ERP TUTORIAL 1

A Brief Introduction to the Use of Event-Related Potentials (ERPs) in Studies of Perception and Attention

Geoffrey F. Woodman

Vanderbilt University Vanderbilt Vision Research Center Center for Integrative and Cognitive Neuroscience

Abstract

Due to the precise temporal resolution of electrophysiological recordings, the event-related potential (ERP) technique has proven particularly valuable for testing theories of perception and attention. Here, I provide a brief tutorial of the ERP technique for consumers of such research and those considering the use of human electrophysiology in their own work. My discussion begins with the basics regarding what brain activity ERPs measure and why they are well suited to reveal critical aspects of perceptual processing, attentional selection, and cognition that are unobservable with behavioral methods alone. I then review a number of important methodological issues and often forgotten facts that should be considered when evaluating or planning ERP experiments.

Electroencephalogram (EEG) technique continue to make it one of the most recordings were the first method developed for widely used methods to study the architecture direct and noninvasive measurements brain of cognitive processing. activity from human subjects (Adrian & The primary goal of this tutorial is to Yamagiwa, 1935; Berger, 1929; Jasper, 1937, introduce researchers who are unfamiliar with 1948). By noting when stimuli were presented ERPs to their use, interpretation, and and tasks were performed, early studies dissemination in studies of sensation, examining the raw EEG sought to characterize perception, attention, and cognition. I hope the changes in the state of electrical activity that the uninitiated readers will become better during sensory processing and the consumers of ERP research. For those who performance of simple-detection tasks (e.g., P. plan to conduct ERP research and add human A. Davis, 1939; Walter, 1938). However, electrophysiology to your methodological when scientists began to take advantage of toolbox, I strongly urge you to read the more signal averaging the event-related potential detailed resources that are afforded the space (ERP) technique quickly became the primary to cover the theoretical and practical issues tool of the cognitive neuroscientist (Cooper, with which practicing electrophysiologist Winter, Crow, & Walter, 1965; H. Davis, should be familiar (Handy, 2005; Hillyard & 1964; Donchin, 1969; Donchin & Cohen, Picton, 1987; Luck, 2005; Nunez & 1967; Spong, Haider, & Lindsley, 1965; Srinivasan, 2006; Regan, 1989; Rugg & Coles, Sutton, Braren, Zubin, & John, 1965; Sutton, 1995).1 Tueting, Zubin, & John, 1967; Walter, Cooper, I will cover three general topics in this Aldridge, McCallum, & Winter, 1964). tutorial. First, I will provide an extremely Despite the rise of modern neuroimaging brief review of the biophysical basis of the methods, several advantages of the ERP EEG and the averaged ERPs which make them ERP TUTORIAL 2 ideal for studying perception and attention. appears to account for a number of Second, I will present advice for conducting observations. Early electrophysiologists and evaluating ERP studies that test specific hypothesized that the EEG and intracranially hypotheses. Third, I will discuss several often recorded field potentials (local-field potentials forgotten characteristics of ERPs that should or LFPs) were due to postsynaptic activity of be considered when designing new neural ensembles (Adrian & Yamagiwa, 1935; experiments and interpreting ERP findings. Li, McLennan, & Jasper, 1952). This view is The topics covered here were selected based widely accepted today (Logothetis, Pauls, on my experience reviewing and publishing Augath, Trinath, & Oeltermann, 2001; Luck, manuscripts reporting findings from ERP 2005; Nunez & Srinivasan, 2006), although experiments. A large number of ERP the biological basis of the EEG and ERPs has experiments, including my own, that have been periodically debated (e.g., Fox & serious trouble during the review process are O'Brien, 1965; Galambos & Jahasz, 1997). in such a state because the waveforms do not This means that instead of recording a afford unambiguous interpretation due to one summation of the action potentials generated or more of a handful of common problems. by individual neurons, we believe that the Before discussing the characteristics that EEG and averaged ERPs are measuring readers should look for in ERP studies, I will electrical potentials generated in the quickly review some basics about this extracellular fluid as ions flow across cell electrophysiological technique. membranes and neurons talk to one another via neurotransmitters. Why ERPs are Well Suited to Study To create electrical fields large enough Perception and Attention? to propagate through the brain, dura, skull, and With their spectacular spatial skin, a large number of neurons must be active resolution, it is reasonable to ask, “Why have simultaneously (i.e., in the ballpark of 107 imaging techniques not made ERPs obsolete?” neurons, see Cooper et al., 1965; Ebersole, The most direct answer to this question is 1997). In addition, this large group of neurons simple: time. By time, I mean that ERPs have not only needs to be active synchronously but a temporal resolution that allows for the also to have a geometry that is perpendicular measurement of brain activity from one relative to the surface of the skull and not millisecond to the next, and many aspects of cancelled out by other neuronal ensembles attention and perception appear to operate on a active at the same time and with an opposite scale of tens of milliseconds. As the brain is orientation (Luck, 2005; Nunez & Srinivasan, essentially a wet electrical device, these 2006). The simultaneously active neurons electrophysiological recordings provide a must have approximately the same orientation direct measure of the currency of the system for the potentials to summate and this means we study. Also, given the nature of electrical that ERPs are primarily generated by the activity and the tissue in which ERPs are postsynaptic potentials of cortical pyramidal generated and propagated, there is no cells (which are perpendicular to the cortical measurable conduction delay between the surface). Given the location and orientation of brain activity generated inside the head and the a specific neural generator in the brain, we can potentials recorded from the scalp (Nunez & predict the pattern of voltage that will be Srinivasan, 2006). observed across the head. This is known as What exactly generates the voltage the forward problem and is easily solved, fluctuations recorded outside the head? We unlike its evil twin, the inverse problem. The operate under a working hypothesis that inverse problem states that if we are given a ERP TUTORIAL 3 distribution of electrical potential across a Sheatz, 1962). This same period saw the volume conductor, like the head, then we development of the 10/20 system for cannot know where it is generated if we do not standardized electrode placement which made know the number of simultaneously active ERP findings far easier to integrate and generators a priori (Helmholtz, 1853). replicate across studies (Jasper, 1958). Practically, this means that we cannot The peaks and troughs of a stimulus- definitively localize the neural generators of locked ERP waveform allow us to visualize ERP effects within the head from the data cognitive processing as it unfolds during a recorded outside of it. Thus, the outstanding trial. Figure 1 shows idealized waveforms temporal resolution of ERPs comes with the time locked to the presentation of a visual cost of living with an unknown degree to stimulus during a target discrimination task. spatial resolution. First things first, you will note that voltage is Although the inability to resolve the plotted with negative going up. Stories about activity of individual neurons may seem like a the origin of this convention abound. It is large drawback of the ERP technique, it clear that this method of presentation has been, appears that the functional unit of analysis that and continues to be, a contentious issue as ERPs measure fortuitously maps on to the efforts have been made to flip the voltage axis cognitive processes that psychologists (Bach, 1998; Luck, 2005). In my own work, I frequently hypothesize about. Specifically, follow the decades-old convention of plotting ERPs allow us to observe a series of cognitive negative up for practical reasons. Specifically, operations that take place from before the the vast majority of existing ERP papers delivery of sensory information to the plotted their waveforms with negative going peripheral nervous system until even after a up and perceptual learning has made such behavioral response is made. The earliest waveforms significantly easier for me to studies showed that stereotyped fluctuations in interpret than those with negative plotted potential were elicited by the presentation of down. sensory stimuli (e.g., P. A. Davis, 1939). The series of voltage fluctuations However, ERP research really gained shown in Figure 1 index a sequence processes popularity when a study demonstrated that the as the brain transforms information from raw cognitive activity related to preparing for a sensory input to the appropriate behavioral task could be measured. The Contingent response. First, we see the C1 component Negative Variation (CNV) was shown to build which flips polarity based in whether the up prior to the onset of a stimulus to which eliciting stimulus appears in the upper or lower participants were required to respond (Walter visual field and is believed to be generated by et al., 1964). It is fitting that this anticipatory activity in primary visual cortex (Clark, Fan, effect was the first of the ERP components & Hillyard, 1995; Clark & Hillyard, 1996; indexing cognitive processes discovered in the Estevez & Spekreijse, 1974; Jeffreys & modern era (see also Kornhuber & Deecke, Axford, 1972). This initial deflection is 1965). This era is marked by the advent of followed by the P1 and N1 components as averaging together potentials time-locked to an information propagates through the visual observable event and recorded on multiple system and perceptual analysis is performed trials to extract the small amplitude voltage (Heinze et al., 1994; Heinze, Mangun, & fluctuations common to each trial from the Hillyard, 1990; Luck, 1995; Vogel & Luck, much larger amplitude EEG ‘noise’ in which 2000).2 Next, we can observe waveforms they are embedded (Dawson, 1954; Donchin, elicited by the deployment of covert attention 1969; Donchin & Heffley, 1975; Galambos & to peripheral targets in the visual field (e.g., ERP TUTORIAL 4 the N2pc, Eimer, 1996; Luck & Hillyard, bottom of Figure 1 illustrates how the small 1994a, 1994b) and components associated ERP components are embedded in the with categorization of the visual stimulus (e.g., background EEG which is at least an order of the N2/P3 complex, Kutas, McCarthy, & magnitude larger. The idealized ERP Donchin, 1977; Pritchard, Shappell, & Brandt, waveforms shown in Figure 1 would take 1991; Sutton, 1979; Sutton et al., 1965). hundreds or even thousands of trials from a Waveforms indexing working memory number of participants to approximate through encoding and maintenance are the next to averaging and do not capture the amount of come online (i.e., the P3 and contralateral- latency jitter that the components and RTs delay activity, Donchin, 1981; Vogel & exhibit. Finally, the term component also has Machizawa, 2004), followed by components deeper meanings that generally refer to the elicited during the selection and preparation of underlying cognitive processes and brain the motor response (i.e., the lateralized- activity indexed by the potential (for detailed readiness potential or LRP, Coles, 1989). discussions see Luck, 2004; Rugg & Coles, Even after the participant completes the 1995). One reason that ERP studies of behavioral response and the trial is ostensibly attention and perception continue to flourish is over, the ERPs show us that cognitive that they rest upon a foundation of decades of processing continues. For example, waveforms worth of basic research on ERP components. elicited after the behavioral response are I provide this brief and simplified related to evaluating performance on the trial overview of a handful of the most prominent that just occurred (e.g., the error-related ERP components to make the point that the negativity and error positivity, or ERN and Pe, foundational work of early ERP researchers respectively, Falkenstein, Hoormann, Christ, has provided the current generation with a & Hohnsbein, 2000). toolbox overflowing with instruments An ERP component can be simply measuring process-specific activity in the defined as one of the component waves of the working brain. Figure 2 illustrates how more complex ERP waveform. ERP auditory stimuli elicit a somewhat different components are defined by their polarity series of ERP components including very early (positive or negative going voltage), timing, waveforms that actually begin with potentials scalp distribution, and sensitivity to task generated in the brain stem (Hillyard & Picton, manipulations. Different ERP component 1987). But the ability of ERPs to show the nomenclatures emphasize different aspects of progression of information processing in the these defining features and to provide a brain is qualitatively similar regardless of jumping off point for literature reviews I sensory modality and task (e.g., Desmedt, describe several in Table 1. It should also be Huy, & Bourguet, 1983; Pratt, in press). This noted that although we frequently discuss a ability to measure the dynamics of processing component as a unitary entity (e.g., the N2 or in the brain through the sequence of ERP P3) it is probably more accurate to describe a components has made this technique a vital given component as belonging to a family of tool for testing theories of perception, components with a similar polarities and attention, and cognition. temporal characteristics (e.g., Johnson, 1986; Pritchard et al., 1991). An average ERP Fundamentals of Rigorous ERP Studies waveform can be time locked to any externally I now turn to a discussion of the observable event with the primary reference features of effective studies of attention and events being the presentation of a stimulus and perception using ERPs. Conversely, the most the execution of a behavioral response. The common problems that I have observed in ERP TUTORIAL 5 submitted manuscripts, conference specific components and failing to take presentations, and published reports are advantage of the wealth of component-based instances of experiments that violate these research is a risky endeavor. As just rules of thumb. I do not pretend that this is a discussed, this can lead to the rediscovery of comprehensive list, or that I have had known components and phenomenon. In sufficient longevity in the field to rank these addition, those that engage in exploratory issues by importance. Instead, I list them here studies sometimes record and compare ERPs in approximately the order of frequency that I elicited by physically different stimuli while have encountered them in my own studies and observers perform different tasks. This those of others. naturally results in ERP waveforms that differ A Long History can be a Blessing and a Curse in many ways, across many different time Given the massive literature on which points. Besides leading to the statistical new ERP studies can draw, the majority of problem of multiple comparisons, it becomes ERP studies are designed to isolate and very difficult to determine the critical locus of measure modulations of specific ERP the behavioral effects. Are the stimuli components. This can mean measuring a processed differently at an early point in time number of different ERP components during with these early differences propagating the same task (e.g., Vogel, Luck, & Shapiro, through the system? Or, does the crucial 1998). More often, studies focus on a specific difference in information processing that ERP component during a variety of hypothesis causes the behavioral effects occur later in the driven task manipulations (e.g., Woodman & trial? Luck, 2003a; 2003b). New ERP components Nothing Should be Happening when Nothing is are still being discovered (e.g., Bach & Happening Meigen, 1992; Klaver, Talsma, Wijers, When I get a new ERP paper to read, Heinze, & Mulders, 1999; Luck & Hillyard, the first thing that I do is flip to the figures and 1994b) and the significance of existing ERP look at the baseline period of the waveforms components reinterpreted (e.g., Vogel & Luck, (e.g., -200 to 0 ms relative to stimulus onset). 2000). However, many papers are submitted I do this because examining the differences in which tout the discovery of a new ERP the waveforms before anything has happened component or novel modulation, but do not provides a quick way to assess the noise level actually report something new. This is the in the averaged potentials. If the stimuli and cost of using a technique with such a long trial types were randomized, then the brain history and rich literature. When digging in response prior to the presentation of the ancient ground, we need to be careful not to stimuli should not differ between types of rediscover the triumphs of previous teams of trials. When there are differences between the archeologists. waveforms before the trial begins this is a A related issue is that a number of ERP clear sign that the signal-to-noise ratio of the studies are conducted in which the researchers averages is low or that some kind of confound do not test predictions about how specific ERP is present.3 components will behave in a certain Many times, papers are submitted and experimental paradigm. Some of these have even published that describe the significance been important and groundbreaking (e.g., of stimulus-elicited activity despite the fact Hillyard, Hink, Schwent, & Picton, 1973; that the amplitudes of the effects of interest are Otten, Quayle, Akram, Ditewig, & Rugg, similar in magnitude to differences in the 2006). However, avoiding the approach of baseline. In Figure 3, I show an example of tailoring experimental designs to focus on waveforms in which the prestimulus noise is at ERP TUTORIAL 6 least as large as the P1 and N1 modulations It is possible that you would like to based on the task relevance of the stimuli record ERPs during a paradigm that involves presented. We have no reason to believe that presenting stimuli in fairly rapid succession. these potential effects are real because they are As I will discuss more below, the waveforms approximately the same size as the differences and effects elicited by a stimulus last for at found before information has even reached the least a second. This means that the baseline retina or left the retina en route to the brain interval immediately prior to the onset of a (i.e., before 30 ms after stimulus onset). critical stimulus might overlap with Consumers of ERP research, or those waveforms elicited by a preceding stimulus. preparing their research for consumption, In this case, it would be prudent to show a should be wary of waveforms showing effects longer epoch that precedes the onset of the that are not bigger than the noise, even if it is stimulus sequence or, at least, before the possible to find a measurement window that critical stimuli are presented that distinguish will produce significance. Moreover, the trial types. In addition, reviewers might reviewers should insist that a prestimulus not only ask you to show a sufficient baseline baseline period of at least 100 ms be shown in period to assess the noise level in your ERP the figures. waveforms but also request that you show I mentioned in the brief introduction to more of your waveforms than just the first ERP components that preparatory activity is several hundred milliseconds following a observed even before an imperative stimulus stimulus to assess the reach of your appears. Researchers interested in how the experimental manipulation. brain perceives new inputs and allocates Collect as Many Trials From Each Participant attention are rarely interested in its ability to as You Can predict when the next trial will begin. For this Imagine you are reading an ERP paper reason, ERP researchers usually take two and the baselines of the ERP waveforms countermeasures to remove such anticipatory contain voltage deflections of the same activity that could contaminate or simply magnitude as the effects of interest described obscure the stimulus-elicited waveforms. The later in the article. How could the researchers first is randomization and the other is baseline have avoided reporting these unconvincing correction. Randomization of the trial types results? The only solution to this problem is to and stimulus sequence, along with sufficient improve the signal-to-noise ratio. This usually signal averaging, removes the possibility that means relying upon signal averaging across a the anticipatory ERP components can be larger number of trials and making an differentially active preceding the different experiment much longer than it would take if trial types. Baseline correction is also only the behavioral data were being collected. ubiquitous in ERP methodology. Simply put, One of the most useful things that can baseline correction means subtracting the be learned through training in an ERP voltage measured during a prestimulus laboratory is that the number of trials that are window (e.g., -200 to 0 ms relative to stimulus typically necessary to accurately measure a onset) from the entire waveform, so that the specific ERP component depends on your waveform reflects the voltage relative to the component of interest. The early visual ERP average prestimulus voltage. This is done so components, like the C1, P1, and N1, are in that the anticipatory effects, like the CNV, will the same frequency range as the largest source not distract readers from the important effects of noise in the human brain. That noise source elicited following the stimulus presentation. is the 8-12 Hz alpha band oscillations that dominate the EEG.4 Alpha waves are ERP TUTORIAL 7 particularly large when an observer is sleepy addition, the more aggressively we filter our or bored (i.e., during an ERP experiment) and ERP waveforms, the more we cause abate significantly when a stimulus is unintended distortions of the amplitude, and presented (Berger, 1929; Pfurtscheller, potentially timing, of the signals (Duncan- Stancák, & Neuper, 1996). As a rule of Johnson & Donchin, 1979). This is thumb, to get a good measure of the C1 from particularly unfortunate given that precise an individual participant it can take over 1000 timing and amplitude measurements are the artifact-free trials per condition (i.e., per cell of very reasons for recording ERPs in the first the experimental design) due to alpha noise, its place. small amplitude, and substantial overlap with There are at least three methods that I the P1 component. The P1 and N1, which are would suggest for eliminating noise at the frequently of interest to perception and front end of data acquisition that can increase attention researchers, can require 300-1000 your signal-to-noise ratio. First, shield your trials per condition to measure reliably. For electrodes from environmental noise by the N2pc component, I try to obtain 250 trials recording in a radio-frequency shielded room per condition per subject. In comparison, the or chamber and placing equipment powered by large and slow P3 component can be measured alternating current in faraday cages (see with only about 35-60 trials per condition from Chapter 8 in Luck, 2005). Second, randomly each subject. I have found these ballpark jitter the exact timing of the inter-trial interval values to be very handy when planning new to ensure that the alpha-wave activity of the ERP experiments that require a modest participant does not become phase locked with number of observers (i.e., < 15). I need to the stimulus presentation rate. Third, you can point out that if you are running an ERP reduce noise in your recordings by keeping experiment that allows you to collect fewer your participants as alert and engaged in the trials from each participant than you would task as possible. This is because alpha-band like, it should be possible to compensate for noise increases when participants are drowsy this by collecting data from a large number of and bored (i.e., precisely the conditions under participants. In addition, if your recording which most of us collect our data). environment is virtually noiseless and your I use three tricks that help encourage subjects are very attentive, then trial numbers participants to remain engaged in the task smaller than these might be workable (see throughout the experiment. One trick is to use more on how to achieve this below). short blocks of trials with ample time for the However, Figure 3 illustrates that the early participants to take breaks during the components used in studies of perception and experiment. The second trick we use is to attention are unlikely to be interpretable with provide participants with refreshing significantly fewer trials from individual caffeinated beverages during these breaks in subjects than the estimates provided above. the tasks. The third trick it to provide an It would be great if it were possible to additional type of stimulation during the filter ERP data and remove noise without experimental trials themselves. In particular, distorting or removing the signal of interest. we encourage our participants to listen to As described above, much of the frequency music during visual experiments. Because the content of the ERP components is in exactly onset of the visual stimuli are jittered and not the same frequency band as the noise. This in phase with the prosody of the music, the means that no filter settings exist that will brain activity generated when processing this remove the noise without wiping out a music is averaged out while reducing the significant portion of the signal itself. In amplitude of alpha noise. Informal analyses ERP TUTORIAL 8 have shown that listening to music improves good at detecting blinks and changes in eye signal-to-noise ratios without changing the position caused by saccades relative to fixation patterns of effects observed during visual ERP but are not sensitive to slow shifts in fixation experiments. Of course, if you are studying that accumulate across trials (e.g., when the the processing of auditory or linguistic stimuli, task-relevant stimuli always appear at the same or the ability to multi-task, playing music for peripheral location). When fixation of an your participants might not be possible. In absolute position is necessary, we must use an summary, these dietary and environmental eye tracker in conjunction with the EEG sources of enrichment reduce noise levels in recording. Eye movements can also be a the raw EEG and keep the number of trials that problem when auditory (or other modality) must be averaged together down to a stimuli are presented because visual attention minimum. is automatically drawn to the source of such Why Blinks and Eye Movements Corrupt Your signals (e.g., McDonald, Teder-Sälejärvi, & Data Hillyard, 2000) and the eyes often follow The largest single electrical dipole in attention’s lead (Hoffman & Subramaniam, the head is the corneoretinal potential, which 1995; Kowler, Anderson, Dosher, & Blaser, points from the back of the eye toward the 1995). front of the eye. This means that when you are One approach to address the problem recording the EEG, and the time-locked ERPs, of trials contaminated by blinks and eye even a fairly small eye movement or blink will movements has been to use artifact-correction cause a massive electrical transient (e.g., a procedures (e.g., Berg & Scherg, 1994). typical eyeblink is over 100 µV, compared to Artifact-correction procedures can be useful in only 1-2 µV for a typical P1 wave). For working with clinical or developmental example, when I began recording EEG from populations who cannot be expected to monkeys, the first test of the new hardware maintain fixation when peripheral stimuli are was to verify that I could see these eye presented (Ille, Berg, & Scherg, 2002). movement artifacts online in the raw EEG. However, it is always best to exclude trials When you are interested in using ERPs containing ocular artifacts from ERP averages to study how stimuli are perceived or attended, and analyses. This can also mean needing to it is critical to ensure that your findings are not replace participants due to excessive eye simply due to contamination by eye movements across the experiment (for a two- movements. The idea that step procedure of artifact rejection and electrophysiological findings can be explained participant exclusion see Woodman & Luck, by oculomotor behavior is not new (Walter, 2003b). 1938), nor does it seem that we have stopped Rejecting trials with artifacts is the finding potentially interesting effects that surest way of avoiding ambiguous data might simply be due to even small eye because movements of the eyes not only shift movements (e.g., Yuval-Greenberg, Tomer, the corneoretinal potential, but also drastically Keren, Nelken, & Deouell, 2008). The change the input to the visual system. If it electrical artifact caused by a blink or saccadic were the case that just the former occurred, eye movement of more than 1-2 degrees is then it would result in voltage changes fairly large (approximately 16 µV per degree localized to the orbits. Because shifting the of eye movement) and can be detected with fovea to a different part of the world electrooculogram (EOG) recordings (Hillyard drastically changes the input to the visual & Galambos, 1970; Lins, Picton, Berg, & system, the activity across at least half of the Scherg, 1993). These EOG recordings are cortex will also differ between trials with and ERP TUTORIAL 9 without eye movements. These visual enhancement of the brain’s response to the differences will not have the same time course white stimulus. However, this is cannot be as the artifact itself due to transmission delays concluded from the ERP results. Among ERP in the visual system and the distributions will researchers who study attention this not be focused on the orbits as some artifact- experimental design is said to have a physical correction algorithms assume. Thus, to be stimulus confound (Hillyard & Picton, 1987; confident in the conclusions you draw, it is Naatanen, 1975). That is, the manipulation of critical to exclude trials and participants from presenting physically different stimuli is ERP averages that exhibit evidence of eye confounded with any potential modulation by movements or blinks. attention. Different Stimuli Inherently Modulate ERP When someone is interested in how Components attention influences the processing of stimuli, Differences in the amplitudes of ERP it is necessary for the stimuli eliciting the components should be expected when different waveforms to contribute equally to the stimuli are presented to observers. This should averages for the attended and unattended particularly be the case for the early conditions. This means that the stimulus components related to sensory and perceptual manipulations need to be orthogonal to the processing (i.e., the C1, P1, or N1). An manipulation of attention. For example, in the example is useful to illustrate this point. If above scenario using white and black squares, you record the ERPs time locked to the participants could switch between blocks of presentation of either a white square or a black trials in which the white squares were task square of equal sizes, presented at fixation on relevant and blocks in which the black squares a gray background, then different neurons are were task relevant. If we take the example of going to be activated by those different the stimulus array shown in Figure 1, stimulus onsets (e.g., those with different observers could alternate between blocks of contrast polarity sensitivities). The differential trials where they search for targets that are responses will be present throughout the visual colored red and those colored green so that we system (Felleman & Van Essen, 1991; could determine that the effects of attending to Ungerleider & Mishkin, 1982). With the target location were due to the task sufficient power, this will result in sensory relevance of the target color and not some low- ERP components that differ in amplitude for level characteristic like its luminance. In the two stimuli. This result would hardly be contrast, if someone is interested in the more surprising given previous reports in the low-level questions of how the brain responds literature (Ellemberg, Hammarrenger, Lepore, to different stimuli without simultaneous Roy, & Guillemot, 2001; Luck, 2005; Pratt, in manipulations of attention, then it is entirely press). However, suppose that you were appropriate to measure and report how the interested in whether white stimuli are more brain responses differ to stimuli with different likely to attract attention to themselves than physical features (e.g., Eimer, 1998; Thierry, black stimuli. A huge number of publications Martin, Downing, & Pegna, 2007). have reported modulations of the P1 and N1 Measuring Voltage Amplitudes and Latencies components due to the allocation of attention Another asset that ERP researchers to locations and stimuli (Luck, 1995; Mangun have gained from decades of previous work is & Hillyard, 1990). It might seem natural to how to quantify the observed effects. As conclude that if the white stimulus elicits a described directly above, when a brighter larger amplitude P1 or N1 component than the stimulus is presented, the early visual ERP black stimulus, that this was due to attentional components exhibit larger amplitude ERP TUTORIAL 10 responses. How do we go about quantifying Meyer, 1992; Renault, Ragot, Lesevre, & the magnitude of such amplitude effects? One Remond, 1982) although no less difficult approach has been to measure the voltage of (Luck, 2004). an ERP component at its peak or trough. The third problem of measuring Similarly, people have sometimes measured component latencies is related to an issue that I the point in time that an ERP component sidestepped when discussing the sequence of reaches its maximum voltage before the ERP components. Despite the fact that ERPs voltage returns back toward zero. The ERP have excellent temporal resolution, the precise literature convincingly demonstrates that measurement of the timing of an individual focusing on these arbitrary local maxima (i.e., ERP component is made difficult by the fact positive or negative voltage ‘peaks’) is that ERP components typically overlap with misleading. I will first discuss why measuring their neighbors. Component overlap is the the peak is difficult, and potentially term used to describe the fact that the voltage uninformative, before presenting the fluctuations of the ERP waveform inherently measurement methods I recommend and use. overlap with each other in time and space. For Confining your analyses of ERP example, the same change in observed voltage components to the metrics of the peaks is between two conditions could be the result of dangerous for a number of reasons. I will an earlier offset of the N2 component or an discuss the three problems that loom largest. earlier onset of the P3 component (e.g., First, the peak of a component is heavily Naatanen, Gaillard, & Mantysalo, 1978; influenced by high-frequency noise. Figure Naatanen & Michie, 1979). This is a special 4A shows how the timing of the peak of a case of the same potential for simultaneous relatively slow waveform (e.g., a 10-Hz wave, activity that makes the generators of ERP like the P1 or N1 component) can be heavily components difficult to localize. The influenced by bursts of higher-frequency component overlap problem has been difficult noise. As a result, measuring the peak of an to adequately address with statistical and ERP component is practically difficult. Even mathematical analysis techniques alone (Luck, after the observed waveform is digitally low- 2005; Rugg & Coles, 1995). The problems of pass filtered, the peak of the wave can be overlap and susceptibility to high-frequency contaminated by the random fluctuations in the noise also result in difficulty in measuring higher-frequency noise. voltage amplitude using the peak of an ERP A second deeper question is why you component. would want to measure the peak of a At this point the reader might feel that component in the first place. The peak is an this discussion has become depressing so I will arbitrary feature of an ERP component that now turn to some positive aspects and the could be argued to be less important than its approaches that appear to be the most beginning or end (Luck, 2004). Rarely does a productive. In some instances it is possible to debate between cognitive scientists hinge on isolate specific ERP components by virtue of when a certain process is approximately their unique scalp distributions (e.g., Coles, halfway completed. However, this is Gratton, & Donchin, 1988; Woodman & Luck, essentially what one is trying to quantify when 1999). Although these tend to be exceptions measuring the timing of the peak of an ERP and not the rule, the use of such measures can component. I could easily argue that the onset significantly simplify interpretation. Even or offset of a component is often a much more without such exotic components, we can use critical measurement (e.g., Miller & Hackley, the scalp distribution of ERP effects to help 1992; Osman, Bashore, Coles, Donchin, & infer the locus of experimental effects. For ERP TUTORIAL 11 example, if it appears that an experimental As a result, I was taught that all analyses of the manipulation led to the reduction in the timing and amplitude of averaged ERP amplitude of the N2 component we could rule components should be performed on the out that the effect is really due to an earlier unfiltered ERP waveforms passed by the onset of the P3 by showing that the scalp amplifier. Although virtually every published distribution of the modulation was consistent ERP paper shows filtered waveforms in their with the known distribution of the N2 and not figures, the analyses and measurements will be the P3. undistorted by this filtering process if they are When it comes to quantifying aspects performed on the waveforms prior to this step of specific ERP components it is best to not of cosmetic enhancement. Finally, one robust focus on the peak and instead to look more and potentially fruitful approach that I would broadly at the component. Specifically, many recommend trying is a fractional-latency influential ERP papers have measured the measure that works backward from an easy to features of ERP components using temporal define feature of an ERP component (Kiesel, windows (Luck, 2005). The widths of these Miller, Jolicoeur, & Brisson, 2008). This windows are set such that they bracket the approach offers the promise of providing the entire ERP component of interest, across all of advantages of measuring fractional-area the individual subject’s waveforms, and are latency without the ambiguities in setting the similar to previous studies measuring the same measurement window. component. Sufficient care must be taken in Voltages are Measured Relative to Reference this step of setting the window because a Sites skewed or narrow setting can taint the To record a voltage from the scalp it is measures of timing and amplitude using the sufficient to have one active and one reference window. In practice, a liberal setting of the electrode. This is because voltage is always size of the window (i.e., broad) makes the the difference in electrical potential between a measurement procedure as conservative as given electrode and the reference. The possible compared to analyses focused on the mathematics of subtraction mean that peak which are driven by selection bias. This electrical activity generated near the reference procedure applies to measurements of latency will appear as an inverted polarity voltage at as well as amplitude. When measuring the active electrode. If we used a two- latency, the most rigorous method is to use a electrode configuration, as Berger (1929) did, fractional-area latency metric. This involves it would not be possible to determine whether measuring when a component achieves some the activity we measured was greatest at the threshold of its total voltage in the window. site of the reference electrode, the active For example, if I wanted to measure the 25% electrode, or somewhere in between. Early in fractional-area latency of waveforms from two the history of human electrophysiology, different conditions, then I would measure at researchers identified the importance of using what time point 25% of the area under the multiple active electrodes to facilitate voltage curve defined by the measurement interpretation of the recorded voltages and window has accrued in each condition. determine the effect of the distance between Figure 4B shows how the fractional- the reference electrode and the active area latency and peak-latency measures could electrodes (Walter, 1938). The practical yield qualitatively different results from a pair importance of referencing for ERP users is that of hypothetical waveforms. Recall that the voltage measured at active electrode sites filtering of ERP waveforms can distort closer to the reference site will necessarily be amplitude and possibly latency measurements. ERP TUTORIAL 12 closer to zero volts, all other things being like measuring the loudness of your voice in equal. the front row of a Metallica concert. An implicit assumption is that the There are situations in which you need reference is at a location that provides a zero- to place the reference near the active voltage baseline. However, there is no perfect electrodes. These are when you are using a reference site because there is no truly bipolar electrode configuration (e.g., Brown & electrophysiologically neutral location on the Norcia, 1997). The logic of this configuration body. In my own laboratory, we use a is to have a reference very close to the active reference on the right mastoid process (the electrode to remove all potentials but those bone behind your right ear), re-referenced occurring between the reference and active offline to an electrode on the left mastoid. electrode (Nunez & Srinivasan, 2006). This re-referencing minimizes spatial However, the monopolar configurations distortion in the distribution of potentials described above are used in virtually all ERP measured across the head (Luck, 2005; Nunez, studies and common methods are very useful 1981) and has the added benefit of being in trying to relate your findings to previous widely used so that our findings can be (and future) research. There are several books compared to those from other labs. The linked that describe why certain reference mastoids and average reference procedures configurations can be problematic or induce significant spatial distortions in the advantageous (Luck, 2005; Nunez, 1981; pattern of potentials measured across the head. Nunez & Srinivasan, 2006) and these should Linked mastoids do this by creating a short be consulted for additional details. circuit between the left and right sides of the head because they are, by definition, linked by Issues We Rarely Discuss low resistance electrical wire. The average There are a number of characteristics reference procedure induces distortions due to of ERP waveforms that were discovered the assumption that the activity across all of during the first several decades of research and the electrodes in the array captures all of the are now rarely the focus of studies or electrical activity generated in the brain (Dien, discussion in papers. I believe it is useful to 1998). Another reference procedure that is discuss these characteristics here because they sometimes used is to place the reference at Cz, often come up in conversations with the electrode location at the top and center of colleagues and while designing new the head (i.e., along the anterior-posterior experiments. midline). Although there is no ERP Refractoriness electrophysiologically neutral location for the One of the first ERP projects I reference, using this location can be collected data for involved measuring the particularly problematic in certain situations. I visual N1 component during detection and have seen a number of papers in which the discrimination tasks (Vogel & Luck, 2000). study was interested in measuring broadly During that educational experience, I was distributed components with fronto- or parieto- informed that early sensory and perceptual central maxima (e.g., the ERN or P3) and Cz components, like the N1, are refractory at was used as the reference. Given we know short inter-stimulus intervals (ISIs). In other that the ERP components of interest should be words, ERP components are reduced in apparent at this electrode location, any effects amplitude when the eliciting stimulus follows measured at the active electrodes would be soon after another stimulus. In this situation, minimized by using the voltage underlying this soon is loosely defined. Specifically, Woods, location as the reference point. This would be Courchesene, Hillyard, and Galambos (1980; ERP TUTORIAL 13

Woods, Hillyard, Courchesne, & Galambos, Boyd, & Towell, 2006). Although 1980) showed that the amplitudes of the N1 documented and known among ERP and P2, but not the P3, elicited by an auditory researchers interested in human vision (e.g., stimulus were reduced even when Crevits, van Lith, & Viifvinkel-Bruinenga, approximately one second had passed since the 1982; Woodman, Arita, & Luck, 2009), the presentation of the last stimulus as compared nature of the ERP response to visual offsets to longer ISIs (see also Lu, Williamson, & has received less attention. Instead, it is Kaufman, 1992). Indeed, it has been proposed recognized as a potential confounding factor in that the refractory period of the auditory N1 that offsets elicit a series of sensory evoked may last tens of seconds (Nelson & Lassman, components (i.e., P1 and N1) similar to those 1973). Although this fundamental observed following the onset of a visual characteristic of the ERP components often stimulus. Methodological sources that discuss used to study perception and attention is this issue recommend presenting visual stimuli known to many ERP experts, this feature so briefly that no distinct offset response is might be unknown to many fairly new users or visible (typically 200 ms or less for visual readers of ERP papers. It means that the use stimuli) or for sufficiently long intervals that of short ISIs that allow for more trials to be the offset response to a stimulus does not collected may reduce the size of the overlap with the onset-elicited components components that the researchers seek to (Luck, 2005). The perceptual and harvest. The rule of thumb that I inherited and neurophysiological underpinnings of the offset use in my own laboratory is to temporally response to sensory stimuli are fertile grounds space stimulus onsets by approximately 1 for investigation. It is possible that a better second, if at all possible. It deserves understanding of these effects could yield tools mentioning that many interesting paradigms in for the investigation of pathway specific the attention and perception literatures involve activity in the visual system or the specificity presenting stimuli at different ISIs (e.g., the of deficits in clinical disorders. psychological refractory period paradigm). Individual Differences This means that ERP researchers may Perhaps the best known, yet least well misinterpret the inherent refractoriness of the documented, facet of participant’s ERP ERP components as being due to more components is the existence of individual interesting or complex phenomena (e.g., the differences. To my knowledge, the best depletion of cognitive resources). Thus, when discussion of this widely known secret is in trying to avoid physical stimulus confounds Steve Luck’s book (Luck, 2005). He describes across conditions we should be careful to how fairly striking individual differences exist equate both the individual stimuli as well as in even the early sensory and perceptual ERP the ISIs between successive stimuli. components, like the P1 and N1. These Offset Transients individual differences can be directly A characteristic of ERPs that I often appreciated if we consult earlier ERP studies discuss with collaborators when talking about in which the data from each participant are potential experimental designs is the fact that shown (e.g., Hillyard et al., 1973). The sudden stimulus offsets elicit ERP components individual differences that we observe are not too. This has been an important topic of simply due to excess noise in our data investigation in the auditory (e.g., Hillyard & acquisition systems or analysis procedures Picton, 1978; Naatanen & Picton, 1987; because these differences are very reliable Picton, Woods, & Proulx, 1978a, 1978b) and across recording sessions with the same somatosensory modalities (e.g., Spackman, observer. ERP TUTORIAL 14

In ERP studies, we seek to average with unique cognitive abilities. An example of together waveforms from a large enough this type of proposal is the recent work of sample of observers so that our results Vogel and colleagues (Fukuda & Vogel, 2009; generalize to the entire population from which Vogel & Machizawa, 2004; Vogel, they are drawn. However, the underlying McCollough, & Machizawa, 2005). Their cause of the observed individual differences work shows that the amplitude of the could either be trivial or integral for our contralateral-delay activity (or CDA) predicts understanding of the cognitive process that a the individual observer’s ability to store given component indexes. information in visual working memory and One trivial explanation of individual avoid distraction from irrelevant stimuli. We differences in the morphology of ERP may be at an exciting time when we have the components is that the observed differences tools and necessary theoretical motivation to are simply due to the underlying pattern of determine how much of the individual cortical folding in each participant. According differences in ERP components are due to to this explanation, all people have uninteresting geometry versus factors that are fundamentally the same ERP components and critical to our understanding of how the mind cognitive processing mechanisms, but the works. voltage patterns that we observe at the scalp The known individual differences in depend heavily on the folding pattern of the ERP component morphology have important cortical tissue. At a fundamental level, we implications at a practical level. When we know that the orientation of the gray matter begin running participants using a new relative to the skull is critically important neuroscience technique or behavioral (Nunez & Srinivasan, 2006). To put this more paradigm, it would be great if we could run a concretely, my N1 component might have a couple of pilot participants and know whether small amplitude relative to yours because the results we are obtaining were going to be some of the critical chunks of cortex that interpretable. However, this is sometimes not generate the N1 are in a sulcus in my brain possible with ERP experiments due to the whereas they are all on gyri in your brain. individual differences described above. One Skull thickness and conductivity also vary or more of my pilot participants might not across individuals and are important for the have the component that I am predicting will morphology of ERP components (Hoekema et be modulated by the task manipulations. The al., 2003). Cortical folding and the metrics of timeless problems associated with small tissue in the head are two examples of a samples and unknown effect sizes can be number of relatively uninteresting possible exacerbated by presenting each participant explanations for the individual differences in with an insufficient number of trials to obtain ERP components that are observed (Nunez & clean data, as discussed earlier. For this Srinivasan, 2006). Essentially these are reason, when working with a number of explanations that propose that the differences unknown factors, I believe that the best pilot are due to geometric ‘noise’ in our anatomy experiment is just to run Experiment 1 of the which are unrelated to how the brain processes study. However, when you are operating with information. questions and tools that you better understand Accounts at the other extreme propose it can be possible to perform pilot studies with that differences in the morphology of ERP a fairly modest number of participants. components across individuals may tell us about the fundamental differences in information processing that endow each of us ERP TUTORIAL 15

these battlefronts utilizes the weapons of Known Unknowns increasingly dense electrode arrays and How can the temporal resolution of computer modeling. The hope is that with ERPs be at the millisecond level but the spatial sufficiently dense arrays of electrodes, and resolution is not even known? This is because models constrained by structural imaging of if the geometry of the cortex generating a the brain, that the number of possible given ERP component is just right relative to generators will be sufficiently small as to be the skull, and all other cortical generators, then tractable. The second front uses combined the spatial resolution may be fairly precise (on imaging and electrophysiological recordings. the order of several cm3 Nunez & Srinivasan, This seems like the best of both worlds. fMRI 2006). Often, we implicitly assume that if a has excellent spatial resolution but slow component has a relatively focused scalp temporal resolution, while ERPs have topography and is maximal at a given excellent temporal resolution but poor to electrode site (e.g., Pz in the standard 10/20 unknown spatial precision. However, even system) that it is generated by the cortex just when these two data sets are collected at the beneath the electrode (e.g., posterior parietal same time, it is still difficult to confidently cortex, see Homan, Herman, & Purdy, 1987; link a fast ERP effect (e.g., 100 ms long) that Koessler et al., 2009; Steinmetz, Fürst, & occurs just after the stimulus appears (> 1 Meyer, 1989). However, given that we do not second), with a slow BOLD response know the number of simultaneously active measured long after the stimulus appeared neural generators that are contributing to a (~10 seconds). We can show support for a given ERP component or effect, then it is functional relationship by correlating the unknown whether the relevant activity we are signals across multiple levels of an measuring is generated near the electrode or independent variable, or on a trial-by-trial or across the entire cortical sheet. For example, block-by-block basis. But the signal-to-noise consider the case in which you have two ratios of the signals can make these analyses neural generators active at the same time but difficult and even in the best circumstances do of opposite polarity. Given this configuration not provide evidence for a causal relationship the voltage measured on the distal scalp will between localized BOLD activity and ERP be zero. Now imagine one large dipole of one component effects. In the third approach, orientation and two smaller but equal dipoles researchers have measured ERPs from patients of the opposite orientation. It should be with specific brain lesions (Knight, 1991). obvious at this point, as it was to Helmholtz This combination of methods has the power of (1853), that there are theoretically an infinite affording causal inferences but also has the number of ways to measure zero volts outside ambiguities inherent to neuropsychological the volume conductor of the head despite studies due to possible reorganization and the abundant electrical activity inside of it. Now fact that lesions can damage critical fibers of imagine how many possible dipole passage. The advent of transcranial magnetic configurations generated by chunks of active stimulation (TMS) has provided a way to cortex could be giving you the voltage perform experiments in humans with virtual, distribution you observe for any ERP reversible lesions while recording ERPs (e.g., component. Driver, Blankenburg, Bestmann, Vanduffel, & The difficulty of solving the problem Ruff, 2009; Fuggetta, Pavone, Walsh, Kiss, & of localizing ERP effects within the brain has Eimer, 2006). In addition, a number of the resulted in researchers battling the problem on technical challenges of combining TMS with at least four fronts simultaneously. One of ERPs appear to have been addressed (see Thut, ERP TUTORIAL 16

Ives, Kampmann, Pastor, & Pascual-Leone, to the feedforward sweep of activity through 2004). Using the fourth approach, researchers the sensory pathway (e.g., Zhang & Luck, have sought to record ERPs from animals that 2008). In contrast, other components have a can be linked to those found in humans and timing and distribution that suggests they are then localize the neural generators of those largely due to feedback. An example is the components in the animal models with N2pc component that occurs approximately invasive techniques (e.g., Cohen, Heitz, 200 ms poststimulus and has a scalp Schall, & Woodman, 2009; Mehta, Ulbert, & distribution which suggests that it might be Schroeder, 2000a; Mehta, Ulbert, & generated by activity in the ventral visual Schroeder, 2000b; Schroeder, Tenke, & Givre, stream. Given that this attention effect occurs 1992; Schroeder, Tenke, Givre, Arezzo, & about 30-130 ms after the first attentional Vaughan, 1991). Personally, I have great hope modulations of ERP components with similar for this approach (Cohen et al., 2009; distributions, Luck and Hillyard (1994b) Woodman, in press; Woodman, Kang, Rossi, proposed that the N2pc is due to feedback & Schall, 2007). The initial paper showing from an attentional control structure, like the that monkeys had a P3 component similar to of pulvinar (or the frontal-eye field, see Cohen et humans identified the potential for such work al., 2009). Beyond using logical arguments to be combined with causal manipulations like this, the methods needed to demonstrate such as lesion and inactivation studies (Arthur that certain ERP effects of interest are due to & Starr, 1984; Pineda, Foote, & Neville, feedback can be technically challenging 1989). (Cohen et al., 2009; Martinez et al., 1999). I The final issue that I will mention has wish that the interesting questions regarding been the topic of frequent discussions with the role of feedback were more easily testable colleagues and in manuscripts. This is the using only the ERP technique. As with question of whether there are distinct ERP behavioral studies of perception and attention, signatures of feedback processes. The role of the best way to effectively use ERP studies to feedback between regions of the brain has answer questions about feedback processes is become increasingly important in theories of with clever logic and experimental designs. attention and perceptual processing (e.g., Di Lollo, Enns, & Rensink, 2000; Lamme & Summary Roelfsema, 2000). Although ERPs have great The goal of this brief introduction to temporal resolution it is very difficult to the ERP technique was to familiarize cognitive determine which ERP components are due to scientists who study perception and attention feedforward processing and which have with the basics of interpreting findings from contributions from feedback. ERP experiments. My approach was to share The primary variable that people use to my perspective on the topics of ERP infer that certain effects are due to feedforward methodology that I deal with most frequently. versus feedback processes in the brain is the I discussed a number of issues that I hope will absolute timing of the modulation. We can help those running ERP experiments, readers reasonably expect that the earliest ERP interpreting ERP findings, and researchers components elicited following the presentation dealing with ERP data for first time. Despite of stimuli are due to feedforward processing of the advent of new technologies, studies of the representations of the stimuli. For attention and perception still rely heavily upon example, both the C1 and P1 components the ERP technique to test hypotheses and occur early enough that ERP researchers theories about how the brain rapidly processes assume that they are wholly or primarily due information. In fact, Figure 5 shows that ERP TUTORIAL 17 during the last 10 years the number of papers reporting or referring to findings from ERP experiments has increased by approximately 500%. It is evident that the impact and volume of research using the ERP technique continues to increase. This makes it particularly important to be a savvy consumer of ERP research, even if it is not a methodology that you utilize in your own work. ERP TUTORIAL 18

Footnotes about and are studying. This would account 1 One of the critical issues covered in these for why many of the ERP components we more comprehensive sources is the study have most of there content arising from relationship between the term event-related the 8-12 Hz frequency band (Makeig et al., potential (ERP) and other terms, such as 2002). Research into the fundamental nature evoked potential (EP), visual evoked potential of the activity underlying ERP components (VEP), steady state visual evoked potential continues (e.g., Palva & Palva, 2007; Shah et (SSVEP), and so forth. al., 2004). It is sufficient to say, that a significant portion of the activity of the brain 2 ERP components are typically named using is not related to processing stimuli in the task a polarity (N for negative and P for positive) the experimenter is interested in. As alpha- and ordinal (1 for first, 2 for second, etc.) band activity is one of the frequency bands nomenclature. The latter convention is due to that dominate the raw EEG, its ubiquitous absolute timing differences being fairly presence is probably both treasure and trash. common in many of the components across participants and paradigms. However, some researchers prefer to use a temporal label following the indication of polarity (e.g., P300 instead of P3). Finally, some ERP components have been named using a more descriptive label and its acronym (e.g., lateralized readiness potential and LRP). As a result, care must be taken to relate findings from different studies using different nomenclatures, but really measuring the same component. Table 1 is provided to make some connections between the same, or similar components, described with different nomenclatures.

3 One sarcastic comment that ERP researchers often utter is that the participants may have had ESP (i.e., extra-sensory perception), when they observe prestimulus activity that discriminates between the stimuli or trial types that have yet to occur. Indeed, the ERP technique could be a good way of testing individuals claiming to have such abilities. In the realm of cognitive neuroscience, in which no solid evidence for ESP exists, such comments are obviously criticisms of the signal-to-noise ratio inherent to the data.

4 Although I describe alpha-band activity as noise, it is possible that the oscillations in this frequency band are what we actually care ERP TUTORIAL 19

Acknowledgements Berger, H. (1929). Ueber das Elektrenkephalogramm des Menschen. I would like to thank Steve Luck for Archives fur Psychiatrie feedback on an earlier draft of this paper. Min- Nervenkrankheiten, 87, 527-570. Suk Kang, Nancy Carlisle and Melonie Brown, R. J., & Norcia, A. M. (1997). A Williams provided useful discussions. Jason method for investigating binocular Arita and Michelle Young aided in the rivalry in real-time with steady-state preparation of the figures. Michael Bach, VEP. Vision Research, 37, 2401-2408. Martin Eimer, and Trafton Drew gave me Brunia, C. H. M., van Boxtel, G. J. M., & excellent comments and suggestions as Böcker, K. B. E. (in press). Negative reviewers. G.F.W. is supported by the NIH slow waves as indices of anticipation: (NEI, RO1EY019882 & P30EY008126). The bereitshaftspotential, the Correspondence should be addressed to contingent negative variation, and the Geoffrey F. Woodman, Department of stimulus preceding negativity. In S. J. Psychology, Vanderbilt University, PMB Luck & E. Kappenman (Eds.), Oxford 407817, 2301 Vanderbilt Place, Nashville, TN handbook of event-related potential 37240-7817, (email) components. New York Oxford [email protected]. University Press. Clark, V. P., Fan, S., & Hillyard, S. A. (1995). Identification of early visually evoked References potential generators by retinotopic and topographic analyses. Human Brain Adrian, E. D., & Yamagiwa, K. (1935). The Mapping, 2, 170-187. origin of the Berger rhythm. Brain, 58, Clark, V. P., & Hillyard, S. A. (1996). Spatial 323-351. selective attention affects early Arthur, D. L., & Starr, A. (1984). Task- extrastriate but not striate components relevant late positive component of the of the visual evoked potential. Journal auditory event-related potential in of Cognitive Neuroscience, 8, 387-402. monkeys resembles P300 in humans. Cohen, J. Y., Heitz, R. P., Schall, J. D., & Science, 223, 186-188. Woodman, G. F. (2009). On the origin Bach, M. (1998). Electroencephalogram of event-related potentials indexing (EEG). In G. K. von Schulthess & J. covert attentional selection during Hennig (Eds.), Functional Imaging: visual search. Journal of Principles and Methodology (pp. 391- Neurophysiology, 102, 2375-2386. 408). Philadelphia: Lippincott-Raven. Coles, M. G. H. (1989). Modern mind-brain Bach, M., & Meigen, T. (1992). reading: Psychophysiology, physiology Electrophysiological correlates of and cognition. Psychophysiology, 26, texture segregation in the human visual 251-269. evoked potential. Vision Research, 32, Coles, M. G. H., Gratton, G., & Donchin, E. 417-424. (1988). Detecting early Berg, P., & Scherg, M. (1994). A multiple communication: Using measures of source approach to the correction of movement-related potentials to eye artifacts. Electroencephalography illuminate human information & Clinical Neurophysiology, 90(3), processing. Biological Psychology, 26, 229-241. 69-89. ERP TUTORIAL 20

Cooper, R., Winter, A. L., Crow, H. J., & Donchin, E. (1981). Surprise!...Surprise? Walter, W. G. (1965). Comparison of Psychophysiology, 18, 493-513. subcortical, cortical, and scalp activity Donchin, E. (1969). Average Evoked using chronically indwelling electrodes Potentials, Methods, Results, and in man. Electroencephalography and Evaluations. Washington, D.C.: U.S. Clinical Neurophysiology, 18, 217-228. Government Printing Office. Crevits, L., van Lith, G., & Viifvinkel- Donchin, E., & Cohen, L. (1967). Averaged Bruinenga, S. (1982). On and off evoked potentials and intramodality contribution to the combined occipital selective attention. on-off response to a pattern stimulus. Electroencephalography and Clinical Ophthalmologica, 184, 169-173. Neurophysiology, 22, 537-546. Crowley, K. E., & Colrain, I. M. (2004). A Donchin, E., & Heffley, E. (1975). review of the evidence for P2 being an Minicomputers in the signal-averaging independent component process: age, laboratory. American Psychologist, 30, sleep and modality. Clinical 299-312. Neurophysiology, 115, 732-744. Driver, J., Blankenburg, Bestmann, S., Davis, H. (1964). Enhancement of evoked Vanduffel, W., & Ruff, C. C. (2009). cortical potential in humans related to a Concurrent brain-stimulation and task requiring a decision. Science, 145, neuroimaging for studies of cognition. 182-183. Trends in Cognitive Sciences, 13, 319- Davis, P. A. (1939). Effects of acoustic stimuli 327. on the waking human brain. Journal of Duncan-Johnson, C., & Donchin, E. (1979). Neurophysiology, 2, 494-499. The time constant in P300 recording. Dawson, G. D. (1954). A summation Psychophysiology, 16, 53-55. technique for the detection of small Ebersole, J. S. (1997). Defining epileptic foci: evoked potentials. past, present, future. Journal of Electroencephalography and Clinical Clinical Neurophysiology, 14, 470-483. Neurophysiology, 6, 65-84. Eimer, M. (1996). The N2pc component as an Desmedt, J. E., Huy, N. T., & Bourguet, M. indicator of attentional selectivity. (1983). The cognitive P40, N60 and Electroencephalography and clinical P100 components of somatosensory Neurophysiology, 99, 225-234. evoked potentials and the earliest Eimer, M. (1998). Does the face-specific N170 electrical signs of sensory processing component reflect the activity of a in man. Electroencephalography and specialized eye processor? Clinical Neurophysiology, 56, 272-282. Neuroreport: an International Journal Di Lollo, V., Enns, J. T., & Rensink, R. A. for the Rapid Communication of (2000). Competition for consciousness Research in Neuroscience, 9(13), among visual events: The 2945-2948. psychophysics of reentrant visual Ellemberg, D., Hammarrenger, B., Lepore, F., processes. Journal of Experimental Roy, M.-S., & Guillemot, J.-P. (2001). Psychology: General, 129(4), 481-507. Contrast dependency of VEPs as a Dien, J. (1998). Issues in the application of the function of spatial frequency: The average reference: Review, critiques, parvocellular and magnocellular and recommendations. Behavior contributions to human VEPs. Spatial Research Methods, Instruments, & Vision, 15, 99-111. Computers, 30(1), 34-43. ERP TUTORIAL 21

Estevez, O., & Spekreijse, H. (1974). Kappenman (Eds.), Oxford handbook Relationship between pattern of event-related potential components. appearance-disappearance and pattern New York: Oxford University Press. reversal responses. Experimental Brain Handy, T. C. (2005). Event-Related Research, 19, 233-238. Potentials: A Methods Handbook. Falkenstein, M., Hoormann, J., Christ, S., & Cambridge, MA: MIT Press. Hohnsbein, J. (2000). ERP components Heinze, H. J., Mangun, G. R., Burchert, W., of reaction errors and their functional Hinrichs, H., Scholz, M., Münte, T. F., significance: A tutorial. Biological et al. (1994). Combined spatial and Psychology, 51, 87-107. temporal imaging of brain activity Felleman, D. J., & Van Essen, D. C. (1991). during visual selective attention in Distributed hierarchical processing in humans. Nature, 372, 543-546. the primate cerebral cortex. Cerebral Heinze, H. J., Mangun, G. R., & Hillyard, S. Cortex, 1, 1-47. A. (1990). Visual event-related Folstein, J. R., & Van Petten, C. (2008). potentials index perceptual accuracy Influence of cognitive control and during attention to bilateral stimuli. In mismatch on the N2 component of the A. G. C. Brunia, A. Kok, G. Mulder, ERP: a review. Psychophysiology, 45, and M. Verbaten (Ed.), 152-170. Psychophysiological Brain Research Fox, S. A., & O'Brien, J. H. (1965). (pp. 196-202): Tilburg University Duplication of evoked potential Press. waveform by curve of probability of Helmholtz, H. v. (1853). Ueber einige Gesetze firing of a single cell. Science, 147, der Vertheilung elektrischer Ströme in 838. körperlichen Leitern mit Anwendung Fuggetta, G., Pavone, E. F., Walsh, V., Kiss, auf die thierisch-elektrischen Versuche. M., & Eimer, M. (2006). Cortico- Annalen der Physik und Chemie, 89, cortial interactions in spatial attention: 211-233, 354-377. A combined ERP/TMS study. Journal Hillyard, S. A., & Galambos, R. (1970). Eye of Neurophysiology, 95, 3277-3280. movement artifact in the CNV. Fukuda, K., & Vogel, E. K. (2009). Human Electroencephalography and Clinical variation in overriding attentional Neurophysiology, 28, 173-182. capture. Journal of Neuroscience, 29, Hillyard, S. A., Hink, R. F., Schwent, V. L., & 8726-8733. Picton, T. W. (1973). Electrical signs Galambos, R., & Jahasz, G. (1997). The of selective attention in the human contribution of glial cells to brain. Science, 182, 177-179. spontaneous and evoked potentials. Hillyard, S. A., & Picton, T. W. (1978). ON International Journal of and OFF components in the auditory Psychophysiology, 26, 229-236. evoked potential. Perception and Galambos, R., & Sheatz, G. C. (1962). An Psychophysics, 24, 391-398. electroencephalographic study of Hillyard, S. A., & Picton, T. W. (1987). classical conditioning. Paper presented Electrophysiology of cognition. In F. at the The American Journal of Plum (Ed.), Handbook of Physiology: Physiology. Section 1. The Nervous System: Gehring, W. J., Liu, Y., Orr, J. M., & Carp, J. Volume 5. Higher Functions of the (in press). The error-related negativity Brain, Part 2 (pp. 519-584). Bethesda, (ERN/Ne). In S. J. Luck & E. MD: Waverly Press. ERP TUTORIAL 22

Hillyard, S. A., Vogel, E. K., & Luck, S. J. Johnson, R., Jr. (1986). A triarchic model of (1998). Sensory gain control P300 amplitude. Psychophysiology, 23, (amplification) as a mechanism of 367-384. selective attention: Kiesel, A., Miller, J. O., Jolicoeur, P., & Electrophysiological and neuroimaging Brisson, B. (2008). Measurement of evidence. Philosophical Transactions ERP latency differences: A comparison of the Royal Society: Biological of single-participant and jackknife- Sciences, 353, 1257-1270. based scoring methods. Hoekema, R., Wieneke, G. H., Leijten, F. S. Psychophysiology, 45, 250-274. S., van Veelen, C. W. M., van Rijen, P. Klaver, P., Talsma, D., Wijers, A. A., Heinze, C., Huiskamp, G. J. M., et al. (2003). H. J., & Mulders, G. (1999). An event- Measurement of the conductivity of related brain potential correlate of skull temporarily removed during visual short-term memory. epilepsy surgery. Brain Topography, Neuroreport, 10, 2001-2005. 16, 29-38. Knight, R. T. (1991). Evoked potential studies Hoffman, J. E., & Subramaniam, B. (1995). of attention capacity in human frontal The role of visual attention in saccadic lobe lesions. In H. S. Levin, H. M. eye movements. Perception & Eisenberg & A. L. Benton (Eds.), Psychophysics, 57, 787-795. Frontal lobe function and dysfunction Homan, R. W., Herman, J., & Purdy, P. (pp. 139-153). New York: Oxford (1987). Cerebral location of University Press. international 10-20 system electrode Koessler, L., Maillard, L., Benhadid, A., placement. Electroencephalography Vignal, J. P., Felblinger, J., and Clinical Neurophysiology, 66, 376- Vespignani, H., et al. (2009). 382. Automated cortical projection of EEG Ille, N., Berg, P., & Scherg, M. (2002). sensors: Anatomical correlation via the Artifact correction of the ongoing EEG international 10-10 system. using spatial filters based on artifact NeuroImage, 46, 64-72. and brain signal topographies. Journal Kornhuber, H. H., & Deecke, L. (1965). of Clinical Neurophysiology, 19, 113- Hirnpotentialanderungen bei 124. Wilkurbewegungen und passiven Jasper, H. H. (1937). Electrical signs of Bewegungen des Menschen: cortical activity. Psychological Bereitschaftspotential und reafferente Bulletin, 34, 411-481. potentials. Pflugers Archives, 284, 1- Jasper, H. H. (1948). Charting the sea of brain 17. waves. Science, 108, 343-347. Kowler, E., Anderson, E., Dosher, B., & Jasper, H. H. (1958). The ten-twenty electrode Blaser, E. (1995). The role of attention system of the International Federation. in the programming of saccades. Vision Electroencephalography & Clinical Research, 35, 1897-1916. Neurophysiology, 10, 371-375. Kutas, M., McCarthy, G., & Donchin, E. Jeffreys, D. A., & Axford, J. G. (1972). Source (1977). Augmenting mental locations of pattern-specific chronometry: The P300 as a measure components of human visual evoked of stimulus evaluation time. Science, potentials. I: Components of striate 197, 792-795. cortical origin. Experimental Brain Lamme, V. A. F., & Roelfsema, P. R. (2000). Research, 16, 1-21. The distinct modes of vision offered by ERP TUTORIAL 23

feedforward and recurrent processing. electrophysiology. Journal of Trends in Neurosciences, 23, 571-579. Experimental Psychology: Human Li, C.-l., McLennan, H., & Jasper, H. (1952). Perception and Performance, 20, Brain waves and unit discharge in 1000-1014. cerebral cortex. Science, 116, 656-657. Makeig, S., Westerfield, M., Jung, T.-P., Lins, O. G., Picton, T. W., Berg, P., & Scherg, Enghoff, S., Townsend, J., Courchesne, M. (1993). Ocular artifacts in EEG and E., et al. (2002). Dynamic brain event-related potentials I: Scalp sources of visual evoked responses. topography. Brain Topography, 6, 51- Science, 25(5555), 690-694. 63. Mangun, G. R., & Hillyard, S. A. (1990). Logothetis, N. K., Pauls, J., Augath, M. A., Electrophysiological studies of visual Trinath, T., & Oeltermann, A. (2001). selective attention in humans. In A. B. Neurophysiological investigation of the S. a. A. F. Wechsler (Ed.), basis of the fMRI signal. Nature, 412, Neurobiology of Higher Cognitive 150-157. Function (pp. 271-295). New York: Lu, Z.-L., Williamson, S. J., & Kaufman, L. Guilford. (1992). Behavioral lifetime of human Martinez, A., Anllo-Vento, L., Sereno, M. I., auditory sensory memory predicted by Frank, L. R., Buxton, R. B., Dubowitz, physiological measures. Science, 258, D. J., et al. (1999). Involvement of 1668-1670. striate and extrastriate visual cortical Luck, S. J. (1995). Multiple mechanisms of areas in spatial attention. Nature visual-spatial attention: Recent Neuroscience, 2, 364-369. evidence from human McDonald, J. J., Teder-Sälejärvi, W. A., & electrophysiology. Behavioural Brain Hillyard, S. A. (2000). Involuntary Research, 71, 113-123. oriention to sound improves visual Luck, S. J. (2004). Ten simple rules for perception. Nature, 407, 906-908. designing ERP experiments. In T. C. Mehta, A. D., Ulbert, I., & Schroeder, C. E. Handy (Ed.), Event-Related Potentials: (2000a). Intermodal selective attention A Methods Handbook. Cambridge, in monkeys. I: Distribution and timing MA: MIT Press. of effects across visual areas. Cerebral Luck, S. J. (2005). An introduction to the Cortex, 10(4), 343-358. event-related potential technique. Mehta, A. D., Ulbert, I., & Schroeder, C. E. Cambridge, MA: MIT Press. (2000b). Intermodal selective attention Luck, S. J. (in press). ERP components in the in monkeys. II: Physiological perception of multiple-element mechanisms of modulation. Cerebral stimulus arrays. In S. J. Luck & E. Cortex, 10(4), 359-370. Kappenman (Eds.), Oxford handbook Miller, J. O., & Hackley, S. A. (1992). of event-related potential components. Electrophysiological evidence for New York: Oxford University Press. temporal overlap among contingent Luck, S. J., & Hillyard, S. A. (1994a). mental processes. Journal of Electrophysiological correlates of Experimental Psychology: General, feature analysis during visual search. 121, 195-209. Psychophysiology, 31, 291-308. Naatanen, R. (1975). Selective attention and Luck, S. J., & Hillyard, S. A. (1994b). Spatial evoked potentials in humans - A filtering during visual search: critical review. Biological Psychology, Evidence from human 2, 237-307. ERP TUTORIAL 24

Naatanen, R., Gaillard, A. W. K., & Perez, V. B., & Vogel, E. K. (in press). What Mantysalo, S. (1978). Early selective- ERPs can tell us about working attention effect on evoked potential memory. In S. J. Luck & E. reinterpreted. Acta Psychologica, 42, Kappenman (Eds.), Oxford handbook 313-329. of event-related potential components. Naatanen, R., & Michie, P. T. (1979). Early New York: Oxford University Press. selective attention effects on the Pfurtscheller, G., Stancák, A., Jr., & Neuper, evoked potential: A critical review and C. (1996). Event-related reinterpretation. Biological synchronization (ERS) in the alpha Psychology, 8, 81-136. band - an electrophysiological correlate Naatanen, R., & Picton, T. (1987). The N1 of cortical idling: A review. wave of the human electric and International Journal of magnetic response to sound: A review Psychophysiology, 24, 39-46. and an analysis of the component Picton, T. W., Woods, D. L., & Proulx, G. B. structure. Psychophysiology, 24, 375- (1978a). Human auditory sustained 425. potentials. I. The nature of the Nelson, D. A., & Lassman, F. M. (1973). response. Electroencephalography and Combined effects of recovery period Clinical Neurophysiology, 45, 186-197. and stimulus intensity on the human Picton, T. W., Woods, D. L., & Proulx, G. B. auditory evoked vertex response. (1978b). Human auditory sustained Journal of Speech and Hearing potentials. II. Stimulus relationships. Research, 16, 297-308. Electroencephalography and Clinical Nunez, P. L. (1981). Electric Fields of the Neurophysiology, 45, 198-210. Brain. New York: Oxford University Pineda, J. A., Foote, S. L., & Neville, H. J. Press. (1989). Effects of Locus Coeruleus Nunez, P. L., & Srinivasan, R. (2006). Electric lesions on auditory, long-latency, fields of the brain: The neurophysics of event-related potentials in monkey. EEG (2nd ed.). Oxford: Oxford Journal of Neuroscience, 9, 81. University Press, Inc. Polich, J. (in press). Neuropsychology of Osman, A., Bashore, T. R., Coles, M., P300. In S. J. Luck & E. Kappenman Donchin, E., & Meyer, D. (1992). On (Eds.), Oxford handbook of event- the transmission of partial information: related potential components. New Inferences from movement-related York: Oxford University Press. brain potentials. Journal of Pratt, H. (in press). Sensory ERP components. Experimental Psychology: Human In S. J. Luck & E. Kappenman (Eds.), Perception and Performance, 18, 217- Oxford handbook of event-related 232. potential components. New York: Otten, L. J., Quayle, A. H., Akram, S., Oxford University Press. Ditewig, T. A., & Rugg, M. D. (2006). Pritchard, W. S., Shappell, S. A., & Brandt, M. Brain activity before an event predicts W. (1991). Psychophysiology later recollection. Nature N200/N400: A review and Neuroscience, 9, 489-491. classification scheme. In J. R. Palva, S., & Palva, J. M. (2007). New vistas Jennings, P. K. Ackles & M. G. H. for alpha-frequency band oscillations. Coles (Eds.), Advances in Trends in Neurosciences, 30, 150-158. psychophysiology (Vol. 4). London: Jessica Kingsley. ERP TUTORIAL 25

Regan, D. (1989). Human Brain cortical evoked responses to visual and Electrophysiology: Evoked Potentials auditory stimuli. Science, 148, 395- and Evoked Magnetic Fields in Science 397. and Medicine. New York: Elsevier. Steinmetz, H., Fürst, G., & Meyer, B.-U. Renault, B., Ragot, R., Lesevre, N., & (1989). Craniocerebral topography Remond, A. (1982). Onset and offset within the international 10-20 system. of brain events as indices of mental Electroencephalography and Clinical chronometry. Science, 215, 1413-1415. Neurophysiology, 72, 499-506. Rossion, B., & Jacques, C. (in press). N170. In Sutton, S. (1979). P300 -- Thirteen years later. S. J. Luck & E. Kappenman (Eds.), In H. Begleiter (Ed.), Evoked Brain Oxford handbook of event-related Potentials and Behavior: Plenum potential components. New York: Publishing. Oxford University Press. Sutton, S., Braren, M., Zubin, J., & John, E. R. Rugg, M. D., & Coles, M. G. H. (Eds.). (1965). Evoked potential correlates of (1995). Electrophysiology of Mind. stimulus uncertainty. Science, 150, New York: Oxford University Press. 1187-1188. Schroeder, C. E., Tenke, C. E., & Givre, S. J. Sutton, S., Tueting, P., Zubin, J., & John, E. R. (1992). Subcortical contributions to the (1967). Information delivery and the surface-recorded flash-VEP in the sensory evoked potential. Science, 155, awake macaque. 1436-1439. Electroencephalography & Clinical Thierry, G., Martin, C. D., Downing, P., & Neurophysiology, 84, 219-231. Pegna, A. J. (2007). Controlling for Schroeder, C. E., Tenke, C. E., Givre, S. J., interstimulus perceptual variance Arezzo, J. C., & Vaughan, H. G. J. abolishes N170 face selectivity. Nature (1991). Striate cortical contribution to Neuroscience, 10, 505-511. the surface-recorded pattern-reversal Thut, G., Ives, J. R., Kampmann, F., Pastor, VEP in the alert monkey. Vision M. A., & Pascual-Leone, A. (2004). A Research, 31, 1143-1157. new device and protocol for combining Shah, A. S., Bressler, S. L., Knuth, K. H., TMS and online recordings of EEG Ding, M., Mehta, A. D., Ulbert, I., et and evoked potentials. Journal of al. (2004). Neural dynamics and the Neuroscience Methods, 141, 207-217. fundamental mechanisms of event- Ungerleider, L. G., & Mishkin, M. (1982). related brain potentials. Cerebral Two cortical visual systems. In D. J. Cortex, 14, 476-483. Ingle, R. J. W. Mansfield & M. A. Smulders, F. T. Y., & Miller, J. O. (in press). Goodale (Eds.), The Analysis of Visual The lateralized readiness potential. In Behavior (pp. 549-586). Cambridge, S. J. Luck & E. Kappenman (Eds.), Massachusetts: MIT Press. Oxford handbook of event-related Vogel, E. K., & Luck, S. J. (2000). The visual potential components. New York: N1 component as an index of a Oxford University Press. discrimination process. Spackman, L., Boyd, S., & Towell, T. (2006). Psychophysiology, 37, 190-123. Identification and characterization of Vogel, E. K., Luck, S. J., & Shapiro, K. L. somatosensory off responses. Brain (1998). Electrophysiological evidence Research, 1114, 53-62. for a postperceptual locus of Spong, P., Haider, M., & Lindsley, D. B. suppression during the attentional (1965). Selective attentiveness and blink. Journal of Experimental ERP TUTORIAL 26

Psychology: Human Perception and perception, and awareness during Performance, 24, 1656-1674. object-substitution masking. Vogel, E. K., & Machizawa, M. G. (2004). Psychological Science, 14, 605-611. Neural activity predicts individual Woodman, G. F., & Luck, S. J. (2003b). Serial differences in visual working memory deployment of attention during visual capacity. Nature, 428(15 April), 748- search. Journal of Experimental 751. Psychology: Human Perception and Vogel, E. K., McCollough, A. W., & Performance, 29, 121-138. Machizawa, M. G. (2005). Neural Woods, D. L., Courchesne, E., Hillyard, S. A., measures reveal individual differences & Galambos, R. (1980). Recovery in controlling access to working cycles of event-related potentials in memory. Nature, 438, 500-503. multiple detection tasks. Walter, W. G. (1938). Critical review: The Electroencephalography and Clinical technique and application of electro- Neurophysiology, 50, 335-347. encephalography. Journal of Woods, D. L., Hillyard, S. A., Courchesne, E., Neurology & Psychiatry, 1, 359-385. & Galambos, R. (1980). Walter, W. G., Cooper, R., Aldridge, V. J., Electrophysiological signs of split- McCallum, W. C., & Winter, A. L. second decision-making. Science, 207, (1964). Contingent negative variation: 655-657. An electric sign of sensorimotor Yuval-Greenberg, S., Tomer, O., Keren, A. S., association and expectancy in the Nelken, I., & Deouell, L. Y. (2008). human brain. Nature, 203, 380-384. Transient induced gamma-band Woodman, G. F. (in press). Homologues of response in EEG as a manifestation of human event-related potential miniature saccades. Neuron, 58, 429- components in nonhuman primates. In 441. S. J. Luck & E. Kappenman (Eds.), The Zhang, W., & Luck, S. J. (2008). Feature- Oxford Handbook of Event-Related based attention modulates feedforward Potential Components. New York: visual processing. Nature Oxford University Press. Neuroscience, 12, 24-25. Woodman, G. F., Arita, J. T., & Luck, S. J. (2009). A cuing study of the N2pc component: An index of attentional deployment to objects rather than spatial locations. Brain Research. Woodman, G. F., Kang, M.-K., Rossi, A. F., & Schall, J. D. (2007). Nonhuman primate event-related potentials indexing covert shifts of attention. Proceedings of the National Academy of Sciences, 104, 15111-15116. Woodman, G. F., & Luck, S. J. (1999). Electrophysiological measurement of rapid shifts of attention during visual search. Nature, 400, 867-869. Woodman, G. F., & Luck, S. J. (2003a). Dissociations among attention, ERP TUTORIAL 27

Table 1 Summary of ERP components using a variety of nomenclatures during a simple visual-manual task similar to that of Figure 1. This list focuses on visual components and neglects components from the auditory, language, and memory literatures. Abbreviations: CNV, Contingent Negative Variation; O- & E-waves, Orienting & Expectancy Waves; C1, component 1; N, negative; P, positive; N2pc, N2-posterior-contralateral; PCN, Posterior Contralateral Negativity; CDA, contralateral-delay activity; SPCN, Sustained Posterior Contralateral Negativity; LRP, Lateralized Readiness Potential; ERN/Ne, Error-Related Negativity/Error Negativity; Pe, Error Positivity.

Nomenclature Ordinal Latency Scalp Task/Stimulus Hypothesized Useful (peak) Distribution Specificity Process(es) Indexed Reference

Components CNV Anticipation, (Brunia, van Preceding a (O- & E-waves) Cognitive & Motor Boxtel, & Stimulus Preparation Böcker, in press)

C1 P/N50-70 Sensory Processing (Pratt, in press)

P1 P90-100 Sensory/Perceptual (Pratt, in press) Processing

N1 N170-200 Posterior N170 for faces Perceptual (Hillyard, versus Processing, Expert Vogel, & Luck, Anterior N1 Recognition, Visual 1998; Rossion Discrimination & Jacques, in press; Vogel & Luck, 2000)

Components Following a Stimulus P2 Not Well Understood (Crowley & Colrain, 2004)

N2 N225-250 Object Recognition, (Folstein & Van Categorization Petten, 2008; Pritchard et al., 1991)

N2pc PCN Deployment of (Luck, in press) Covert Attention

P3 P300 P3a/P3b P3a/P3b Stimulus Evaluation (Polich, in Time, Categorization, press) Context (Working Memory) Updating, Cognitive Load

SPCN CDA Maintenance in (Perez & Visual Working Vogel, in press) Memory ERP TUTORIAL 28

LRP Response Preparation (Smulders & Miller, in press)

Medial ERN/Ne & FBN Error Processing, (Gehring, Liu, Frontal Reinforcement Orr, & Carp, in Negativity Learning or Response press) Components Conflict Signal Following a Response Pe Affective or (Falkenstein et Conscious al., 2000) Assessment of Task Performance

instead havefairlybroad scalpdistributionswithcentralmaxima. either becorrectorincorrect. NotethattheP3andERNcomponentsare not typicallylateralizedbut observed duringtheperformance ofthetaskrequiringaresponsewithfinger onthelefthandthatcould at occipital-temporal-parietal electrodesites. The top-rightpanelshows thesequenceofERP components shows theclassicsequenceofERP components elicitedduringavisualtaskwithleftfieldtargets illustrates howERPsareextractedfromtheraw electroencephalogram (EEG). The middle-rightpanel human headwiththeplacementofasubsetelectrodes fromthe10/20system. The bottom-rightpanel left panelshowsabilateralstimulusarraytowhich thewaveformsaretimelockedandamodelof Figure 1.Idealizedevent-relatedpotential(ERP)waveformsingavisualsearchtask. eliciteddur The -30 µV

Raw EEG -3 µV -3 µV Stim 1... ERPs ERPs Contralateral tothetarget Ipsilateral tothetarget C1 P1 N1 200ms Incorrect response Ipsilateral toresponsehand Contralateral tothe P2 N2pc responsehand N2 LRP Stim N Behavioral Response P3 ERN 10ms 100ms 1000ms -3 µV Na Nb N1 N0 N2

P0 II III VI I Pa

Auditory ERPs IV V P1 P2

P3 Figure 2. Idealized event-related potential (ERP) waveform evoked by a brief auditory stimulus beginning with the early brain-stem responses (waves I-VI). Time scale is logarithmic to show these early responses. Waveforms shown would be expected from a central electrode cite (i.e., Cz)(Adapted from Hillyard & Picton, 1987). Amplitude Differences O1 Similar to Noise

Prestimulus noise

-3.0 µV

Task Irrelevant Stimulus Task Relevant Stimulus

0 400 800 Time Poststimulus (ms)

Figure 3. Example of a waveform in which the prestimulus noise is equal in amplitdue to the potential effects of interest. These are actual data recorded from electrode O1 and averaged across two subjects with approximately 150 artifact-free trials elicited by each type of stimulus to illustrate the need for sufficient power and clean baselines. Observed waveform Observed low-pass A filtered waveform

True underlying ERP component waveform

B Peak Voltage Measure Measurement Window

25% Fractional-Area Latency Measure

Figure 4. Hypothetical waveforms illustrating the difficulty of measuring the latency of an ERP component from the moment of peak voltage. A) Demonstration of how high-frequency noise can bias the measurement of latency based on the peak. B) Illustration of how measuring peak latency of waveforms in two conditions can lead to qualitatively different patterns than the less biased method of measuring factional-area latency. Gray region shows the measurement window. 6000

5000

4000

3000

2000 Number of Publications 1000 1990-91 1991-92 1992-93 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 Year of Publication

Figure 5. Number of event-related potential publications by year. Data derived from PsychInfo searches for the terms event-related potential, ERP, or evoked potential.