THE COMPONENTS OF AGING. David Friedman. Cognitive Electrophysiology Laboratory, New York State Psychiatric Institute

Transcription

1 Friedman chapter 1 THE COMPONENTS OF AGING David Friedman Cognitive Electrophysiology Laboratory, New York State Psychiatric Institute Columbia University Medical Center New York, NY In press, in: Luck & Kappenman (2008). Oxford Handbook of Event-Related Potential Components. New York, Oxford University Press.

2 Friedman chapter 2 THE COMPONENTS OF AGING David Friedman Normal aging carries certain risks, not the least of which is change in those aspects of cognition, such as top-down, executive-control and mnemonic processes, that are critical for navigating everyday life and, therefore, successful aging. However, not all aspects of cognition are impaired as we age. The age-related pattern of spared and relatively compromised cognition in aging populations appears to be reflected in a pattern of spared and relatively compromised ERP components. Unfortunately, the amount of research work devoted to higher-order cognitive processes as opposed to putatively sensory processes is disproportionately weighted towards the endogenous components (Friedman, Kazmerski, & Fabiani, 1997; Polich, 1996). Endogenous components are relatively insensitive to the stimulus s physical properties, but their amplitudes and latencies are intimately linked to the nature of the task in which those stimuli are embedded (Sutton, Tueting, Zubin, & John, 1967). As a result, the current review of age-related change in ERP components will, of necessity, be slanted towards these components. Within the endogenous domain, the greatest amount of research attention has been paid to the P300 family of components (see chapter 5, this volume), primarily using the ubiquitous oddball paradigm (Donchin & Coles, 1988). Despite what could be considered an overemphasis on the oddball task, P300s have also been recorded in an extremely wide variety of other task paradigms including working memory (McEvoy, Pellouchoud, Smith, & Gevins, 2001) episodic memory (Friedman, 2007; Friedman, Nessler, & Johnson, 2007), task switching (Friedman, Nessler, Johnson, Ritter, & Bersick, 2007; Themanson, Hillman, & Curtin, 2006), repetition priming (Hamberger & Friedman, 1992; Rugg, Mark, Gilchrist, & Roberts, 1997), semantic

3 Friedman chapter 3 priming (Kutas & Iragui, 1998), and selective attention (Gaeta, Friedman, & Ritter, 2003; Karayanidis, Andrews, Ward, & Michie, 1995). By virtue of the fact that a large number of studies on aging have involved the auditory oddball paradigm (Iragui, Kutas, Mitchiner, & Hillyard, 1993) and to a lesser extent its visual counterpart (Beck, Swanson, & Dustman, 1980), there are a number of investigations of the aging brain s response to the sensory information inherent in the standard, target and novel stimuli typically presented during this task (manifested in the middle- and longer-latency components between 50 and 100 ms post-stimulus), such as P1 or P50 and N1 or N100 (Reinvang, Espeseth, & Gjerstad, 2005). These relatively early components are thought to reflect the arrival of sensory information at midbrain and higher, modality-specific, cortical processing centers. Extremely little is known about the impact of aging on the very early-latency brainstem responses (ABRs) occurring within the first 10 milliseconds following stimulus onset (see Chapter 4, this volume). These ABRs reflect the arrival of auditory information in the nuclei of the brain stem. Several ERP modulations are defined on the basis of a difference between electrical activity in one condition compared to that in another, including the mismatch negativity or MMN (Naatanen, Paavilainen, Rinne, & Alho, 2007), the negative difference or Nd waveform in studies of selective attention (Hansen & Hillyard, 1980), the N400 in linguistic and semantic memory assessments (Chapters 12 and 14, this volume), the lateralized readiness potential or LRP (Coles, Gratton, & Donchin, 1988; Chapter 9, this volume), and the error-related (Falkenstein, Hoormann, & Hohnsbein, 2001) and medial frontal (Gehring & Willoughby., 2002) negativities in studies of executive control (Chapter 10, this volume). With the exception of the Nd (Gaeta et al., 2003; Karayanidis et al., 1995; Woods, 1992) and LRP (e.g., Falkenstein,

4 Friedman chapter 4 Yordanova, & Kolev, 2006; Zeef & Kok, 1993), for which only a small number of studies exist, age-related changes in these remaining components will be reviewed below. I turn first to a review of aging effects on the components between 50 and 200 ms poststimulus recorded primarily in passive and active auditory-oddball paradigms. Although similar studies have been conducted with comparable findings in the visual modality (see De Sanctis et al., 2007 for a review), a much greater number have used auditory stimuli. Four components have been assessed, P50, N1 (N100), P2 (P200) and N2 (N200). The MMN is part of the N2 complex (Naatanen et al., 2007), which includes the N2b. Relatively Early-Latency Components Because some of these ERP activities appear to be affected by attention, it is safest to measure them when attention is not an issue, as in the passive oddball paradigm, in which participants are instructed to ignore the stimuli. One can also measure the frequent standards within the active oddball paradigm, as these presumably do not recruit the same degree of attentional processing as targets (which typically require a reaction time, RT, response) or novels; the latter, if sufficiently surprising, typically capture attention involuntarily (Friedman, Cycowicz, & Gaeta, 2001). To illustrate the basic phenomena that will be reviewed in this section, Figure 16.1 (modified from Gaeta, Friedman, Ritter, & Cheng, 2001a) depicts the grand-averaged young- (18-30 years old) and older-adult (65-85 years old) ERPs elicited by first-position standards occurring in a train of 8 identical standards each separated by a 500-ms inter-stimulus-interval (ISI). Participants watched a silent movie while ignoring the trains of stimuli. There are prominent P50 (also known as P1), N1 and P2 components present in both the young and older adult grand mean data. The P50 appears larger in the older adult waveforms, whereas the N1

5 Friedman chapter 5 component is of similar magnitude in the two groups. The P2 is larger in the young but only at the 8-sec inter-trial-interval (ITI; i.e., the interval between the last tone in one train and the first tone in the next train). Furthermore, there is a marked effect of ITI on the N1 and P2 components, both of which are larger after an 8-sec interval. By contrast, the P50 does not appear to be influenced by the amount of time between stimulus trains. The greater reduction in the N1 and P2 components following a 1-s compared to an 8-s ITI is most likely due to the refractoriness or fatigue of the neural generators. This is explained a few paragraphs below in greater detail. The effects of ITI appear to be similar in both age groups because both show larger N1 and P2 components at the longer ITI. Some of these observations are mirrored in the age-related studies reviewed below. In reviewing the evidence, there appears to be a consensus that the P50 is larger in older than young adults when it has been measured directly and reported in the text of the published reports (Amenedo & Diaz, 1998b; Bennett, Golob, & Starr, 2004; Bertoli, Smurzynski, & Probst, 2005; Chao & Knight, 1997; Czigler, Csibra, & Csontos, 1992; Fabiani, Low, Wee, Sable, & Gratton, 2006; Smith, Michalewski, Brent, & Thompson, 1980; Snyder & Alain, 2005) or observed by visual inspection of the published waveforms by the current author (Bertoli & Probst, 2005; Gaeta, Friedman, Ritter, & Cheng, 1998; Golob, Irimajiri, & Starr, 2007; but see Bertoli, Smurzynski, & Probst, 2002). Interestingly, there does not appear to be any evidence that P50 is of longer latency in older relative to young adults The enhanced P50 amplitudes observed in older adults might reflect a deficit in inhibitory function. For example, Knight and colleagues (Knight, Scabini, & Woods, 1989) recorded P50 components to auditory clicks in patients with dorsolateral prefrontal lesions. P50 amplitudes were reliably larger in the patients compared to age-matched controls. A highly similar result

6 Friedman chapter 6 was reported by Alho, Woods, Algazi, Knight, & Naatanen (1994). Knight and colleagues (Alho et al., 1994; Knight, Scabini, & Woods, 1989) interpreted this to mean that the prefrontal patients had lost inhibitory control over thalamically-mediated gating of inputs to sensory cortex. Thus, extrapolating this result to normally-aging older adults might implicate a deficit in prefrontal inhibitory control as a cause of age-related enhancement of P50 amplitudes (see also Chao & Knight, 1997; see also the section on N1 amplitude below). Similarly, based on P50 magnitude findings, there is some support for the possibility that probable Alzheimer s disease patients (PAD) may be characterized by deficits in inhibitory control. Further, there is limited evidence that some individuals within the aging population who show poor inhibitory control may eventually develop PAD. In healthy persons, the second of a two-click pair separated by a 500-ms interval elicits smaller P50 than the first click (i.e., the P50 to the second click is suppressed), which has been labeled sensory gating (Freedman et al., 1987). Reduced suppression of P50 relative to normal controls was originally reported in schizophrenia patients (i.e., the response to the second click was less suppressed in the patients; Adler, Waldo, & Freedman, 1985; Siegel, Waldo, Mizner, Adler, & Freedman, 1984; Chapter 17, this volume). The reduced suppression was interpreted as indexing a deficit in inhibition or sensory gating (i.e., a failure to filter irrelevant input) and a proclivity to sensory overload, consistent with some schizophrenia symptomatology (Braff & Geyer, 1990; Freedman et al., 1987). Based on these findings, there is some very limited evidence that altered P50 suppression might be a marker of probable Alzheimer s disease or PAD (Ally, Jones, Cole, & Budson, 2006; Cancelli et al., 2006; Jessen et al., 2001), because, like schizophrenia patients, the suppression has been reported to be smaller in PAD patients than age-matched controls. Applying the interpretation proffered for schizophrenia patients would suggest that PAD patients may also be

7 Friedman chapter 7 characterized by inhibitory or sensory gating difficulties. However, the results of the studies of P50 suppression in PAD are difficult to evaluate because in two of them the waveforms from PAD patients and controls were not depicted, precluding a determination of the quality of the data and the measurement technique (Donchin et al., 1977). Other studies have indicated that the P50 component itself is larger in patients diagnosed with mild cognitive impairment (MCI), a putative precursor stage to PAD, compared to age-matched, healthy controls. Hence, poor inhibitory control (i.e., enhanced P50 magnitudes relative to age-matched controls) may be a predictor of conversion to AD (Golob et al., 2007; Golob, Miranda, Johnson, & Starr, 2001; Irimajiri, Golob, & Starr, 2005). In similar fashion to the P50, N1 has also been reported to be larger in older than young adults (Alain & Woods, 1999; Amenedo & Diaz, 1999; Anderer, Semlitsch, & Saletu, 1996; Chao & Knight, 1997; Gaeta, Friedman, Ritter, & Hunt, 2002; Karayanidis et al., 1995; Kisley, Davalos, Engleman, Guinther, & Davis, 2005; Snyder & Alain, 2005), although there is little evidence for age-related prolongation of N1 latency (Anderer et al., 1996; Goodin et al., 1978; Iragui et al., 1993). On the other hand, there are also data suggesting that young, relative to older, adults show larger amplitudes (Bennett et al., 2004; Bertoli & Probst, 2005; Cooper, Todd, McGill, & Michie, 2006; Ford et al., 1995; Golob et al., 2001). Confusing the picture further, other studies have reported equivalent amplitudes between young and older adults (Bertoli et al., 2002; Czigler et al., 1992; Ford et al., 1995; Friedman, Simpson, & Hamberger, 1993; Gaeta et al., 1998; Gaeta, Friedman, Ritter, & Cheng, 2001b; Gaeta et al., 2003; Iragui et al., 1993; Pekkonen et al., 1996; Picton, Stuss, Champagne, & Nelson, 1984; Woods, 1992). One possibility for these disparate results is the quite different paradigms under which the N1 has been measured, including passive and active oddballs, startle-noise oddball, delayed match-to-

8 Friedman chapter 8 sample, selective- and cued-attention tasks. As noted, N1 is modulated by attention, e.g., it is larger to targets (deviants) than standards and attended compared to unattended standards during selective attention tasks (Hillyard, Hink, Schwent, & Picton, 1973; Naatanen & Picton, 1987; Chapter 11, this volume). To the extent that these attentional effects differ for young and older adults, some of these differences could be accounted for on this basis. However, as shown by Amenedo & Diaz (1999), older adults demonstrated larger N1 magnitudes whether the standards were attended or unattended (see Selective Attention section below). Hence, although the data are clearly limited, age-related differences in attentional function may not explain these discrepancies (see also Gaeta et al., 2003). An alternative explanation might lie in age-related differences in the recovery cycle or refractory period of N1. That is, if the time between successive auditory events is short (e.g., 500 ms), the N1 to the second in a series of tones will be markedly smaller than that to the first. As the interval between successive stimuli becomes longer, the N1 generators recover some of their amplitude, producing equivalent magnitude to the first tone at an interval between 6 and 10 sec (Davis, Mast, Yoshie, & Zerlin, 1966). There is some recent evidence that older adults do show different N1 recovery functions than their young adult counterparts (Fabiani et al., 2006), although the underlying mechanism may not be a simple age-related change in the refractory period of N1 generators (Sable, Low, Maclin, Fabiani, & Gratton, 2004). Fabiani et al. (2006; see also Gaeta et al., 2001a) presented trains of 5 auditory events (400-ms ISIs), with each train separated by either 1- or 5-sec ITIs. Fabiani et al. (2006) reported that the recovery rate for N1 was similar for young and old (i.e., N1 to the first standard in the train was larger for the 5- compared to the 1-sec ITI in both older and young adults; see also Figure 16.1). However, relative to the first tone of a train, older adults showed reliably less N1 suppression to the

9 Friedman chapter 9 remaining tones compared to the young, suggesting the possibility, as noted earlier, of agerelated inefficiency in inhibitory control or sensory gating (Chao & Knight, 1997; Lynn Hasher, Stoltzfus, Zacks, & Rypma, 1991). Because N1 amplitude to the first tone in the 1-sec ITI was equivalent in both age groups and was as large in older as young adults under the 5-sec ITI condition, the authors concluded that sensory memory was intact in older adults. On the other hand, evidence based on the mismatch negativity (MMN), suggests that sensory memory decays faster in older adults (Pekkonen, 2000), but this interpretation is open to question. I will consider this issue further in the section on the mismatch negativity (MMN). There is a dearth of studies of the N1 in PAD and MCI. Intriguingly, however, in a study in which MCI patients and controls listened passively to tones at two ISIs (2/ and 1.5/sec), the two groups had equivalent N1s at the longer ISI, whereas MCI subjects showed greater N1 at the shorter ISI (Irimajiri et al., 2005). On the other hand, there is, to my knowledge, no evidence that this is also the case in PAD. In fact, one study (Pekkonen, Jousmaki, Kononen, Reinikainen, & Partanen, 1994) found equivalent N1 magnitudes in PAD patients and controls at both 1- and 3- sec ISIs. In another, Yamaguchi, Tsuchiya, Yamagata, Toyoda, & Kobayashi (2000) reported reliably smaller N1 magnitudes in PAD relative to age-matched controls. Hence, these results must be interpreted cautiously. Nonetheless, they suggest that this phenomenon should be explored further in PAD, MCI and controls, in the eventuality that reduced suppression may help identify those older persons at greatest risk for the development of PAD. One could use N1 amplitude (or, for that matter, any component s amplitude) as a biomarker for a given disease process without knowledge of the underlying processes. However, it would arguably be more useful diagnostically if those processes were known. N1, like the earlier P50, is an obligatory component of the auditory ERP because it is elicited

10 Friedman chapter 10 whether the stimuli are attended or ignored, although its magnitude can be affected by attention. According to Naatanen and Picton's (1987) exhaustive review, there are potentially three sets of processes contributing to the true N1 component recorded at the scalp (i.e., those processes not reflecting aspects of the mismatch negativity and the Nd). Two of these are thought to reflect the activity of cerebral generators in and around primary and secondary auditory cortex. Hypotheses advanced by Naatanen and Picton (1987) suggest that one or both of these cortical generators could reflect the representation of sensory information and/or the formation of a sensory memory within the auditory cortex (see also Naatanen, Jacobsen, & Winkler, 2005; Chapter 4, this volume). Based on the Fabiani et al. (2006) data detailed above, it does not appear as if aging adversely affects the representation or formation of auditory sensory memories, but these mechanisms might be disrupted in those aging individuals with PAD or those with a diagnosis of MCI who progress to PAD. However, there are simply too few data to determine whether or not this is the case. Relative to P50 and N1, age-related change in the P2 component has not been as well studied, and only a handful of investigators have directly measured it. Czigler and colleagues (Czigler et al., 1992) used ISIs of 800, 2400 and 7200 ms in a passive oddball task. While P2 magnitude increased with ISI (as would be expected based on a refractory-period explanation, just as for the N1), this increment was much less dramatic for older adults. A similar finding was also reported in the Fabiani et al. (2006) study described earlier and is reminiscent of their N1 finding. Amenedo & Diaz (1999) reported larger P2 amplitudes in older compared to younger adults, again regardless of whether the standards were attended or unattended. Similar agerelated enhancements have been described by Anderer et al. (1996), although one study found

11 Friedman chapter 11 larger P2 magnitude in young compared to older adults (Bertoli et al., 2002) and another (Iragui et al., 1993) reported no difference between young and older adults. Early-latency components in the visual modality analogous to P50 and N1 have also been reported to show age-related change. However, this literature is as mixed in its results as the data on age-related change in the auditory modality described above. For the sake of brevity, I refer the reader to a very recent paper in which these studies are reviewed (De Sanctis et al., 2007). In this paper, De Sanctis and colleagues observed that the amplitude of the visual N1 (which was significantly larger in older adults) was much more variable in older than young adults. Whereas young adults showed a fairly tight distribution of amplitude values, older adults showed a bimodal distribution 10 older adults with lower-amplitude N1s showed values similar to the young, whereas the remaining 9 produced the highest-magnitude N1s. This potentially important observation might explain some of the variability among studies in both the auditory and visual modalities. To summarize, relative to N1 and P2, the age-related enhancement of auditory P50 (P1) seems to be better established in the literature. The lack of suppression of N1 observed by Fabiani et al. (2006) is intriguing and deserves follow up. Both of these phenomena may be indicative of an age-related inhibitory deficit and could provide additional evidence for the frontal-lobe and inhibitory-control deficit hypotheses of cognitive aging (West, 1996). In addition, there is some, albeit limited, evidence that P50 enhancement may be a marker of risk status in some older adults who progress to PAD. Hence, this phenomenon may be worthy of pursuit in large samples of PAD, MCI and age-matched controls.

12 Friedman chapter 12 The Mismatch Negativity. A reasonably large, age-related literature has accumulated on the automatic deviance detection system, as reflected by the MMN (Pekkonen, 2000). Figure 16.2 illustrates data recorded during the ignore paradigm described earlier (Gaeta et al., 2001a). In the figure, the ERPs to first-position standards and deviants from the 1-sec ITI condition are depicted. In addition, grand averaged preliminary data from 5 PAD patients are shown. Two basic age-related phenomena are reflected in these data the MMN is somewhat smaller and of longer latency in healthy older and PAD participants compared to young adults, and PAD patients appear to produce MMN magnitudes and latencies similar to those of healthy controls. Despite the overall age-related difference in MMN amplitude, its scalp distribution appears similar in all three groups. The current source density (CSD) maps are consistent with bilateral generators in and around auditory cortex, which is thought to be a primary contributor to the MMN recorded at the scalp (Giard, Perrin, Pernier, & Bouchet, 1990; Naatanen, 1992; Naatanen et al., 2007). As described by Naatanen in this volume (chapter 8) and elsewhere (Naatanen et al., 2007), the relatively-automatic, preattentive, MMN-based deviance detection system is recruited under a wide variety of situations, including simple changes in pitch, intensity, duration, phonetic characteristics and spatial location, as well as by more abstract changes in the acoustic background, such as whether two tones are rising or falling in pitch (Paavilainen, Jaramillo, & Naatanen, 1998). Moreover, on the basis of this latter finding, the MMN deviance detection system cannot reflect a simple sensory (echoic) memory of the acoustic past but rather appears to reflect a memory of the regularities or invariances in the acoustic stream (whether it is for pitch or the abstract nature of the regularity). Because of these properties, it has been natural

13 Friedman chapter 13 for investigators to ask whether aging affects any aspects of the MMN deviance detection system. At the early stages of research into the MMN s characteristics, it was thought that it reflected a short-lived sensory memory (Sams, Hari, Rif, & Knuutila, 1993) and, on this basis, several of the early studies of aging compared short (e.g., 200 ms) and long (e.g., 4000 ms) stimulus-onset-asynchronies (SOA) to determine whether the memory on which the MMN was based had a shorter duration in older than young adults. However, as noted above, the MMN is not thought to reflect a simple sensory memory. There are also methodological limitations in assessing effects of SOA even in young adults (see Ritter, Sussman, Molholm, & Foxe, 2002 for a discussion). Therefore, these investigations are equivocal with respect to age-related changes in the duration of the memory on which the MMN is based at short and long SOAs using the typical oddball paradigm (Gaeta et al., 2001a; Ritter, Gomes, Cowan, Sussman, & Vaughan, 1998; Winkler et al., 2002). Hence, the review below only considers the results of experiments in which auditory stimuli were ignored and a short SOA was used (between 0.5 and 2 sec) even if a long SOA was also included. When considered in this fashion, although there are exceptions, there appears to be a consensus that the MMN is reduced in older adults. This suggests the possibility that sensitivity to regularities in the acoustic environment decreases as we age. In one of the first of these investigations, Czigler et al. (1992) used infrequent changes in pitch. They found that older adults showed smaller-amplitude MMNs than young adults. Similarly, Cooper et al. (2006) reported smaller MMNs in older compared to young adults whether deviance was defined by pitch or duration. Furthermore, two reports suggested that when the MMN was recorded to duration deviants in an unattended channel (i.e., ear of input) during selective attention paradigms, it was smaller in older compared to younger adults

14 Friedman chapter 14 (Karayanidis et al., 1995; Woods, 1992; see also Gaeta et al., 2001b for frequency deviants; but see Pekkonen et al., 1996) even at very short ISIs between 200 and 400 ms. Gaeta et al. (1998) also reported smaller MMNs in older relative to young adults whether elicited by small- (50 Hz difference from standard) or large- (300 Hz difference) frequency deviants or novel, environmental sounds. In fact, older adults did not show a reliable MMN to the small-frequency deviant. Alain & Woods (1999) assessed MMN magnitudes to both frequency changes (small- [122 Hz] and large- [414 Hz]) and pattern deviants (tones of two different pitches alternated and interrupted infrequently by a repeat). They observed reduced MMNs to small- and largefrequency as well as pattern deviants. In another study from this same group, Alain, McDonald, Ostroff, & Schneider (2004) assessed the sensitivity of the older adult s deviance detection system by varying systematically the gap between tone pips constituting a deviant event. Tones with gaps occurred infrequently and continuous tones served as standards. MMN amplitudes were smaller in older than young adults. Importantly, when MMNs were computed to nearthreshold gap deviants, thereby matching the performance of young and older subjects, young but not older adults showed reliable MMNs. In a very similar paradigm, Bertoli et al. (2002) varied the gap duration of deviant stimuli between 6 and 24 ms. MMNs were smaller in older compared to young participants. Moreover, similar to the results of Alain et al. (2004), it took a longer duration gap (15 ms) to produce a reliable MMN in older compared to young (9 ms) adults. The results of these investigations suggest that the preattentive deviance detection system of older adults is less sensitive than that of young adults. Employing rule-based auditory features to create invariances in the acoustic environment, the results of a study by Gaeta et al. (2002) support this hypothesis. Stimuli were either a frequent ascending tone pair or an infrequent

15 Friedman chapter 15 descending tone pair. Tone-pairs were presented under three conditions: 1) physical feature monaural (1 tone pair), 2) abstract feature monaural (10 tone pairs of different pitches), and 3) abstract feature binaural (10 tone pairs of different pitches; the first presented to the left ear and the second to the right). Relative to young adults, older adults showed smaller-magnitude MMNs, which were elicited under all three conditions for the young, but only in the monaural conditions for older adults. Thus, rule-based neural representations were created by both age groups under monaural conditions, but only by the young in the binaural condition. Reliable MMNs in the rule-based conditions were present despite the fact that behavioral discrimination (after the MMN recordings) fell to near chance levels for both age groups, suggesting an agerelated decline in the efficacy of integrating multiple sources into a single auditory stream. By contrast with the studies reviewed above, there are some reports of age-equivalent MMN amplitudes. For example, with pitch deviants and a 1-sec ISI, Pekkonen, Jousmaki, Partanen, & Karhu (1993) reported that MMN magnitude was similar in young and older adults, a finding also reported by Gunter, Jackson, & Mulder (1996) with a 1-sec ISI. In follow up of their earlier study, Pekkonen et al. (1996) found age-equivalent MMN magnitudes to pitch deviants at.5- and 1.5-sec ISIs. Using frequency deviants and a similar selective-attention procedure to those mentioned earlier with 600-ms ISIs, Amenedo & Diaz (1998a) did not observe age-related differences in MMN magnitudes. However, these latter results are equivocal because there appeared to have been no effect of selective attention on the ERP waveforms, questioning the sensitivity of the paradigm. Using the design described above in the section on early-latency components, Gaeta et al. (2001a; see Figure 16.2), showed that at a 1-sec ITI, the vast majority of young and older subjects showed a robust MMN. However, at the 8-sec ITI, only 6 young and 5 older adults

16 Friedman chapter 16 showed robust MMNs. Although this might suggest that the memory on which the MMN was based had decayed at the 8-sec ITI for both age groups, Gaeta et al. (2001a) argued that it was the nature of the perceptual grouping of the trains that was modulating MMN magnitude. That is, when the SOA is constant between the standards within a train and the interval between standards and deviants is short (1-sec ITI), all tones are treated as belonging to the same perceptual group. Therefore, a deviant is detected as a departure from invariance, resulting in a robust MMN. When the SOA between standards within a train is short and the interval between standards and deviants long (8-sec ITI), the perceptual group that comprises the invariance is the train of standards and the deviant lies outside the frame of temporal relevance for some subjects. Hence, presentation of a deviant is not perceived by the MMN system as a divergence from regularity and an MMN is not elicited for these subjects. A similar interpretation is possible to account for the data of Fabiani et al. (2006), who also reported age-invariant MMNs using a highly-similar design with 1- and 5-sec ITIs. Hence, although these two studies are ostensibly at odds with much of the data reviewed above, it is likely that the age-equivalent MMNs in these studies is peculiar to the types of stimuli used (trains rather than single events) and reflect the maintenance with age of perceptual grouping of acoustic stimuli. With respect to abnormal aging, less than a handful of investigations exist. In the earliest of these, Pekkonen et al. (1994) reported that the MMN was of similar amplitude in PAD and controls at a 1-sec ISI. This was confirmed in a subsequent study by the same group using the magnetoencephalographic analog of the electrical MMN (Pekkonen et al., 2001). Similarly, Gaeta, Friedman, Ritter, & Cheng (1999) observed that MMN magnitudes were similar in PAD and age-matched controls.

17 Friedman chapter 17 In summary, it seems fairly clear that the system upon which the MMN relies is relatively less sensitive in older compared to young adults and, based on very limited evidence, is similar in PAD patients and controls, at least in the fairly simple paradigms used to date. What is less clear is whether there is an age- and/or PAD-related reduction in the duration of the memory upon which the MMN is based. This will certainly require further work. The Novelty P3 (P3a) Although the MMN reflects the detection of a change in the invariance of the acoustic environment, it does not reflect the capture of attention by that change. The involuntary capture of attention is reflected by the novelty P3 or P3a, which is elicited if the change in the background is sufficiently deviant (Friedman et al., 2001; see also Chapter 7, this volume). The novelty P3, therefore, reflects an aspect of the orienting response, a fundamental biological mechanism necessary for survival (Sokolov, 1990). An example of the dissociation between the MMN and the novelty P3 is depicted in Figure 16.3, which illustrates the averaged ERPs elicited by standards and small (50 Hz) and large (i.e., novel, environmental sound) deviants recorded while young and older adults watched a silent movie and ignored the background auditory events (Gaeta et al., 1998). While a MMN is elicited by both small and large deviants in young as well as older adults (although the MMN was not reliable in the latter), only the environmental-sound deviants elicit the novelty P3. Note also that the novelty P3 is reduced in older compared to young adults, an age-related phenomenon that has been replicated most often with auditory stimuli, but has also been observed in the visual and somatosensory modalities (Czigler, Pato, Poszet, & Balazs, 2006; Fabiani & Friedman, 1995; Friedman, Simpson et al., 1993; Friedman & Simpson, 1994;

18 Friedman chapter 18 Friedman, Kazmerski, & Cycowicz, 1998; Gaeta et al., 1998; Gaeta et al., 2001b; Knight, 1987; Weisz & Czigler, 2006; Yamaguchi & Knight, 1991; but see Daffner et al., 2006b). Although the majority of studies have employed active oddball designs, in which the participant is asked to respond to targets via RT and withhold responding to standards (and novels), it is perhaps more ecologically valid to assess the ERP concomitants of the orienting response while participants ignore the background events. Using this method, one can obtain a truer assessment of how and to what extent deviant events involuntarily capture attention, because the participant is engaged in the primary task of reading, watching a silent movie or performing a visual discrimination (Alain & Woods, 1999; Gaeta et al., 1998; Friedman et al., 1998; Gaeta et al., 1998; Yago, Escera, Alho, Giard, & Serra-Grabulosa, 2003; see Figure 16.3). Like older adults of the Gaeta et al. (1998) study, PAD patients only showed a novelty P3 in response to the environmental-sound deviants which did not differ in amplitude from that of age-matched controls (Gaeta et al., 1999). A similar finding was reported by Yamaguchi et al. (2000), also in the auditory modality, although with an active paradigm. The results of these latter two studies contrast with those from an investigation by Daffner et al. (2001) using visual stimuli. They reported that the novelty P3a was reliably smaller in PAD patients than agematched controls. The methods used by Daffner et al. (2001) are quite different than those typically employed to assess either age- or dementia-related changes in the processing reflected by the novelty P3. For example, these investigators collect looking times (i.e., how long a participant spends viewing a given standard, target or novel visual event), as well as RTs to predesignated targets. Participants are allowed to view the stimuli for as long as they deem necessary. Although an intriguing method for assessing the extent to which novel events are processed, this technique may change the oddball task significantly, such that brain systems

19 Friedman chapter 19 other than those reflecting the involuntary capture of attention (novelty P3) may be recruited. It may be one or more of those very systems, such as controlled attention (effortful processing resources), that are dysfunctional in PAD (Parasuraman & Haxby, 1993) leading to the reduction in the P3 elicited by the novel objects. A critical aspect of the orienting response is its habituation over time (Lynn, 1966). Accordingly, several investigators have shown that, like other ubiquitous markers of the orienting response such as the galvanic skin response, novelty P3 amplitude diminishes in young adults as more and more novel events are experienced or the same novel event is repeated (Czigler et al., 2006; Friedman & Simpson, 1994; Kazmerski & Friedman, 1995; Knight, 1984; Knight, 1996; Yamaguchi & Knight, 1991; reviews by Friedman et al., 2001 and Ranganath & Rainer, 2003). By contrast, in older adults, the novelty P3 has been shown not to habituate (Czigler et al., 2006; Fabiani & Friedman, 1995; Friedman & Simpson, 1994; Kazmerski & Friedman, 1995; Weisz & Czigler, 2006), whether those events are attended or ignored (Friedman et al., 1998). Because the scalp-recorded novelty P3 receives contributions from prefrontal cortex (Daffner et al., 2000; Halgren, Marinkovic, & Chauvel, 1998a; Knight, 1984) and patients with prefrontal damage do not show habituation to these types of stimuli (Knight, 1984; Woods & Knight, 1986), this type of finding has been interpreted by some to indicate support for the frontal-lobe deficit hypothesis of cognitive aging (Friedman et al., 1998; see also Buckner, 2004 and West, 1996; but see Greenwood, 2000; Greenwood, 2007). This hypothesis has also been motivated by the well-replicated finding that older, relative to young, adults show more frontally-oriented scalp distributions to both oddball targets and deviant, novel events (Fabiani & Friedman, 1995; Fjell & Walhovd, 2003; Friedman, Simpson et al., 1993; Friedman & Simpson, 1994; Kutas, Iragui, & Hillyard, 1994; Pfefferbaum, Ford,

20 Friedman chapter 20 Wenegrat, Roth, & Kopell, 1984; Yamaguchi & Knight, 1991; see Figure 16.4), suggesting that older adults may call on frontal lobe resources to a greater extent than young adults even for events that should have been well encoded and categorized. However, recent evidence suggests that some older adults may call on these resources to a greater extent than others (Daffner et al., 2006a; Fabiani, Friedman, & Cheng, 1998; Riis et al., 2008). An open question is whether the change in scalp distribution (and, presumably, the underlying generators) reflects a compensatory mechanism in high-functioning older adults whose brains might have the capacity to produce this change. That is, relative to young adults, the topographic change ought to be associated with equivalent or near-equivalent performance that, without such compensatory activity, would have been lower. On the other hand, the change could represent an attempt on the part of low-functioning older adults performing poorly on the task to compensate for the deleterious effects of brain aging (Raz, 2000). Adjudicating between these alternatives is difficult and, in fact, examples of both interpretations have been advanced. For example, Fabiani and colleagues (Fabiani et al., 1998) divided their older participants into good- and poor- frontallobe test performers. The poor performing subgroup showed target P3 scalp distributions that had frontal maxima (the more typical finding when averages are computed across all participants data within the older age group) whereas those categorized as good performers showed parietal-maximal scalp topographies. In this case, the data suggested that the frontal scalp topography reflected less efficient processing. On the other hand, in a series of investigations, Daffner and coworkers, using the visual oddball paradigm described earlier, have come to the opposite conclusion (Daffner et al., 2005; Daffner et al., 2006a, 2006b; Riis et al., 2008), more frontal P3 activity indicates greater compensatory recruitment of frontal resources and reflects successful cognitive aging. Notwithstanding the differences in design and stimulus

21 Friedman chapter 21 materials between the Daffner experiments and those predominating in the literature (for discussion see Fjell & Walhovd, 2005 and Daffner et al., 2005), the compensation hypothesis is currently highly controversial (Colcombe, Kramer, Erickson, & Scalf, 2005; Friedman, 2003; Greenwood, 2007; Zarahn, Rakitin, Abela, Flynn, & Stern, 2006). Determining whether a pattern of brain activity is compensatory, inefficient, detrimental or unrelated to performance will require more precise definitions of what is meant by compensation (see Davis, Dennis, Daselaar, Fleck, & Cabeza, 2007 and Stern, 2002 for examples), and under which conditions such compensation might or might not be expected. To summarize, with the exception of the studies by Daffner and colleagues, the majority of investigations of age-related change in the novelty P3 have shown it to be reduced and its topography more frontally oriented in older compared to young adults. Consistent with this, some data suggest that only highly deviant events are likely to capture the older adult s attention. Moreover, unlike young adults who show a reduction in amplitude after the first few presentations of unexpected, novel events, older adults do not, suggesting that they repeatedly recruit prefrontal cortical mechanisms for stimulus events which should no longer involuntarily capture attention. Finally, the extent to which compensatory mechanisms are responsible for the recruitment of frontal resources in response to novel events by some older adults is currently equivocal and requires further experimentation and documentation. N2b and P3b. By contrast with the MMN, which appears to reflect a relatively automatic, preattentive mechanism, the N2b (at approximately ms post stimulus), like the P3b, is typically observed only when participants are focusing attention on the sequence of events in order to make a task-relevant decision (Ritter, Simson, Vaughan, & Friedman, 1979; Chapter 11, this

22 Friedman chapter 22 volume). The N2b is also elicited in the passive oddball paradigm when highly deviant stimuli, such as environmental sounds, involuntarily capture attention (Kazmerski, Friedman, & Ritter, 1997). The P3b component is arguably the most well-studied of the ERP components. It is, therefore, not surprising that much more research effort has been expended in assessing agerelated changes in its amplitude, latency and scalp topography compared to the N2b. To illustrate some of the effects of aging on the N2b and P3b components, Figure 16.4 depicts the grand mean ERPs elicited by targets and, for comparison, novel, environmental sounds in young (N=10) and older (N=10) adults recorded during a version of the novelty oddball task (Friedman, unpublished data). The scalp distribution maps to the right of the waveforms depict, as noted earlier, one of the most consistent findings in the ERP/aging literature relative to young adults, the older adults novel (P3a) and target (P3b) topographies are more frontally oriented. Another ubiquitous finding evident in the figure is the age-related prolongation of the latency of both novel and target P3 components (e.g., Brown, Marsh, & Agenath, 1983; Fjell & Walhovd, 2003; Iragui et al., 1993; Picton et al., 1984). The most comprehensive assessment to date of age-related effects on P3b latency has been published by Polich (1996), and the reader is referred to that publication for more details. A third wellreplicated finding observable in Figure 16.4 is the smaller amplitude P3b component in older compared to young adults. This phenomenon has been reported in a wide variety of tasks in addition to the oddball paradigm (e.g., Ford et al., 1997; Friedman et al., 1997; Friedman, Nessler, Johnson et al., 2007; Gaeta et al., 2003; Iragui et al., 1993; McEvoy et al., 2001). The N2b also shows age-related variation. For example, its latency increases with age in highly similar fashion to that demonstrated for the P3b (Anderer et al., 1996; Enoki, Sanada, Yoshinaga, Oka, & Ohtahara, 1993; Goodin, Squires, Henderson, & Starr, 1978; Iragui et al.,

23 Friedman chapter ). This can be seen in Figure 16.4 where, relative to young adults, a prominent, though prolonged-latency N2b component is evident in the target ERPs of the older adults. By contrast with latency, one of the difficulties in assessing the consistency of age-related change in N2b magnitude is that, although it is typically recorded in a wide range of paradigms, it has often not been measured directly. Nonetheless, when N2b amplitude has been calculated there are a somewhat greater number of studies whose results show larger N2b magnitudes in older compared to young adults (Anderer et al., 1996; Czigler et al., 2006; Friedman, Simpson et al., 1993; Gaeta et al., 2003; Woods, 1992) than indicate the converse (Bertoli et al., 2005; Czigler & Balazs, 2005; Karayanidis et al., 1995). With respect to pathological aging, the results are quite mixed. Although, as might be expected, PAD patients sometimes show prolonged-latency P3b components relative to agematched controls (e.g., Bennys, Portet, Touchon, & Rondouin, 2007; Golob & Starr, 2000; O'Donnell, Friedman, Squires, & Maloon, 1990; Patterson, Michalewski, & Starr, 1988; Williams, Jones, Briscoe, & Thomas, 1991), other research groups have reported a failure to distinguish PAD participants from controls on the basis of this metric (Gordon, Kraiuhin, Stanfield, & Meares, 1986; Kazmerski & Friedman, 1998; Kraiuhin et al., 1990; Pfefferbaum, Wenegrat, Ford, Roth, & Kopell, 1984). Whether the P3b latency prolongation has sufficient sensitivity and specificity for clinical diagnosis is, therefore, open to question (Gordon et al., 1986; Patterson et al., 1988). Similarly, P3b amplitudes are sometimes (Frodl et al., 2002; Goodin, Squires, & Starr, 1978; Saito et al., 2001), but not always (Golob & Starr, 2000; Kazmerski & Friedman, 1998) reported to be smaller in PAD patients compared to controls. Moreover, the scalp distribution of P3b does not appear to differ between PAD patients and controls, at least in the auditory modality (Ford et al., 1997; Kazmerski & Friedman, 1998).

24 Friedman chapter 24 Hence, as for P3b latency, the utility of amplitude and topography for clinical diagnosis is equivocal. This conclusion is as further supported by the results of a very recent study by Golob et al. (2007), who reported that neither P3b latency prolongation nor amplitude reduction were able to predict which MCI patients converted to PAD (see Taylor & Olichney, 2007 for a review of ERPs in dementia). In sum, the evidence to date is fairly conclusive that the mental operations indexed by N2b and P3b increase in latency as individuals age, consistent with the phenomenon of general slowing reported in the behavioral literature (Salthouse, 1991). Whether P3b latency slowing is exacerbated in pathological aging is not yet certain. The tenuousness of this finding in PAD may be due to the fairly simple oddball paradigms that have most often been used to assess P3b latency. N2b is thought to be the first brain event indicating that a conscious, sensory discrimination has been made (Ritter et al., 1979). P3b, therefore, must reflect a subsequent stage of processing, although a consensus on its functional significance has yet to be reached (see below). One of the major theoretical positions advanced to account for age-related changes in cognition postulates that even relatively early stages of processing are slowed (Salthouse, 1996). If this hypothesis is valid, then the early latency ERP components (e.g., P50, N100) should be of longer latency in older compared to young-adult controls. The prolongation of these relatively early-latency components could engender a cascade of slowing that might be manifested in the age-related RT retardation that has been ubiquitously observed in a wide variety of cognitive paradigms (Salthouse, 1996) and prolonged P3b latencies, one of the most often-replicated findings in the ERP aging literature (Polich, 1996). However, there is very little evidence for retardation in the early-latency components (Anderer et al., 1996; Goodin et al., 1978; Iragui et

25 Friedman chapter 25 al., 1993). By contrast, the ERP data suggest that the slowing is restricted to the later stages of information processing, as indexed by N2b and P3b. Hence, because of this fact, the retardation in N2b and P3b latencies in older adults cannot be due simply to information loss from slowed operations at earlier stages in the processing stream. Like P3 latency, the evidence for an age-related reduction in P3b amplitude is compelling. By contrast, the evidence for further reduction by PAD and MCI is equivocal. However, the implications of the age-related reduction in P3b amplitude for understanding the underpinnings of cognitive aging phenomena are not clear at this time. Complicating the picture further is the age-related change in P3b scalp topography. This is somewhat problematic for interpretation of the age-related significance of the changes in cognition that underlie the P3b recorded at the scalp because it is well known that it receives contributions from a widespread intracranial network (Halgren, Marinkovic, & Chauvel, 1998b), any aspect of which could be altered with aging. Hence, these alterations could be functionally driven reflecting, for example, an age-related change in the source regions that contribute to the P3b or, less interestingly, to age-related structural changes within the brain. For example, age-related shrinkage in brain volume, which is well documented (Raz et al., 2004), might alter the orientation of the brain generators, thereby modifying the older adults P3b scalp topography even though the processes reflected by the P3b do not change with age. However, for this account to be viable, the topographic change would most likely have to be highly similar across tasks (e.g., oddball, Sternberg short-term memory, repetition priming etcetera). Although this proposition has, to my knowledge, not been tested directly, there is some evidence that, although older adults typically produce more frontally-oriented P3b distributions in a variety of cognitive paradigms, those topographies can be modified by task demands (Friedman et al., 1997). Then again, the Friedman

26 Friedman chapter 26 et al. (1997) topographic comparisons were made between independent samples of subjects who had participated in similar, though not identical, experiments. Thus, the evidence from the Friedman et al. (1997) investigation is limited and needs to be bolstered by within-subject, taskrelated comparisons of P3b topography in older adults. One influential theory of the functional significance of the P3b posits that it reflects updating when the subject s model of the environment requires revision (Donchin & Coles, 1988), a key aspect of working memory (Baddeley, 1992). Furthermore, a major hypothesis used to account for cognitive decline in aging is that working memory, which has been conceptualized as the amount of resources available to process information online, is reduced (Craik & Byrd, 1982; Park, 2000a). Therefore, it could be the case that older adults call upon this type of general-purpose, working memory/attentional resource much more often than young adults, thereby accounting for the ubiquity and prominence of the frontal aspect of the P3b distribution in a wide variety of tasks (Fabiani & Friedman, 1995; Ford et al., 1997; Friedman, Nessler, Johnson et al., 2007; see Friedman et al., 1997 for discussion). However, as for the novelty P3, whether this topographic change reflects a compensatory modification of brain activity to counteract the deleterious effects of brain aging is unknown at this time and clearly requires further research effort (Friedman, 2003; Friedman, 2007). N400. The presentation of a phrase- or sentence-ending pictorial or verbal concept that is incongruent with the meaning of the preceding material produces a large-amplitude N400 effect in young adults (Kutas & Hillyard, 1980; Chapter 14, this volume). This incongruity effect, defined by subtracting the ERP to incongruous from that to congruous endings, has been assessed in a number of age-related investigations (reviews by Federmeier, 2007; King & Kutas,