Behavioural Brain Research 322 (2017) 329–338

Contents lists available at ScienceDirect

Behavioural Brain Research

jou rn al hom epage: www.elsev ier .com/ locate /bbr

Research report

Attentional updating and monitoring and affective shifting areimpacted independently by aging in macaque monkeys

Daniel T. Graya,b, Anne C. Smitha, Sara N. Burkec, Adam Gazzaleyd, Carol A. Barnesa,b,e,∗

a Evelyn F. McKnight Brain Institute, University of Arizona, Tucson, AZ 85724, USAb Division of Neural System, Memory & Aging, University of Arizona, Tucson, AZ 85724, USAc Evelyn F. McKnight Brain Institute, Department of Neuroscience, University of Florida, Gainesville, FL 32611, USAd Department of Neurology and Psychiatry, University of California, San Francisco, CA 94158, USAe Department of Psychology, Neurology and Neuroscience, University of Arizona, Tucson, AZ 85724, USA

h i g h l i g h t s

• Macaques show age impairments on both shifting and attentional control.• Monkeys show evidence of retroactive and proactive interference.• Age-related deficits in attentional monitoring and shifting were independent.

a r t i c l e i n f o

Article history:Received 12 March 2016Received in revised form 2 June 2016Accepted 27 June 2016Available online 28 June 2016

Keywords:Executive functionInterferenceCognitive agingNonhuman primate

a b s t r a c t

One hallmark of the normal cognitive aging process involves alterations in executive function. Execu-tive function can be divided into at least three separable components, including set shifting, attentionalupdating and monitoring, and inhibition of prepotent responses. The ability to study the neural basis ofcognitive aging has been enriched by the use of animal models such as the macaque monkey. In agedmacaques, changes in attentional updating and monitoring systems are poorly understood compared tochanges in shifting and inhibition. A partial explanation for this is the fact that the tasks designed tostudy executive function in aged monkeys, to date, primarily have probed shifting and inhibition pro-cesses. Here we examine how aging impacts attentional updating and monitoring processes in monkeysusing an interference task designed after a paradigm used to examine multi-tasking in older humans.Young and aged macaque monkeys were tested on this interference task as well as on an object reversallearning task to study these processes in the same animals. Relative to the young monkeys, aged animalswere impaired on both tasks. Proactive and retroactive interference did not differ between age groupson an array of 40 object pairs presented each day in the object reversal learning task. The levels of per-formance on the interference task were not correlated with levels of performance in the object reversaltask. These results suggest that attentional updating and monitoring and affective shifting are separablefunctions in the macaque, and that normal aging affects these mental operations independently.

© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Goal-directed behaviors allow individuals to deliberately selectactions that accomplish specific objectives, differentiating themfrom habitual behaviors, which are executed regardless of con-text. The mental processes that guide the proper execution ofgoal-directed behaviors, collectively referred to as executive func-

∗ Corresponding author at: Evelyn F. McKnight Brain Institute, University of Ari-zona, Life Science North, Room 355, Tucson, AZ 85724-511, USA.

E-mail address: carol@nsma.arizona.edu (C.A. Barnes).

tion, are thought to be supported by neural networks withinthe mammalian prefrontal cortex [7,15,17]. Declines in executivefunction can be detected in early adulthood, and are among thefirst cognitive impairments to emerge in normative, healthy aging[45,46,51,52,58]. Recent extensions in average lifespans world-wide have led to increases in the number of aged individuals [41],highlighting the importance of understanding the neurobiologicalchanges that underlie age-related declines in executive function.

A number of theories have been proposed to account for thediversity of mental operations considered to be involved in execu-tive function [44,24,12]. For example, Miyake et al. [44] suggestthat executive processes can be divided into at least three sep-

http://dx.doi.org/10.1016/j.bbr.2016.06.0560166-4328/© 2016 Elsevier B.V. All rights reserved.

330 D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338

Fig. 1. Depictions of the delayed nonmatching-to-sample (DNMS) interference paradigm and the object reversal learning paradigm. (A) Monkeys learned a DNMS task ina Wisconsin General Testing Apparatus (WGTA) using trial-unique objects, ensuring that the monkeys did not develop a bias towards any particular object. In the samplephase of the task, a single object was presented over the middle of the three wells. A 30 s delay period followed the sample phase. During this delay a wooden guillotinedoor separated the animal from the wells in the testing apparatus. The test phase followed the delay. In this phase two objects, the sample object and a novel object, werepresented over the lateral two wells of the apparatus. Only the novel object was baited, and the monkey must displace this object to receive food reward. After animals reach

D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338 331

arable functions, including attentional updating and monitoring,set shifting, and inhibition of prepotent responses [23,44]. Lesionand functional imaging studies clearly show that different regionsof the prefrontal cortex operate independently to give rise toeach of these executive functions [18,31,60]. For example, innon-human primates, lesions to the orbitofrontal cortex result inperformance deficits on an object reversal learning task but not anextra-dimensional set-shifting task, and lesions to the dorsolateralprefrontal cortex result in the exact opposite pattern of impair-ment [18]. In the context of normative aging, there is evidencethat distinct executive processes change at different rates withinand between people [22,30], suggesting that distinct prefrontalnetworks experience age-related neurobiological alterations inde-pendently.

Animal models of cognitive aging have enriched our under-standing of the brain alterations that underlie lifespan changes incognition (for reviews see Refs. [7,20,33,63]). The macaque monkeyprovides a particularly valuable model for examining the neuralbasis of cognitive decline during normal aging since: 1) macaquesexperience age-related cognitive decline similar to that of humans(e.g., [20,25,33,58,63]), 2) comparative studies of the cytoarchi-tectonic organization of dorsolateral and ventromedial prefrontalcortical regions in macaques and humans suggest that the twospecies possess numerous homologous frontal cortical fields, withdiscrepancies existing only in the size and delineation of these areas(e.g., [55,54]), and 3) macaques do not develop dementing neu-rodegenerative diseases, allowing for studies of normal aging to becarried out at molecular, anatomical and electrophysiological levelsof analysis (e.g., [33,47,53]).

When cognitive tests similar to those used in humans areadministered to monkeys, like humans, older macaques showclear individual differences, with some monkeys showing minimalimpairment, some showing severe impairment, and most fallingbetween the two ends of this continuum (e.g., [45,46,53,57]). Inter-estingly, tasks designed to assess disparate mental skills within thesame group of aged monkeys suggest that distinct aspects of cogni-tion can be impaired differently within an individual monkey [66].For example, older animals that are impaired on an object-basedtask may not be impaired on spatial versions of the same task,and vice versa [66]. Similarly, Bizon and colleagues have shownthat aged rats with impairments on a spatial working memorytask can perform normally on a set-shifting task, and rats impairedon a set shifting task can perform spatial working memory taskswithout impairment [4]. Together, these results suggest that nor-mative aging alters different cognitive systems independently ofone another within an individual, and the pattern in which differentcognitive systems change with age varies between individuals.

Whether different components of the executive function net-work age independently in non-human primates is not known,and studying this question requires the same individual agedmonkeys to perform multiple executive tasks. To this end, theexperiments reported here examine the performance of youngand aged macaques on two separate tasks that probe the execu-tive processes of attentional updating and monitoring and affectiveshifting. The first task is an interference task adapted from a mul-titasking paradigm in humans used to probe attentional updatingand monitoring [15,13]. This task is designed to deliver differentforms of interference in the delay period of a delayed nonmatching-to-sample task, requiring animals to switch attention from theprimary object recognition task to a secondary task and back.

Basile and Hampton [2] used this paradigm with a matching ruleinstead of a nonmatching rule to show that interference signifi-cantly reduces performance when an active encoding strategy isused. In humans, similar tasks have been shown to engage the lat-eral and medial frontal cortices, but not ventromedial prefrontalareas [15,14,13]. The second task used in this study is an objectreversal learning task, which tests the ability of an animal to alterbehavior following changes in the emotional significance of stim-uli, a process referred to as affective shifting. Lesion and imagingstudies have shown that orbitofrontal cortical, striatal and amyg-dalar networks underlie affective shifting [18,21,38,61,67,39]. Theexecution of both the attentional updating and monitoring andaffective shifting tasks in the same set of monkeys allow us toevaluate whether these separate executive processes are affectedindependently by the aging process or whether there is a com-mon age-related pattern of decline across individual monkeys andexecutive processes.

2. Materials and methods

2.1. Subjects

Six young (mean: 10.5 years) and seven aged (mean: 23.4years) female bonnet macaques (Macaca radiata) participated inthe current study. Every monkey received a semiannual healthassessment from the veterinary staff at the University of Arizona(Tucson, AZ), and no animal included in this study presented withhealth concerns before or during the time of testing. Animalswere paired-housed in a humidity- and temperature-controlledvivarium with a 12 h light-dark cycle and ad libitum food andwater. Prior to testing, all animals underwent behavioral shapingto tolerate restraint in a specialized non-human primate trans-port box (50.8 cm × 31.1 cm × 40 cm), which was used to transportthe monkeys from their home vivarium to the behavioral testingbox (below). The experiments described followed guidelines estab-lished by the National Institutes of Health and were approved bythe Institutional Animal Care and Use Committee at the Universityof Arizona.

2.2. Testing apparatus and task stimuli

A modified Wisconsin General Testing Apparatus (WGTA; [34])was used to acquire all behavioral data (based on Ref. [3]). TheWGTA consists of a box where animals reside during the testingprocedure. At one end of the box vertical bars separate the ani-mal from a tray with three equally spaced wells where stimuluspresentation and reward delivery occur. Two partitions could bemanipulated by the experimenter to control the animal’s accessto and visibility of the stimulus objects and rewards. The firstpartition consisted of a wooden guillotine door, and served toblock the monkeys from both visualizing and interacting withthe objects/rewards. A transparent Plexiglas door was used asthe second partition, which allowed the animals to visualize theobjects/rewards, but not to interact with them. A one way mirrorseparated the experimenter from the animals, allowing their per-formance to be monitored without detection. Stimuli consisted ofplastic toy objects of comparable size (∼8 cm3), and the rewardsused were dry and fresh fruit, vegetables, and sugar free gummybears.

a 90% performance criterion over 5 days, three interference test conditions were implemented during the delay period of the DNMS task. (B) An ‘Interruption Condition’in which the monkey had to perform an object discrimination task during the delay, (C) a ‘Relevant Distraction Condition’ in which an object could be moved to obtain areward, and (D) an ‘Irrelevant Distraction Condition’ in which an object was presented but could not be touched. (E) The object reversal learning task required monkeys tolearn 40 novel object pairs presented sequentially in the same order every session until they reached a 90% performance criterion (Object Discrimination). After reachingcriterion, the object-reward associations were switched (Reversal Learning), and the monkeys re-learned the new associations to the same 90% criterion.

332 D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338

2.3. Behavioral paradigms

2.3.1. Delayed nonmatching-to-sample task with differing levelsof interference

The delayed nonmatching-to-sample (DNMS) task [57] beginswith a single object presented over the center well of the WGTA(Fig. 1A) – this is the “sample phase”. Animals are allowed to dis-place the object to receive a food reward, ensuring that the sampleobject is encoded. Following the sample period, the wooden guillo-tine door is closed for a delay period of 30 s. After the delay period,the sample object is presented alongside a novel object on eachof the lateral wells (this is the “test phase”). Only the novel objectis baited, and the monkey must learn this rule through trial anderror. Objects were presented in a trial-unique manner, ensuringthat monkeys only encountered each object once over the courseof the experiment to avoid any biases that may arise from repeatedobject presentation. Each animal performed 20 trials daily untilreaching a 90% performance criterion over 5 consecutive days. Oncecriterion performance on the DNMS task had been achieved, themonkeys were trained on an object discrimination task with 4 pairsof objects (Fig. 1B). In the object discrimination task the animalsare presented two objects on each of the lateral wells, one baitedand one not baited, and through trial and error the monkey mustlearn the object-reward association of each pair. To maintain a highlevel of performance on the DNMS task during object discrimina-tion training, monkeys performed 20 discriminations daily withthe 4 pairs of objects (presented 5 times for each pair) and 10 trialsof the DNMS task each day, alternating which task came first. Cri-terion performance was reached when the monkey achieved 90%performance on both tasks over 5 days.

After reaching criterion, three forms of interference (describedbelow) were presented during the 30 s delay period of the DNMS:1) an interruption condition (Fig. 1B) in which the monkeys wererequired to complete an object discrimination problem during thedelay interval, forcing animals to shift attention to a secondary taskprior to making the final choice in DNMS, 2) a relevant distrac-tion (Fig. 1C) condition that required animals to displace a singleobject for food reward during the delay period prior to completingthe DNMS, and 3) an irrelevant distraction (Fig. 1D) condition inwhich a transparent Plexiglas barrier was lowered and the monkeycould view a single object behind this barrier, but could not interactwith it nor receive a food reward. The relevant distraction condi-tion served as a motor and food reward control for the interruptioncondition, as both conditions require a motor response and deliverfood reward. Unlike the interruption condition, however, the rel-evant distraction condition does not require attention to objectfeatures for a correct response. The irrelevant distraction conditionin which animals could view the distracting object, but were pre-vented from interacting with it or receiving a food reward, testedthe animal’s ability to ignore task irrelevant stimuli. The stimuli forthe interruption, distraction and irrelevant distraction conditionswere all presented 10 s after the presentation of the DNMS sampleobject. These three stimuli were kept in front of the animal for 10 sbefore being removed, and the DNMS test phase came 10 s after theinterference stimuli were removed. Thus, the total delay period ofthe DNMS with interference task was 30 s in all conditions. Fig. 1Athrough D present pictorial depictions of the DNMS interferenceparadigm.

2.3.2. Object reversal taskThe training protocols and description of the object reversal

learning task are described in detail elsewhere [9]. Briefly, in thefirst phase of the object reversal task, monkeys learned 40 twoobject discrimination problems. Each day the animals were pre-sented with 40 object pairs that each began as novel and werepresented in the same order. Through trial and error the monkeys

determined which object in the pair would always be associatedwith a food reward. At the beginning of each trial, the objectswere hidden from the animals by a wooden partition. Once theguillotine door was lifted the animals were required to displace asingle object and receive the food reward if they selected the cor-rect object. The rewarded and unrewarded objects did not changefor the duration of the object discrimination learning period, andmonkeys performed this task daily until they reached a 90% per-formance criterion over 5 consecutive days (180/200 trials). Afterreaching criterion, the reversal learning component of the task waspresented. During this phase of the task, the rewarded and unre-warded objects in each pair were switched, and the animals wererequired to learn this novel association rule. Again, the monkeysperformed this reversal task until reaching the same 90% criterionover 5 consecutive days after which testing on the object reversaltask was complete. Fig. 1E presents a schematic depiction of theobject reversal paradigm.

2.4. Data analysis

2.4.1. DNMS task with differing levels of interferenceFor each animal, 4 different conditions were administered

within the delay portion of the DNMS task: 1) no interference(delay alone, standard condition; Fig. 1A); 2) interruption withinthe delay (performing a simple object discrimination task, oneobject rewarded, Fig. 1B); 3) relevant distraction (retrieving rewardfrom a single object, this object was always rewarded, Fig. 1C); and4) irrelevant distraction (viewing a single object behind a transpar-ent partition, no reward, Fig. 1D). For each animal, the proportion ofcorrect responses and reaction time in the four separate conditionswere used as performance metrics. To test for the specific effect ofage on the different tasks, a linear mixed-effects model (LMM) wasfit to the data. The proportion of correct responses, Yi,t, for animali and task t was modeled using

Yi,t = b0 + b1 Agei + b2 Taskt + b3 Agei × Taskt + ri (1)

where Agei was a binary categorical variable corresponding toyoung and old age groups and Taskt was a categorical variablecorresponding to the use of no distraction, irrelevant distraction,relevant distraction and interruption conditions, respectively. Arandom effects intercept term, ri ∼N(0, sigmab

2), was includedto account for inter-animal variability. The model was fit using amaximum likelihood approach in MATLAB (MATLAB and Statis-tics Toolbox Release 9.0.0.341360, R2016a; the MathWorks, Inc.,Natick, MA). To check appropriateness of the model, a post-fittingassessment was made such that a plot of the residuals versus fittedvalues showed no trend. To assess the necessity of adding the ran-dom effects intercept term, we also fit a model with no randomeffects term. Comparison of the two models using the theoreti-cal likelihood ratio test indicated that inclusion of random effectswas necessary (p < 2.1e-9). In the text, this model’s statistics arereported in terms of t-statistic, degrees of freedom and a p-value.The same approach was used for the reaction time data (on aslightly reduced data set as there was no reaction time data fortwo of the monkeys). Reaction time was defined as the time fromobject presentation to the time that the animal touched the cho-sen object, and was measured using infrared beams that detectedthe monkeys reach from the holding box to interact with theobjects.

2.4.2. Object reversalFor the object reversal task (Figs. 4–6), the temporal progres-

sion of learning of each object was tracked for each animal as afunction of day, generating 40 learning curves per animal. A state-space modeling approach was used to compute the learning curvesusing Bernoulli observation models [64]. For each animal a learning

D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338 333

curve, its 90% confidence bounds and a learning trial was estimatedfor acquisition of the object discrimination task and its reversal. Thelearning day was defined as the day for which the learning curve (itslower confidence boundary) exceeded and remained above chancefor the remainder of the experiment (Fig. 4). It is this learning daythat is compared across age groups (Figs. 5 and 6). Learning curvemodel estimation was performed using a Bayesian approach [65] inPython using the Monte Carlo package, pymc3, assuming the initialcondition is unknown and applying an exponentially-distributedprior for the variance of the underlying state with parameter,lambda = 50. The results were estimated based on 2000 Monte Carlosamples after a 1000 sample burn-in period.

Once learning day was computed, between-group comparisonswere made using two linear mixed models. The first model forlearning day, Lj, is identical to the model in Eq (1). In the secondmodel, the 40 objects shown per day were grouped into three timeepochs based on their order of presentation. Specifically, the firstfourteen objects were in the first epoch, the middle twelve were inthe second epoch, and the final fourteen were in the third epoch.The average learning trials of each object within these three epochswere then compared on an individual animal basis (i.e. the learningday Li,t,j for animal, i, on task, t, for epoch, j, was modeled by thelinear mixed model)

Li,t,j = b0 + b1 I(Taski,2,j) + b2 I(Epochi,t,2) + b3 I(Epochi,t,3)

+ b4 I(Taski,2,j) × I(Epochi,t,2) + b5I(Taski,2,j)

× I(Epochi,t,3) + ri (2)

where ri ∼N(0, sigmab2) is a random effects term that accounts for

inter-animal variation and bi (i = 0,.,5) are constants estimated bymaximum likelihood. The indicator variable I(Taski,2,j) is a dummyvariable taking the value zero when the OD task is applied and onewhen the RL is applied. The dummy indicator variables I(Epochi,t,2)and I(Epochi,t,3) are zero except during epochs 2 and 3 when theytake the value one, respectively.

2.4.3. Relative impairment scoresFor the DNMS task, the extent of impairment for a given animal

is approximated by the percentage correct in the test phase of theDNMS with interruption, divided by the percentage correct in thetest phase of the DNMS task alone (no interference). For the reversaltask, impairment is measured by subtracting the number of daysrequired for original acquisition of the object discrimination fromthe number of days required to learn the reversal component of thetask (i.e., reversal learning day – object discrimination day for eachanimal).

3. Results

3.1. Results for DNMS task with differing levels of interference

3.1.1. Performance of the DNMS taskYoung and aged monkeys were trained on the DNMS task to

a performance criterion of 90% over 5 consecutive days. Agedmonkeys required more trials to reach this performance crite-rion than did young monkeys, although this difference did notreach statistical significance (aged mean: 1300 trials; aged stan-dard deviation: 590 trials; young mean: 812 trials; young standarddeviation: 13 trials; ANOVA p = 0.13). This pre-training ensuredthat young and aged monkeys exhibited the same levels of per-formance on the DNMS alone condition (LMM, main effect of age,t-stat = 0.70, df = 44, p = 0.49; Fig. 2). Thus, any age differences inperformance of the DNMS task during the interference conditionscould be attributable to effects of those intervening tasks ratherthan to performance differences on the DNMS task itself.

Fig. 2. Performance on the DNMS alone condition in young and old monkeys. Boxesindicate the interquartile range (IQR) with whiskers extending to the most extremedata points that are no more than 1.5 × IQR from the edge of the box. Red line indi-cates median values. Grey circles are jittered so that individual animal performancescores can be more easily seen. There is no difference between age groups afterreaching asymptotic behavioral levels. (For interpretation of the references to colourin this figure legend, the reader is referred to the web version of this article.)

3.1.2. Irrelevant distraction condition impaired DNMSperformance in young and aged monkeys

Performance on the DNMS task in the irrelevant distractioncondition was worse than performance on the DNMS task withno interference (LMM, DNMS-ID, t-stat = 2.6, DF = 44, p = 0.014) inboth age groups (LMM, AGE*DNMS-ID interaction, t-stat = −0.46,DF = 44, p = 0.640; Fig. 3A). For both age groups, reaction timeswere faster in the DNMS alone condition than in the DNMS withirrelevant distraction condition (LMM, DNMS-ID reaction time,p = 0.0001; Fig. 3B).

3.1.3. Relevant distraction condition did not impair DNMSperformance in young and aged monkeys

Performance on the DNMS task in the relevant distraction con-dition was not different from performance on the DNMS taskwith no interference (LMM, DNMS-RD, t-stat = 1.3, DF = 44, p = 0.21),and was not affected by age (LMM, AGE*DNMS-RD interaction, t-stat = −1.7, DF = 44, p = 0.10; Fig. 3C). Reaction times in the DNMSwith relevant distraction condition were not different from thosein the DNMS condition with no interference (LMM, DNMS-RD, t-stat = −1.5, df = 36, p = 0.15; Fig. 3D) and were not affected by age(LMM, AGE*DNMS-RD interaction, t-stat = 1.3, DF = 36, p = 0.21).

3.1.4. Interruption condition impaired aged and young monkeys’performance on the DNMS task

Performance on the DNMS task in the interruption condi-tion was worse than performance on the DNMS task with nointerference (LMM, DNMS-INT, t-stat = 3.9, DF = 44, p = 0.0004),and aged monkeys performed worse than did young monkeys(LMM, AGE*DNMS-INT interaction, t-stat = −2.2, DF = 44, p = 0.032;Fig. 3E). Reaction times in the DNMS with interruption conditionwere not different from the DNMS condition with no interference(LMM, DNMS-INT, t-stat =0.0.043, df = 36, p = 0.96; Fig. 3F) and werenot affected by age (LMM, AGE*DNMS-INT interaction, t-stat = 1.3,DF = 36, p = 0.21).

3.2. Results for object reversal learning task

3.2.1. Aged monkeys are significantly impaired on the objectreversal task

Fig. 4 shows the performance of representative young andaged monkeys on the object discrimination and reversal learn-

334 D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338

Fig. 3. Effects of irrelevant distraction, relevant distraction and interruption conditions on accuracy of choice performance on the DNMS task (Panels A, C, and E) and onreaction time (Panels B, D and F) for all animals. In all plots black circles represent individual young monkeys and light grey squares represent individual aged monkeys.All diagonal lines represent the unity line, where performance or reaction times from both conditions are equal. Note that performance and reaction time after the relevantdistraction condition does not significantly deviate from the unity line in either age group (C, D). For the irrelevant and interruption conditions, performance of both youngand old animals significantly deviated from the unity line (poorer performance; A, E). Reaction times in the irrelevant distraction condition were slower than in the DNMSalone condition (B), whereas reaction times in the interruption condition did not differ from the DNMS alone (F).

Fig. 4. Example performance for one young (top row) and one aged (bottom row) monkey for the object discrimination task (left column) and reversal learning task (rightcolumn). Raw proportion of correct responses per day are shown as grey circles. Blue shaded area indicates the 90% confidence bounds for the learning curves as estimated bya state-space model. The probability of a correct response by chance is indicated by the horizontal red dashed line at 0.5. The estimated learning day (the day that performanceis estimated to be more than 95% greater than chance) is indicated by numbers inside in each figure. (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

ing tasks. Both young and aged animals required more trials toreach the learning criterion during the reversal learning task com-pared to the object discrimination task (LMM, main effect of task,t-stat = −8.7, df = 24, p = 7.6e-9, Fig. 5). Older animals required moretrials to reach learning criterion on both tasks (LMM, main effectof age, t-stat = −3.7, DF = 24, p = 0.001, Fig. 5). The degree of this ageimpairment in the object discrimination and reversal learning taskswere not different (LMM, age*task interaction, t-stat = 0.46, df = 24,p = 0.65).

3.2.2. Evidence for proactive and retroactive interference onobject discrimination and reversal learning

To determine whether objects at the beginning, middle or endof the association pair list were learned differentially, objects weregrouped into three time epochs based on their order of presenta-tion. Specifically, the first fourteen objects were in the first group,the middle twelve were in the second group, and the final fourteenwere in the third group. The average learning trials of each objectwithin these three groups were then compared on an individual

D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338 335

Fig. 5. Estimated learning days for young and aged animals for the object discrimi-nation and reversal learning tasks. Green circles are jittered so that individual animalperformance scores can be more easily seen. Boxes indicate the interquartile range(IQR) and whiskers are extended to the most extreme data points that are no morethan 1.5 × IQR from the edge of the box. Red line indicates median values. Both youngand aged monkeys required more trials to reach the learning criterion in the reversallearning component of the task compared to the object discrimination component.Aged animals required more trials to reach learning criterion in both componentsof the task. (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

Fig. 6. Comparison of learning days for old and young monkeys combined for thefirst 14 objects, the second 12 objects and the final 14 objects of the object pair array(40 pairs total). (A) object discrimination and (B) reversal learning tasks. Box andwhiskers indicate the interquartile range (IQR) and the most extreme data pointsthat are no more than 1.5 × IQR from the edge of the box, respectively. Red linesindicates median values. Grey circles are jittered so that individual animal perfor-mance scores can be more easily seen. Note that in both cases, the middle third ofobjects were acquired faster than objects at the beginning or end of the list. (Forinterpretation of the references to colour in this figure legend, the reader is referredto the web version of this article.)

animal basis. During both tasks, the second and third group ofobjects were learned significantly faster than were the first objects(LMM, main effects of second (third) epoch, tstat = −5.4(−3.8),DF = 1114, p = 9.9e-8 (1.3e-4); Fig. 6). Additionally, the middle thirdof objects were learned significantly faster than the last third ofobjects (LMM, Coef. Test, F = 1.6e + 01, DF = 1108; p = 7.1e-5; Fig. 6).There was no difference between age groups in any of these mea-surements. These results suggest that young and aged monkeys did

not show either a ‘primacy’ or ‘recency’ effect on learning theseassociations, but rather experience some level of proactive andretroactive interference during associative and reversal learning(see discussion).

3.3. Performance on the DNMS interruption condition wasindependent of performance on the object reversal task

To examine whether those monkeys with poor performance onthe DNMS interruption condition also were particularly impairedon the object reversal task, we obtained relative impairment scoresfor every individual animal on both tasks (described in Materialsand methods). There was no significant relationship between theseperformance measures in either young or aged animals (young:r = −0.22 p = 0.67, aged: r = −0.16 p = 0.73; Fig. 7).

4. Discussion

4.1. Novel findings

The main novel findings of the present study are that 1) agedmacaques exhibit greater performance deficits than do youngmacaques in an object recognition task when interference requir-ing attentional shifts intervene; 2) both young and aged monkeysexhibit proactive and retroactive interference during an affectiveshifting task; and 3) deficits in attentional updating and monitor-ing are independent of deficits in affective shifting. Together theseresults indicate that aging impacts distinct components of execu-tive function independently.

4.2. Interruptions during the delay of a DNMS task negativelyimpact aged monkeys more than young monkeys

Previous studies [2] have shown that young monkeys areimpaired on a delayed matching-to-sample task with interferenceonly when the choice is actively maintained in memory, but not

Fig. 7. Performance on the reversal learning task compared with performance onthe DNMS with interruption task. Relative impairment scores for the object reversallearning task were obtained by subtracting the learning day on the object discrim-ination task from the learning day on the reversal learning task (y-axis). Relativeimpairment scores for the DNMS with interruption task were obtained by divid-ing the performance on the DNMS with interruption condition by performances onthe DNMS alone condition (x-axis). Young and old animals’ data are indicated withgrey squares and black circles, respectively. Trend lines for the young and old dataare in grey and black, respectively. Note that there is no relationship between theimpairment scores on these two tasks.

336 D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338

when a passive maintenance strategy is used. The present studyused the nonmatching rule with interference to test the ability ofmonkeys to disengage from and re-engage with a memory rep-resentation during conditions akin to multitasking (Fig. 1A). Bothyoung and aged monkeys were impaired on the DNMS task wheninterruptions (Fig. 3E) and irrelevant distractions (Fig. 3A) werepresented during the delay, suggesting that the animals may haveused an active maintenance strategy in this task. Aged animals werenegatively affected by interruptions more than were young mon-keys, but equally affected by irrelevant distractions. Differencesin the attentional requirements in the interruption and irrele-vant distraction conditions may explain these effects (discussedbelow).

Gazzaley and colleagues have shown that different forms ofattention are altered uniquely during normative aging. For exam-ple, unimodal sensory processing and attention are degradedin aged human subjects, while similar demands presented in acrossmodal context do not elicit age-related impairments [15,43].Completion of the DNMS with interruption condition and DNMSwith irrelevant distraction condition require different aspects ofattentional updating and monitoring to complete. The interruptioncondition requires animals to shift attention from the DNMS taskto the object discrimination task and back. The irrelevant distrac-tion condition does not require a shift in attention, but does requireanimals to suppress attention towards the task irrelevant stimuli.The current data suggest that, in macaques, attentional shiftingis negatively impacted by aging, while attentional suppression isnot.

In humans, the neural structures implicated in attention shiftingand suppression involve distinct regions of the prefrontal cor-tex [15,10,26]. Clapp et al. [15] demonstrated that the middlefrontal gyrus of the dorsolateral prefrontal cortex functionally con-nects with different visual cortical regions during attentional shiftsbetween tasks. Conversely, top-down modulation of sensory infor-mation appears to arise from the medial prefrontal cortex, anteriorcingulate cortex, parietal cortex, and the medial temporal lobe (e.g.,[10,16,6,5,8,26,59].

Functional connectivity analyses in human subjects perform-ing a task similar to the one described here, indicate that agedsubjects, compared to young, do not show dynamic switchingbetween functional brain networks in lateral frontal cortex andhigher-order visual cortical regions [15]. While the brains of olderindividuals were able to switch to higher-order visual networksrepresenting the interruption condition, their brains effectivelybecame stuck there, and failed to switch back in a network statethat represented the primary task. This deficit might be explainedby the concept of backward inhibition, or the inability to dis-engage from a recently completed task. In fact, aged humanshave been shown to be impaired on tasks requiring this process[35,36,42].

4.3. Implications for cohort effects in human studies

Older humans tend to have less exposure to technology thando younger individuals, and it has been proposed that cohortdifferences in technological abilities contribute to age-related dif-ferences in the performance on cognitive tasks (for review see Ref.[11]). One strength of the current study is the ability to control forgenerational effects by ensuring that young and old age groups areexposed to similar testing and living conditions over their lifespan.The detrimental effects of interference on object recognition mem-ory in older nonhuman primates argues that multitasking deficitsobserved in older humans do not simply reflect cohort effects.Rather, the observed multitasking deficits most likely arise from

dysfunction of prefrontal cortical circuits that occurs during nor-mative biological aging (e.g. [15,13]).

4.4. Aged macaques are impaired at affective shifting

Reversal learning paradigms have been used to study execu-tive function in macaques for decades (e.g., [1,19,56]). The mentalprocess primarily probed in the reversal learning task is affec-tive shifting, which requires unlearning an initial stimulus-rewardassociation and replacing it with a new association when the rein-forcement is switched. Performance on this task is thought to reflectthe ability to alter behavior based on changes in the emotionalvalence of stimuli, and has been suggested to provide a measureof cognitive flexibility [33]. Compared to young monkeys, agedmonkeys have been shown to require more trials to reach learningcriterion during the reversal learning component of the task, but areunimpaired in learning the initial associations during the object dis-crimination component of the task [1,19,33,56]. The current groupof aged monkeys performed worse than young monkeys on boththe object discrimination and reversal learning components of thetasks (Fig. 5), suggesting an impairment in both associative learn-ing and affective shifting. Species differences (bonnet macaque vsrhesus macaque), more sensitive analytic methods used here, orprior exposure to complex tasks in the present animals (e.g. DNMSwith interference task) could account for the discrepancy betweenthe current data and previous findings [1,19,33,56].

While the full extent of the neural networks underlying objectreversal learning are still being elucidated, both imaging and lesionstudies performed in humans and nonhuman primates have con-sistently implicated the orbitofrontal cortex (Walker’s areas 11,13,and 14) and the medial prefrontal cortex (e.g., [21,38,61,67]). Inaddition to these cortical regions, the dorsal striatum, ventralstriatum and basolateral nucleus of the amygdala have also con-sistently been implicated as non-cortical brain areas involved inreversal learning (for reviews see Ref. [27,39]). Importantly, it isbecoming clear that these individual brain regions do not sup-port reversal learning in isolation. Rather, it is the interaction ofthis cortical-striatal-amygdalar network that is required. For exam-ple, Rudebeck and colleagues demonstrated that fiber-sparinglesions of the orbitofrontal cortex in macaques does not impairobject reversal learning [62], suggesting that the white matter,but not grey matter in this region is necessary for reversal learn-ing. Whether these white matter tracts connect subcortical regionswith the orbitofrontal cortex or are fibers of passage remains to beclarified.

Interestingly, the volume of the orbitofrontal cortex was exam-ined in the same cohort of macaques used in the present study,and the number of perseverative errors on the reversal learningtask did not correlate with orbitofrontal cortex grey matter volume[9]. Orbitofrontal cortex grey matter volume did, however, corre-late with performance on a reward devaluation test. This indicatesthat the orbitofrontal cortex plays a role in updating the value ofstimuli, which is a key component of the reversal learning task,but a separate cognitive operation from affective shifting. Underthis view, the neural networks that underlie affective shifting,perhaps located in non-cortical areas such as the striatum or baso-lateral nucleus of the amygdala, use the stimulus-associated rewardinformation provided by the orbitofrontal cortex during reversallearning.

4.5. Proactive and retroactive interference during associative andreversal learning

In the object reversal learning paradigm, 40 total object pairswere presented daily in the exact same order, allowing for an anal-ysis of serial position effects on the acquisition and reversal of

D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338 337

these associations. For example, there is a well-known bias towardsremembering items both at the beginning (primacy effect) andat the end (recency effect) of a list (e.g., [48,28]). Somewhat sur-prisingly, in the present study, we did not observe a primacy or arecency effect in either the acquisition or reversal components ofthe task. Rather, objects encountered in the middle third of trials oneach day were learned faster than were objects encountered in thefirst and last third of trials. This may suggest over-riding effects ofproactive interference (when learned information disrupts the abil-ity to learn subsequently presented information) and retroactiveinterference (when newly formed associations disrupt the recall ofpreviously encoded information) in these types of associative learn-ing and affective shifting tasks. In the present study, associationsformed at the beginning of a session likely experience the great-est amount of retroactive interference since the majority of objectpairs follow them. The opposite trend would be expected for theeffects of proactive interference on the array of object pairs. Thesetwo forms of interference are thought to arise through differentneural mechanisms and would be expected to exert their effectsindependently of one another (e.g., [37,49,40,50]). Interestingly,the effect sizes of the ‘middle-list advantage’ were not differentbetween young and aged monkeys, suggesting that there is a simi-lar susceptibility to proactive and retroactive interference in theseanimals.

To explain the present data, the effects of proactive and retroac-tive interference cannot simply be additive – rather, they mustexert stronger influence at the beginning and ends of the lists,and this may interact with the number of object-pair associationsencountered to produce the middle-list advantage. The finding thatold and young monkeys are equally prone to the effects of proactiveand retroactive interference in the context of associative learningand its reversal, is in contrast to the age deficit observed on theDNMS task with interruption (Fig. 3). This dissociation provides fur-ther support for the idea that the networks that underlie the DNMSwith interference task are separable from those involved in objectreversal learning. This could either imply completely independentcircuits, or that redundant cortical systems exist to compensate forage-related brain changes in the latter task but are not recruited inthe former.

4.6. Deficits in attention shifting and affective shifting do notcorrelate

Performance on the DNMS task with interruption did not corre-late, within an individual monkey, with performance levels on theobject reversal task. While these results require replication, theyare consistent with the idea that attentional updating and monitor-ing and affective shifting are independently affected by normativeaging [23,44]. From a neurobiological perspective, dissociationssuch as these suggest that the pattern of network dysfunction isvariable between individuals, and such variability can be noted atmolecular, anatomical, electrophysiological and behavioral levels[32]. Further work will be required to specifically define the cir-cuits underlying different components of executive function. Thepresent results suggest, however, that aged monkeys with affec-tive shifting deficits should have distinct patterns of age-relatednetwork alteration compared to animals with attentional updatingand monitoring impairments. This is reminiscent of the suggestionmade by Glisky and colleagues that a major challenge in cognitiveaging research is to identify the unique behavioral and neurobio-logical profiles of individuals [22,30,29]. A better understanding ofthe unique brain-behavior relationship of each individual result-ing in memory loss during normative aging may facilitate thedevelopment of more targeted treatment approaches for neurode-generative diseases such as Alzheimer’s and Parkinson’s diseases.

Acknowledgements

The authors would like to thank Kojo Plange for animal hus-bandry and testing, Kate Andersh and Wonn Pyon for data entry,and Luann Snyder and Michelle Carroll for administrative support.This work was supported by NIH Grant R01 AG050548 and McK-night Brain Research Foundation.

References

[1] R.T. Bartus, R.L. Dean, D.L. Fleming, Aging in the rhesus monkey: effects onvisual discrimination learning and reversal learning, J. Gerontol. 34 (1979)209–219.

[2] B.M. Basile, R.R. Hampton, Dissociation of active working memory and passiverecognition in rhesus monkeys, Cognition 126 (2013) 391–396.

[3] M.G. Baxter, A. Parker, C.C.C. Lindner, A.D. Izquierdo, E.A. Murray, Control ofresponse selection by reinforcer value requires interaction of amygdala andorbital prefrontal cortex, J. Neurosci. 20 (2000) 4311–4319.

[4] B.S. Beas, B. Setlow, J.L. Bizon, Distinct manifestations of executivedysfunction in aged rats, Neurobiol. Aging 34 (2013) 2164–2174, http://dx.doi.org/10.1016/j.neurobiolaging.2013.03.019.

[5] M. Behrmann, J.J. Geng, S. Shomstein, Parietal cortex and attention, Curr. Opin.Neurobiol. 14 (2) (2004) 212–217.

[6] J.M. Birrell, V.J. Brown, Medial frontal cortex mediates perceptual attentionalset shifting in the rat, J. Neurosci. 20 (11) (2000) 4320–4324.

[7] J.L. Bizon, T.C. Foster, G.E. Alexander, E.L. Glisky, Characterizing cognitiveaging of working memory and executive function in animal models, Front.Aging Neurosci. 4 (2012), http://dx.doi.org/10.3389/fnagi.2012.00019.

[8] T.J. Buschman, E.K. Miller, Top-down versus bottom-up control of attention inthe prefrontal and posterior parietal cortices, Science 315 (5820) (2007)1860–1862.

[9] S.N. Burke, A. Thome, K. Plange, J.R. Engle, T.P. Trouard, K.M. Gothard, C.A.Barnes, Orbitofrontal cortex volume in area 11/13 predicts rewarddevaluation but not reversal learning performance, in young and agedmonkeys, J. Neurosci. 34 (2014) 9905–9916, http://dx.doi.org/10.1523/JNEUROSCI.3918-13.2014.

[10] J.Z. Chadick, T.P. Zanto, A. Gazzaley, Structural and functional differences inmedial prefrontal cortex underlie distractibility and suppression deficits inageing, Nat. Commun. 5 (2014) 4223, http://dx.doi.org/10.1038/ncomms5223.

[11] N. Charness, W.R. Boot, Aging and information technology use potential andbarriers, Curr. Dir. Psychol. Sci. 18 (5) (2009) 253–258.

[12] C.H. Chatham, S.A. Herd, A.M. Brant, T.E. Hazy, A. Miyake, R. O’Reilly, N.P.Friedman, From an executive network to executive control: a computationalmodel of the n-back task, J. Cogn. Neurosci. 23 (November) (2011)3598–3619, http://dx.doi.org/10.1162/jocn a 00047.

[13] W.C. Clapp, A. Gazzaley, Distinct mechanisms for the impact of distractionand interruption on working memory in aging, Neurobiol. Aging 33 (2012)134–148, http://dx.doi.org/10.1016/j.neurobiolaging.2010.01.012.

[14] W.C. Clapp, M.T. Rubens, A. Gazzaley, Mechanisms of working memorydisruption by external interference, Cereb. Cortex (2009), http://dx.doi.org/10.1093/cercor/bhp150.

[15] W.C. Clapp, M.T. Rubens, J. Sabharwal, A. Gazzaley, Deficit in switchingbetween functional brain networks underlies the impact of multitasking onworking memory in older adults, Proc. Natl. Acad. Sci. U. S. A. 108 (2011)7212–7217, http://dx.doi.org/10.1073/pnas.1015297108.

[16] C.L. Colby, M.E. Goldberg, Space and attention in parietal cortex, Annu. Rev.Neurosci. 2 (1999) 319–349.

[17] M. D’Esposito, J.A. Detre, D.C. Alsop, R.K. Shin, S. Atlas, M. Grossman, Theneural basis of the central executive system of working memory, Nature 378(1995) 279–281, http://dx.doi.org/10.1038/378279a0.

[18] R. Dias, T.W. Robbins, A.C. Roberts, Primate analogue of the Wisconsin cardsorting test: effects of excitotoxic lesions of the prefrontal cortex in themarmoset, Behav. Neurosci. 110 (5) (1996) 872–886.

[19] I. Divac, H.E. Rosvold, M.K. Szwarcbart, Behavioral effects of selective ablationof the caudate nucleus, J. Comp. Physiol. Psychol. 63 (1967) 184–190.

[20] C.A. Erickson, C.A. Barnes, The neurobiology of memory changes in normalaging, Exp. Gerontol. 38 (2003) 61–69.

[21] L.K. Fellows, M.J. Farah, Ventromedial frontal cortex mediates affectiveshifting in humans: evidence from a reversal learning paradigm, Brain J.Neurol. 126 (2003) 1830–1837, http://dx.doi.org/10.1093/brain/awg180.

[22] M.A. Fernandes, P.S.R. Davidson, E.L. Glisky, M. Moscovitch, Contribution offrontal and temporal lobe function to memory interference from dividedattention at retrieval, Neuropsychology 18 (2004) 514–525, http://dx.doi.org/10.1037/0894-4105.18.3.514.

[23] J.E. Fisk, C.A. Sharp, Age-related impairment in executive functioning:updating, inhibition, shifting, and access, J. Clin. Exp. Neuropsychol. 26 (2004)874–890, http://dx.doi.org/10.1080/13803390490510680.

[24] N.P. Friedman, A. Miyake, R.P. Corley, S.E. Young, J.C. DeFries, J.K. Hewitt, Notall executive functions are related to intelligence, Psychol. Sci. 17 (2) (2006)172–179.

[25] M. Gallagher, P.R. Rapp, The use of animal models to study the effects of agingon cognition, Annu. Rev. Psychol. 48 (1997) 339–370, http://dx.doi.org/10.1146/annurev.psych.48.1.339.

338 D.T. Gray et al. / Behavioural Brain Research 322 (2017) 329–338

[26] A. Gazzaley, M. D’esposito, Top-down modulation and normal aging, Ann. N.Y. Acad. Sci. 1097 (2007) 67–83, http://dx.doi.org/10.1196/annals.1379.010.

[27] A. Genovesio, S. Tsujimoto, From duration and distance comparisons to goalencoding in prefrontal cortex, Adv. Exp. Med. Biol. 829 (2014) 167–186,http://dx.doi.org/10.1007/978-1-4939-1782-2 10.

[28] M. Glanzer, A.R. Cunitz, Two storage mechanisms in free recall, J. Verbal Learn.Verbal Behav. 5 (4) (1966) 351–360.

[29] E.L. Glisky, M.R. Polster, B.C. Routhieaux, Double dissociation between itemand source memory, Neuropsychology 9 (1995) 229–235, http://dx.doi.org/10.1037/0894-4105.9.2.229.

[30] E.L. Glisky, S.R. Rubin, P.S. Davidson, Source memory in older adults: anencoding or retrieval problem? J. Exp. Psychol. Learn. Mem. Cogn. 27 (2001)1131–1146.

[31] O. Godefroy, M. Cabaret, V. Petit-Chenal, J.P. Pruvo, M. Rousseaux, Controlfunctions of the frontal lobes. Modularity of the central-supervisory system?Cortex J. Devoted Study Nerv. Syst. Behav. 35 (1999) 1–20.

[32] D.T. Gray, C.A. Barnes, Distinguishing adaptive plasticity from vulnerability inthe aging hippocampus, Neuroscience 309 (2015) 17–28, http://dx.doi.org/10.1016/j.neuroscience.2015.08.001.

[33] Y. Hara, P.R. Rapp, J.H. Morrison, Neuronal and morphological bases ofcognitive decline in aged rhesus monkeys, Age 34 (2012) 1051–1073, http://dx.doi.org/10.1007/s11357-011-9278-5.

[34] H.F. Harlow, J.A. Bromer, A test apparatus for monkeys, Psychol. Rec. 19(1938) 434–436.

[35] L. Hasher, R.T. Zacks, T.A. Rahhal, Timing, instructions, and inhibitory control:some missing factors in the age and memory debate, Gerontology 45 (1999)355–357, 22121.

[36] M. Hübner, G. Dreisbach, H. Haider, R.H. Kluwe, Backward inhibition as ameans of sequential task-set control: evidence for reduction of taskcompetition, J. Exp. Psychol. Learn. Mem. Cogn. 29 (2003) 289–297.

[37] K. Irlbacher, A. Kraft, S. Kehrer, S.A. Brandt, Mechanisms and neuronalnetworks involved in reactive and proactive cognitive control of interferencein working memory, Neurosci. Biobehav. Rev. 46 (Pt. (1)) (2014) 58–70,http://dx.doi.org/10.1016/j.neubiorev.2014.06.014.

[38] S.D. Iversen, M. Mishkin, Perseverative interference in monkeys followingselective lesions of the inferior prefrontal convexity, Exp. Brain Res. 11 (1970)376–386.

[39] A. Izquierdo, J.L. Brigman, A.K. Radke, P.H. Rudebeck, A. Holmes, The neuralbasis of reversal learning: an updated perspective, Neuroscience (2016),http://dx.doi.org/10.1016/j.neuroscience.2016.03.021, pii:S0306-4522(16)00244-X.

[40] J. Jonides, D.E. Nee, Brain mechanisms of proactive interference in workingmemory, Neuroscience 139 (1) (2006) 181–193, Epub 2005 Dec 5.

[41] G.W. Leeson, G.W. Leeson, Increasing longevity and the new demography ofdeath, increasing longevity and the new demography of death, Int. J. Popul.Res. Int. J. Popul. Res. 2014 (2014) e521523, http://dx.doi.org/10.1155/2014/521523,10.1155/2014/521523.

[42] U. Mayr, S.W. Keele, Changing internal constraints on action: the role ofbackward inhibition, J. Exp. Psychol. Gen. 129 (2000) 4–26.

[43] J. Mishra, A. Gazzaley, Preserved discrimination performance and neuralprocessing during crossmodal attention in aging, PLoS One 8 (2013), http://dx.doi.org/10.1371/journal.pone.0081894.

[44] A. Miyake, N.P. Friedman, M.J. Emerson, A.H. Witzki, A. Howerter, T.D. Wager,The unity and diversity of executive functions and their contributions tocomplex frontal lobe tasks: a latent variable analysis, Cognit. Psychol. 41(2000) 49–100, http://dx.doi.org/10.1006/cogp.1999.0734.

[45] T.L. Moore, R.J. Killiany, J.G. Herndon, D.L. Rosene, M.B. Moss, Executivesystem dysfunction occurs as early as middle-age in the rhesus monkey,Neurobiol. Aging 27 (2006) 1484–1493, http://dx.doi.org/10.1016/j.neurobiolaging.2005.08.004.

[46] T.L. Moore, R.J. Killiany, J.G. Herndon, D.L. Rosene, M.B. Moss, Impairment inabstraction and set shifting in aged rhesus monkeys, Neurobiol. Aging 24(2003) 125–134.

[47] J.H. Morrison, M.G. Baxter, The ageing cortical synapse: hallmarks andimplications for cognitive decline, Nat. Rev. Neurosci. 13 (2012) 240–250,http://dx.doi.org/10.1038/nrn3200.

[48] B.B. Murdock, The serial position effect of free recall, J. Exp. Psychol. 64 (5)(1962) 482–488.

[49] D.E. Nee, J. Jonides, Dissociable interference-control processes in perceptionand memory, Psychol. Sci. 19 (2008) 490–500, http://dx.doi.org/10.1111/j.1467-9280.2008.02114.x.

[50] D.E. Nee, J. Jonides, M.G. Berman, Neural mechanisms of proactiveinterference-resolution, Neuroimage 38 (2007) 740–751, http://dx.doi.org/10.1016/j.neuroimage.2007.07.066.

[51] D.C. Park, G. Lautenschlager, T. Hedden, N.S. Davidson, A.D. Smith, P.K. Smith,Models of visuospatial and verbal memory across the adult life span, Psychol.Aging 17 (2002) 299–320.

[52] D.C. Park, P. Reuter-Lorenz, The adaptive brain: aging and neurocognitivescaffolding, Annu. Rev. Psychol. 60 (2009) 173–196.

[53] A. Peters, D.L. Rosene, M.B. Moss, T.L. Kemper, C.R. Abraham, J. Tigges, M.S.Albert, Neurobiological bases of age-related cognitive decline in the rhesusmonkey, J. Neuropathol. Exp. Neurol. 55 (1996) 861–874, http://dx.doi.org/10.1097/00005072-199608000-00001.

[54] M. Petrides, D.N. Pandya, Dorsolateral prefrontal cortex: comparativecytoarchitectonic analysis in the human and the macaque brain andcorticocortical connection patterns, Eur. J. Neurosci. 11 (1999) 1011–1036.

[55] M. Petrides, F. Tomaiuolo, E.H. Yeterian, D.N. Pandya, The prefrontal cortex:comparative architectonic organization in the human and the macaquemonkey brains, Cortex J. Devoted Study Nerv. Syst. Behav. 48 (2012) 46–57,http://dx.doi.org/10.1016/j.cortex.2011.07.002.

[56] P.R. Rapp, Visual discrimination and reversal learning in the aged monkey(Macaca mulatta), Behav. Neurosci. 104 (1990) 876–884.

[57] P.R. Rapp, D.G. Amaral, Recognition memory deficits in a subpopulation ofaged monkeys resemble the effects of medial temporal lobe damage,Neurobiol. Aging 12 (1991) 481–486.

[58] P.R. Rapp, D.G. Amaral, Evidence for task-dependent memory dysfunction inthe aged monkey, J. Neurosci. 9 (1989) 3568–3576.

[59] J. Rissman, A. Gazzaley, M. D’Esposito, The effect of non-visual workingmemory load on top-down modulation of visual processing,Neuropsychologia 47 (2009) 1637–1646, http://dx.doi.org/10.1016/j.neuropsychologia.2009.01.036.

[60] T.W. Robbins, D. Weinberger, J.G. Taylor, R.G. Morris, Dissociating executivefunctions of the prefrontal cortex [and discussion], Philos. Trans. R. Soc. Lond.B Biol. Sci. 351 (1996) 1463–1471, http://dx.doi.org/10.1098/rstb.1996.0131.

[61] R.D. Rogers, T.C. Andrews, P.M. Grasby, D.J. Brooks, T.W. Robbins, Contrastingcortical and subcortical activations produced by attentional-set shifting andreversal learning in humans, J. Cogn. Neurosci. 12 (2000) 142–162.

[62] P.H. Rudebeck, R.C. Saunders, A.T. Prescott, L.S. Chau, E.A. Murray, Prefrontalmechanisms of behavioral flexibility, emotion regulation and value updating,Nat. Neurosci. 16 (8) (2013) 1140–1145, http://dx.doi.org/10.1038/nn.3440,Epub 2013 Jun 23.

[63] R.D. Samson, C.A. Barnes, Impact of aging brain circuits on cognition, Eur. J.Neurosci. 37 (2013) 1903–1915, http://dx.doi.org/10.1111/ejn.12183.

[64] A.C. Smith, L.M. Frank, S. Wirth, M. Yanike, D. Hu, Y. Kubota, A.M. Graybiel,W.A. Suzuki, E.N. Brown, Dynamic analysis of learning in behavioralexperiments, J. Neurosci. 24 (2004) 447–461, http://dx.doi.org/10.1523/JNEUROSCI.2908-03.2004.

[65] A.C. Smith, S. Wirth, W.A. Suzuki, E.N. Brown, Bayesian analysis of interleavedlearning and response bias in behavioral experiments, J. Neurophysiol. 97(2007) 2516–2524, http://dx.doi.org/10.1152/jn.00946.2006.

[66] M.L. Voytko, Impairments in acquisition and reversals of two-choicediscriminations by aged rhesus monkeys, Neurobiol. Aging 20 (1999)617–627.

[67] E.Z. Wheeler, L.K. Fellows, The human ventromedial frontal lobe is critical forlearning from negative feedback, Brain J. Neurol. 131 (2008) 1323–1331,http://dx.doi.org/10.1093/brain/awn041.