG: Oscillationsreflect different sensory and cognitive processes and playan important role for the timing of neural processes alsofor the event-related EEG response.Klimesch et al., 2007
29Event-Related Phase Reorganization (ERPR) Model a= amplitudew = frequencytx = xth time pointy = trial indexNo noise termAn ERP generated by ERPR can be understood as thesum of instantaneous amplitudes (a) of different taskrelevant frequencies o at time point t , averaged over trials k :Unlike the evoked model, ERP components generatedby ERPR lack an additive noise component. Noise may bedue to the influence of two different factors: (i) to thenumber of frequencies that are not task relevant and/or (ii)to the extent task relevant frequencies do not align inabsolute phase within certain time windows. Both factorswill reduce the amplitude of the ERP.Thus, for optimal processing of astimulus, phase reorganization is obligatory.For optimal processing of a stimulus, phase reorganization is obligatory.This doesn’t mean a phase reset, but instantaneous phase alignment (IPA)This means: Event-related alignment in phase between (task relevant) frequenciesKlimesch et al., 2007
30How is noise embedded in the ERPR Model? Input from non-task relevant and thus not aligned oscillationsImperfect alignement between phases of task relevant oscillations
31Oscillations and ERPs Klimesch et al., 2007 oscillatory model assumes that evoked components are generated by a combination of three processesPhase reorganization of ongoing oscillations,evoked oscillationsIPAKlimesch et al., 2007
32ER ComponentsEarly processing of a stimulus in subcortical regions, oscillatory activity might be induced in the cortex (much earlier then the first event-related components) widely distributed neuronal processEvoked components are localized processes. Distribution can be explained through volume conductionIt should also be noted that the alignment of phase of different assemblies may lead to a large and transient increase in neural activity during the excitatory period of alignment that may resemble at least in part an evoked component. But in a physiological sense there would be a large difference in thefunctional meaning of evoked components and ERPRdriven components. Although ERPR-driven components might also appear as a serial sequence of components, theyare most likely generated by parallel processes. As an example, after very early processing of a stimulus insubcortical regions (such as the reticular formation and thalamus) oscillatory activity might be induced in the cortex much earlier than the first event-related componentsappear. Dynamic interactions between oscillations and respective cell assemblies may be important for the latency and anatomical site of IPA and the appearance of an eventrelated component. As illustrated in Fig. 3 , oscillatory activity might be a very distributed neural process, andsome components such as the P1 might be generated simultaneously—or with a latency shift—at different sites (cf. sites 1 and 2 in Fig. 3(A) ). Recent empirical evidence, showing that the P1 can be described as traveling alpha wave propagating from occipital to parietal sites suggests that early ERP components may be generated by a widely distributed neural process (Klimesch et al., 2007b ). Evoked components, on the other hand, are more compatible with a localized process and the appearance of this component at different scalp sites might be explained by volume conduction due to a large dipole at a particular region in the brain.Klimesch et al., 2007
33What would you consider to be positive evidence for the ERPR Model?
34Implications of the ERPR Model (p1) EEG oscillations are associated with specific functions Positive evidence(p2) Interaction between pre- and poststimulus EEG Positive evidence(p3) Phase reset of task relevant ongoing oscillations, generation of task relevant evoked oscillations, IPA between task relevant oscillations Positive evidence(p4) ERP components are determined by IPA; reset and/or IPA takes place not necessarily at pos. or neg. peak Positive evidence(p5) ERP latencies/ interpeak latencies and evoked power reflect frequency characteristics of functionally relevant oscillations Positive evidence(p6) Dipole source analysis may not be considered an adequate method Not investigated(p7) Evoked components reflect parallel distributed neural processes Not investigated
36Voltage Peaks are not special Rule 1. It doesn’t make sense to measurepeak amplitude and peak latency to measurethe magnitude and timing of ERP components.Luck, 2005
37Peak Shapes Are not the Same as Component Shapes Rule 2. It is impossible to estimate the time course or peak latency of a latent ERP component by looking at a single ERP waveform.Rule 3. It is dangerous to compare an experimental effect (i.e.,the difference between two ERP waveforms) with the raw ERP waveforms.Luck, 2005
38Peak amplitudes are different from component sizes… Rule 4. Differences in peak amplitude do not necessarily correspond with differences in component size, and differences in peak latency do not necessarily correspond with changes in component timing.Luck, 2005
39Averaging changes your data Rule 5. Never assume that an averaged ERP waveform accuratelyrepresents the individual waveforms that were averaged together.In particular, the onset and offset times in the averaged waveformwill represent the earliest onsets and latest offsets from the individualtrials or individual subjects that contribute to the average.Luck, 2005
40The way to go… Strategy 1. Focus on a Specific Component Strategy 2. Use Well-Studied Experimental ManipulationsStrategy 3. Focus on Large ComponentsStrategy 4. Isolate Components with Difference WavesStrategy 5. Focus on Components That Are Easily IsolatedStrategy 6. Component-Independent Experimental DesignsAvoiding Confounds and Misinterpretations