Oily fish are the major dietary source of the n-3 LC PUFAs, EPA and DHA. A consistent finding across studies is that higher intake of fish is related to less cognitive decline 9 10 11 12 and incidence of dementia 10 13 14 in prospective studies, and associated with better cognitive performance in non-clinical samples 11 15 16 17 , in cross-sectional analyses. However, variable associations have been found between dietary intake levels of n-3 LC PUFAs and cognitive outcomes; only a handful of the aforementioned studies that also examined relationships between cognitive outcomes and dietary intake levels of n-3 LC PUFAs have found significant positive relationships 9 14 15 .

Stronger evidence of a positive relationship between cognitive outcomes in older age and levels of n-3 LC PUFAs has been found in studies assessing plasma or erythrocyte levels of n-3 PUFAs. In these studies, DHA and EPA levels, either individually or in combination, have been associated with better cognitive function in normal older adults in cross-sectional analyses 18 19 , and in prospective studies, with better cognitive outcomes over time 19 or reduced risk of cognitive decline 20 , and lower risk of developing dementia 21 ; in contrast, Laurin et al 22 reported a negative relationship between cognition and biomarkers of n-3 LC PUFAs.

There have been six published double-blind RCTs of n-3 LC PUFAs on cognitive functioning in the elderly within both clinical and non-clinical populations 23 24 25 26 27 28 . Only two studies have been conducted with cognitively healthy populations, screened for dementia at baseline, and no overall effect of supplementation was found. The "Older People and n-3 Long-chain polyunsaturated fatty acids" study 24 supplemented 867 cognitively healthy participants aged 70-79 years, with either 200 mg of EPA and 500 mg of DHA, or olive oil, daily for 24 months. After adjustment for confounders, there was no significant difference between the groups in performance on the California Verbal Learning Test, the main cognitive outcome measure, or any secondary cognitive outcomes (a global cognitive score, and domain scores on memory, processing speed, executive function, and delayed recall). In a 26-week RCT of fish oil on cognitive performance in 302 cognitively healthy adults aged 65+ years 27 , no main effect of supplementation was seen for any of the cognitive domains (attention, sensorimotor speed, memory, and executive function), assessed by 5 tests. However, subgroup analysis revealed that carriers of the Apolipoprotein E (ApoE) ε4 allele, in both the low (226 mg EPA, 176 mg DHA) and high dose (1093 mg EPA, 847 mg DHA) fish oil groups, showed an improvement after 26 weeks compared with the placebo on the Digit-Span Forward task, their marker of attention.

In three of the studies with cognitively impaired (including Alzheimer's disease (AD)) samples, the interventions had a beneficial effect on cognitive outcomes but only in participants with mild cognitive dysfunction 23 25 28 . This finding was not replicated in a large RCT conducted over 18 months of 402 participants with mild to moderate AD. However, in this study in ApoE-ε4 non-carriers only, there was a positive effect on two cognitive outcome measures 26 .

The presence or absence of the Apolipoprotein E-ε4 allele appears to be a potentially important modifier of relationships between n-3 LC PUFAs and older-age cognitive outcomes but there are conflicting findings regarding the direction of this interaction. Epidemiological studies that have reported a differential relationship between cognitive outcomes and measures of n-3 LC PUFAs for carriers versus non-carriers of the ApoE-ε4 allele have found a beneficial relationship present only in non-carriers 19 29 . However, the two RCTs that have included examination of the effect of ApoE-ε4 allele carriage in sub-analyses have reported a beneficial effect of supplementation dependent on ApoE-ε4 allele status, but differ on which group benefits from supplementation 26 27 . Nonetheless, ApoE polymorphisms have been found to influence lipid responses to fish oil supplementation 30 and in a small group of males, those without ApoE-ε4 showed a greater increase in DHA and EPA after fish-oil supplementation 31 . Further, possession of the ApoE-ε4 allele is associated with worse cognitive performance in old age 32 and greater risk of cognitive decline and dementia 33 , hence, even in cognitively healthy samples screened for dementia, the genesis of any cognitive change may differ between carriers and non-carriers, providing another potential avenue for differential effects of supplementation on cognitive outcomes due to ApoE genotype.

These studies suggest that n-3 LC PUFA intake, in particular DHA, has a beneficial effect on cognition in older adults. However, the majority of evidence for the benefits of n-3 PUFA on cognitive functioning in adults is associational. Furthermore, work to date in this area has often not comprehensively assessed cognitive function or cognitive decline. Without a priori expectations of specific cognitive domains that will be affected, a comprehensive operationalisation of cognition is required in order to ascertain whether a null result is due to the absence of a relationship or effect, or the relevant cognitive domains were not assessed. Given the potential importance of ApoE-ε4 allele carriage as a determinant of n-3 LC PUFAs influence upon cognitive outcomes in older age, more randomised controlled trials are needed that incorporate consideration of its impact in order to determine whether n-3 LC PUFAs can influence age-related cognitive change in older adults.

The trial was a parallel, 18-month, randomised, double-blind, placebo-controlled trial with repeated measures every 6 months, totalling 4 measurement points. Because the primary hypothesis related to change in a continual developmental process, multiple repeated measures enable determination of the rate of change on the measures, allowing assessment of an intervention effect on a trajectory, rather than effects at a specific time point. Decline of the cognitive abilities vulnerable to ageing effects begins by the early 30s so although participants were screened for cognitive impairment, all were expected to experience cognitive decline to some extent; for instance, meta-analyses of cross-sectional data suggest that processing speed declines, on average, by approximately 20% at age 40 and by 40-60% at age 80 34 . In terms of supplementation duration, studies have detected significant cognitive changes due to n-3 PUFA supplementation after 3-4 months 35 36 . However, these studies focused on clinical populations; change in normal populations would not be expected to be as rapid, or of the same magnitude and the longer duration allowed for the multiple testing points required to best test the hypothesis and conferred greater power to detect an effect.

The trial was co-ordinated from CSIRO Food and Nutritional Sciences, Adelaide, Australia. The protocol was explained to participants and written informed consent was obtained prior to study commencement. All experimental procedures were approved by the Human Research Ethics Committee of CSIRO and in compliance with the Helsinki declaration. The study follows Good Clinical Research Practice. Participants were encouraged to inform the trial research officer of any adverse events; these were recorded by the officer. Adverse effects suspected as being related to supplement use were reported to the trial clinical doctor for follow-up. Participants attended the centre every 3 months; every 6 months, to undergo their assessments, and 3-months after each assessment to obtain additional capsules and be followed-up. At these visits, adverse events were recorded, along with any changes in medication, supplement use, or general health.

After the group information session, participants were scheduled one visit to the laboratory for assessment, every 6 months, for 18 months. Self-report questionnaires (see following section on study instruments) were mailed out to participants up to four weeks prior to each assessment visit, with instructions to complete and return them between one to two weeks prior to their assessment. Missing data or incorrectly completed questionnaires were thus identified beforehand and addressed during their assessment sessions. All physical measurements and cognitive testing were conducted at the same time in the morning at each session.

For each assessment visit, participants attended the clinic after an overnight fast, and a blood sample and physical measurements (height (at baseline only), weight, and blood pressure) were undertaken by trained phlebotomists and staff. Participants were then given a standardised breakfast in the centre, prior to undertaking the cognitive test battery. At baseline and final assessment, blood was taken for measurement of the following. Carotenoids, folate, and vitamin B12, were assessed, all having been associated with cognitive functioning in the elderly; this enables examination of cross-sectional associations between these measures and outcomes, and if necessary in the intervention analyses, to control for any potential differences between groups at baseline or any changes in these measures that differ between groups (despite randomisation). Glucose, glycated haemoglobin (HbA1c), lipid concentrations (total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides), and homocysteine were also measured at baseline and final assessment, because of associations between abnormalities in these parameters of metabolism and cognitive dysfunction 30 . Inclusion of these measures allows us to control for possible baseline differences between groups and potentially assess whether any changes on these measures due to the intervention mediate changes on the main outcome variables. Finally, our blood-derived measures included a high sensitivity assay for C-reactive protein (CRP), an inflammatory marker, plasma malondialdehyde (MDA), a marker of oxidative stress, and telomere length, a marker of oxidative stress and DNA damage. Oxidative stress and inflammation have been associated with cognitive decline or impairment 40 41 and are two postulated mechanisms through which n-3 LC PUFAs may exert an effect on cognitive functioning 34 35 ; thus, we will examine cross-sectional associations between these measures and our cognitive outcomes, be able to control for potential differences between groups at baseline, and assess whether any changes due to fish oil supplementation mediate any change on the primary outcomes. Apolipoprotein E (ApoE) genotyping was also conducted in order to examine any differential effect of treatment due to carriage of the ApoE-ε4 allele. Blood was taken at the second and third assessment points in order to assess erythrocyte membrane fatty acid composition at each time point. n-3 LC PUFA status will be used to assess compliance/uptake post-hoc and we will also assess whether changes in n-3 LC PUFA status are associated with any changes on our outcomes. EPA levels in erythrocytes reflect intake over the past month or two, change linearly with dosage (up to 9 g/day), and increase is evident after 3 days of supplementation; EPA and DHA are incorporated equally effectively 42 . Dietary data for current n-3 and n-6 fatty acid intake was gathered at all time-points, to assess stability of dietary LC PUFA intake over the trial, a food frequency questionnaire (FFQ) was administered as a measure of typical nutrient intake over the preceding one-year period at baseline and at Time 4, and historical seafood consumption was also assessed. Cross-sectional associations between these measures and outcomes will be investigated, and if there are differences between groups at baseline or changes in n-3 or overall dietary intake which differ significantly between the active and control groups, then we will control for any potential effects of this on outcomes. Table 1 summarises the data collected at each time point.

No previous definitive evidence suggests which cognitive domains may be amenable to change via n-3 LC PUFAs, thus the test battery used in this study comprehensively assessed all relevant domains that might benefit from a nutritional intervention. The focus of the battery was on constructs that are sensitive to age-related changes and that are most strongly associated with functional status, functional decline, and dementia.

Most of the cognitive domains were assessed using tasks in which speed of response was the main outcome measure. The psychometric properties of speed tasks make them more sensitive to subtle cognitive changes and obviate the need for parallel forms for repeated assessment. These properties, together with evidence that processing speed is integral to higher-order processes such as reasoning 43 , may be a key indicator of changes in neural efficiency 44 and functional outcomes 45 , and is an important mediator of age-related cognitive changes 46 , make speed tasks ideal as outcome measures to test the effect of a nutritional intervention on cognitive performance in older-age.

With the exception of two individual tasks, at least two measures of each cognitive construct were used to indicate a latent construct for statistical analyses, thereby reducing task-specific variance and increasing construct validity and the reliability of assessment. The cognitive constructs assessed were Reasoning, Working Memory, Short-term Memory, Retrieval Fluency, Inhibition, Simple and Choice-Reaction Time, Perceptual Speed, Odd-man-out Reaction Time, Speed of Memory Scanning, Psychomotor Speed, Inspection Time, and Knowledge. The latter two were measured by only one task each and the knowledge measure was included as a 'control' task in that the construct it assesses does not decline with age nor would it be expected to be modifiable by intervention. Excepting the tasks measured by speed of response, parallel versions of each task were presented at each assessment visit unless otherwise specified. For a detailed description of the test battery see Appendix 1.

All measures were completed in small groups of up to 7 people, in CSIRO's Cognitive Laboratory, and administered by 2 trained research officers. The testing sessions lasted approximately four hours and consisted of four sections of around 50 min duration, with a break of approximately 10 min between each section. During this break participants were provided with standard refreshments, including decaffeinated tea/coffee. The speed tasks were interspersed between the accuracy based tasks and to also help minimise fatigue, paper and pencil tasks were interspersed with computerised tasks, which were administered in an adjacent room. The order of tasks was fixed across participants and testing sessions. Instructions and practice items preceded each test, and questions were clarified before commencement.

At study commencement, a questionnaire ascertained the following demographic and health information: sex, date of birth, and occupation (prior main occupation if retired); diagnosed medical conditions and regularly taken medications; supplement use and duration; and, for women, the use of hormone replacement therapy. A smoking questionnaire assessed current smoking status, the number of cigarettes/cigars/pipes smoked per day, and smoking history, from which pack years was derived (number of pack years = (number of cigarettes smoked per day × number of years smoked)/20) to measure lifetime direct smoking exposure. Changes to smoking habits were captured at study end.

Information was updated regarding medical and health details during the first and then subsequent visits to the CSIRO clinic.

During the assessment sessions, questionnaires were administered to gather other relevant demographic information: highest level of education achieved, country of birth and if relevant, the age of arrival in Australia and years of speaking English, parents' birth country and occupations, perceived ethnicity, cultural determinants of family diet in childhood, familial history of diagnosed dementia, current income level (as per the Australian census, 2006), and relationship status. Frequency of computer use, handedness, and possible hearing and vision impairments (including colour blindness) were also recorded. Participants' level of engagement in day-to-day mental or cognitive activities was assessed by self-reported minutes per week doing challenging puzzles, self-directed learning, managing finances, or teaching/attending classes. At study end participants stated which experimental group they thought they were in, to assess adequacy of blinding.

To assess changes in functional capacity and well-being, and potentially control for any effect of changes in mood or physical activity on cognitive functioning, validated questionnaires assessing everyday functioning (The Late Life Function and Disability Instrument 47 48 ), depression (Centre for Epidemiology Studies Depression Scale 49 ), self-reported cognitive functioning (Cognitive Failures Questionnaire 50 , Prospective and Retrospective Memory Questionnaire 51 ), subjective well-being (mood: Positive and Negative Affect Scale-Extended 52 ; life satisfaction: Diener's Satisfaction with Life Scale 53 , the Personal Wellbeing Index 54 ), health-related quality of life (SF-36 55 , the sleep scale from the Nottingham Health Profile 56 ), and physical activity (The Yale Physical Activity Survey 57 ) were compiled as a general health and wellbeing questionnaire, which participants completed for each assessment. See Appendix 2 for descriptions of the questionnaires and derived scores.

The Victorian Cancer Council FFQ 58 59 was used to measure average daily dietary intake, and a Polyunsaturated Fatty Acid FFQ 60 was used to assess usual dietary intake of n-3 and n-6 PUFAs. The PUFA FFQ was administered in paper format and scored using the electronic version. Lifetime seafood intake was assessed via the Lifetime Diet Questionnaire 61 .

Height was measured at baseline, using a wall-mounted stadiometer (SECA, Hamburg, Germany) and weight was measured at each assessment session using a calibrated digital precision scale (UC-321PBT; A&D Medical, Sydney, Australia). Body Mass Index was calculated as weight (kg)/height (m)² to examine as a potential covariate. Blood pressure was measured in the left arm and after the subject had been seated supine for 5 minutes, using an automated sphygmomanometer (Sure Signs V3; Philips Medical Systems, Andover, MA), and the average of three consecutive readings was used as the measured value for each assessment.

Blood was collected into Vacutainer tubes containing EDTA, cooled to 4°C and separated into plasma (for malondialdehyde and carotenoids analyses) and erythrocytes (for analysis of fatty acids) by centrifugation. Erythrocytes were washed twice with isotonic saline and reconstituted to the original volume in water to lyse the cells. For glucose analyses, plasma was separated from blood collected into Vacutainer tubes containing EDTA + NaF. Serum was separated from clotted blood for lipid analyses. Aliquots of plasma, serum and lysed erythrocytes were stored at -80°C until analysis.

Serum vitamin B12 and folate, plasma homocysteine, and HbA1c were measured at an accredited pathology laboratory (IMVS, Adelaide, Australia). Plasma glucose and serum total cholesterol, triglycerides, HDL, and CRP were measured on a Hitachi 902 clinical analyser using diagnostic kits (Roche Diagnostics, Australia). LDL cholesterol was calculated using the Friedewald Equation 62 . Plasma MDA was measured by a fluorescence high performance liquid chromatography (HPLC) method as detailed by Belobradjic et al 63 .

Carotenoids were measured on a Shimadzu LC10 HPLC fitted with a refrigerated auto sampler and a SPD-MV10 Avp photodiode array detector with a class LC10 chromatography workstation. The chromatographic separations of the fat soluble vitamins and carotenoids were carried out on a Varian Microsorb-MV reverse phase C18 column, 4.6 mm ID × 250 mm length, 5 micron spherical particles 64 .

The analysis of erythrocyte membrane fatty acids was based on a method reported by Ridges et al. 65 , which uses a one-step extraction and transesterification procedure 66 . Solid phase extraction on florisil (Sigma-Aldrich Pty Ltd, Castle Hill, NSW Australia) was used to clean up the toluene extract containing the fatty acid methyl esters (FAME) after the transesterification procedure. Briefly, the toluene was dried under nitrogen, residue redissolved in hexane and loaded onto florisil. The FAME were eluted from the florisil with 10% ether in hexane, dried under nitrogen and re-dissolved in isooctane. An aliquot was injected onto a bonded-phase vitreous silica BPX70 column 30 m × 0.53 mm × 0.5 micron (SGE, Australia) in an Agilent 6890N gas chromatograph equipped with a cool on-column injector. Peak identification was based on a comparison of retention times with Supelco 37 component FAME mix 47885-U (Sigma-Aldrich Pty Ltd, Castle Hill, NSW Australia) and peak areas were measured using ChemStation software.

ApoE genotyping was performed using Polymerase Chain Reaction (PCR) - Restriction Fragment Length Polymorphism method, as described by Hixson et al. 67 . Absolute telomere length was measured by determining the number of TTAGGG hexamer repeats using quantitative real-time PCR as described by O'Callaghan et al 68 .

Treatment and analysis of reaction time data was conducted only on trials with correct responses. Unfeasibly long latencies on single trials and latencies less than 200 ms or greater than 3.5 intraindividual standard deviations above the individual's mean on a particular task were set to missing, and individuals' mean latencies then recomputed. Considering the above factors, if a participant had less than 60% of valid trial data on a particular task, their mean reaction time for the task was set to missing. Participants' scores were also set to missing on a particular speed task if their accuracy was less than 70% on that task. For the speed tasks and excepting the inhibition tasks' difference scores, the inverse of mean correct latencies were used, to normalise the distribution 70 and to express performance as a work rate. For all tasks (both speed and accuracy based), total or mean reaction time scores for data for tasks not completed, or not completed according to correct procedure, were considered missing.

Confirmatory factor analyses were planned on the baseline data (N = 391) to assess the relationships between the tasks and constructs and, in particular, to ensure no empirical redundancy between any of the constructs. The cognitive test battery was potentially to be abbreviated for future assessments, on the basis of the outcomes.

CFA was performed separately on the baseline data for the tasks assessed by accuracy and speed of response, respectively, using full information maximum likelihood estimation in AMOS 7, to deal with the small amount of missing data. For the accuracy based tasks, latent variables were specified for Fluid Intelligence, Working Memory, Short-term Memory, Long-term Memory, and Retrieval Fluency, indicated by the respective tasks specified in Appendix 1 under each construct. The latent variable covariances were free to vary and the residual variances of the immediate and delayed versions of the two memory tasks were allowed to covary to account for their shared methods. All loadings were of at least moderate magnitude, significant, and in the expected direction and the model fit the data well (χ² = 66.42, df = 42, p = .010; Root Mean Square Error of Approximation (RMSEA) = .039, 90%CI [.019, .056]; Comparative Fit Index (CFI) = .984; Akaike Information Criterion (AIC) = 162.4); however, the covariance matrix among the factors was not positive definite because the correlation between Long-term Memory and Short-term Memory was near unity (r = .975). To examine whether both factors were necessary, a model was specified without Long-term Memory and with the delayed memory tasks loading on Short-term Memory (χ² = 71.5, df = 46, p = .009; RMSEA = .038, 90%CI [.019, .054]; CFI = .983; AIC = 159.5). This model fit the data better. Because an independent Long-term Memory factor could not be identified, the indicators of Long-term Memory were deleted and the model re-estimated (χ² = 53.8, df = 29, p = .003; RMSEA = .047, 90%CI [.027, .066]; CFI = .979; AIC = 125.8); this model also fit the data well and to abbreviate the test battery the long-term memory indicators, that is, the delayed word recall and delayed face recognition tasks, were dropped from the test battery for future assessment sessions. This final model was taken as the measurement model for the accuracy-based tasks and is depicted in Figure 1.

For the speed-based tasks, the model specified latent variables for Simple/Choice Reaction Time, Odd-man-out Reaction Time, Inhibition, Speed of Memory Scanning, and Psychomotor Speed, with the indicators for each factor as specified under each construct in Appendix 1. The model fit was acceptable (χ² = 171.4, df = 104, p < .001; RMSEA = .041, 90%CI [.030, .051]; CFI = .979; AIC = 303.4) and, except for Inhibition, the indicator loadings were of moderate to high magnitude, in the expected direction, and significant; the loadings of the Inhibition tasks were generally lower but all were significant except Flanker, which was near zero (-.02). The magnitude of the factor correlations ranged from .19 to .70, all were in the direction indicating positive relationships, and none was so high as to imply redundancy of any factor. The Flanker task was deleted and the model re-estimated (χ² = 155.9, df = 89, p < .001; RMSEA = .044, 90%CI [.032, .055]; CFI = .979; AIC = 281.947); all coefficients were significant and the fit statistics for this model were similar but the AIC indicated better model fit. Although the loading of Colour Stroop on inhibition was also low (.18), given the more experimental nature of this factor (ie., not an established unitary construct in the literature) only the Flanker task was dropped from the test battery for future assessments. Figure 2 depicts the final speed measurement model.

In total, on the basis of the above analyses, three tasks from the initial test battery were dropped from the battery for the subsequent assessments.

Vocabulary (#17). 20 items from forms 1 and 2 of the standard, advanced, and extended range vocabulary tests from ETS Kit of Factor-Referenced Tests 71 comprise this multiple-choice task. A 4-minute time limit was imposed.

Raven's (Standard Plus) Progressive Matrices (#10) 72 . Stimuli are a three by three array of symbols, with the bottom right hand symbol missing. Participants choose, from options given below the matrix, which symbol logically completes the matrix. Twenty items, selected to maintain progressive difficulty, were administered with a ten-minute time limit. The same version was administered at each assessment.

Letter Sets (#13). This task is from the ETS Kit of Factor-Referenced Tests 71 . Stimuli consist of a series of five, four-letter items, which follow a predetermined alphabetical ordering. The task is to determine the rule or pattern linking four of the items and cross-out the item not fitting the rule. Fifteen items comprised this test, administered with a seven minute time limit. The same version was administered at each assessment.

Everyday Problems Test (#19) 73 . This test is an objective measure of adults' ability to solve problems of daily living, presented as printed material, in seven domains considered critical for independent living. The seven domains, similar to those covered by the Instrumental Activities of Daily Living (IADL 74 ), are health (medications), meal preparation/nutrition, phone usage, shopping, financial management, household management, and transportation. Stimulus material is in three formats - directions, charts, and forms - and number of questions is balanced across domains and format. For each item, a stimulus (e.g., prescription medication label, phone fee chart, mail-order form) is presented from which the participant must derive the answer through reasoning, to the accompanying item question. The test at each time point comprised 21 multiple-choice items (selected from the full-test 84-item pool), with a 15 min time limit imposed.

Word Endings (#18). This test is derived from the ETS Kit of Factor-Referenced Tests 71 . The task is to write down as many words ending with a specific set of letters (e.g., 'in'). The score is the number of valid words written within 2 minutes.

Retrieval Fluency (#21). This test is derived from the Woodcock-Johnson III Tests of Cognitive Ability 75 . Participants have to write as many words falling within a given category as possible, in two minutes. The score is the number of things written, bar repeats.

Face Memory - immediate recall (#1). This task, derived from Herzmann, Danthiir, Schacht, Sommer, and Wilhelm 76 , is computerised. 20 photographic portraits are presented with 1.5 mins to study them. Each trial in the subsequent test period presents one of the studied faces next to a completely new face and the task is to indicate which face was the studied face.

First and Last Names (#11). In this paired associates test, derived from the ETS Kit of Factor-Referenced Tests 71 , participants are given 2.5 mins to study a list of 15 first and last name pairs. They are then shown only the last names and allowed 1.5 mins to write down the associated first names.

Word Memory - immediate recall (#2). Word list A from the Rey Auditory Verbal Learning Test 77 was presented via computer at the first assessment and parallel word lists were presented for subsequent sessions. 15 words (nouns) were presented, one at a time, on screen for one sec duration, with an inter-trial interval of one sec. Immediately after presentation of the word-list, participants write down, in any order, as many of the words recalled.

Face Memory - delayed recall (#29). Approximately 3.5 hrs after presentation of the initial Face Memory task, participants were tested again on the faces studied in that task. Each trial presents one of the studied faces next to a completely new face. The task is to indicate the face that was studied in the initial study phase of the Face Memory task.

Word Memory - delayed recall (#6). Approximately 30 mins after presentation of the initial Word Memory task, participants write down as many words as they can recollect from the list, in any order.

Digit-Symbol Substitution (#12); from the revised version of the Wechsler Adult Intelligence Scale 78 . A code with nine symbols, each uniquely associated with one of the digits "1" through to "9", is presented at the top of the page, with rows of only the digits below. The task is to fill in beneath each digit, in order of presentation, as many of the corresponding symbols as possible, within 90 secs. The score is the number of correct answers.

Finding As (#9); from the ETS Kit of Factor-Referenced Tests 71 . Participants are required to examine columns of words; each column has five words containing the letter "A." The task is to place a line through words that contain "A" and ignore those that don't. The score is the number of words marked correctly within the one-minute time limit.

Number Comparison (#20); from the ETS Kit of Factor-Referenced Tests 71 . This task consists of two columns of number-pairs, the numbers varying from 3 to 12 digits in length. Participants are required to place a cross between pairs of numbers if they are not identical. The score is the number marked correctly within the one and a half minute time limit.

The following tasks are all computerised tasks.

Each of the following tasks presents a processing task prior to the stimulus to be remembered. A reminder of the response keys for the processing task is present on the screen throughout. Each trial consists of two to five screens, depending on the difficulty level of the trial, prior to the cue to recall the stimuli. 12 items comprise the test, with three trials of set-size two, and two each of set-sizes three to five. The score is the mean of the proportion of numbers recalled in the correct serial position for each item.

Counting Span (#15) 79 80 . Participants must remember digits while concurrently undertaking a counting task. For each screen, a number of blue circles and squares, and green circles are shown. The participant must count and remember the number of blue circles in each screen, and then press a key indicating whether the number was odd or even. After a response, the next set of figures is presented and after the last set, participants have to serially recall the number of blue circles in each of the preceding displays, responding using the keyboard and mouse. If a participant does not correctly recall the serial position of a minimum of one digit in a set-size, the task ends.

Operation Span (#4) 79 80 . In this task, participants remember words (nouns) while concurrently solving a simple arithmetic problem. Each screen presents a maths equation followed by a word, of the form: "DOES (6 × 2) - 5 = 7? CLASS". Participants look at the word then press a key indicating whether the equation was correct or not. After the response, the next equation/word screen appears and after the final screen, participants write down the words recalled in the correct serial position. Parallel word sets were used in each session.

A windows based version of Inspection Time (#31) 81 was administered. Participants repeatedly view a target stimulus consisting of two vertical lines, joined at the top by a horizontal line, for varying durations. One vertical line is always shorter, with equiprobability. For each trial, the participant indicates by mouse-press whether the shorter line was on the left or right side. Using an adaptive staircase algorithm 82 , a participant's inspection time is defined as the duration of stimulus exposure (in ms) which has an associated probability of 79% of making a correct response.

For the following reaction time tasks, unless otherwise specified, a practice session of 12 trials precedes the start of each test and each task consists of 30 trials. Feedback (accuracy) is presented after each practice trial. The stimuli in both the practice trials and the test proper are balanced such that an equal number of each potential response is required, and for tasks with more than one condition (e.g., Colour Stroop), task-relevant stimulus characteristics are also balanced across conditions.

A simple (#3) and two- (#5) and four-choice (#8) reaction time tasks are included, modified from Danthiir, Wilhelm, and Roberts (Further evidence for a multifaceted model of mental speed: Factor structure and validity of computerised tasks, submitted). The stimuli in these tasks are digits, each associated with arrows pointing in different directions. In all tasks, the stimulus-response coding is present at the top of the screen. Participants must press the appropriate arrow response key, depending on which number is displayed. The number of possible responses increases from one in the simple reaction time task to four in the 4-choice reaction time task.

Two variants of the OMO task (Letter (#14) and Number (#23) OMO, modified from Danthiir et al. (Danthiir, Wilhelm, & Roberts: Further evidence for a multifaceted model of mental speed: Factor structure and validity of computerised tasks, submitted) are included. The OMO task requires an additional discrimination compared to typical CRT tasks. A string of eight equidistant stimuli are presented on screen. In each string, three stimuli are identical (ie., targets) and different from the other five. Two of the targets are always closer together in relation to the position of the third target. The task is to press a key to indicate if the odd-man-out (i.e., the target furthest from the two closer target stimuli) is to the left or right of the other targets.

Two variants of Sternberg's memory scanning paradigm 83 were devised, measuring speed of scanning items in short-term memory; one task using numbers (#7) as stimuli and the other, letters (#16). A string of two to five stimuli are presented simultaneously on the screen for 2200 ms. A probe stimulus is then presented and remains onscreen until participants press a key to indicate whether the probe was in the preceding string.

The following tasks that were devised tap inhibitory processes, also known as interference control or inhibition of prepotent responses. In all tasks the stimuli vary by a task-relevant and -irrelevant characteristic. The task-irrelevant characteristic is linked (e.g., spatially) via automatic prepotent response to the response dimension, providing two or three stimulus-response compatibility (SRC) conditions depending on the task; congruent, incongruent, and sometimes neutral. Trials are balanced in number across stimulus response congruency (SRC) condition and stimulus. During the test, if an incorrect response was given twice in a row for a particular trial-type (e.g., a green left oriented stimulus) a reminder of the correct response keys was presented for 5 sec.

Simon Task (#30). The typical Simon effect refers to the interference (manifesting both in delayed RT and increase in errors) experienced when the response required by a task is spatially opposite to the location of the stimulus, creating a stimulus-response conflict 84 . Stimuli in this task were a red or green circle, presented on the left or right side of the screen. Participants press the right SHIFT key if the stimulus is green and the left SHIFT key if it is red. Two conditions are present in this task; stimulus-response congruent, where the location of the circle is on the same side as the required response (e.g., a green circle on the right side of the screen) and incongruent, where the location of the circle is on the opposite side of the required response (e.g., a green circle on the left side of the screen). The task comprises 60 trials. The dependent variable in analyses is the difference in RT between the congruent and incongruent conditions.

Colour Stroop Task (#22). This task is based on the well-known Stroop effect 85 . Stimuli are the words blue, yellow, note, or engine, coloured yellow or blue, presented one at a time on the computer screen. The task is to press the appropriate key indicating the colour the word is displayed in. Three conditions are present in this task: stimulus-response congruent, where the word matches the colour it is presented in (e.g., the word "YELLOW, " coloured yellow); incongruent, where the word does not match the colour it is presented in (e.g., the word "YELLOW, " coloured blue); and neutral, where the word is not a colour word (e.g., the word "ENGINE, " coloured yellow). 90 trials are presented and the dependent variable in analyses is the difference in RT between the neutral and incongruent conditions.

Spatial Stroop Task (#25). This task is a measure of the so-called spatial Stroop effect (e.g., 86 ). The word LEFT or RIGHT is presented on the left or right side of the screen. Participants press the Right SHIFT key for the word "RIGHT" and the Left SHIFT key for the word "LEFT." There are two conditions in this task; stimulus-response congruent, where the word matches the side it is presented on (e.g., the word LEFT presented on the left side of the screen) and incongruent, where the word does not match the side it is presented on (e.g., the word LEFT presented on the right side of the screen). 60 trials comprise this task and the dependent variable in analyses is the difference in RT between the congruent and incongruent conditions.

Flanker Task (#24). The flanker effect refers to the influence of flanking distracter stimuli on target identification, whereby identification is facilitated by response-congruent flankers and inhibited by response-incongruent flankers 87 . The target in this case was either the letter "S" or the letter "H, " presented in the centre of the screen, flanked simultaneously by two letters displayed either side of the target letter. If the target letter is an "H" participants press the left SHIFT key and if the target letter is an "S" participants press the right SHIFT key. There are three different conditions: the congruent condition where the flanking letters are all the same as the target letter, the neutral condition where the flanking letters (the letter "P") are different from either of the target letters, and the incongruent condition where the flanking letters are the alternative target letter (i.e. if the target letter is "H" the flanking letters are "S"s and vice-versa). 90 trials are presented. The dependent variable in analyses is the difference in RT between the neutral and incongruent conditions.

For these tasks, participants use only the index finger of their preferred hand. For each task, ten practice trials and 20 test trials are given. The dependent variable is the time taken (in ms) between key presses. Following the prompt to begin, participants press either one key, repeatedly (Simple Movement Time task (#26)), or two keys (vertically adjacent - Up Movement Time task (#27); diagonally adjacent - Diagonal Movement Time task (#28)) as quickly as possible.

The Australian SF-36 version 2 55 (SF36V2) health survey standard form was used as a measure of health-related quality of life. The SF36V2 comprises 36 items from which are derived eight norm-based scale scores: Physical Function, Role Physical, Bodily Pain, General Health, Vitality, Social Function, Role Emotion, and Mental Health. Two summary scale scores are derived from sums of the eight scale scores, weighted by factor score coefficients, representing physical and mental health (physical component score (PCS) and mental component score (MCS)). The eight scale scores were normed according to 2004 data from a large South Australian population-based health survey 88 89 and the PCS and MCS were weighted using factor score coefficients computed in this population. On all scales, a higher score indicates better health.

The SF-36 was supplemented with the sleep scale from the Nottingham Health Profile 56 , a generic validated health-related quality of life measure, to assess quality of sleep. The scale consists of five items weighted for severity, which have a maximum sum of 100. Participants answer yes or no to each item and the weights are added on items with a positive response; the higher the score the greater the health problem.

The Yale Physical Activity Survey is a validated 57 90 , interviewer-administered questionnaire specifically designed to be sensitive enough to capture the lower-intensity range of physical activities of older adults. It assesses the level of activity during a typical week in the last month. Minor changes to instructions were implemented to adapt the survey to a format suitable for self-completion. The questionnaire assesses work-, yardwork-, exercise-, caretaking-, and recreational-related physical activity, expressed as minutes per week, as well as current participation in several different types of activity categories believed to reflect physical activity dimensions. A total of eight indices are derived, three summary and five subscale indices, and a seasonal adjustment score can be calculated: 1) Total Time Summary Index, expressed as hours per week; 2) Energy Expenditure Summary Index, expressed as kcal/week^-1; and 3) Activity Dimensions Summary Score (sum of five subscale indices, vigorous activity, leisurely walking, moving, standing, and sitting, each derived by multiplying frequency score by duration score for each activity type and then multiplied by a weighting factor (based on relative intensity of activity), expressed as a unit score), expressed as total units.

Depression was measured by the Centre for Epidemiology Studies Depression Scale (CES-D) 49 , a 20-item tool measuring depressive symptoms in the general population; it has been extensively validated, including on older adults 91 92 93 and is widely used as a screening measure. Frequency of occurrence of symptoms is reported for the past week (as in original) and a higher score is indicative of more depressive symptoms, with a maximum score of 60; scores of 16 to 26 are considered indicative of mild depression and scores of 27 or more indicative of major depression. Because the CES-D uses a 4-point response scale it is more likely to be sensitive to change than other commonly used measures with a 2-choice response format.

Self-reported cognitive functioning in everyday situations was assessed by the Cognitive Failures Questionnaire (CFQ) 50 and the Prospective and Retrospective Memory Questionnaire (PRMQ) 51 .

The CFQ comprises 25 questions measuring self-reported failures in everyday perception, memory, and motor function. Participants rated on a 5-point Likert scale the frequency of such events over the last 4 weeks. A total score is typically derived, measuring proneness to everyday slips and errors, with higher scores indicating greater perceived frequency of mistakes with a maximum score of 100. However, an investigation of the latent structure of the CFQ, using a large representative adult sample, suggests its items assess three constructs - Forgetfulness, Distractibility and False Triggering - which are sensitive to age-effects 95 .

The PRMQ is a validated 51 96 97 98 self-report measure, comprising 16 questions, divided equally to assess the frequency of occurrence of prospective and retrospective memory errors in everyday life. Participants were asked to assess the frequency of these occurrences over the last 4 weeks, on a 5-point Likert scale. The PRMQ has been found to have a tripartite latent structure consisting of a general memory factor plus orthogonal specific factors of prospective and retrospective memory 98 99 , thus the PRMQ total scale can be used as a measure of general self-rated memory or specific memory scores can be derived. The maximum total score is 80, with higher scores equating greater perceived frequency of memory mistakes.

The Late Life Function and Disability Instrument 100 101 (LLFDI) was designed to assess meaningful change in functioning and disability in older community dwelling adults. Concurrent and predictive validity of the LLPDI has been assessed 102 .

The Physical Function component of the LLFDI 101 comprises 32 items measuring the level of difficulty experienced in performing specific activities as part of daily routines. The activities are divided into three domains: basic lower extremity functioning, advanced lower extremity functioning, and upper extremity functioning, with additional questions for users of assistive devices. Respondents rate the difficulty experienced on a five-point Likert scale; all items can be added to derive a total functioning score, with a maximum of 160 and higher scores indicative of better physical functioning.

The Disability component of the LLFDI 100 comprises 16 items measuring two dimensions: the frequency of performance of specific activities as part of daily routines, and the extent of limitation experienced in performing the activities. Respondents rate the frequency of performance and the extent of limitation experienced on a five-point Likert-type scale. Two sub-domains are identified within each dimension: Frequency dimension - 1) Social role (performance of various social and community tasks) and 2) Personal role (performance of various personal tasks); Limitation dimension - 1) Instrumental role (activities at home and in the community) and 2) Management role (tasks that involve minimal mobility or physical activity). Distinct summary scores can be calculated to represent the Limitation and the Frequency dimension, each with a maximum score of 80 and higher scores indicating less limitation and greater frequency, respectively.