Given that ERPs (a) provide precise information about the timing of brain events and some information about their spatial distribution, and (b) are rela- tively easy to acquire even from participants with limited verbal or motor abil- ities, they are a potentially attractive method for studying the development of visual functions. However, when considering applying ERPs to a question of visual development, it is important to consider not only these advantages, but also some potential limitations/challenges related to the typical methods used to acquire and analyze visual ERPs. These issues are considered in the next section, with illustrations of these points from particular visual devel- opmental ERP experiments.
Experimental Design
As with any experiment, some careful thought is required before beginning one involving ERPs. The typical way in which ERPs are recorded places some constraints both on the types of situations or tasks that are amenable to study with ERPs and the type and amount of information that is obtainable even in a suitable paradigm.
Brief, Repeatable Events that Can Be Defined in Time
Event-related potentials are, by definition, brain activity related to a particular event that occurs at an identifiable point in time (e.g., onset of a stimulus, or execution of a behavioral response). Thus, when designing a visual ERP study, it is important to keep in mind that the experimenter must be able to precisely define, on the order of milliseconds, when the event of interest occurs. This timing places some limitations on the types of settings in which ERPs can be acquired. For example, if one wishes to examine the ERPs elicited by familiar compared to novel toys, this task would be extremely difficult to do in free-play situation as it would not be easy to document the precise moment of “onset” when the stimulus of interest came in to the infant’s view (a free-play situation might be more appropriate to investigate with EEG, because it does not have the same time-locking requirements as ERPs).
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
ERP Measures in Visual Development Research 107
Typically, this timing requirement is met by presenting images on a com- puter screen1 or via slide projection with a shutter. For example, one study investigating recognition of familiar faces and toys presented 6-month-olds with images of their own toy and a novel toy or the mother’s face and a stranger’s face on a computer monitor. The results showed that the Nc, a fronto-centrally distributed negative component, is larger to familiar than unfamiliar items (de Haan & Nelson,1999). In addition to documenting this basic recognition effect, the use of ERPs also provided information about the timing and spatial distribution of the brain activity related to recognition. The results showed that recognition occurred by approximately 600 millisec- onds (ms) after stimulus onset and that the recognition effect was bilateral for toys but over the right hemisphere for faces. These results suggest that the right hemisphere “bias” for face processing often observed in adults may be present by 6 months of age (see de Schonen, Gil de Diaz, & Mathivet,1986) for similar findings using a different procedure.
Usually in ERP studies, the events of interest are presented briefly because (a) when the stimulus duration is short, the time-locking is more precise because the perception is constrained to a narrow time window, and (b) the longer the stimulus presentation and recording interval, the increased like- lihood of artifacts such as eye movement artifact from scanning the display or blinking. If the event that one is interested in studying is prolonged, then it is likely more suitable for EEG than ERP study.
As mentioned above, events in ERP studies are also usually presented repeatedly. This allows the time-locked response elicited by the stimulus of interest to be extracted from the ongoing EEG. Since the ERP signals are small in amplitude relative to the EEG, it is necessary to average the recorded brain activity over repeated presentations in order to extract the ERP signal from the EEG. Thus, events that are not easily repeatable are not ideal for ERP study.
Consider two examples of studies using different approaches for investi- gating the neurophysiological basis of sustained attention. One study (Orkee- hova, Stroganova, & Posikera, 1999) investigated this question using a paradigm in which 8- to 11-month-old infants were tested in three phases: (a) sustained attention to an object, (b) anticipation of the person in a peak- a-boo game, and (c) attention to the re-appeared person in a peek-a-boo
1Note that when images are presented by computer screen there can also be issues related to timing (e.g., with respect to the time to draw the image, the uncertainty in vertical refresh, etc.)
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
108 Michelle de Haan
game. These experimental conditions are not ones that are necessarily brief or easily repeatable many times, and thus are more suitable for study by EEG. Using EEG recordings, the authors of this study found that frontal theta activity was related to anticipatory attention differently depending on age but in a manner suggestive of a maturational shift in the functioning of the anterior attention system. A second study (Richards,2003) used a different approach, and recorded heart rate as a measure of attention while 4.5-, 6-, and 7.5-month-old infants viewed briefly presented geometric patterns that were repeated or novel. As this studied employed briefly presented visual stimuli that could be repeated, it was suitable to recording ERPs. The author of this study found that the Nc was larger during periods of sustained attention, as defined by heart rate, than during periods of inattention.
The need for repeated presentations also has another implication that is particularly important for infant studies: to some extent it limits the number of different conditions that can be tested. In most infant visual ERP studies, at most 2–4 different stimulus conditions are tested (e.g., familiar vs. novel; happy vs. fearful expression), in order to maximize the likelihood that enough good (i.e., artifact-free) data can be collected in each condition before the infant stops cooperating. Although older children may be more tolerant, there still is a practical limit to how many different conditions can be tested in a session and still obtain sufficient useable data for all conditions. Awake, Attentive Infant or Child
Although some visual brain electrical responses can be recorded from sleep- ing infants (e.g., flash VEPs; Apkarian, Mirmiran, & Tijssen,1991; Shepherd, Saunders, & McCulloch,1999), visual ERPs typically require that the child be awake, with eyes open and attending to the stimulus of interest. These requirements contrast with auditory ERPs, where components such as the mismatch negativity (MMN) can be obtained even if the participants are not attending or are asleep (Matrynova, Kirjavainen, & Cheour,2003; Naatanen et al.,1993).
Though with older children, the required behavioral state can usually be obtained relatively easily with the instruction (“Sit as still as you can and keep your eyes on the screen”) and/or by giving a task to perform during the ERP (e.g., “Press this button when you see a familiar picture”), with infants and very young children it can be more challenging. Young children may find it difficult to appropriately carry out even a relatively simple button press task in the context of an ERP experiment because, for example, they often tend to look down at their hand as they make the response and thereby introduce eye/body movement artifact.
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
ERP Measures in Visual Development Research 109
Most visual ERP studies with infants and very young children have there- fore used “passive tasks,” where the only requirement is to look at the stimuli (although some studies have used different approaches due to their ques- tion of interest, such as to study EEG events related in time to the onset of responses, e.g., saccades, rather than related in time to stimulus onset, e.g., Csibra, Tucker, & Johnson,1998; Richards,2000). Though this task is something most infants and young children are able to do, it typically is not the case that they will continuously look at the screen for the whole series of stimuli. Typically, infants may look for a while, but then look away from the stimulus display, for example, to the ceiling or back to the caregiver on whose lap they may be sitting. Ideally, the testing area is made as uninterest- ing as possible to encourage sustained looking at the stimulus display (e.g., by having monochrome screens or curtains around the room, windows shut with shade, etc.). But even with these precautions, infants invariably at some point look away from the display before all the stimuli are presented. One factor that may contribute to this tendency is that stimuli are often presented for relatively brief exposures (e.g., 500 ms), followed by blank screens during inter-trial intervals that may last longer than the stimulus itself (e.g., 1000– 2000 ms). From the experimenter’s point of view, this opportunity allows the brief presentation to occur followed by a period where brain activity elicited by that stimulus continues to be recorded even in its absence. This period helps to ensure that brain activity returns to baseline (i.e., stimulus- elicited processing is “finished” before the next stimulus is presented). The most common solution to the problem of looking away is to present visual and/or auditory stimuli to re-attract the participant’s attention to the screen. These might be colorful, small moving patterns on the screen where the stimulus is displayed and/or brief sounds from that location, or calling the infants name, and so on. Other experimenters have presented the experimen- tal stimuli during brief streams interleaved in a more interesting children’s video (Richards,2003). In both cases, the experimenter then deletes/ignores the EEG recorded during the re-orienting stimuli or video and retains only the data from the trials in which the participant was looking appropriately at the experimental stimuli. Although these approaches are usually effective at a practical level for retaining the infant’s or child’s interest in the ongo- ing experiment, it is important to remember that the distinction between the “real” experimental stimuli and re-orienting stimuli made by the experi- menter is not necessarily made by the participant. The infant or child simply observes a series of visual images, and the occasional occurrence of interven- ing stimuli that re-attract their attention are a part of this series and could influence their response to the “real” experimental stimuli. For example, in
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
110 Michelle de Haan
an oddball paradigm where one stimulus is presented frequently and the other (the oddball) presented infrequently, the occasional occurrence of the re-orienting stimuli might influence the perceived novelty or frequency of the oddball stimuli. A different approach that to some extent circumvents the disadvantage of presenting re-orienting patterns is to present stimuli without any inter-trial interval, so a stimulus is always on the screen (e.g., Gliga & Dehaene-Lambertz,2005). However, this approach may affect the latency of the observed responses and could have other disadvantages, for example, if the presentations are brief, brain activity may not have time to return to baseline, or if presentations are longer than is typical, there may be increased likelihood of eye movement artifacts due to scanning of the visual display.
Even for infants and young children who attend well to the visual display, it is important to keep in mind that the “passive” task is in essence still a task that requires sustained attention. In order to provide useable data, the participant must be able to sustain attention to the visual display for a period of time long enough to see sufficient trials for an average ERP waveform to be extracted. It is possible that the increased ease with which participants can maintain their attention in a passive visual ERP task might contribute to the reported disappearance of the Nc component by adolescence (Courchesne, 1978), as this component is believed to reflect processes involved in orienting and/or sustaining attention.
Investigations that have related the Nc to behavioral measures of attention provide some support that it may be linked to the ability to sustain attention. For example, one study reported that the amplitude of the Nc at right frontal electrodes in toddlers correlates with their ability in a separate task to sustain visual attention on an intermittent audio-visual display (Early Childhood Vigilance Task; Goldman, Shapiro, & Nelson,2004; see also Richards,2003). It is possible that this study found a relation between looking time measures and the Nc because the looking time task used presents very similar attentional requirements to the typical ERP test session. Investigators who have instead attempted to correlate the size of the Nc with the length of infants’ looking to novel or salient stimuli in separate behavioral tests using the visual paired comparison test have failed to find a link between the two. For example: (a) although at 6 months of age the Nc is larger for the mother’s face than an unfamiliar face, infants do not look longer at the mother’s face (de Haan & Nelson,1997); (b) although infants show a larger Nc and longer looking times to fearful compared to happy faces (de Haan & Nelson,1999; de Haan et al.,2004), there is no correlation between the amplitude of the Nc and the length of looking (de Haan et al,2004); (c) in visual paired comparison
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
ERP Measures in Visual Development Research 111
tests of novel and standard stimuli given after two-stimuli or three-stimuli oddball tests, there is no consistent evidence that infants look longer at the novel stimuli (Karrer & Monti,1995; Nelson & Collins,1991,1992) and no correlation between the Nc and the duration of looking to novelty (Karrer & Monti, 1995); and (d) although both Nc amplitude and the duration of infants’ fixations decrease over a study session, the decrease in fixation appears to happen more rapidly (Nikkel & Karrer,1994). Together, these findings suggest that there is not a close relation between Nc amplitude and length of looking to novel/salient stimuli (Nikkel & Karrer,1994).
Adding Recording Electrodes and Requirement to Sit Still
Not only must infants and children be awake and attending to the brief, repeated events, they must do so while wearing the electrodes and sitting relatively still. Good descriptions of different systems of electrodes and their application can be found in de Boer and colleagues (2007) and Johnson and colleagues (2001). Methods in which electrodes are placed within electrode caps or nets have the advantages that inter-electrode distances are not variable and that they allow a larger number of electrodes to be placed accurately in a relatively short amount of time. The time savings is a considerable advantage, in assuring that infants and children are not so bored or tired from the electrode application procedure that they are no longer willing to cooperate with the experiment once the electrodes are placed. This issue is important as the main contributors to participant attrition tend to be failure to view sufficient trials and insufficient data due to movement artifact, both things that are likely to occur if the participant is already bored or tired from the electrode preparation period.
For infants and young children, there may be some apprehension about wearing an electrode cap or net or some urge to grab it or the electrode wires. For children, a warm-up period can help where the child has a chance to look at the net and perhaps see the mother or a favorite toy wearing it. Having an entertaining video to watch can also help the child “forget” about the net once it is in place. For infants, out of sight can be out of mind, so it can be helpful for the experimenter placing the cap or net and preparing it to work from behind while another experimenter entertains the infants with toys or blowing bubbles. Distraction with this type of entertainment also often helps calm the infant if s/he begins to cry when the cap or net is placed on the head. For infants with a tendency to grab the cap/net or wires, and if keeping these out of sight as much as possible does not help, placing infant mitts on the hands, keeping the hands occupied with toys, biscuits, or simply instructing the individual holding the child to keep control of the hands can help.
P1: JZP
9780521821063c04 CUFX163/Schmidt 978 0 521 82106 3 August 9, 2007 14:9
112 Michelle de Haan
Infants and children ideally should be seated in a comfortable chair or infant seat rather than in an adult’s lap. The reason is that the adult may influence how the child reacts, and also to prevent additional movement artifact (e.g., due to bouncing them on the knee or turning to look at the adult). However, for some young infants who have poor head control, this method may not be the most appropriate procedure. If infants rest their heads on the back of their seat, this position may cause artifacts in the electrodes over the back of the head (occipital area), a region that may be of particular importance in visual studies. In this case, some method for supporting the head might be needed, either with a neck support in the infant seat or with an adult holding the infant and supporting the neck.
Equal Measures of Task Performance
In cases where a task is used and different groups, such as different ages or typical versus atypical children are assessed, one consideration is whether group differences are expected in behavioral performance on the task. This issue is an important consideration for at least two reasons. First, in ERP studies with tasks, it is typically the case that only data from those trials with correct responses are included in the averages. Therefore, the task used must be one that all groups can perform at a level well enough for there to be enough correct response trials with good ERP data to create the ERP average for each subject. This issue is important to consider because often an impairment in a particular domain may be of particular interest in a study of children with developmental disorders. A second reason to consider behavioral performance is that, even if all groups are able to complete the task to some extent, the results will be more difficult to interpret if the groups differ in performance. The reason is that it is not necessarily clear in these situations to what extent any differences in the ERPs reflect developmental differences in the neural correlates of the processes being studied and to what extent they reflect more generic influences of task difficulty. This difficulty in interpretation is not solved by averaging only correct responses, as the process by which the correct answer was generated may differ for groups who find the task more versus less difficult regardless of age/stage of development.
One useful aspect of including ERPs in studies of developmental disorders is that they can reveal atypicalities even for domains in which a patient pop-