INT J LANG COMMUN DISORD, XXXX 2020, VOL. 00, NO. 0, 1–13

Review Understanding event-related potentials (ERPs) in clinical and basic language and communication disorders research: a tutorial

Sean McWeeny† and Elizabeth S. Norton†‡ †Roxelyn and Richard Pepper Department of Communication Sciences and Disorders, Northwestern University, Evanston, IL, USA ‡Institute for Innovations in Developmental Sciences, Northwestern University, Chicago, IL, USA

(Received January 2020; accepted March 2020)

Abstract Background: Event-related potentials (ERPs), which are electrophysiological neural responses time-locked to a stimulus, have become an increasingly common tool in language and communication disorders research. They can provide complementary evidence to behavioural measures as well as unique perspectives on communication disorders. ERPs have the distinct advantage of providing precise information about the timing of neural processes and can be used in cases where it is difficult to obtain responses from participants, such as infants or individuals who are minimally verbal. However, clinicians and clinician–scientists rarely receive training in how to interpret ERP research. Aims: To provide information that allows readers to better understand, interpret and evaluate research using ERPs. We focus on research related to communication sciences and disorders and the information that is most relevant to interpreting research articles. Method: We explain what ERPs are and how ERP data are collected, referencing key texts and primary research articles. Potential threats to validity, guidelines for interpreting data, and the pros and cons using of ERPs are discussed. Research in the area of paediatric language disorders is used as a model; common paradigms such as the semantic incongruity and auditory are used as tangible examples. With this foundation of understanding ERPs, the state of the field in terms of how ERPs are used and the ways they may inform the field are discussed. Main Contribution: To date, no review has focused on ERPs as they relate to clinical or communication research. The main contribution of this review is that it provides practical information geared toward understanding ERP research. Conclusions: ERPs offer insights into neural processes supporting communication and can both complement behaviour and provide information that behavioural measures cannot. We encourage readers to evaluate articles using ERPs critically, effectively pushing the field forward through increased understanding and rigor. Keywords: event-related potential (ERP), neuroscience, mismatch negativity, N400.

What this paper adds r ERPs have become more prevalent in research relevant to communication sciences and disorders. In order for clinicians to review and evaluate this research, an understanding of ERPs is needed. This review adds to the field by providing an accessible description of what ERPs are, a description of what ERP components are, and the most relevant commonly used components, as well as how ERP data are recorded and processed. With this foundational understanding of how ERPs work, guidelines for the interpretation of ERP data are given. Though few ERP studies currently have direct implications for clinical practice, we discuss several ways through which ERPs can impact clinical practice in future, by providing information that cannot be obtained by behaviour alone about the aetiology of disorders, and as potential biomarkers of disorder or treatment response.

Address correspondence to: Elizabeth S. Norton, Northwestern University, Evanston, IL 60208, USA; e-mail: [email protected]

International Journal of Language & Communication Disorders ISSN 1368-2822 print/ISSN 1460-6984 online C 2020 Royal College of Speech and Language Therapists DOI: 10.1111/1460-6984.12535 2 Sean McWeeny and Elizabeth S. Norton Introduction clinical relevance, using work in the area of paediatric language as an example. Communication requires a host of complex neural pro- Note that several methods similar or related to ERPs cesses, many of which are still not fully understood. are outside the scope of this review. Electroencephalog- When the is processing a stimulus, such as a spo- raphy (EEG) is the continuous signal from which ERPs ken or written word, it produces characteristic electri- are derived, and often involves measuring the oscillatory cal signals that can be recorded from sensors on the activity of the brain (for a review of EEG in cognitive scalp. These neural signals in response to stimuli, called development, see Bell and Cuevas 2012; for a review event-related potentials (ERPs), can index various sen- of EEG and language, see Weiss and Mueller 2003). sory, cognitive and motor processes such as detecting a (MEG) is a measure of mag- stimulus, evaluating or categorizing it, detecting an er- netic changes in the brain, similar to EEG (for a review ror or anomaly in the stimulus, and making a motor re- of the MEG/EEG methods, see Puce and Ham¨ al¨ ainen¨ sponse such as pressing a button or articulating a word. 2017). (ECoG) records electrical ERPs are essentially measurements of small electrical activity directly from the brain’s surface (for a review, voltage changes produced by the brain, with very pre- see He 2015). Also outside the current scope are the cise timing. They can provide insights that behavioural auditory brainstem response (ABR) and frequency fol- measures cannot, such as what happens just millisec- lowing response (FFR); although both can be considered onds after a stimulus is presented (before a behavioural ERPs, their interpretation and characteristics differ suffi- response can occur), or processing that happens subcon- ciently to exclude from this review (for a review, see Skoe sciously (without explicit effort or ). They can and Kraus 2010). Because the ABR’s neural origins are also provide information into the aetiology of disorders; known, and it is highly reliable/replicable, it has been for example, if a group of patients with a certain dis- extremely clinically useful; building on years of research, order look like typical controls in terms of their ERPs newborn hearing screening using ABR is common and that reflect early sensory processing but different in their required across most of the United States (National Cen- later appraisal of stimuli, that indicates that the disorder ter for Hearing Assessment and Management 2011). may cause disruption after the phase of early sensory processing. ERPs are also very useful for studying pop- ulations who may not be able to make overt responses What is an event-related potential (ERP)? or follow directions in behavioural tasks, such as infants Recall that an ERP is a measurement of the brain’s and young children or patients who have motor, lan- electrical response to a stimulus recorded at the scalp. guage or cognitive deficits. Assessing the brain can also However, to understand how to collect those measure- provide information on the aetiology of communication ments, we must first consider how electrical activity is disorders and may reveal biomarkers of a disorder that recorded from the living human brain. Electrodes (elec- behavioural measures alone cannot. For these and other trical sensors) are placed on the scalp to record EEG, reasons, ERP studies have become increasingly common which is a continuous measure of the electrical activ- in speech, language and hearing research. ity of the brain, similar to an electrocardiogram (EKG) ERPs can be challenging to understand and inter- of the heart. ERPs represent average activity from the pret without some background in this area, which is not continuous EEG signal that is synchronized (i.e., time- typically part of clinical education and training in com- locked) to a certain stimulus (e.g., an auditory tone or a munication sciences and disorders. The primary goal printed word). of this review is thus to provide speech, language, and The electrical activity recorded with EEG comes hearing clinicians and scientists with a foundation from from neurons in the brain, which function by creating which they can understand and evaluate research using electrical signals. The electrical signals from a neuron are ERPs. As with any method, ERP research needs to be tiny in magnitude, so the signals that can be measured rigorously assessed not only by the peer-review process from the scalp reflect thousands of neurons changing but also by the ‘consumers’ of scientific research. In the their electrical activity in similar patterns. ERPs mostly same way that a reader needs to understand an interven- reflect a certain type of activity from one type of neuron tion approach or statistical procedure to evaluate fully a arranged in a certain layout in the cortex (postsynaptic research paper that uses it, understanding the basics of potentials of cortical pyramidal neurons oriented per- ERPs can help readers evaluate the conclusions of papers pendicular to the scalp; for more details on the neural that use them. The goals of this review are to describe: origin of ERPs, see Buzsaki´ et al. 2012 and Luck 2014), (1) what ERPs are and where they come from; (2) sug- but understanding this is not paramount to understand- gestions to evaluate the quality and validity of an ERP ing most applications of ERPs. publication’s methods and interpretation of data; and The amplitude of electrical activity recorded over (3) how ERPs can be effectively used in research with time is displayed as brainwaves (typically with a measure ERPs in clinical and basic language and communication disorders research 3 of voltage on the y-axis and time on the x-axis) and have been successfully utilized in numerous ERP exper- serves as the basis for EEG and ERP signals. To get iments with children (e.g., Henderson et al. 2011). from continuous EEG to the brain’s electrical response to an event (i.e., an ERP), the EEG signal must be time- Examples of ERP paradigms to guide the review locked to a stimulus. However, the signal from any one stimulus, or event, captures brain signals that resulted Throughout this review, we will use two classic ERP from that event, but also a substantial amount of noise paradigms in the area of language processing to give (this includes unrelated neural processes, measurement concrete examples of concepts. We will first introduce error and environmental electrical noise). Even in perfect these experimental paradigms and then how the ERP recording conditions with no noise from measurement that they elicit is measured. error or the environment, unrelated brain signals are captured in the recording. Thus, presenting the stimulus Semantic incongruity paradigm and N400 component many times and averaging the responses together isolates The N400 semantic incongruity paradigm based on a the activity that is related to the event of interest. landmark paper in which Marta Kutas and Steve Hill- yard established that ERPs can index detection of a semantic incongruity in language (Kutas and Hillyard ERP components and paradigms 1980). In this now widely used paradigm, a participant What are ERP components? reads a sentence that is presented one word at a time on a computer screen; most of the sentences are semantically Experimental paradigms for ERPs use stimuli that elicit appropriate or expected, but some have an incongru- a series of components, which are sensory, perceptual, ous or unexpected final word. An example semantically cognitive or motor processes that appear as a series of appropriate sentence would be ‘It was his first day at positive and negative changes in the waveform (the plot- work.’ In the original experiment, 25% of the sentences ted ERP voltage) over time. Put simply, a specific ERP had a semantic incongruity, that is, the final word of the component is the brain’s electrical response to a specific sentence is unexpected from the context, such as ‘He stimulus feature. Most components are named for the spread the warm bread with socks.’ The brain’s response direction of their voltage deflection (P for positive, N for to the final word is very different in this semantically negative) and the approximate timing (e.g., N400 com- incongruous condition than in the expected condition. ponent peaks around 400 ms) or order of the peak (N1 This ERP response to semantic incongruity is called the for first negative-going peak), though others are abbre- N400 component, as there is a more negative voltage viations of their names (such as the mismatch negativity, for the unexpected word around 400 ms after the word MMN). Components have an onset, a peak or plateau, is presented. The N400 component also occurs when and an offset. For example, the onset of a component a word is semantically acceptable but not expected. For might be when the first few thousand neurons in pri- example, the sentence ‘Do not touch the wet dog’ would mary auditory cortex are systematically responding to a elicit an N400, even though it is not a semantic error; sound, and the offset would be when auditory cortex ‘Do not touch the wet paint’ is the semantically ex- quiets to baseline levels. pected ending (Kutas and Hillyard 1984). The N400 ERP components reflect how processing in the brain occurs regardless of whether the stimuli were presented unfolds over time, from early sensory processing com- visually (e.g., Kutas and Hillyard 1980) or auditorily ponents precede later cognitive processing components (e.g., McCallum et al. 1984) and thus specifically relates that require attention and evaluation of the stimulus. to lexical–semantic processing. The N400 may be used Early components reflecting sensory and perceptual pro- to answer questions about the strength of semantic rela- cessing of a stimulus are often called ‘pre-attentive’, tionships, or whether clinical populations have different meaning they occur without or before conscious at- patterns of semantic processing. For example, studies tention from the participant. Later components typi- have used the N400 component in a modified semantic cally reflect cognitive processing and require conscious paradigm to examine whether receptive semantic pro- attention, though some cognitive ERPs can appear with- cessing differs in children with autism spectrum disorder out the subject consciously perceiving a stimulus (Luck who are minimally verbal or non-verbal (e.g., Cantiani et al. 1996). Passive or unattended paradigms are well et al. 2016). suited to paediatric populations, as they do not re- quire any overt responses or compliance with direc- Auditory and mismatch negativity tions, and they can be collected while the child is at- (MMN) component tending to something more engaging, such as watch- ing a movie. Nonetheless, well-designed and engaging The second example is an oddball paradigm in which paradigms that require active participation or attention one stimulus is presented often and one or more other 4 Sean McWeeny and Elizabeth S. Norton

Figure 1. Typical plots present in ERP research: (A) group grand average waveforms (averaged from about 100 child participants) showing two conditions (frequent and rare syllables, in black and red) and the difference or subtraction between them (blue); and (B) group grand average scalp plot showing the spatial distribution of the mean amplitude of the difference wave depicted in (A) within a time window of interest from 300 to 500 ms. The negativity shown in blue extends bilaterally across fronto-central electrode sites. Data are from the authors’ laboratory. [Colour figure can be viewed at wileyonlinelibrary.com] stimuli are presented rarely. For example, the syllable it exists for visual stimuli as well (Kimura et al. /ba/ is played 90% of the time and the syllable /da/ is 2009). played the remaining 10% of the time while the par- ticipant watches a silent video. The difference between Other common ERP components the frequent stimulus (/ba/) and the oddball or deviant stimulus (e.g., /da/) is a component known as the MMN Although it is not necessary to have in-depth knowl- (Na¨at¨ anen¨ and Kreegipuu 2011). The rare sound elicits edge of established components to read ERP research, a larger negative voltage response from about 100–250 understanding these components will give a better sense ms; this difference between the rare and frequent stim- of how they may be applied to research. These descrip- uli, shown in figure 1A, is the MMN. If the stimuli tions are designed to give a cursory overview of what probabilities were switched (e.g., /ba/ played 10% and types of stimuli and paradigms may elicit components /da/ played 90%), we would see a highly similar pat- related to communication. There are many more com- tern of rare versus frequent responses as in the original ponents that are not covered here, such as ones related paradigm, indicating that the differences are not solely to syntactic incongruity () or error detection (the driven due to acoustic differences in the two syllables, error-related negativity, ERN). Note that some auditory but depend on more general change detection processes. and visual components share the same name, despite The MMN can be used to index automatic auditory dis- originating from completely different areas of the brain crimination of different auditory features (such as tone and responding to different stimulus modalities (i.e., pitch or duration) and how the brain tracks stimulus hearing versus vision). For further detailed reading, see probability, which are skills important for language, es- Luck and Kappenman (2011). pecially language learning. If the difference between the two sounds is too small for the participant to perceive, Auditory sensory components the MMN is not observed. This is only one example of an oddball paradigm, In adults, cortical auditory responses begin with a posi- which is a commonly used ERP paradigm that can tive wave around 50 ms called the P1 or component. be modified with multiple deviants, different stimulus The P50 is generated in auditory cortex and reflects sen- probabilities (e.g., 25% deviants), or a task where the sory gating (Pratt 2011), the process through which the participant actively responds to categorize the stimuli. brain filters out irrelevant information. P50 is quickly The MMN allows researchers to investigate populations followed by a larger negative wave called the N1, or in which auditory discrimination may be impaired. . The N100 reflects a change in auditory stim- Because it does not require overt attention or responses, ulation, such as the offset and onset of sounds (Pratt it has been widely used in infants and children (Bishop 2011). Some evidence suggests that N1 may be used to 2007) and shown to differ in groups with or at risk index speech segmentation, such as in a statistical learn- for language and reading disorders (Bishop 2007, ing paradigm (Sanders et al. 2002). In young children Bruder et al. 2011). Although it is most commonly these components are different, with P1 being larger used in auditory research, some evidence exists that in children than adults (Ceponienˇ e´ et al. 2002); across ERPs in clinical and basic language and communication disorders research 5 development, they tend to shift in time and change in ERP data are recorded and introduce key terms that will shape. show up in most ERP research.

Visual sensory components Recording ERPs Although brief flashes of visual stimuli can evoke ERPs as To record ERPs, a cap or net holding the electrodes soon as 70 ms after stimulus onset (Allison et al. 1977), is placed on the scalp. Saline solution or gel is typi- the main visual ERPs described in research manifest as a cally applied to conduct the electrical signal from the series of three peaks. The first of these components, C1, scalp to the electrodes. This typically takes 5–20 min- changes polarity based on where in the visual field the utes depending on the system and participant. (Newer stimulus is presented (hence, its name having a C rather ‘dry’ electrode systems that do not require conductive than a P or N). It is followed by P1 and N1, which gel or liquid are becoming available, but it is not yet overlap in time (in other words, P1 has not completely clear whether these systems can provide signal-to-noise offset by the time N1 begins); this a similar pattern ratio that is on par with existing research-quality EEG to auditory P1 and N1, though from different neural systems.) Net and cap systems measure the same signals, sources. Although C1, N1 and P1 components each but have some advantages and trade-offs. Net systems peak at different times, they are all generated in visual can be quicker to set up as the saline solution can be cortex in short succession (Pratt 2011). All are affected applied to all electrodes at once and then just placed on by luminance (i.e., brightness) change, such as a white the participant’s head. Electrodes in a cap system need stimulus appearing on a black background. These three to be gelled individually, which takes more time, but the components nicely demonstrate that components can data are often cleaner (less noisy), especially with active have temporal overlap in their offsets and onsets and electrode systems that have signal amplifiers at each site. thus potentially influence the peak latency of any of the In the past, the scalp was typically abraded (scraped) three components (figure 2). to remove a layer of skin to improve signal, but this is not typical with modern systems, which makes EEG recording safer and more comfortable for participants. /P3/P3a/ Additional ‘external’ electrodes are typically placed near Like the MMN, the P3 (also called P300) is typically the eyes to detect eye movements and blinks, as the as- elicited by an oddball paradigm. However, unlike the sociated muscle movements generate an electrical signal MMN, it requires the participant to pay attention to all which can contaminate the data of other nearby elec- the stimuli and have an expectation or categorization of trodes. Very young children may not tolerate electrodes the difference between the standard and the rare target on their face, and so experimenters may use electrodes stimuli (Polich 2007). It can be elicited using stimuli from the forehead to detect blinks. We discuss how to from visual, auditory and other sensory domains (e.g., clean blinks (i.e., reducing their effect on the data) in the tactile; Brouwer and Van Erp 2010, Polich 2007). The Preparing Data for Analysis section. External electrodes P3 wave can be broken up into two unique subcompo- may also be placed to record a reference or comparison nents: P3a and P3b. These two subcomponents reflect signal, such as on the earlobe or mastoid bones behind different aspects of the stimulus; P3b occurs only when the ears. All electrodes are typically plugged in to a small the target stimulus is presented (the one the participant box that amplifies and/or converts the signal to a format is trying to detect), but P3a appears in response to an that can be recorded by a computer. Unlike magnetic infrequent stimulus, regardless of whether the partici- resonance imaging (MRI) or MEG, this equipment is pant was looking for it or not (Polich 2011). The exact small and can be portable. cognitive processes that P3b represents (e.g., working Electrodes are typically named and labelled based versus a more general process) remains an ac- on their location using what is called the 10–20 system, tive area of research and debate (Polich 2011). This which provides a standardized map for locating an elec- illustrates how a single large wave may have multiple trode on the head across people and studies (Homan generators or underlying subcomponents. The P3 com- et al. 1987). Electrode locations in the 10–20 system ponent may be useful in cases in which participants are are labelled with a letter indicating where they are on classifying objects, for example, based on certain object the head (e.g., Fp, frontal pole; F, frontal; FC, fronto- features (Deveney et al. 2019). central; C, central; CP, centro-parietal; P, parietal; PO, parieto-occipital; O, occipital; T, temporal; FT, fronto- temporal; and TP, temporo-parietal) and number (even ERPs from recording to analysis numbers on the right, odds on the left, with larger num- A familiarity with the basics of the experimental set-up bers indicating farther distance from the midline, which can help contextualize ERP data. Here, we explain how uses z instead of a number). For example, Fz is located 6 Sean McWeeny and Elizabeth S. Norton

Figure 2. How different underlying component structures can manifest as identical waveforms: (A) a theoretical ERP waveform with three peaks; (B) one possibility for the underlying component structure that when summed together creates the waveform in (A); and (C) another possibility for the component structure of the waveform in (A). Notice the shorter duration of X2 as compared with X’2 and the larger amplitude of X3 as compared with X’3 in (B). [Colour figure can be viewed at wileyonlinelibrary.com] over frontal cortex, equidistant from the ears, and F1 is not necessarily always better; Luck, 2014: ch. 5.) Voltage slightly to the side of Fz, over the left hemisphere. at each electrode is compared with activity at another Once the electrodes are placed and connected, the site such as the nose, earlobe, mastoid processes or an experiment can begin. Researchers typically use a com- average of all the scalp electrodes. This comparison pro- puter program to present stimuli to the participant. cess is known as referencing, and the site of the reference Everytimeavisualorauditorystimulusispresented electrode(s) greatly affects how the data appear. For ex- (though ERPs exist for other sensory modalities, they ample, using Cz (the electrode at the very top of the are rare), a code is sent from the presentation software head) as the reference would produce different looking to the EEG acquisition software that is recording data ERPs than the average of all electrodes as a reference, from the electrodes. This code in the EEG data indi- as the reference is subtracted from each electrode. Stud- cates the exact moment a stimulus was presented, and ies with similar paradigms typically will use the same different codes are typically used to indicate the type reference locations as previous work, meaning that one of stimulus that was presented. Using the semantic in- can compare directly across publications; however, when congruity task example, if we know exactly when each comparing results across publications, it is important to semantically appropriate word and when each semantic check if the studies use the same reference electrodes. incongruity was presented, we can isolate the electrical activity to each type of stimulus (i.e., incongruous or Preparing data for analysis not) by averaging the trials of each type together. ERP analysis involves many steps and many choices, and so understanding the main aspects of analysis is Time and voltage: The units of ERP measurements important for interpreting the results from papers. There ERPs reflect how electrical voltage changes over time, are established best practices for making these decisions typically over several hundred milliseconds. Voltage and and guidelines for what information about ERP pro- time also characterize the two primary measurements of cessing that studies need to report (Keil et al. 2014). ERPs: amplitude and latency. Amplitude is how large Studies that do not follow these guidelines could re- an ERP voltage is, and latency is when an ERP hap- port insufficient information for replication, inaccurate pens; we discuss how to best measure these below. ERPs or biased data, or inaccurate conclusions about neural measurevoltageovertimewithincredibletemporalres- differences. Thus, this section gives readers some tools olution, often hundreds or even thousands of times per to understand better the steps and decisions that go into second. This temporal precision is perhaps the greatest making inferences about ERPs and the relative method- strength of ERPs. Because electrical current travels very ological strength of a given publication. quickly and the brain likewise processes certain aspects Several steps are needed to go from an EEG record- of a stimulus very quickly, the human auditory brain- ing to a set of ERP waveforms. Here, we describe a stem can register sound about 1.5 ms after it is played. typical processing ‘pipeline’ for generating ERP mea- A cognitive event such as detection of a semantic incon- surements from EEG data. The first is to filter the EEG gruity occurs much later, 400 ms post-stimulus onset, data so that only activity that is likely to be relevant to though that may not seem long! brain processes remains. Much of the activity of neurons Most research laboratories use systems that record is rhythmic, or oscillatory. Thus, a first step is to filter voltage from many electrodes simultaneously; 32 and out oscillatory activity that is slower than about 0.01– 64 electrode caps are common in research, and nets of- 0.3 Hz (high-pass filter), or faster than about 100 Hz ten include 128 or 256 electrodes. (More electrodes are (low-pass filter). High-pass filters at 0.3 Hz and above ERPs in clinical and basic language and communication disorders research 7 can systematically change the shape and timing of ERPs terval and reject any interval in which data that exceed (Tanner et al. 2015), so that is something to watch out a given voltage threshold (e.g., 150 µV). This process is for. Often, a notch filter, which knocks out a single fre- then repeated for every 200-ms interval in each epoch. quency or small frequency range, is applied to remove Some researchers choose apply the same threshold for electrical noise (e.g., at 50 Hz in Europe and Asia or 60 artefact detection to each subject’s data, which can speed Hz in North America), as utility electricity is carried at up this process; others choose to use individualized cri- those frequencies and electrodes may pick up that fre- teria, as ‘clean data’ look different in each person, which quency. Electrical noise, also called line noise, is a type is an acceptable and often necessary approach for pae- of artefact. diatric data (Luck 2014). The data marked as artefacts An artefact is best described as a non-event-related are then excluded from further analysis. It is common signal, or noise, that can arise for many reasons. To ob- in studies of children for up to 50% of epochs to have tain the cleanest ERPs, artefacts must first be detected, artefacts. Thus, it is important to consider how many and then either corrected or rejected from the data. usable trials are averaged together; too few trials may Perhaps the most common artefact comes from the par- lead to noisy averages that do not reflect the real neural ticipant blinking, which creates a large voltage change activity accurately. If different conditions or groups are due to the muscle activity required to blink. Blinks can being compared, it is ideal to make sure that conditions be corrected (i.e., their effect subtracted from the data) or groups do not have disparate numbers of included through a computational process called independent trials (see Number of Trials section). component analysis (ICA). An alternative approach is After the data have been cleaned of artefacts, epochs to just exclude (‘reject’) moments of data with blinks; within each condition (e.g., all incongruous trials) are however, ICA can increase the number of usable trials, averaged together. As discussed above, averaging to- which is typically desirable. gether multiple epochs allows one to see the components Once ICA is complete, the data are segmented into of interest, because random noise not associated with epochs, which are periods around an event or stimulus processing the stimulus will cancel out. Once epochs of onset. Epochs are created based on the codes that are sent a given type are averaged together, the resulting wave- when a stimulus is presented, described in the Record- form is considered an ERP. Average ERPs of a whole ing ERPs section. Typically, within an experiment, each group of participants or comparisons of ERPs between epoch is the same length. Epochs typically begin 50– groups are often the main interest in ERP research; thus, 200 ms before the stimulus and last until sometime publications do not often include the individual partici- before the next stimulus is presented. The time before pant waveforms calculated in the previous step. Instead, each stimulus is called the pre-stimulus baseline, or just they typically present grand average waveforms which baseline. The average voltage over each trial’s baseline is are calculated by averaging the waveforms from every subtracted from the post-stimulus signal in an effort to individual in a group. For example, if we were interested reduce noise. in whether early auditory processing differs for toddlers Next, any remaining artefacts must be detected in at risk for dyslexia compared with those not at risk, we the data and rejected. Artefact rejection and correction might compare the grand average for the at-risk group is necessary, as large artefacts will distort the final ERP with the grand average for the not-at-risk group (e.g., waveforms with large voltage changes due to noise. (The Plakas et al. 2013). These smooth grand average wave- saying ‘a few bad apples can spoil the bunch’ applies forms are not necessarily what any individual’s waveforms in this case.) Many artefacts in EEG data come from look like, nor what any given trial might look like for motion by the participant; this is particularly true for any individual. This is because it takes many trials to children, who tend to move more during recording. cancel out unrelated activity (i.e., noise), and many par- Another common type of artefact appears when an elec- ticipants sufficiently to cancel out noise and smooth the trode’s connection to the scalp was poor or lost during group’s response (i.e., signal). Grand averages can also the experiment. The information from these electrodes be compared across different experimental conditions can be deleted and re-created using an average of the within-subjects, such as semantically expected versus electrodes surrounding it in a process known as inter- incongruous words, to understand better basic neural polation. Ideally, this occurs for very few electrodes, as processes of semantic comprehension. interpolating too many adjacent electrodes will result in an interpolated channel that is likely very different Plotting ERPs and interpreting figures than the actual brain activity that was occurring at that electrode site. After any bad electrodes are interpolated, Plotting ERPs allows one to visualize the differences the data are examined for remaining artefacts. There between conditions (e.g., semantic incongruity versus are lots of ways to do this: one way is to measure how semantically appropriate word) or between groups (e.g., much the measured voltage changes over a 200-ms in- typical development versus autism). Typically, just 8 Sean McWeeny and Elizabeth S. Norton before plotting, a 30-Hz low-pass filter is applied to depth in the Localization section. Nonetheless, differ- remove high-frequency noise, which makes the wave- ences in scalp maps between conditions or groups can form appear smoother. Figure 1A shows a typical ERP provide meaningful information. waveform plot using the auditory oddball paradigm (/ba/ presented 90% of the time and /da/ presented 10% Quantification or measurement of ERPs of the time). Time (in ms) is plotted on the x-axis, and voltage (in µV) is plotted on the y-axis. Here, negative Although looking at the waveforms provides a great values are plotted up; some figures use this convention, deal of information, quantifying or measuring an ERP which has been typical in the history of ERP research, component is necessary to compare statistically be- although others plot positive up, which is more intuitive tween groups or conditions. Understanding the gen- for most readers. It is important to pay attention to the eral strengths and weaknesses of different measurement axis’ labels to confirm which direction is plotted up. approaches helps readers to understand better the rel- The x-axis spans from –100 to 500 ms relative to the ative strength of a given paper. As mentioned above, stimulus onset. As discussed above, the average voltage ERPs can be quantified in terms of amplitude and la- over the baseline from –100 to 0 ms has been subtracted tency; most papers measure one or both of these char- from the whole waveform, and the baseline section is acteristics of a given component. Though many stud- thus relatively flat. Readers can check for pre-stimulus ies measure features about the peak of a component, baselines to be relatively flat compared with the wave- there are some drawbacks to doing so. There is nothing forms that occur after stimulus onset; if they are not, inherently meaningful about the time and exact volt- this could indicate noisy data or inaccurate processing. age at which a component reaches a maximum voltage Figure 1A has three ERP waveforms: the waveform (Luck 2014). Furthermore, the exact time and volt- to the /ba/ (90%) condition is in black and the ERP age at which peaks occur (i.e., peak latency and peak waveform to the /da/ (10%) condition is in red. Notice amplitude) can change based on how much noise is that both the black and red waves have a positive-going in the EEG signal. This is because noise distorts the wave around 100 ms. Before stimulus onset and early waveform, which may create a false peak that does after onset, the two conditions do not differ, which not accurately represent the component. Although peak suggests that neural responses are similar between con- amplitude and peak latency may provide information ditions, as we would expect. From 100 to 200 ms and about a component, some less-biased alternatives are from 250 to 500 ms, there are negative-going peaks in mean amplitude between two time points. Finding an this wave, indicating that one condition had a stronger alternative to peak latency that is robust to noise is negative response than the other. The blue waveform is more challenging; the approach that is most valid (from a difference wave; is a subtraction between the two con- options such as fractional area latency, or relative cri- ditions (deviant minus standard waves), which allows terion technique) may vary depending on the com- readers to visualize more easily the difference between ponent of interest (Kiesel et al. 2008). Overall, peak the conditions. The waveforms being compared in a measures should be regarded with these considerations difference wave should have an approximately equiv- in mind. alent number of trials (Luck 2005). From our odd- Even for noise-robust measurements such as mean ball paradigm example, comparing every frequent sound amplitude, the measurement window and electrode(s) with every rare (deviant) sound would result in signal to measure and analyse must be chosen carefully. Ide- that is much noisier for the rare sound, because there ally, all publications would use measurement windows are nine times fewer trials of this kind presented. In this selected in advance of analysis based on closely related case, researchers can use a subset of the frequent sounds previous literature so as to be unbiased; however, the (e.g., average together and analyse only some standards). literature can differ in exactly what time window they ERPs can also be plotted using scalp maps, which quantified their components, or a paradigm or popula- show how positive or negative the ERP signal is across tion may have never been studied before. When certain thescalpatacertaintimeortimeinterval.Infigure1B, ERP components begin can depend on the duration the mean amplitude (voltage) of the difference wave be- of a stimulus, the time between stimulus onsets (i.e., tween 300 and 500 ms is plotted at each electrode and stimulus onset asynchrony), as well as the time between then smoothed in space to create a continuous map of stimulus offset and the next stimulus onset (i.e., inter- voltage. The colour bar indicates that negative values stimulus interval). Thus, the time when the component are plotted in blue; meaning that over the frontal and occurs will vary from study to study depending on the central part of the scalp, the difference wave is negative. stimuli used. Even though we see an ERP over a certain part of the Given the concerns about rigor and reproducibil- scalp, this does not necessarily mean that the process ity in neuroscience (e.g., Loken and Gelman 2017), happened in that brain area; this will be discussed in a major concern for all ERP research is the potential ERPs in clinical and basic language and communication disorders research 9 for experimenter bias, either conscious or unconscious. other, known as lateralization. Lateralization is relevant Ideally, analytical decisions such as which electrode site to questions related to language because language func- and time window to test, and how to exclude artefacts tions are typically left-hemisphere lateralized; thus, one and outliers, are made in advance of data collection, might expect to see larger amplitudes in left versus right so that the data cannot influence the results. Practices hemisphere electrodes. Many studies have reported lat- that should be avoided are ‘fishing’ for a time window eralization being weaker or absent in language-related in which there is a difference and then conducting sta- disorders. Lateralization can be calculated in numer- tistical tests only in that window, or testing many time ous ways such as using centroids (spatial means) of windows and electrodes and not reporting these tests scalp maps (e.g., Maurer et al. 2009) or comparing sin- and appropriately controlling for the multiple statistical gle or groups of electrodes from left and right hemi- comparisons. Luck and Gaspelin (2017) illustrate this spheres (e.g., Monjauze et al. 2011). Although there is nicely with an example where statistically significant re- no consensus on the best way to study lateralization, it sults were obtained from real ERP data by pure chance is nonetheless a valid practice in ERP research. by using analytical practices that are common in the field, but which are not rigorous. Their message is clear: in highly multidimensional ERP data, it is extremely Considerations for experimental design and easy to find statistically significant results, even though interpretation nothing meaningful exists, and peer review may not Number of trials always catch these instances of false-positive findings. Readers can look for whether authors made analytical Researchers face many trade-offs in designing experi- choices based on previous literature or a priori hypothe- ments. Ideally, many trials (often, hundreds per con- ses, rather than seeming to test many possibilities until dition) are collected so that the cleanest possible aver- one reached significance. As more authors pre-register age ERP can be achieved. However, this is not always their analytical plans in advance of data collection and practical for many reasons. The length of an experi- journals encourage this rigorous practice, it may become ment paradigm is an important consideration because easier to know which findings were truly robust. children (and even sleep-deprived undergraduates) may have a hard time sitting still and maintaining attention to the experiment over long periods. On the one hand, Localization increasing the number of trials can help increase statis- Perhaps the most notable limitation of ERPs is their poor tical power for between- and within-group analyses by spatial resolution; because of the way electricity moves improving reliability (Boudewyn et al. 2018). On the and spreads through the brain, skull and scalp, voltage other hand, limiting the length of the paradigm may at a given electrode does not simply reflect activity that allow more participants to tolerate the duration of the is generated directly below its location. Quantifying the paradigm. Furthermore, some ERPs may habituate or precise anatomical region in the brain that produces an disappear entirely over time (McGee et al., 2001). There ERP component, also known as localization, is often im- is not a clear rule or guideline of how many trials are possible without making several assumptions about the ‘enough’ to include, as this depends on how clean the locations and numbers of generators of that process in data are and how large is the component. Efforts should the brain that are often unknown. Making claims about thus be made to determine the reliability of a given ERP, a certain ERP’s precise location in the brain, or differ- or refer to published research that has determined an ap- ences by conditions or groups, requires an extremely propriate number of trials (e.g., Foti et al. 2013). This high burden of proof. Even with analysis technology is challenging, as there seems to be no clear answer for such as LORETA (Pascual-Marqui et al. 2002), the use how many trials are needed to measure any given ERP of high-density electrode arrays, and using each individ- component reliably (e.g., anywhere from 10 to several ual’s MRI scans to assist in analysis, poor localization hundred) across all EEG systems or stimuli parameters, remains a fundamental limitation of EEG and ERPs. but the number of trials presented and used/accepted that The sources of ERP signals can be estimated these pro- are compared per participant group and/or condition cesses, but we suggest proceeding with caution regarding should be reported (Boudewyn et al. 2018). claims about truly ‘localizing’ an ERP. Differences across conditions and motor movements Lateralization Motor responses are another important consideration in Because localization of ERPs is so limited, researchers in- the design of ERP experiments. For example, if the goal terested in spatial questions will sometimes test whether of an experiment is to have participants actively catego- an effect is stronger on one side of the brain than the rize a rare stimulus and researchers want to be sure that 10 Sean McWeeny and Elizabeth S. Norton this is being done accurately, they may ask participants rather than a negative-going wave in infants, and is thus to press a button to each rare stimulus. Importantly, the sometimes called the mismatch response (MMR) (van ERP response to the rare stimulus will also contain the Leeuwen et al. 2006). This can make comparisons be- motor activity associated with the button press. This un- tween different age groups very difficult when discussing derscores an important feature of design that readers can each component on its own, especially considering the again watch for; everything possible should be the same problem presented in figure 2 of overlapping compo- between two types of trials being compared. If trials in nents that affect each other’s shapes. It is also typical one condition have a button press (or other response), for the overall amplitude of ERPs to decrease with age. the comparison trials should as well, or the source of any Readers should consider these factors when studies com- differences between trials cannot be determined. Some pare across ages. potential solutions would be to have a button press on every trial (one button for rare and one for frequent, pressed using the same hand or finger), to add a third Why use ERPs? type of trial that is excluded from analysis in which par- Given the complexities and challenges of ERP research, ticipants press a button to indicate their attention, or to what are the advantages of this technique? Perhaps the have no button press at all. Closely matched conditions clearest reason to use ERPs is that ERPs and behaviour thus are necessary to isolate the activity of interest with do not have a one-to-one relationship; if they did, there no confounding factors. would be no reason to study ERPs when behaviour could be used. For example, the MMN may be a more sensi- tive indicator of auditory discrimination than behaviour Interpreting peaks and components (Stoodley et al. 2006). This means that we are measuring Even though components follow the general pattern of something unique with ERPs, that can add information timing of different processes in the brain, they must to our understanding of typical and atypical behaviours. be interpreted carefully. A common misconception is Here, we discuss the primary reasons researchers use that the peak (highest or lowest point between onset ERPs. and offset) indicates something meaningful about the component or underlying brain processes (Luck, 2014). The same observed waveform (i.e., the plot of voltage When behaviour is hard to obtain changeovertime)couldbemadeupofmultiplesetsof One of the best reasons to use ERPs is when behavioural components that overlap, so the final waveform peak’s paradigms would be ill-suited to, or difficult to obtain amplitude (voltage at the peak) and latency (timing at from, the population of interest. A clear example of the peak) may not necessarily be important (figure 2). this is with auditory discrimination among infants. The This is important to keep in mind; for example, it is MMR ERP,an infant corollary of the MMN, is elicited incorrect to assume that because you see the peak of an by an oddball paradigm and does not require the atten- ERP in condition A earlier than the peak in condition B, tion of the infant (van Leeuwen et al. 2006). Like the that this is because something about condition A causes MMN, the component is believed to represent an objec- the component to happen earlier. There could be other tive measure of auditory discrimination, which roughly components’ onsets or offsets occurring simultaneously corresponds to behavioural discrimination thresholds in that would shift the peaks despite the component of adults (Na¨at¨ anen¨ and Kreegipuu 2011). As behavioural interest remaining stable. tests cannot otherwise assess this skill in very young infants, ERPs offer a window into how this process de- velops over time. Furthermore, as behavioural auditory Developmental changes discrimination is impaired in some (but not all) indi- A particular challenge for paediatric and longitudinal viduals with paediatric language delays and disorders ERP research is that components change over the course (Benasich and Tallal 2002), one might expect this dif- of development. A pattern of auditory responses, the P1- ference to be present in the brain from birth. N1-P2 complex, is very well established and studied in adults, is not found even by age 9; in children, P1 is the most prominent of these components, whereas N1 As a biomarker when behaviour cannot tell us is most prominent in adults (Ceponienˇ e´ et al. 2002). enough Indeed, there is considerable change in the morphology Another major interest of many researchers is whether (i.e., shape) and timing of these and other components ERPs could practically aid in clinical screening, from infancy to childhood (Ceponienˇ e´ et al. 2002, Wun- diagnosis or progress monitoring for individuals with a derlich et al. 2006). The MMN can actually be a positive disorder. An ERP used in this way would be considered ERPs in clinical and basic language and communication disorders research 11 a ‘biomarker’; a biomarker is a biological substance or the hypothesis and ERP components used, should dis- process that is objectively quantifiable, reproducible cuss how their results tested their hypothesis as explicitly and related to underlying biology implicated in a as possible. disorder. Because so many disorders in the field of paediatric communication disorders, such as autism, Furthering basic understanding of neural processing language disorder and dyslexia cannot be diagnosed until children fail to achieve typical milestones, early ERPs have also been applied at a more basic level to and objective neural markers of these disorders have understand how humans process different types of in- been a focus of much research. Using ERPs in this formation. The semantic incongruity paradigm has been way has the potential to be an extremely powerful used extensively to describe how the brain processes lin- tool, but it requires highly reliable measures. Many guistic information. Other examples of basic applica- studies have found ERP components that relate to tions of ERPs are how the brain form auditory objects specific language impairment/developmental language (e.g., Sussman and Steinschneider 2009), how it inte- disorder (e.g., Friedrich et al. 2004) or dyslexia (Bruder grates information from multiple senses (Brett-Green et al. 2011, Halliday et al. 2014, Maurer et al. 2009). et al. 2008) and how it processes emotion in faces (Blau Researchers have also measured the MMR at birth in et al. 2007). Because these uses are not particularly ap- at-risk individuals and followed them through to school plicable clinically, we will not discuss them further, but age to see how well ERPs predicted who would develop many of the applied uses of ERPs were born out of these dyslexia (Jyvaskyl¨ a¨ Longitudinal Study of Dyslexia; e.g., basic research applications. Guttorm et al. 2010). Associations with later language outcomes from these early measures show some promise, When and why not to use ERPs but much more research is needed to determine the generalizability of these results (Norton et al. 2019). Despite the advantages of ERPs in many situations, To date, these biomarker studies have yet to be taken there are also several situations where it is most appro- beyond the initial stages of discovery to more clinically priate to not use this technique. When a behavioural meaningful statistical validation and implementation. experiment can answer the same question, there is often Once preliminary associations are found, candidate ERP no reason to ‘tack on’ ERPs. That is, just because a biomarkers must be subjected to cross-validation to test process can be measured using an ERP does not mean it their generalizability and controlled replication studies should; a study should use ERPs when they are the right (for a model of bringing brain measures to clinical utility, tool to answer a given question. If the research questions see Gabrieli et al. 2015). Despite these challenges, ERPs focus on where a process occurs in the brain, ERPs remain an enticing biomarker because they are relatively may not be a good technique relative to other methods fast, easy and inexpensive to obtain (compared with, such as functional magnetic resonance imaging (fMRI) say, a detailed cognitive and language assessment or an or MEG. ERP studies are less well-suited to certain MRI), and can be obtained early in life and thus lead to populations, such as those using cochlear implants, earlier identification and intervention. as the electrical activity and placement of the implant receiver may affect the signal. Uncovering aetiology of disorders Clinical implications and conclusions Most language disorders are neurologically based, but their causes are not fully understood; there is hope that In our view, research using ERPs is promising, but has ERPs may be able to help elucidate causes of disorders, not yet made a direct impact on how clinical diagnosis or rather than just understand their behavioural symptoms. treatment of speech or language disorders proceeds. Al- Even if we cannot tell much about an individual owing though ERPs have been used in basic research to under- to low reliability of measures, we may still be able to stand cognitive processes for decades, their applications understand the nature of neural processing in groups of to clinical questions has been growing as technology be- people with language differences or disorders with large comes more affordable and accessible. Using ERPs to samples and rigorous practices. understand clinical disorders is still somewhat in its in- For example, one hypothesis about the cause of fancy, and it is important to recognize that ERP research dyslexia is that there is a neural impairment in the way overall has evolved and been refined over decades. It will that sounds are processed and discriminated. Compar- likely take decades more of rigorous research to bring ing the MMN ERPs from two groups that are carefully ERPs to clinical application, as we learn more every matched (e.g., in age, number of included trials, etc.) year about the most valid and reliable ways to measure can indicate whether this automatic auditory process is ERPs and technology makes them more feasible to col- altered in language disorders. All studies, regardless of lect (including that they are more participant-friendly, 12 Sean McWeeny and Elizabeth S. Norton quick to set up and affordable). At the same time, we BROUWER,A.-M.andVAN ERP, J. B. F., 2010, A tactile P300 are also learning more about the heterogeneity of speech brain-computer interface. Frontiers in Neuroscience, 4:19. and language disorders, and how we can best address this https://doi.org/10.3389/fnins.2010.00019. BRUDER,J.,LEPPANEN¨ ,P.H.T.,BARTLING,J.,CSEPE´ , heterogeneity both clinically and scientifically. Nonethe- V., D EMONET´ ,J.-F.andSCHULTE-KORNE¨ , G., 2011, less, we believe it is important for clinicians to interpret Children with dyslexia reveal abnormal native lan- ERP results (1) because informed readers will push re- guage representations: Evidence from a study of searchers to improve the state of science, (2) because mismatch negativity. Psychophysiology, 48, 1107–1118. understanding the shortcomings in existing research can https://doi.org/10.1111/j.1469-8986.2011.01179.x. BUZSAKI´ ,G.,ANASTASSIOU,C.A.andKOCH, C., 2012, The ori- provide a road map of what to look for in the years to gin of extracellular fields and currents — EEG, ECoG, come in ERP experiments, and (3) lastly, if ERPs move LFP and spikes. Nature reviews neuroscience, 13, 407–420. into clinical practice in future, clinicians can be ready https://doi.org/10.1038/nrn3241. to apply their knowledge of the data to their patients CANTIANI,C.,CHOUDHURY,N.A.,YU,Y.H.,SHAFER,V.L., as if it were standardized behavioural assessment data. SCHWARTZ,R.G.andBENASICH, A. A., 2016, From sensory perception to lexical-semantic processing: An ERP study in Although ERPs are not used in the clinic to assess speech non-verbal children with autism. PLOS ONE, 11, e0161637. and language disorders at this time, we believe that sci- https://doi.org/10.1371/journal.pone.0161637. entific literacy on the topic is key to facilitating research Cˇ EPONIENE´,R.,RINNE, T. and NA¨AT¨ ANEN¨ , R., 2002, Matura- that would help achieve this goal. tion of cortical sound processing as indexed by event- related potentials. Clinical Neurophysiology, 113, 870–882. https://doi.org/10.1016/S1388-2457(02)00078-0. Acknowledgments DEVENEY,C.M.,BRIGGS-GOWAN,M.J.,PAGLIACCIO, D., ES- TABROOK,C.R.,ZOBEL,E.,BURNS,J.L.,NORTON,E. We acknowledge Steve Luck, whose writing and teaching has guided S., PINE,D.S.,BROTMAN,M.A.,LEIBENLUFT,E.and many of the topics covered in this tutorial. We thank Katherine WAKSCHLAG, L. S., 2019, Temporally sensitive neural mea- Wolfert for comments on EEG net setup and Brittany Manning sures of inhibition in preschool children across a spectrum CCC-SLP for discussion on clinical considerations. E.S.N. was sup- of irritability. Developmental Psychobiology, 61, 216–227. ported by the National Institute on Deafness and Other Communi- https://doi.org/10.1002/dev.21792. cation Disorders (NIDCD) under award number R01DC016273. FOTI, D., KOTOV, R. and HAJCAK, G., 2013, Psychometric consid- The content is solely the responsibility of the authors and does not erations in using error-related brain activity as a biomarker necessarily represent the official views of the National Institutes of in psychotic disorders. Journal of Abnormal Psychology, 122, Health. 520–531. https://doi.org/10.1037/a0032618. FRIEDRICH,M.,WEBER,C.andFRIEDERICI,A.D., 2004, Electrophysiological evidence for delayed mis- References match response in infants at-risk for specific lan- guage impairment. Psychophysiology, 41, 772–782. ALLISON,T.,MATSUMIYA,Y.,GOFF,G.D.andGOFF, W. R., 1977, https://doi.org/10.1111/j.1469-8986.2004.00202.x. The scalp topography of human visual evoked potentials. GABRIELI,J.D.E.,GHOSH,S.S.andWHITFIELD-GABRIELI, and Clinical Neurophysiology, 42, 185– S., 2015, Prediction as a humanitarian and pragmatic 197. https://doi.org/10.1016/0013-4694(77)90025-6. contribution from human cognitive neuroscience. Neuron, BELL,M.A.andCUEVAS, K., 2012, Using EEG to study 85, 11–26. https://doi.org/10.1016/j.neuron.2014.10.047. cognitive development: Issues and practices. Journal of GUTTORM,T.K.,LEPPANEN¨ ,P.H.T.,HAM¨ AL¨ AINEN¨ ,J.A.,EKLUND, Cognition and Development, 13, 281–294. https://doi.org/ K. M. and LYYTINEN, H. J., 2010, Newborn event-related 10.1080/15248372.2012.691143. potentials predict poorer pre-reading skills in children at risk BENASICH,A.A.andTALLAL, P., 2002, Infant discrimi- for dyslexia. Journal of Learning Disabilities, 43, 391–401. nation of rapid auditory cues predicts later language https://doi.org/10.1177/0022219409345005. impairment. Behavioural Brain Research, 136, 31–49. HALLIDAY,L.F.,BARRY,J.G.,HARDIMAN,M.J.andBISHOP,D. https://doi.org/10.1016/S0166-4328(02)00098-0. V., 2014, Late, not early mismatch responses to changes BISHOP, D. V. M., 2007, Using mismatch negativity to study in frequency are reduced or deviant in children with central auditory processing in developmental language dyslexia: an event-related potential study. JournalofNeurode- and literacy impairments: Where are we, and where velopmental Disorders, 6, 21. https://doi.org/10.1186/1866- should we be going? Psychological Bulletin, 133, 651–672. 1955-6-21. https://doi.org/10.1037/0033-2909.133.4.651. HE, J. B., 2015, Electrocorticogram (ECoG). In D. Jaeger and R. BLAU,V.C.,MAURER,U.,TOTTENHAM, N. and MCCANDLISS,B. Jung (eds), Encyclopedia of Computational Neuroscience (New D., 2007, The face-specific component is modulated by York, NY: Springer Publishing Company), pp. 1070–1074. emotional facial expression. Behavioral and Brain Functions, HENDERSON,L.M.,BASELER,H.A.,CLARKE,P.J.,WAT- 3, 7. https://doi.org/10.1186/1744-9081-3-7. SON,S.andSNOWLING, M. J., 2011, The N400 ef- BOUDEWYN,M.A.,LUCK,S.J.,FARRENS,J.L.andKAPPENMAN, fect in children: Relationships with comprehension, vo- E. S., 2018, How many trials does it take to get a signif- cabulary and decoding. Brain and Language, 117, 88–99. icant ERP effect? It depends. Psychophysiology, 55, e13049. https://doi.org/10.1016/j.bandl.2010.12.003. https://doi.org/10.1111/psyp.13049. HOMAN,R.W.,HERMAN, J. and PURDY, P., 1987, Cerebral lo- BRETT-GREEN,B.A.,MILLER,L.J.,GAVIN,W.J.andDAVIES, cation of international 10–20 system electrode placement. P. L., 2008, Multisensory integration in children: A pre- Electroencephalography and Clinical Neurophysiology, 66, 376– liminary ERP study. Brain Research, 1242, 283–290. 382. https://doi.org/10.1016/0013-4694(87)90206-9. https://doi.org/10.1016/j.brainres.2008.03.090. ERPs in clinical and basic language and communication disorders research 13

KEIL,A.,DEBENER,S.,GRATTON,G.,JUNGHOFER¨ ,M.,KAPPEN- NA¨AT¨ ANEN¨ ,R.andKREEGIPUU, K., 2011, The Mismatch Negativity MAN,E.S.,LUCK,S.J.,LUU,P.,MILLER,G.A.andYEE, (MMN). In The Oxford handbook of event-related potential C. M., 2014, Committee report: Publication guidelines and components (Oxford University Press). recommendations for studies using electroencephalography NATIONAL CENTER FOR HEARING ASSESSMENT AND MANAGEMENT, and magnetoencephalography. Psychophysiology, 51, 1–21. 2011, State early hearing detection and intervention/universal https://doi.org/10.1111/psyp.12147. newborn hearing screening mandates (Utah State University). KIESEL,A.,MILLER,J.,JOLICŒUR,P.andBRISSON,B., NORTON,E.S.,GABRIELI,J.D.E.andGAAB, N., 2019, Neural pre- 2008, Measurement of ERP latency differences: dictors of dyslexia. In L. Verhoeven, C. Perfetti and K. Pugh A comparison of single-participant and jackknife- (eds), Developmental Dyslexia across Languages and Writing based scoring methods. Psychophysiology, 45, 250–274. Systems: A Handbook (Cambridge University Press), pp. 253– https://doi.org/10.1111/j.1469-8986.2007.00618.x. 276. KIMURA,M.,KATAYAMA,J.,OHIRA,H.andSCHROGER¨ ,E., PASCUAL-MARQUI, R. D., ESSLEN,M.,KOCHI, K. and LEHMANN, 2009, Visual mismatch negativity: New evidence from D., 2002, Functional imaging with low resolution brain elec- the equiprobable paradigm. Psychophysiology, 46, 402–409. tromagnetic tomography (LORETA): a review. Clinical Phar- https://doi.org/10.1111/j.1469-8986.2008.00767.x. macology, 12. KUTAS, M. and HILLYARD, S. A., 1980, Reading senseless sentences: PLAKAS,A.,VAN ZUIJEN,T.,VAN LEEUWEN,T.,THOMSON,J.M. brain potentials reflect semantic incongruity. Science, 207, and VAN DER LEIJ, A., 2013, Impaired non-speech auditory 203–205. https://doi.org/10.1126/science.7350657. processing at a pre-reading age is a risk-factor for dyslexia KUTAS, M. and HILLYARD, S. A., 1984, Brain potentials during but not a predictor: An ERP study. Cortex, 49, 1034–1045. reading reflect word expectancy and semantic association. https://doi.org/10.1016/j.cortex.2012.02.013. Nature, 307, 161–163. https://doi.org/10.1038/307161a0. POLICH, J., 2007, Updating P300: An integrative theory of VAN LEEUWEN,T.,BEEN,P.,KUIJPERS,C.,ZWARTS,F.,MAASSEN,B. P3a and P3b. Clinical Neurophysiology, 118, 2128–2148. and VAN DER LEIJ, A., 2006, Mismatch response is absent in 2- https://doi.org/10.1016/j.clinph.2007.04.019. month-old infants at risk for dyslexia. NeuroReport, 17, 351. POLICH, J., 2011, Neuropsychology of P300. In S. J. Luck and E. https://doi.org/10.1097/01.wnr.0000203624.02082.2d. S. Kappenman (eds), The Oxford handbook of event-related LOKEN, E. and GELMAN, A., 2017, Measurement error potential components (Oxford, England: Oxford University and the replication crisis. Science, 355, 584–585. Press), pp. 159–188. https://doi.org/10.1126/science.aal3618. PRATT, H., 2011, Sensory ERP components. In S. J. Luck and LUCK, S. J., 2005, Ten simple rules for designing ERP experiments. In E. S. Kappenman (eds), The Oxford handbook of event- T. C. Handy (ed), Event-related potentials: a methods handbook related potential components (Oxford University Press), pp. 89– (Cambridge, MA: MIT Press), pp. 17–32. 114. LUCK, S. J., 2014, An introduction to the event-related potential tech- PUCE,A.andHAM¨ AL¨ AINEN¨ , M. S., 2017, A review of issues related nique. 2nd ed. (Cambridge, MA: MIT Press). to data acquisition and analysis in EEG/MEG studies. Brain LUCK,S.J.andGASPELIN, N., 2017, How to get sta- Sciences, 7, 58. https://doi.org/10.3390/brainsci7060058. tistically significant effects in any ERP experiment SANDERS,L.D.,NEWPORT, E. L. and NEVILLE, H. J., 2002, Seg- (and why you shouldn’t). Psychophysiology, 54, 146–157. menting nonsense: an event-related potential index of per- https://doi.org/10.1111/psyp.12639. ceived onsets in continuous speech. Nature Neuroscience, 5, LUCK,S.J.andKAPPENMAN, E. S., 2011, The Oxford handbook of 700–703. https://doi.org/10.1038/nn873. event-related cotential components (Oxford, England: Oxford SKOE, E. and KRAUS, N., 2010, Auditory brainstem response to University Press). complex sounds: a tutorial. Ear and hearing, 31, 302–324. LUCK,S.J.,VOGEL, E. K. and SHAPIRO, K. L., 1996, Word meanings https://doi.org/10.1097/AUD.0b013e3181cdb272. can be accessed but not reported during the attentional blink. STOODLEY,C.J.,HILL,P.R.,STEIN,J.F.andBISHOP,D. Nature, 383, 616–618. https://doi.org/10.1038/383616a0. V. M., 2006, Auditory event-related potentials differ in MAURER,U.,BUCHER,K.,BREM,S.,BENZ,R.,KRANZ,F., dyslexics even when auditory psychophysical performance SCHULZ,E.,VAN DER MARK,S.,STEINHAUSEN,H.-C.and is normal. Brain Research, 1121, 190–199. https://doi.org/ BRANDEIS, D., 2009, Neurophysiology in preschool im- 10.1016/j.brainres.2006.08.095. proves behavioral prediction of reading ability through- SUSSMAN, E. and STEINSCHNEIDER, M., 2009, Attention ef- out primary school. Biological Psychiatry, 66, 341–348. fects on auditory scene analysis in children. Neu- https://doi.org/10.1016/j.biopsych.2009.02.031. ropsychologia, 47, 771–785. https://doi.org/10.1016/ MCCALLUM,W.C.,FARMER, S. F. and POCOCK, P. V., 1984, j.neuropsychologia.2008.12.007. The effects of physical and semantic incongruities on audi- TANNER, D., MORGAN-SHORT, K. and LUCK, S. J., 2015, tory event-related potentials. Electroencephalography and Clin- How inappropriate high-pass filters can produce artifac- ical Neurophysiology/Evoked Potentials Section, 59, 477–488. tual effects and incorrect conclusions in ERP studies of https://doi.org/10.1016/0168-5597(84)90006-6. language and cognition. Psychophysiology, 52, 997–1009. MCGEE,T.J.,KING,C.,TREMBLAY,K.,NICOL,T.G.,CUNNING- https://doi.org/10.1111/psyp.12437. HAM, J. and KRAUS, N., 2001, Long-term habituation of WEISS, S. and MUELLER, H. M., 2003, The contribu- the speech-elicited mismatch negativity. Psychophysiology, 38, tion of EEG coherence to the investigation of lan- 653–658. guage. Brain and Language, 85, 325–343. https://doi.org/ MONJAUZE,C.,BROADBENT,H.,BOYD,S.G.,NEVILLE,B.G. 10.1016/S0093-934X(03)00067-1. R. and BALDEWEG, T., 2011, Language deficits and altered WUNDERLICH,J.L.,CONE-WESSON, B. K. and SHEPHERD, R., 2006, hemispheric lateralization in young people in remission from Maturation of the cortical auditory in in- BECTS. Epilepsia, 52, e79–e83. https://doi.org/10.1111/ fants and young children. Hearing Research, 212, 185–202. j.1528-1167.2011.03105.x. https://doi.org/10.1016/j.heares.2005.11.010.