Neuropsychologia 51 (2013) Contents lists available at SciVerse ScienceDirect. Neuropsychologia

Transcription

1 Neuropsychologia 51 (2013) Contents lists available at SciVerse ScienceDirect Neuropsychologia journal homepage: Cognitive function predicts neural activity associated with pre-attentive temporal processing $ Shannon M. Foster a, Michael A. Kisley a, Hasker P. Davis a, Nathaniel T. Diede a, Alana M. Campbell b, Deana B. Davalos b,n a University of Colorado at Colorado Springs, Department of Psychology, 1420 Austin Bluffs Pkwy, Colorado Springs, CO 80918, USA b Colorado State University, Department of Psychology, Campus Box 1876, Fort Collins, CO 80523, USA article info Available online 26 September 2012 Keywords: Temporal processing Cognition Aging Mismatch negativity (MMN) Event-related potential (ERP) abstract Temporal processing, or processing time-related information, appears to play a significant role in a variety of vital psychological functions. One of the main confounds to assessing the neural underpinnings and cognitive correlates of temporal processing is that behavioral measures of timing are generally confounded by other supporting cognitive processes, such as attention. Further, much theorizing in this field has relied on findings from clinical populations (e.g., individuals with schizophrenia) known to have temporal processing deficits. In this study, we attempted to avoid these difficulties by comparing temporal processing assessed by a pre-attentive event-related brain potential (ERP) waveform, the mismatch negativity (MMN) elicited by time-based stimulus features, to a number of cognitive functions within a non-clinical sample. We studied healthy older adults (without dementia), as this population inherently ensures more prominent variability in cognitive function than a younger adult sample, allowing for the detection of significant relationships between variables. Using hierarchical regression analyses, we found that verbal memory and executive functions (i.e., planning and conditional inhibition, but not set-shifting) uniquely predicted variance in temporal processing beyond that predicted by the demographic variables of age, gender, and hearing loss. These findings are consistent with a frontotemporal model of MMN waveform generation in response to changes in the temporal features of auditory stimuli. & 2012 Elsevier Ltd. All rights reserved. 1. Introduction Temporal processing refers to the act of processing timerelated information and is thought to play a significant role in a variety of cognitive and psychological processes including speech, language, and even one s subjective well-being (Wittmann & Paulus, 2008). The construct of temporal processing is complex and associated with a variety of processes from simple sensory perceptions of time to more elaborate higher-level concepts of temporal reproduction, time estimation, and even processes needed to remember and engage in future activities (e.g., timebased prospective memory) (Barkley, Murphy, & Bush, 2001; Mäntylä, Carelli, & Forman, 2007). Given the plethora of tasks subsumed under the construct of temporal processing, it is not surprising that there have been conflicting findings regarding the neural substrates of this construct and the relationships between $ This research was supported by funding from the National Institute on Aging (1 R15 AG ). n Corresponding author. Tel.: þ ; fax: þ address: Davalos@ColoState.Edu (D.B. Davalos). temporal processing and other cognitive processes (Coull, Vidal, Nazarian, & Macar, 2004; Harrington & Haaland, 1999). One of the main confounds to assessing the basic neural underpinnings of temporal processing is that most measures of timing involve other supporting cognitive processes, such as decision making or working memory (Paul et al., 2011). The recruitment of these other cognitive processes has made identifying neural substrates associated with timing difficult. One method of assessing temporal processing that has emerged over the past decade is the use of pre-attentive electrophysiological methods. Specifically, a pre-attentive event-related brain potential (ERP) waveform, the mismatch negativity (MMN), has been utilized to measure neural activity elicited by a temporal deviant in order to assess timing abilities in the absence of attention demands (Davalos, Kisley, Ellis, & Freedman, 2003; Grimm, Widmann, & Schröeger, 2004; Kisley et al., 2004; Näätänen, Jiang, Lavikainen, Reinikainen, & Paavilainen, 1993; Näätänen, Syssoeva, & Takegata, 2004; Pakarinen, Takegata, Rinne, Huotilainen, & Näätänen, 2007; Tse & Penney, 2006). More specifically, neural responsivity to a deviation in either the duration of a stimulus or in the interstimulus interval (ISI) between stimuli is compared to neural activity associated with a standard duration stimulus or standard /$ - see front matter & 2012 Elsevier Ltd. All rights reserved.

2 212 S.M. Foster et al. / Neuropsychologia 51 (2013) ISI, respectively, which is presented with much greater frequency. The difference between these measures of neural activity is calculated and known as the MMN waveform. It is measured passively by engaging the participant in a distraction task, such as watching a movie, while auditory stimuli are presented and brain responses are measured via scalp electrodes. While the MMN is not entirely insusceptible to certain types of attentional effects (Woldorff, Hackley, & Hillyard, 1991), it is an ERP that allows for the measurement of neural responses to timing stimuli in the absence of motor response, motivational demands, decision making and other potential cognitive demands that may affect temporal processing performance. Temporal deviation MMN paradigms have been used with a variety of clinical populations for whom traditional behavioral timing tasks may be difficult (e.g., schizophrenia) (Davalos, Kisley, Polk, & Ross, 2003; Michie, 2001; Todd, Michie, & Jablensky, 2003). In addition, these paradigms have been used to study development of temporal processing abilities from infancy to late adulthood (Brannon, Libertus, Meck, & Woldorff, 2008; Brannon, Roussel, Meck, & Woldorff, 2004; Cheour, Kushnerenko, Ceponiene, Fellman, & Näätänen, 2002; Gomes et al., 1999). Findings within these populations suggest, albeit indirectly, that pre-attentive temporal processing may be dependent on intact frontal lobe functioning. While temporal deviance MMN studies have been quite useful for assessing perceptual timing abilities, there has been relatively little research directly addressing either the specific neural substrates associated with these event-related potentials or the relationship between the MMN and cognitive variables. It has been argued that the earliest neural process involved in the MMN relates to detection of the change in stimuli by the auditory cortices. This early perceptual or pre-attentive process is proposed to precede a frontal-cortex activation associated with directing conscious attention to the change in the stimuli (e.g., a tone duration or ISI which is different than the preceding tone duration or ISI) (Alho et al., 1998; Näätänen, Jacobsen, & Winkler, 2005; Rinne, Alho, Ilmoniemi, Virtanen, & Näätänen, 2000; reviewed by Deouell, 2007; Näätänen, Kujala, & Winkler, 2011; Winkler, 2007). Neuroimaging studies that have sought to identify the neural substrates of temporal processing utilizing stimuli presented in a manner similar to the temporal deviance MMN paradigms have generally found early processing in the superior temporal gyrus (STG) followed by frontal activation, specifically the inferior, medial, and superior frontal gyri in addition to the dorsolateral prefrontal cortex (DLPFC) (Molholm, Martinez, Ritter, Javitt, & Foxe, 2005; Rinne et al., 2000; Rinne, Degerman, & Alho, 2005; Tregellas, Davalos, & Rojas, 2006). Other studies using lesion methodology (Alain, Woods, & Knight, 1998), positron emission tomography (Müller, Jüptner, Jentzen, & Müller, 2002), and event-related optical imaging (Tse, Tien, & Penney, 2006; Tse & Penney, 2007) suggest involvement of a frontotemporal network in the generation of the MMN in response to temporal deviation. The recruitment of the prefrontal cortex in the temporal deviation MMN suggests that pre-attentive or perceptual timing may share neural substrates with higher order cognitive processes that are also dependent on the frontal lobe (i.e., executive functions ). Studies that have explored the potential relationship between perceptual timing and executive functioning have typically focused on the duration MMN within populations that exhibit executive dysfunction or have known frontal cortical dysfunction (e.g., schizophrenia) (Light & Braff, 2005; Toyomaki et al., 2008). The question remains, therefore, whether perceptual timing is associated with executive functioning and/or other cognitive functions in healthy individuals. There appears to be a growing consensus that timing abilities, or at least temporal processing skills assessed behaviorally, are necessary for a number of tasks described as executive functioning. Mäntylä et al. (2007), among others, argue that most goaldirected activities require some form of temporal integration and monitoring of action (Fuster, 1993, 2002; Norman & Shallice, 1986). Research examining this question, however, has been limited to pediatric populations primarily, with only a few studies including young adults. The findings have been mixed, with only select executive tasks relating to time processing skills. Specifically, study findings have suggested that those types of executive tasks that depend primarily on the prefrontal cortex appear to be more closely related to time processing abilities. For example, tests of inhibition and updating show robust relationships to behavioral indices of timing (Barkley, 1997; Mäntylä et al., 2007). A number of investigators have noted a relationship between increased errors on response inhibition tasks and deficits in time reproduction (Gerbing, Ahadi, & Patton, 1987; Montare, 1977). Specifically, those with poor inhibition generally make more timing errors than those with good inhibition. However, it has been argued frequently that attentional abilities may be at the root of both inhibition and timing deficits (Barkley, Edwards, Laneri, Fletcher, & Metevia, 2001; Brown, 1985; Zakay, 1992). It is therefore questionable whether the relationship between timing and response inhibition would persist if the attentional demands were removed from the timing task (e.g., through the use of a preattentive measure, such as temporal deviance MMN). In contrast, executive tasks focusing on shifting within the task (e.g., from color to number) have not demonstrated the same relationship with timing abilities. Mäntylä et al. (2007) have argued that shifting is different from other executive functioning tasks as the parietal cortex plays a larger role in shifting than the prefrontal cortex. This argument has been supported both via neuropsychological studies and neuroimaging studies (Collette & Van der Linden, 2002; Gehring & Knight, 2002). The question remains whether the relationship between executive functioning and temporal processing reflect shared neuroanatomical substrates only when timing is assessed behaviorally. Given that most behavioral timing tests engage working memory, decision making, and other cognitive processes, it is unclear whether it is those processes that drive the relationships between executive functioning and temporal processing or basic timing skills themselves. The current study seeks to examine whether temporal processing, measured in the absence of other cognitive demands (via ISI-based MMN), is also significantly related to cognitive tasks dependent on the prefrontal cortex in a healthy population. To date, only one study has addressed the potential relationship between temporal processing, as measured by the MMN waveform, and cognitive abilities directly. This study (Kisley, Davalos, Engleman, Guinther, & Davis, 2005) found that in a sample of older adults auditory-verbal acquisition, planning time on a tower task, and accuracy of a conditional reaction time measure were related to MMN amplitude in response to temporal deviation (specifically ISI-deviance). The direction of these relationships was such that more intact cognition was related to greater neural activity associated with pre-attentive temporal processing. However, in addition to utilizing only a portion of the available older adult group for that analysis, the study did not control for demographic variables that may impact MMN amplitude and cognitive function, including gender and sensory loss (e.g., hearing threshold). Given the paucity of research exploring the relationship between pre-attentive measures of temporal processing and cognitive variables in a healthy adult sample, this study sought to expand upon the findings of Kisley et al. (2005) by studying the relationship between pre-attentive temporal processing and cognitive functions associated with a frontotemporal neural network

3 S.M. Foster et al. / Neuropsychologia 51 (2013) in a larger sample of older adults given their innate heterogeneity in neural and cognitive functioning (Gaeta, Friedman, Ritter, & Cheng, 1998). By utilizing a large sample of older adults we were able to explore the ability of cognitive variables to predict variance in pre-attentive temporal processing, as measured by the MMN elicited by deviation in the ISI between auditory stimuli. Specifically, we used a hierarchical regression analysis to explore how various cognitive abilities interact to predict neural processing of time-based features. Given evidence that temporal processing is related to a frontotemporal neural network (Alain et al., 1998; Müller et al., 2002), we hypothesized that cognitive abilities associated with the prefrontal cortex and temporal lobe (i.e., executive functions and verbal memory, respectively) would be associated with, and serve as strong predictors of, preattentive temporal processing. In addition, we predicted that measures more commonly associated with other neural networks (e.g., subcortical white matter, parietal lobe) would not predict pre-attentive temporal processing. 2. Methods Participants gave written informed consent and were compensated for their time monetarily. The Institutional Review Board of the University of Colorado, Colorado Springs, approved all procedures Participants Participants were recruited from a local memory clinic. At the memory clinic, individuals underwent a neuropsychological screening evaluation and were subsequently placed into one of three diagnostic categories based on objective memory performances, daily functioning, and the clinical judgment of a licensed psychologist. Those who appeared to be aging as expected without objective evidence of memory impairment or functional decline were categorized as normal. Those who displayed objective evidence of memory impairment and functional declines suggestive of a dementia process were categorized as possible dementia. Finally, those who displayed mild memory difficulties and/or presented with subjective memory complaints, but who continued to function effectively in daily life, were asked to follow-up at the clinic in 1 year and were classified as follow-up. To ensure a wide range of cognitive ability within the sample, individuals categorized as either normal or follow-up were asked to participate in the study. Individuals categorized as possible dementia were excluded to ensure that all participants were able to provide informed consent and comply with study procedures. A total of 66 participants were recruited for the study. Of these, seven were excluded from analyses due to an insufficient number of acceptable MMN trials (see Section ) and two more due to incomplete data, leaving a total of 57 participants. Approximately two-thirds of the participants were female (n¼37, 64.9%). Age ranged from 53 to 89 years (M¼67.63, SD¼8.23) and education level ranged from 8 to 23 years (M¼15.82, SD¼3.01). Participants included 35 individuals (61.4%) categorized as normal and 22 individuals (38.6%) categorized as follow-up. Time elapsed between the memory clinic screen and the present study ranged from 10 to 25 months (M¼15.51, SD¼2.16), with most participants (94.74%) completing the present study 12 or more months after the initial screening (substantially decreasing the possibility of practice effects in repeated measures) Procedures Prior to the electrophysiological recording, hearing was tested at 1000 Hz (binaural, method of limits) to allow for individual adjustment of auditory stimuli to compensate for hearing ability. Included participants all exhibited less than 15 db hearing loss at 1000 Hz, the frequency at which auditory tones were presented for the MMN task. Recording of electrophysiological signals occurred while participants passively watched a silent close-captioned movie. Movies were rated appropriate for a general audience. The MMN waveform was elicited by deviation in inter-stimulus interval (ISI) during an ongoing auditory stimulus train (Davalos, Kisley, Polk, 2003; Kisley et al., 2004; Näätänen et al., 1993), described in Section Administration order was counterbalanced such that half of the participants completed cognitive assessment prior to the electrophysiological recording and the other half completed the procedure in the reverse order. Cognitive testing was completed in a quiet room free of distractions. Participants were given instructions and a chance to ask questions before each test was administered. For computerized tasks the experimenter stayed in the room with the participant through the first few trials of each task to verify that the task was being completed accurately Electrophysiology Disposable Ag/Ag Cl electrodes (Vermed; Bellows Falls, VT) were placed at Fz, Cz, and Pz; lateral and superior to the left eye; and on both mastoids. The ground electrode was placed on the forehead. Electrophysiological signals were collected and digitized with a NuAmps multi-channel recording system and Scan 4.2 software (Neuroscan; Sterling, VA). Signals were amplified 5000 times, filtered between 0.05 and 100 Hz, and sampled at 1000 Hz. Impedances were kept at or below 10 ko. Signals were referenced to the left mastoid during recording, but later re-referenced to the average of both mastoids during waveform analysis (see Section ). The electrodes near the eyes were used to detect eye movements and blinks artifacts that can overwhelm the measurements of neural activity from sites Fz, Cz and Pz (see Section ). Stimuli were presented to both ears through headphones while participants who were seated in a large reclining chair watched a silent close-captioned movie Mismatch negativity (MMN) paradigm and analysis. The paradigm used to elicit the MMN utilized temporal deviation from the standard interstimulus i- nterval (ISI) of 500 ms (Ford & Hillyard, 1981; Kisley et al., 2004; Näätänen et al., 1993; Nordby, Roth, & Pfefferbaum, 1988). The ISI during the continuous stream of tones (1000 Hz, 50 ms duration, 5 ms rise/fall, 45 db hearing level) was randomly replaced by a deviant ISI of 250 ms, on average every 20th trial. In total, 2850 standard and 150 deviant ISIs were presented during a 30-min recording session. Single-trial evoked responses were epoched from 100 ms before to 300 ms after tone onset and baseline corrected from the pre-stimulus average voltage. Trials for which any channel exceeded 775 mv were discarded. Participants that had fewer than 75 deviant trials remaining after artifact rejection were excluded from analysis (n¼7). Standard and deviant average waveforms were computed separately from the remaining, artifact-free trials. A band-pass filter was applied both forward and reverse, to eliminate phase distortion, from 1 to 30 Hz (48 and 96 db/ octave slope at these corner frequencies). Amplitude of the MMN waveform was calculated as the average amplitude from 100 to 160 ms post-stimulus, as this was the range over which the MMN waveform exhibited maximal deflection. This was computed at Fz and Cz. As these average MMN amplitudes were strongly correlated (r¼0.95, po.001), we only completed analyses using the average amplitude at Fz (AvgFzAmp) to minimize the chance of a Type I error (see also Kisley et al., 2005). We present and analyze the absolute value amplitudes, so MMN amplitudes are shown as positive values despite the negative-going direction of this waveform (as captured in the bottom panel of Fig. A1). For the present investigation the MMN waveform was computed against a reference composed of the average of left and right mastoids channels. The MMN is often measured against a nose reference in order to ensure that waveform polarity reverses between central sites and the mastoids (Näätänen et al., 1993). However, we have demonstrated in previous investigations that the MMN elicited in this specific paradigm reverses polarity at the mastoids (Kisley et al., 2004), and further that only at central sites (i.e., Fz and Cz, not at the mastoids) does MMN amplitude exhibit variation that is related to age and cognitive capacity (Kisley et al., 2005). Thus we decided to utilize a mastoid reference because it is less awkward and distracting for research participants than a nose reference, especially when watching a movie. Due to possible differences in ongoing activity at time of stimulus onset for standard and deviant stimuli, an analytical correction was also applied (Fig. A1). A time-shifted average of an evoked waveform caused by the standard ISI was subtracted from the deviant ISI waveform to eliminate overlap of responses due to the sudden decrease of ISI. For each participant, the average standard waveform was shifted forward by 250 ms and subtracted from the deviant waveform. We also analyzed non-corrected waveforms to test whether this analytical manipulation affected the analyses. The overall pattern of results, including the regression analyses presented below, was virtually the same and thus is not reported here (see also Kisley et al., 2005) Cognition Several behavioral tests were administered to measure cognitive function. Tests previously administered at the memory clinic included Trail Making, Benton Visual Retention Test (BVRT), and the Rey Auditory Verbal Learning Test (RAVLT). With the exception of the BVRT and Trail Making, tests were administered on a computer using the Colorado Assessment Tests software (CATs) (Davis & Keller, 1998; Davis & Klebe, 2001; Davis et al., 2003; Feldstein et al., 1999) Attention. Trail making A and B (Trails A and B) were used to measure basic and divided attention, respectively (Mitrushina, Boone, & D Elia, 1999). Trails A and B are sensitive to early cognitive decline (Botwinick, Storandt, Berg, & Boland, 1988). Trails A is composed of 25 numbers that participants must connect in order, while Trails B has participants switch between connecting numbers and letters in order (e.g., 1, A, 2, B). Participants were encouraged to complete the task as quickly

4 214 S.M. Foster et al. / Neuropsychologia 51 (2013) total number of words recalled following the delay. For both measures, higher scores were associated with better performances. BVRT (Sivan, 1992) was used to assess visuospatial memory. This task is also sensitive to early cognitive decline (Coman et al., 1999; Lezak et al., 2004). The BVRT consists of 10 cards depicting one or more figures. In accordance with Administration A procedures, participants were shown a card for 10 s and then asked to draw the stimulus from memory as accurately as possible once it was removed (Sivan). Form E was used to minimize practice effects. The dependent measure was number of cards reproduced accurately; as such, higher scores reflected better visual recall Executive functions. The CATs Tower of London (TOL; Davis & Keller, 1998) was used to assess planning, an aspect of executive functioning. Regional cerebral blood-flow studies have shown performance on this task to be associated with engagement of the frontal cortex and individuals with frontal lobe damage display deficits on this task (Morris, Ahmed, Syed, & Toone, 1993; Newman, Carpenter, Varma, & Just, 2003; Shallice, 1982). On this computerized version of the task, participants must move beads on the left side of the screen to match those displayed on the right side of the screen in as few moves as possible. Participants completed 21 increasingly difficult problems (i.e., 7 trials of each type: 3 beads/3 pegs, 4 beads/4 pegs, and 5 beads/5 pegs). The dependent measure was total n- umber of excess moves, with higher scores reflecting poorer executive function. The CATs Wisconsin Card Sorting Test (WCST; Davis & Keller, 1998) is an executive functioning measure that captures an individual s ability to categorize stimuli and shift sets based on use of feedback from prior moves (Heaton, Chelune, Talley, Kay, & Curtiss, 1993). Participants were presented with cards and asked to match them to one of four key cards. After each move they were told whether they were correct or incorrect based on a previously established matching criteria that was unknown to the participant. Without warning, the matching criteria changed after 10 correct matches. Continuing to match based on an incorrect pattern is considered a perseverative response and is associated with an inability to inhibit a pre-potent response (Amos, 2000; Everett, Lavoie, Gagnon, & Gosselin, 2001); a function commonly attributed to the frontal lobes (cf. Rieger, Gauggel, & Burmeister, 2003). As such, number of perseverative errors (WCST pers ) was recorded as a dependent measure, where higher scores reflected poorer performance. The number of categories a participant completed (WCST cat ) was also recorded and used as a dependent measure. Higher WCST cat scores captured the ability to categorize stimuli in a variety of ways and to shift between cognitive sets (i.e., stronger executive functioning). Fig. A1. MMN to deviantshortened ISI compared to standard. Grand-averaged waveforms computed by combining the neural responses from the entire sample studied (N¼57). These waveforms are not analyzed, but rather presented here to demonstrate the general pattern of neural responding to temporal stimulus features. Shown are the grand-averaged ERP responses to standard (500 ms ISI) and deviant (250 ISI) stimuli, as well as the point-by-point difference wave (i.e., deviant minus standard). The prominent negative-going peak between 100 and 150 ms in the difference wave (bottom panel) is the MMN. The three panels illustrate the steps in the analytical correction described in Section 2. Top, the shifted standard response is the same as the standard response, but moved forward in time by 250 ms. Middle, when the deviant stimulus occurs a prominent response from the preceding standard stimulus is still present in the waveform, especially evident in the pre-stimulus period. The contribution of this ongoing response to the deviant waveform is reduced by subtracting the shifted standard waveform. Bottom, the corrected difference waveforms were used for analysis in the present study, but an uncorrected wave is also shown for comparison. As mentioned in the text, the results from this study were substantively similar regardless of whether the corrected or uncorrected difference waves were utilized Motor speed, processing speed, and inhibition. The CATS Reaction Time task (Davis & Keller, 1998) was used because it includes measures of pure motor speed, processing speed, and conditional inhibition, which have been associated with age-related decline (Lezak et al., 2004; Salthouse, 1996). The simple task (Rxn Motor), a measure of alertness and motor speed, involved pressing a control button when a symbol appeared on the screen. The choice task (Rxn Choice), a processing speed measure, required participants to press one of two control buttons depending on the symbol that appeared (i.e., press left control button if left arrow appears, press right control button if right arrow appears). The conditional task (Rxn Cond.), an executive functioning measure of response inhibition, required participants to vary their responses to directional symbols (i.e., right and left arrows) by a conditional modifier ( þ or ). A þ (positive trial) presented with a directional arrow required a response in the same direction as the arrow, whereas a (negative trial) required a response in the direction opposite to that of the arrow (e.g., with left arrow required the participant to press the right control button). Rxn Cond. was calculated as the absolute difference between the negative and positive trials during the conditional task (Rxn Cond.). The absolute difference was used because several participants engaged atypical strategies during the conditional task that resulted in greater time for positive trials than negative trials. For all measures, higher scores reflected poorer performances. and accurately as possible. If a mistake was made the examiner stopped the participant and redirected him/her to the last correct answer. For both tasks, time to complete the task was used as the dependent measure. Higher scores reflected slower performance and therefore poorer basic and divided attention Memory. The CATs RAVLT was included to measure verbal learning and memory because of its sensitivity to age-related cognitive decline (Davis et al., 2003; Lezak, Howieson, & Loring, 2004). Both learning and delayed memory measures have been associated with left temporal lobe functioning (Foster et al., 2008; Giménez et al., 2004; Grammaldo et al., 2006; Kilpatrick et al., 1997; Sziklas & Jones-Gotman, 2008). RAVLT List B was used to prevent practice effects from the initial memory clinic screening. Participants were asked to recall 15 concrete n- ouns, in any order, across five learning trials (immediate recall) and then again after a 20-min delay (delayed recall). Words were presented at a fixed rate (1 word/2 s) and the order of presentation differed across the five learning trials. The total number of words recalled across the five learning trials (Rey Total) was used as a measure of verbal learning. Verbal memory was assessed using, Rey Delay, the 3. Results Given multiple analyses, alpha is set at.01 to reduce the chance of a type I error. Effect sizes are estimated for correlations (r) and associated percentage of variance accounted for (r 2 ) using the criteria established by Cohen (1992). As such, the correlations associated with small, medium, and large effects were those with absolute values from.10 to.29,.30 to.49, and.50 to 1.0, respectively. Although data were collected at several neural locations, only data from site Fz is analyzed to be consistent with the work previously published by our lab (Kisley et al., 2005). Table A1 provides the means, standard deviations, and range of values for each variable. Fig. A1 displays the grand-averaged MMN waveform for the entire sample.

5 S.M. Foster et al. / Neuropsychologia 51 (2013) Table A1 Descriptive statistics. Min. Max. M SD Age Hearing loss (db) Trails A a (s) Trails B a (s) Rey Total Rey Delay BVRT TOL a a WCST pers WCST cat Rxn Motor a (ms) Rxn Choice a (ms) Rxn Cond. a (ms) AvgFzAmp (mv) Note: Trails A¼basic attention; Trails B¼divided attention; Rey Total¼verbal learning; Rey Delay¼verbal memory; BVRT¼visual memory; TOL¼planning, executive function; WCST pers ¼perseverative errors/inhibition, executive function; WCST cat ¼switching, executive function; Rxn Motor¼simple reaction time; Rxn Choice¼choice reaction time (processing speed); Rxn Cond.¼conditional reaction time (conditional inhibition, executive function); AvgFzAmp¼average amplitude of MMN waveform. a Higher scores on these variables represent poorer performance. Table A2 Correlations of independent variables with MMN amplitude (before and after controlling for age, gender, and hearing). AvgFzAmp (before) Age.30 n Gender b.35 nn Hearing.26 Trails A a.29 n.12 Trails B a.27 n.15 Rey Total.54 nnn.38 nn Rey Delay.52 nnn.36 nn BVRT TOL a.50 nn.40 nn a WCST pers WCST cat.31 n.21 Rxn Motor a Rxn Choice a.28 n.28 n Rxn Cond. a.33 n.28 n AvgFzAmp (after) Note: Trails A¼basic attention; Trails B¼divided attention; Rey Total¼verbal learning; Rey Delay¼verbal memory; BVRT¼visual memory; TOL¼planning, executive function; WCST pers ¼perseverative errors/inhibition, executive function; WCST cat ¼switching, executive function; Rxn Motor¼simple reaction time; Rxn Choice¼choice reaction time (processing speed); Rxn Cond.¼conditional reaction time (conditional inhibition, executive function); AvgFzAmp¼average MMN amplitude at site Fz. a Higher scores on these variables represent poorer performance. b Gender coding: 1 (male) and 2 (female). n po.05. nn po.01. nnn po.001. Correlations were used to examine the relationships between each independent variable and the dependent variable. Partial correlations were then used to control for the effects that the demographic variables may have had on both the cognitive variables and the dependent variable. All assumptions for correlational analyses were met. Correlations between the average amplitude of the MMN waveform and the cognitive variables are depicted in Table A2. In general, significant relationships emerged between MMN amplitude and measures of auditory memory and executive functions. Specifically, stronger response to temporal deviation (i.e., higher MMN amplitude) was related to stronger auditory memory and executive functions. As further depicted in Table A2, after controlling for age, gender, and hearing in a partial correlation analysis, these relationships were persevered. Although the correlations between MMN average amplitude and the demographic variables (i.e., age, gender, and hearing) were relatively small, these factors are frequently correlated with the MMN measure and for this reason we controlled for them during the regression analysis Regression analysis To assess the relative strength of these variables in predicting average amplitude of the MMN waveform we conducted a hierarchical regression analysis. We controlled for the variance associated with demographic variables in the first step. All of the cognitive variables were entered into the second step and a stepwise analysis was used to identify the best set of predictor variables. Probability to enter the model was set at.05 and probability to remove was.10. Given findings that the MMN waveform is associated with activation of a frontotemporal network (Alain et al., 1998; Müller et al., 2002), we expected the most predictive variables to include those associated with frontal (i.e., executive functions) and temporal (i.e., memory) functioning. All assumptions of multiple regression were met, with the exception of sample size. Results of this analysis are shown in Table A3. Semi-partial correlations (sr) were used to assess the unique variance associated with each variable. The variance accounted for by the demographic variables (22.1%) was significant, F(3, 53)¼5.01, po.01, and constituted a medium effect. Gender bordered on making a significant unique contribution to explaining the variance in amplitude at this step (sr¼.32). It accounted for 10.11% of the variance, a medium effect. TOL was selected for entry in the second block and produced a significant change in the predictive ability of the variables, F change (1, 52)¼9.62, po.01, explaining an additional 12.2% of the variance, a medium effect. At this step, gender no longer uniquely accounted for variance in amplitude, and TOL (sr¼.35) uniquely accounted for 12.18% of the variance. In the third step, Rey Delay was selected for entry into the model and it explained an additional 7.8% of the variance, F change (1, 51)¼6.88, p¼.01. This finding represents a small effect and merely bordered on significance at the.01 level. At this step, TOL continued to make a unique contribution (sr¼.31), explaining 9.80% of the variance (medium effect). The unique contribution of Rey Delay to the model (sr¼.28, 7.84% of variance, small effect) bordered on significance at p¼.011. Overall, the final model was significant, F(5, 51)¼7.41, po.001, and explained 42.1% of the variance in MMN amplitude, a large effect. Executive functioning explained the largest unique portion of the variance and memory bordered on a significant unique contribution. To highlight this finding, Fig. A2 visually depicts the direct proportionality between MMN waveform amplitude and TOL performance. 4. Discussion We provide evidence that pre-attentive temporal processing of auditory stimuli is related to executive function and verbal memory. We predicted this relationship because the presumed neural substrates of these functions overlap, specifically within the prefrontal cortex and the temporal lobe (Molholm et al., 2005; Rinne et al., 2000; Tregellas et al., 2006; Tse et al., 2006). Previous studies looking at possible relationships between temporal processing and other cognitive functions have employed exclusively behavioral tasks, making it difficult to determine whether their correlation resulted from specific co-variation of time-based processing or from coincidental co-variation in the cognitive components common to both types of tasks (e.g., working

6 216 S.M. Foster et al. / Neuropsychologia 51 (2013) Table A3 Hierarchical regression of demographic variables entered directly and cognitive variables entered stepwise on MMN amplitude. Model 1st Block 2nd Block 3rd Block B SE b t B SE b t B SE b t (Const.) Gender b n Age Hearing TOL a nn nn Rey Delay n Note: B¼unstandardized coefficient; SE¼standard error; b¼standardized coefficient; TOL¼planning, executive function; Rey Delay¼verbal memory. a Higher scores on these variables represent poorer performance. b Gender coding: 1 (male) and 2 (female). n po.05. nn po.01. Avg. Response (µv) 3 Responses sorted by TOL performace Top Third -2 Mid Third -3 Bottom Third Time Post-Stim. (ms) Fig. A2. MMN to deviant shortened ISI as a function of TOL performance. Grandaveraged waveforms across three groups of 19 participants each: those in the top third of TOL performers, those in the middle third, and those in the bottom third. Note that waveform amplitude increases with better TOL performance. memory) (Barkley et al., 2001; Wittmann, Leland, Churan, & Paulus, 2007). In this study, we avoided this potential confound by assessing temporal processing pre-attentively, in the absence of behavioral responding, through measurement of the MMN waveform elicited by time-dependent stimulus features (Davalos, Kisley, Polk, 2003). The particular pattern of findings described here can be considered relatively robust because they replicate and extend another study conducted with a related approach. Within a subsample of older adults that were part of a larger study (Kisley et al., 2005), we previously described significant correlations between MMN amplitude elicited by deviation in ISI (from a 500 ms standard to a 250 ms deviant) and behavioral measures of executive function and memory. In the current study, we extended these findings in a number of important ways. Because we specifically targeted a larger sample for which all individuals would complete the entire battery of relevant measures, we can be more confident that this pattern of findings does not depend on some property or properties of that original, smaller subsample. Further, we used a more nuanced approach to screening out individuals with potential pathological cognitive decline for the present investigation (a combination of several specific measures of cognitive function here versus the use of only a brief screening measure in the previous study), thus helping to rule out the possibility that these correlations are being driven by individuals with severe cognitive impairment (i.e., outliers ). We also accounted for the potential contribution of demographic variables that may affect MMN amplitude (including gender and hearing loss) in the present study, and found that doing so did not fundamentally alter the general pattern of results. Finally, we employed a more sophisticated analysis (hierarchical regression) that allowed for the separation of unique variance contributed by different cognitive variables, thus increasing our confidence in the statement that pre-attentive time processing is specifically related to executive function separately from its relationship to memory, and vice versa Common neural correlates One potential explanation for the specific pattern of findings here is that the shared variance in executive measures, verbal memory measures, and pre-attentive processing of time arose because of variation in the function of prefrontal cortex across the older adult participants in this study. This would be consistent with past research demonstrating that performance on the RAVLT and related tasks may be dependent on the DLPFC (Chang et al., 2011; Xie et al., 2011), perhaps through executive, taskenhancement strategies associated with the frontal lobe (Higginson et al., 2003; Torralva et al., 2011). Executive planning associated with the TOL task is strongly associated with prefrontal cortex activity as well (Cazalis et al., 2003). This can be compared to the neural recruitment of prefrontal areas, including the DLPFC, during temporal deviance paradigms designed for fmri studies that are similar to the MMN paradigm used in the present study (Molholm et al., 2005; Rinne et al., 2000). In a closely related study, Tregellas et al. (2006) assessed temporal perception using an ongoing train of standard tones (200 ms in duration) occasionally interrupted by deviant tones, which varied from the standard by 15 65% longer or shorter durations ( ms). The stimuli were selected to be brief enough to minimize the recruitment of working memory and are very similar to stimuli used in other duration MMN studies. In this fmri study, they found that the patterns of activation suggested early processing in the STG followed by recruitment of frontal and striatal regions during duration intervals that were perceived as more difficult to detect (e.g., o65% different from the standard). The authors argued that the robust activation of the DLPFC observed in the context of minimal working memory supported involvement of this brain region in time processing. Although the critical role of the prefrontal cortex in all relevant measures included for the present study is well supported by the literature, we cannot rule out the potential importance of variation in temporal cortex function for some of our findings as well. This explanation seems particularly relevant given that Rey Delay explained a significant amount of the variance in MMN amplitude that was not shared with executive functioning (TOL). In addition to its relationship to prefrontal circuits described above, performance on the RAVLT task also depends critically on left temporal lobe integrity (Elger et al., 1997; Foster et al., 2008; Giménez et al., 2004; Grammaldo et al., 2006; Kilpatrick et al., 1997; Sziklas & Jones-Gotman, 2008). Just as for the RAVLT, MMN neuroimaging studies have shown also shown evidence that the temporal

7 S.M. Foster et al. / Neuropsychologia 51 (2013) lobe (in addition to the prefrontal cortex) plays a vital role in preattentive temporal processing when the stimuli are auditory, as they were in the present study (Molholm et al., 2005; Rinne et al., 2000; Tregellas et al., 2006; Tse et al., 2006). Other studies have provided evidence to support the involvement of both temporal and prefrontal networks in pre-attentive processing of auditory stimuli as well. Alain et al. (1998) suggested that a memory trace for auditory stimuli is held in the temporal parietal cortex, and that once a novel stimulus is detected the DLPFC is responsible for involuntarily switching attention to the stimuli. Müller et al. (2002) furthered this work by using positron emission tomography to assess neural activity associated with pre-attentive temporal processing. Regions of significant brain activation in response to deviant tones involved primarily temporal and prefrontal cortex regions (STG, inferior frontal gyrus, middle frontal gyrus, parahippocampal gyrus). These authors also concluded that a frontotemporal network is responsible for processing temporal deviance. It is also worth noting those cognitive measures included in this study that were not correlated with temporal processing. Specifically, the relationship between WCST performance and MMN amplitude was not significant after accounting for the variance associated with the demographic variables. This replicates our previous study (Kisley et al., 2005) and is also consistent with past research using other measurement modalities (Mäntylä et al., 2007). This pattern may arise because executive function tasks with prominent shifting components, such as the WCST, may hinge more critically on parietal lobe function compared to prefrontal cortex function (Collette & Van der Linden, 2002; Gehring & Knight, 2002; Matsui et al., 2007). The relationship between Trails A and B and MMN amplitude was also not significant. Given that Trails A measures attention, it was not expected that there would be a robust relationship between the task and our pre-attentive measure of timing. Trails B is thought to be a measure of divided attention and task switching. Similar to the findings with WCST, it is suspected that Trails B likely tapped in to attentional resources and set shifting skills, both of which appear to tap in to different neural networks than temporal processing. Recent research has associated Trails A and B with white matter functioning, particularly in the right anterior cingulate and left precentral gyrus (Konrad et al., 2012). The reaction time measures also did not uniquely explain variance in preattentive time processing. At the most basic level, there is evidence to suggest that a primarily right hemisphere network, supplemented by medial areas, is involved in maintaining an alert state and responding to visual stimuli (Langer et al., 2012). It may be that this underlying component of all three tasks, including the choice measure associated with processing speed and the conditional measure associated with executive functions, dominated the variance in these tasks decreasing the amount of variance remaining to be associated with temporal processing. Overall, the results of this study suggest that pre-attentive temporal processing of auditory stimuli is related to a frontotemporal network, primarily in the left hemisphere. It does not appear to be strongly influenced by right hemisphere, parietal, or white matter functioning. It should be noted that a limitation of the current study is that an additional MMN paradigm, focusing on a non-temporal deviant, was not included. While the inclusion of such a paradigm would help elucidate the specific relationship of timing stimuli to the frontotemporal network, we drew upon past studies suggesting (a) that there is a distinction between duration-deviant MMN compared to other deviants (e.g. pitch) in terms of areas of the brain that are activated and (b) that duration-deviant MMN are likely more sensitive to detection of disorders characterized by executive dysfunction (Magno et al., 2008; Molholm et al., 2005; Salisbury & McCarley, 2009). Future research utilizing multiple deviant type MMN paradigms would, however, be helpful in further understanding the potentially unique role that timing has in executive functioning Common cognitive correlates The novel results of the current study strongly suggest a link between executive functioning, memory, and temporal processing that cannot be attributed to a simple attentional load shared across all measures, as the neural activity associated with the MMN waveform is pre-attentive (Davalos, Kisley, Ellis, 2003; Kisley et al., 2004; Näätänen et al., 1993; Winkler, 2007). In the behavioral timing literature, past studies have independently suggested links between temporal processing and planning/ sequencing (Mangels & Ivry, 2001), verbal processing/episodic verbal memory (Head, Rodrigue, Kennedy, & Raz, 2008; Montgomery, 2005); and cognitive inhibition (Valko et al., 2009). These studies have, however, utilized disparate measures of temporal processing that arguably involve other cognitive processes (e.g., attention, vigilance, working memory). The current study supported these relationships and allowed for exploration of the relationship between temporal processing and cognition in a considerably larger sample than included in previous studies. The use of a healthy adult populations in the current study was important, as it limits the confounds that often exist when using a clinical sample such as effects of chronicity of disease, possible effects of medication, and, as noted previously, the fact that clinical populations may have other, unmeasured cognitive deficits that affect timing performance. The current study suggests that the relationship between temporal processing and higher-level cognitive processes may exist in the general population as well. These findings are also potentially informative in the context of a recent suggestion that temporal processing may reflect a cognitive primitive (Head et al., 2008). Cognitive primitives are basic neuropsychological processes that have broad influence on other cognitive functions, but cannot be separated into component processes themselves (Salthouse, 1985; Verhaeghen & Salthouse, 1997; Zacks & Hasher, 1994). Head et al. (2008) have argued that temporal processing may be one of the core cognitive primitives, but there has been limited research assessing temporal processing in this role, and to date has only involved behavioral timing measures that likely can be divided into other component processes (e.g., decision making). The current study employed a relatively simple measure of temporal processing independent of other supporting cognitive functions, or at least minimized to a great extent. While additional research is necessary to address these issues, the results of the current study are consistent with the idea that basic temporal processing could be a cognitive primitive in verbal memory and executive functions, including planning and inhibition Summary and conclusion Temporal processing, assessed here by measurement of the pre-attentive MMN waveform elicited by deviation in the ISI, was significantly related to and predicted by cognitive functions associated with temporal and frontal cortices. This supports the frontotemporal model for neural processing of time-based stimulus features. It is also consistent with a more fundamental relationship between temporal processing and other cognitive functions, including verbal memory and executive functions, although further research will be required to determine if these functions are inextricably linked (e.g., if temporal processing is a cognitive primitive for these other functions).