The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing

Demographic information

Interviewers record basic participant information including date of birth and gender, and measure participant characteristics that may affect experimental measures including handedness, as assessed by the Edinburgh Handedness Inventory [10]. Detailed demographic information is gathered from participants including information on marital status, accommodation, employment, income, education, birthplace, ethnicity, and English language history. Additional detailed information about education and training, travel, hobbies, and social activities is gathered using a self-completed questionnaire using items from The Lifetime of Experiences Questionnaire (LEQ) [11] and the European Prospective Investigation into Cancer Study-Norfolk Physical Activity Questionnaire (EPIC-EPAQ2) [12]. Different versions of the questionnaire are used for young (18-29), middle-aged (30-64), and older (over 65) participants. The versions for middle-aged and older participants also ask for employment, education, and other information from earlier periods in their lives.

Cognitive functioning

The cognitive assessment for all participants includes the MMSE [8], the Addenbrooke's Cognitive Examination (ACE-R) [13], logical memory from the Weschler Memory Scale Third UK edition (WMS-III UK) [14], Spot the Word [15], and The Cambridge Memory Questionnaire (the Cambridge 10MQ), a set of 10 questions probing whether participants have memory problems. Spot the Word was time-limited to 5 minutes due to interview time restrictions but this was not indicated to the participants and interviewers could be flexible. Taken together, these measures assess overall cognitive health as well as a number of targeted processes including memory, verbal fluency, premorbid IQ, basic language function, and visuospatial abilities.

Response time measurements

A "simple" response time task (SRT) and a "choice" response time task (CRT) assess basic aspects of speeded responses. In the SRT, participants view an image of a hand with blank circles above each finger, while resting their right hand on a response box with four buttons, one for each finger. When the index finger circle turns black on the image, they press with their index finger as quickly as possible. On pressing the button (or after maximum 3 seconds), the circle becomes blank again, and the variable inter-trial interval (ITI) begins. The ITI varies pseudo-randomly with positively skewed distribution, minimum 1.8 seconds, mean 3.7 seconds, median 3.9 seconds, and maximum 6.8 seconds. There are 50 trials, and the principle outcome measure is the reaction time from stimulus onset to button press. In the CRT, timing parameters are the same as for SRT, but on each trial any one of the four circles above the fingers could become black, and the participant must press the corresponding finger as quickly as possible (maximum 3 seconds response time). There are 67 trials, and the principle outcome measures are the reaction time from stimulus onset to button press (averaged across all four fingers) and the rate of commission errors (pressing an incorrect button).

Social contact

In addition to questions elsewhere about social relationships (e.g. marital status), and activities (e.g. hobbies), interviewers ask a number of targeted questions about social support including communications with friends and relatives, participation in community, religious, or social organisations, and whether or not participants have children.

Measures relevant to physical health, disease, and frailty

A number of measures are taken to assess the physical health of participants, as well as behaviours that are likely to affect their health. Participants provide self-rated assessment of their general health, family history of specific health problems (e.g. heart disease, stroke, and diabetes), and a self-report of their own experience of specific health conditions including those relevant to cardiovascular health (high blood pressure, high cholesterol, etc), mental health (e.g. depression, insomnia), cancer, and a range of other serious health conditions.

Participants are also asked about their history of falls, and perform two tests of balance: first they balance on one leg for 30 seconds, with their eyes open and then with their eyes closed; second, they perform five chair rises from a seated position.

In addition to self-reports of hearing and vision problems, participants receive screening measures of hearing and vision. The Siemens HearCheck Screener tests participants' hearing for three sound pressure levels (75 dB SPL, 55 dB SPL, and 35 dB SPL) at two frequencies (1000 Hz and 3000 Hz). The test is performed without hearing correction (i.e. hearing aids) to mimic conditions in the MRI and MEG sessions of Stage 2 and Stage 3. In order to test near vision, a version of the Snellen test [16] is performed with corrected vision (i.e. glasses) if necessary.

To assess specifics of sleep disturbances, participants complete the Pittsburgh Sleep Quality Index (PSQI) [17]. This measure was designed for purposes of assisting diagnosis of sleep disorders, and provides seven subcomponents of sleep quality as well as an overall score of sleep disturbance.

Finally, participants provide details of medication that they are currently taking, including all prescription and over-the-counter medication.

In addition to objective and self-report measures of physical health, participants answer questions about a number of health-related behaviours. These include questions about current and past alcohol and tobacco use, as well as drug use as assessed by the Drug Abuse Screening Test (DAST-20) [18]. Participants provide information on aspects of their diets (e.g. servings of vegetables per day, frequency of eating processed meats), and a detailed description of their habitual physical activity using the EPIC-EPAQ2 [12].

Mental health

Mental health is assessed by participants' self-report of whether they had been diagnosed and treated for anxiety or depression and at what age. Participants also complete the Hospital Anxiety and Depression scale (HADS) [19], which provides an aid to diagnosis and continuous measures of the severity of symptoms of anxiety and depression.

Measures for older adults

Most measures are taken from all participants across the age range. Two additional metrics are taken to allow for direct comparison with the older cohort of participants in MRC-CFAS [9]. First, participants older than 65 years are asked a series of nine questions probing their historical memory knowledge, with questions taken from the Cambridge Cognitive Examination (CAMCOG) [20]. Second, participants older than 65 years complete a series of 33 questions about activities of daily living.

Questionnaire consistency with other UK Population-Based Studies

The questionnaire is designed to overlap with other population-based studies. Demographic information, diet and working experiences are the same as used in the UK Census for 2011 and UK Biobank, and the cognitive battery for the older population and the activities of daily living is the same as in the MRC-CFAS I and II, augmented with a larger number of activities from the Cambridge City over 75 Study (CC75C) and EPIC-Norfolk. Additional information on substance usage is taken from the ROOTS project [21] as well as other published scales as detailed above.

Be cognitively healthy, with MMSE scores above 24.

Not have MRI safety, comfort, or medical contraindications, such as having various kinds of non-MRI compatible medical implants (cardiac pacemaker, cochlear implants, etc.) or magnetic foreign objects (e.g. shrapnel) close to the head, being pregnant, being claustrophobic, or being unable to lie still for the necessary length of time (approximately one hour). Criteria for MRI and MEG participation are defined by Standard Operating Procedures set by the Medical Research Council Cognition and Brain Sciences Unit (MRC-CBSU).

Not have MEG contraindications that could affect data collection, such as extensive dental work (e.g. permanent brace).

Not have other conditions including serious head injury, current drug abuse, or current serious psychiatric condition (e.g. bipolar, schizophrenic).

Not have poor hearing which could affect the ability to participate in experiments (failing to hear 35 dB in either ear).

Not have poor English or English which is extremely subordinate to another language, which could affect participation in experiments (i.e. those whose native language is not English or who are not bilingual English-speakers from birth).

Emotion expression recognition

This task was developed to tap into processes underpinning the recognition of emotional expressions [22]. This ability represents a potentially dynamic interaction between age-related declines in recognizing facial expressions of emotion [23],[24] and age-related stability or even improvement in many aspects of emotional processing [25],[26]. Likely neural regions contributing to the decline in facial expression recognition include the ventral prefrontal regions, which are connected to other regions involved in facial expression processing, such as the amygdala.

The experiment involves morphed facial continua ranging between the following six facial expression pairs included in the Ekman and Friesen [27] pictures of facial affect series: happiness-surprise, surprise-fear, fear-sadness, sadness-disgust, disgust-anger, and anger-happiness (see [22] for description of materials construction). The stimulus set consists of 15 practice images and 30 experimental images which are repeated in random order across five experimental blocks. For each trial an individual morphed picture is presented on a computer monitor for three seconds. Participants then have as much time as needed to choose which of the six emotion labels (happy, sad, anger, fear, disgust, or surprise) best describes each facial expression. Performance is based only on trials where the morphed image is biased 70% or 90% towards one expression, which provides total accuracy maxima of 20 for each of the six expressions (happiness, surprise, fear, sadness, disgust, and anger).

Emotional memory

This task provides a multidimensional assessment of several aspects of both explicit and implicit memory, and how they are affected by emotional valence. Memory problems are amongst the best-documented declines in normal ageing, but not all aspects of memory are equally affected, with explicit recollection declining most precipitously, recognition-based familiarity judgments being less seriously affected, and aspects of repetition priming often being preserved across the lifespan [28]-[30]. These different aspects of memory are also underpinned by different neural systems, with medial temporal lobe (MTL) critical for explicit memory, and sensory systems such as occipitotemporal cortex involved in (visual) priming.

Furthermore, it is well-established that memory is generally superior for emotional relative to neutral stimuli, which is often attributed to modulation of MTL by the amygdala. This emotional memory advantage may be affected by age, in particular if older adults are increasingly biased towards positively-valenced material, as has been suggested [26]. The task employs a 3 - 3 factorial design, with three types of memory (priming, familiarity, recollection) crossed with three types of valence (positive, neutral and negative).

Like many studies of memory encoding, this task employs a Study phase followed by a Test phase. The Study phase includes 120 trials, and for each trial participants see a background picture for 2 seconds, after which a foreground picture of an object is superimposed. Participants are instructed to imagine a "story" linking the background and foreground picture, and after an 8 second presentation, the next trial begins. The emotional valence manipulation affects only the background image, which is negative, neutral, or positive. Background images are selected from the International Affective Pictures set (IAPS) [31] while all foreground objects are neutral. The Test phase is unannounced (i.e. encoding is incidental), and begins approximately 10 minutes after the end of the Study phase. This phase consists of 160 trials, where 120 trials involve the studied foreground objects, and 40 trials involve new, unstudied objects. Each trial assesses three aspects of memory in succession: visual priming, recognition, and recollection of the background image, including identity and valence. Participants first identify a visually-degraded object presented for 1000 milliseconds (to index perceptual priming). They then see an intact version of the same object which remains on the screen while they indicate their confidence about whether or not they had seen the object in the Study phase (to index familiarity, or "item memory"). Finally, while the object is still displayed, they are required to say whether the valence of the background image against which that object was seen in the Study phase was positive, neutral or negative, and then to provide a verbal description of that background image (to index recollection, or "associative memory").

Priming is assessed by comparing accuracy of naming degraded objects that were studied versus not studied. Recognition confidence is transformed to an accuracy measure using signal-detection theory. Background valence recall is scored as accurate or not by the experimenter, and recollection of the background identity is scored on a 4-point scale by the experimenter (-1 = incorrect, 0 = no information given, 1 = gist correct, 2 = detail correct). Importantly, priming, recognition (when collapsing confidence) and valence recall can also all be scored by the same metric of "Pr", the probability of hits minus probability of false alarms (where 0 is chance and 1 is perfect discrimination), which renders the three memory scores commensurate.

Emotional reactivity and regulation

This task measures processes involved in regulating emotional responses, a critical everyday function which relies on extensive cortical and subcortical networks, including prefrontal cortex, amygdala, insula, and ventral striatum [32]. Functional imaging studies implicate prefrontal cortex in the top-down regulation and modulation of subcortical regions (e.g. amygdala) involved in the core emotion response [33]. While there is evidence for age-related reduction in prefrontal cortex [6], there is behavioural evidence that emotion regulation is preserved or even improves in older adults [25],[26],[34]. Thus, emotion regulation is an important domain for examining age-related flexibility, especially because emotions interact reciprocally with other mental processes, such as attention and memory.

In this task, participants view positive, neutral, and negative film clips and rate their emotional responses after each. For some of the negative films they are asked to reappraise the film content by reinterpreting its meaning in order to reduce the emotional impact. The experiment consists of eight experimental blocks, each containing four experimental trials from one of the four main conditions: (1) watch neutral, (2) watch positive, (3) watch negative, or (4) reappraise negative. For each trial, participants received a prompt to indicate the valence and how they should observe the film (e.g. "WATCH NEUTRAL" or "REAPPRAISE NEGATIVE"). This is followed by a 30 second film clip. Following the film clip, participants provide ratings on three scales. First, they rate how negative they felt during the film from 0 to 100. Second, they rate how positive they felt during the film from 0 to 100. Finally, they use a continuous scale to provide a "strategy" rating, indicating the degree to which they had been watching versus reappraising the film. Each scale appeared for 10 seconds. For blocks containing positively or negatively valenced films, after the four experimental trials there is a "washout" clip where participants watch a calming 30-second film clip before the next block begins.

This task provides measures of how the positive and negative films affect the viewer during the "watch" condition. These "emotional reactivity" scores are calculated by comparing the ratings during the positive and negative films to those from the neutral films. Additionally, this task provides a measure of how well people regulate their negative emotions in the negative reappraise condition. This "emotional regulation" score is calculated by comparing the watch and reappraise conditions for the negative films.

Face recognition: familiar faces

The Face recognition task employs familiar faces (of public figures) to assess the ability of participants to recognize people from their pictures. Previous research has shown evidence of an age-related decline in face recognition [35],[36], which may be associated with age-related memory declines. While the Face recognition test of unfamiliar faces examines novel face processing, this task measures the ability to recognize known (famous) faces.

Recognition of familiar faces is assessed with pictures of 30 celebrities' faces intermixed with 10 unfamiliar face foils. The faces are presented individually in a pseudo-random order. For each face, participants are asked first whether the person is familiar; if so they are asked to provide identifying information such as their occupation, nationality, or work for which they are known. Third, participants provide the name of each familiar person. After the set of 30 faces is complete, the participants are re-presented with any public figures that they did not recognise as "familiar" or any "familiar" people for whom they could not provide identifying information. Participants are provided with the picture and name, and again asked to try to provide their occupation and additional identifying information, in order to provide an additional measure of familiarity.

Separate scores are calculated for the proportion of famous faces where the participant (1) recognises the face as familiar, (2) provides unique identifying information, and (3) provides their correct full name. In addition, a score out of 10 is given to the number of unfamiliar faces that each participant correctly identifies as unfamiliar. Proportional scores are calculated using each participant's individual maximum: the total maximum of 30 familiar faces is adjusted to remove any faces that were persistently unfamiliar in the second phase where the picture is re-presented with the name.

Face recognition: unfamiliar faces

The Benton Test of Facial Recognition [37] assesses the ability to match pictures of unfamiliar faces. Previous research has shown evidence of an age-related decline in face recognition [35],[36]. While the Face recognition test of familiar faces assesses recognition of public figures, this task assesses the ability to recognize a newly-seen face.

We use the short form of the Benton Test of Facial Recognition [38]. There are 27 trials, and on each trial the participant is shown a target face and array of six faces. The task is to find one or more examples of the target face amongst the array of six. For the first six trials the participant has to find one example of the target face in the array of six, in the following seven trials he/she is required to find three examples of the target. Changes in head orientation and lighting can occur between the target and array faces. Each correct response is assigned a score of 1, so the total score is recorded out of a possible score of 27.

Fluid intelligence

Fluid intelligence is an important central cognitive measure because of its broad positive correlations with other cognitive tests. The hypothesis is that, in large part, fluid intelligence reflects the function of the frontoparietal multiple-demand system (see [39]) in constructing the mental control program for any form of complex activity.

We use the standard form of the Cattell Culture Fair, Scale 2 Form A [40],[41]. The test contains four subtests with different types of nonverbal "puzzles": series completion, classification, matrices, and conditions. Each subtest is timed although participants are not informed about precise timings beforehand, with 3 minutes for the first subtest, 4 minutes for the second, 3 minutes for the third, and 2.5 minutes for the final subtest. Before each subtest, instructions are read from the manual and participants are given examples.

The Cattell test is a pen-and-paper test where the participant chooses a response on each trial from multiple choices, and records responses on an answer sheet. Correct responses are given a score of 1 for a total maximum score of 46.

Force matching

The Force Matching task examines the integration of predictive signals in motor control. Normal motor control relies on an integration of predictive signals (the predicted result of one's own action) with low-level sensory signals (sensory feedback from the moving body part). This integration leads to an `attenuation' of the perceived sensory intensity, such that the result of an action that is self-caused is perceived as less intense than a similar sensory event that is externally caused.

The Force Matching task measures the extent of attenuation in order to look into whether and how sensorimotor integration changes during normal ageing. A significant contribution to the morbidity and mortality of the healthy ageing population comes from the gradual deterioration of motor behaviour, and this may reflect both peripheral changes in muscles and joints, and deterioration in the function of the central nervous system.

During this task participants experience a target force on their left index finger and then use their right index finger to match (reproduce) the target force. On each trial, a lever attached to a torque motor applies a target force to the left index finger, which rests under the lever. The target force is applied for 2.5 seconds, with 2.5 seconds of ramping up and ramping down before and after the application. An auditory beep sounds and a visual prompt appears on the computer screen, after which participants match the target force for a period of 4 seconds. There are two ways that participants attempt to match the target force, creating the two main experimental conditions for this task. In the Direct condition, participants press directly on top of the lever with their right index finger, mechanically transmitting the force to the left finger resting below the lever. In the Slider condition, participants move a slider to indirectly transmit force to the left finger via the torque motor and lever. A force sensor at the end of the lever measures both the target and matched forces applied to the left finger. All participants perform both the Direct and Slider conditions and the order is counterbalanced across participants.

For each condition, an initial familiarization of eight trials (two cycles of the four target forces) is performed. The main experiment for each condition consists of 32 trials (eight cycles). Data from each condition of the main experiment is analysed by calculating the average difference between the target force and the matched force on each trial (measured by the force sensor) across the four target force levels. This is referred to as overcompensation, and is positive if participants produced larger matching forces relative to the target forces.

Hotel task

This task examines aspects of executive function that are important for complex planning and multitasking [42]. These abilities are linked to anterior frontal cortex, and have been dissociated from another aspect of executive function, namely fluid intelligence (see Fluid intelligence task description above).

The Hotel task is so named because participants are asked to imagine that they are a manager of a hotel with several tasks to perform. This is a table-top task, and employs props to allow participants to perform a set of five fictionalized tasks: writing out customer bills, sorting money from a charity collection, proofreading an advertising leaflet, sorting mixed playing cards, and alphabetizing name labels for a group of conference attendees. Materials for the five tasks are laid out on a table.

Participants are asked to spend 10 minutes engaged in the tasks, dividing their time between all five tasks. Critically, there is insufficient time to complete any task so that the participant must spontaneously organize their time to ensure they sample all tasks. There is a clock available so participants can check the time when they wish, and if a participant spends more than 5 minutes on the first task they are given a single verbal reminder that the aim of the experiment is to try all five of the tasks. Experimenters record the time spent on each task.

Optimal multitasking would mean that during the 10 minute experiment, all five tasks were attempted, for 2 minutes each. Key outcome measures include how many of the five tasks were attempted and the total deviation from optimal time allocation of 2 minutes per task.

Motor learning

This task taps into motor adaptation, the process of learning new kinematic control in response to deviations in a voluntary action. This online control requires integrating predictive signals for an expected outcome with the actual movement outcome. Learning occurs in response to an unexpected outcome of an action, by updating the model the brain has about the dynamic properties of the environment. This model updating occurs throughout life and is a hallmark of neuroplasticity.

In this task participants use a stylus to make movements to targets. Targets are displayed on a monitor which is projected horizontally and stylus movements are recorded using a digitising touch pad. On each trial participants view the horizontal display and see a yellow target disc 5cm from a central point and in one of four positions. The participants' task is to move the stylus to hit the target within 800 milliseconds, or they receive an error tone and "Too slow" displays. After practicing with their hands visible, during the main experiment a cardboard occluder prevents participants from seeing their hand, but they can see the position of the stylus represented on the display as a red cursor.

The main experiment consists of a total of 192 trials divided into three phases. During the pre-exposure phase, participants perform 24 trials in which the red cursor accurately represents the position of the stylus (veridical condition). During exposure phase, participants perform 120 trials in which the position of the cursor is rotated 30° clockwise relative to the central position. Participants must adapt in this perturbed condition in order to hit the target. During the final post-exposure phase, participants perform 48 trials in the veridical condition; to the extent that participants adapt during the expose phase, they must now adapt back in the post-exposure phase in order to hit the target.

There are two main performance measures, movement time to hit the target, and movement trajectory error. The trajectory error is calculated as the difference between the target angle and the angle of the initial movement trajectory. Both measures are averaged across five distinct periods of the experiment: the pre-exposure phase, late and early in the exposure phase, and late and early in the post-exposure phase. Examining how movement time and movement trajectory error change across the five periods provides measures of initial performance, adaptation rate, and re-adaptation rate.

Picture-picture priming

The aim of this task is to assess core processes involved in word production by measuring the effect of phonological and semantic priming on object naming speed and accuracy. Normal ageing is associated with declines in word finding, which have been variously interpreted as due to age-related declines in semantic memory or phonological access. Aspects of word production are also associated with core aspects of attention, so that we expect naming ability to reflect the interaction between both language-specific and domain-general neural systems.

The experiment involves two phases, a "baseline" phase and a "priming" phase. In the baseline phase participants name aloud a series of 200 pictures of common objects with short (one or two syllable) names, presented in a pseudorandom order. On each baseline trial, a fixation point is presented for 500 milliseconds, followed by an object for 750 milliseconds, followed by a blank screen for 1000 milliseconds.

In the priming phase, 100 of the baseline pictures are repeated, each preceded by a unique prime object that is either unrelated, phonologically-related, or semantically-related to its target. Phonologically-related pairs overlap in their initial phonemes, having either high overlap (first two phonemes, e.g. pencil-penguin) or a low overlap (first phoneme, e.g. robin-ruler) between the prime and target phonology. Likewise, semantically-related primes are category coordinates of the target that had been previously rated as having either higher semantic relatedness (e.g. rabbit-squirrel), or lower relatedness (e.g. box-shelf). Finally, unrelated pairs are neither phonologically- nor semantically-related (e.g. frog-kite). On each priming trial, a 500 milliseconds fixation is replaced by the prime picture for 750 milliseconds and 1000 milliseconds of blank screen, followed by the target picture for 750 milliseconds and a blank screen for 2500 milliseconds.

For both baseline and priming phases, participants are instructed to name every picture as quickly and accurately as possible. Their responses are recorded to a digital sound file as well as scored for accuracy by an experimenter. The primary measures of interest from the baseline phase are correct naming speed and naming accuracy scored out of 200 total possible. Priming effects are calculated by comparing baseline and priming naming times across the unrelated, phonologically-related, and semantically-related conditions.

Proverb comprehension

This task assesses aspects of executive function including abstraction, which are linked to anterior frontal cortex function, and have been dissociated from another aspect of executive function, namely fluid intelligence (see Fluid intelligence task description above).

This simple task has been included in previous batteries of bedside patient assessments [43]. We used a modified version with materials presented on a computer screen and recorded digitally to improve large-scale data collection. In this task participants provide the meaning of three common proverbs in English: "One swallow does not make a summer", "Still waters run deep", and "A bird in the hand is worth two in the bush." On each trial, a proverb appears written on the screen and participants provide a meaning, in their own words, and then press a key to advance to the next trial.

participants' responses are scored by experimenters as incorrect or a "don't know" response (0), partly correct but literal rather than proverbial (1), or fully correct and abstract (2). Total scores for each participant are therefore out of six.

Sentence comprehension

The aim of this experiment is to investigate the on-line comprehension of spoken sentences, focussing on syntactic and semantic processing. Previous research suggests age-related preservation of core aspects of language comprehension, which are largely underpinned by a left-lateralised fronto-temporal network.

The sentence comprehension task is a modified version of similar tasks that have been used with healthy participants and patients with acquired brain damage [44],[45]. This task examines core aspects of sentence comprehension by using syntactic and semantic ambiguity, which occurs frequently and naturally in spoken language. Syntactically-ambiguous phrases such as "landing planes" have at least two different syntactic structures, and semantically ambiguous phrases such as "injured calves" have at least two different meanings.

Although many words and phrases in English are ambiguous in isolation, during comprehension ambiguities are typically disambiguated by the surrounding context. In this experiment the ambiguous phrases are biased, so that there is one interpretation that is most frequent. For example, the dominant meaning of "injured calves" refers to young cows, and the subordinate interpretation refers to leg muscle. Comparing processing during dominant versus subordinate resolutions reveals key processes involved in the activation, integration, and re-evaluation of semantic and syntactic representations. For example, in previous research, healthy participants are slower to accept and more likely to reject subordinate compared to dominant interpretations of syntactic ambiguity, while patients with syntactic impairment are insensitive to the difference between subordinate and dominant interpretation [45].

Experimental materials include 126 grammatically correct sentences, and 98 grammatically incorrect sentences, which act as control stimuli. Of the grammatically correct sentences, 42 are unambiguous, 42 contain syntactically-ambiguous phrases and 42 contain semantically-ambiguous phrases. For the sentences with ambiguities, each ambiguous phrase is followed by a disambiguating word (e.g., "injured calves moo…" or "landing planes are…"). Half of the disambiguating words are consistent with dominant interpretation and half with the subordinate interpretations.

Participants listen to each sentence read in a female voice up to and including the key ambiguous or unambiguous phrase, e.g. "Tom noticed that landing planes…". Two hundred milliseconds after this, participants hear the disambiguating continuation word read in a male voice (e.g. "are") and have 4 seconds to respond with a button press indicating whether they judge the continuation as acceptable or unacceptable.

The total set of 224 sentences is presented in one pseudo-random order in two blocks. Participants have six practice trials before the main experiment, and each block also begins with two lead-in trials that are not analysed. Key dependent measures include the proportion of "unacceptable" judgments and the response times in each condition.

Tip-of-the-tongue task

The aim of this experiment is to examine the common word finding failure known as tip-of-the-tongue states (TOTs). Older adults report word finding problems as one of their main concerns in getting older [46],[47], because of the perceived link between "forgetfulness" and general cognitive decline. However, previous research suggests that TOTs are a linguistic problem, reflecting temporary failures to map active semantic representations onto phonological representations during lexical production e.g. [48]. Neuroimaging data also suggests a role of domain-general systems important for error monitoring [49], suggesting that successfully avoiding or resolving word finding problems relies on an interaction between language-specific and domain-general processing.

The current experiment employs pictures of public figures to elicit TOTs. The task includes 50 faces of people from various walks of life (e.g. actors, musicians, politicians, etc), presented in a single pseudorandom order. Participants are instructed to name each person if they can. If they do not know the name of the person (even if the face is familiar), they respond "Don't Know". If they are sure they know the name but cannot retrieve it, they respond "TOT" to indicate they are in a tip-of-the-tongue state. For each trial, participants view a 1000 milliseconds fixation which is replaced by a picture which remains on the screen for 5000 milliseconds. Participants either name the person (Know response), say they do not know the name of the person (Don't Know response), or say they are having a tip-of-the-tongue (TOT response).

Key measures include the proportion of correct Know responses, Don't Know responses, and TOT responses. Other response categories include incorrect Know responses, null responses (where the participant gives no response), and trials where participants provide information other than the name (such as semantic information).

Visual short-term memory

This task assesses the processes underpinning visual short-term memory (VSTM). Only a handful of objects can be simultaneously attended to or held in VSTM, and this capacity limit declines with normal ageing [50]. Moreover, VSTM capacity is predictive of domain-general cognitive abilities e.g., [51]. Neurally, the capacity limit is reflected in posterior parietal cortex [52],[53], although VSTM tasks activate a frontal-parietal "multiple demand" network [54], and visual memories can be decoded from occipital cortex [55],[56].

This task is a modified version of a previous experiment that separates measures of VSTM quantity and quality [57]. Short term memory for colours is tested using a continuous colour report paradigm [57]. Participants try to remember the colour of circular discs that are presented briefly on a computer screen. After a brief delay, they report the colour of a cued disc, by selecting from a colour wheel that displays a rainbow of hues.

On each trial, participants see a display for 250 milliseconds which contains a central fixation and one to four coloured discs, with the colours chosen at random. The locations of the discs on the screen are randomly selected from eight points equidistant from a central fixation. Following the brief encoding display there is a 900 millisecond blank screen, and then one of the disc locations is highlighted with a border and the response colour wheel appears. On half of trials, any uncued discs also reappear, to provide the context within which the disc was encoded.

Participants try to remember the colour of the disc in the highlighted location, and use a touch screen to press their choice on the colour wheel. They indicate their confidence in the selected colour by the length of time they hold down their finger: as they hold their finger down for longer, white confidence intervals spread out around the selected point indicating more uncertainty about their selection. The response interval does not have a time limit: after participants have confirmed their response there is an 830 milliseconds fixation period before the next trial begins.

After a brief practice, participants complete two blocks of 112 trials, with set-size and probe context being counterbalanced and randomly intermixed within each. Participants also complete a perceptual control block of 56 trials, where single discs are presented at fixation along with the colour wheel, until the participant matches their hue by selecting the appropriate point on the surrounding wheel.

Measures of VSTM quantity and quality can be estimated by fitting the error distribution with a mixture model approach developed by Zhang and Luck [57]. Estimated parameters include VSTM capacity (K), the accuracy of the reported hues (precision), and the probability of mistakenly reporting an un-cued item [58].

T1-weighted structural image

A high resolution 3D T1-weighted structural image is acquired using a Magnetization Prepared RApid Gradient Echo (MPRAGE) sequence with the following parameters: Repetition Time (TR) =2250 milleseconds; Echo Time (TE) =2.99 milliseconds; Inversion Time (TI) =900 milliseconds; flip angle =9 degrees; field of view (FOV) =256mm x 240mm x 192mm; voxel size =1mm isotropic; GRAPPA acceleration factor =2; acquisition time of 4 minutes and 32 seconds.

T2-weighted structural image

A high-resolution 3D T2-weighted structural image is acquired with a SPACE sequence [59] with the following parameters: TR =2800 milliseconds; TE =408 milliseconds; FOV =256mm × 256mm × 192mm; resolution =1mm isotropic; GRAPPA acceleration factor =2; acquisition time of 4 minutes and 30 seconds.

Diffusion-Weighted Images (DWI)

Diffusion-Weighted Images (DWIs) are acquired with a twice-refocused spin-echo sequence, with 30 diffusion gradient directions for each of two b-values: 1000 and 2000 s/mm², plus three images acquired with a b-value of 0. These parameters are optimised for estimation of the diffusion kurtosis tensor and associated scalar metrics, as well as the traditional diffusion tensor. Other parameters are: TR = 9100 milliseconds, TE = 104 milliseconds, voxel size =2 mm isotropic, FOV =192 mm × 192 mm, 66 axial slices, number of averages = 1; acquisition time of 10 minutes and 2 seconds.

Magnetisation Transfer Ratio (MTR) structural image

An MTR image is constructed from two 3D, MT-prepared Spoiled Gradient (SPGR) sequences with either TR =30 milliseconds or TR =50 milliseconds (the TR =50 milliseconds sequences are used when the participant's SAR estimation for the TR = 30 milliseconds sequences exceeds the stimulation limits); TE =5 milliseconds; flip angle =12 degree; FOV =192 mm × 192 mm; voxel-size =1.5 mm × 1.5 mm; bandwidth =190Hz/px; acquisition time of 2 minutes and 36 seconds per sequence for TR = 30 milliseconds, and 4 minutes and 19 seconds per sequence for TR = 50 milliseconds. A Gaussian shaped RF pulse with an offset frequency of 1950Hz (bandwidth =375 Hz, 500 degree flip angle, duration =9984 microseconds) is applied to one of the sequences, and the MTR calculated as MTR = (M0 - Ms)/M0, where Ms and M0 are mean signal intensities with and without the saturation pulse, respectively.

Resting state

To assess intrinsic (passive) aspects of neural connectivity, T2*-weighted fMRI data are acquired while participants rest with their eyes shut using a Gradient-Echo Echo-Planar Imaging (EPI) sequence. A total of 261 volumes are acquired, each containing 32 axial slices (acquired in descending order), slice thickness of 3.7 mm with an interslice gap of 20% (for whole brain coverage including cerebellum; TR =1970 milliseconds; TE =30 milliseconds; flip angle =78 degrees; FOV =192 mm × 192 mm; voxel-size =3 mm × 3 mm × 4.44 mm) and acquisition time of 8 minutes and 40 seconds.

Movie watching

To assess stimulus-driven (active) aspects of neural connectivity across a range of networks, participants watch an excerpt of a compelling but unfamiliar film. A black-and-white television drama previously used in an fMRI study [60], Alfred Hitchcock's "Bang! You're Dead", was edited from a running time of 30 minutes to 8 minutes, while maintaining the plot. A total of 193 volumes are acquired using a multi-echo, T2*-weighted EPI sequence (TR =2470 milliseconds, five echoes [TE =9.4 milliseconds, 21.2 milliseconds, 33 milliseconds, 45 milliseconds, 57 milliseconds], flip angle =78 degrees, 32 axial slices of thickness of 3.7 mm with an interslice gap of 20%, FOV =192mm × 192 mm, voxel-size =3 mm × 3 mm × 4.44 mm) with an acquisition time of 8 minutes and 13 seconds.

Sensorimotor task

To assess basic sensorimotor neural responses, fMRI data are acquired while participants perform a simple audio/visual sensorimotor task. In this task, participants respond to 129 trials consisting of an initial practice trial, 120 bimodal audio/visual trials, and eight unimodal trials included to discourage strategic responding to one modality (four visual only and four auditory only). The timing of trials is optimised for estimation of the fMRI impulse response by generating a sequence of stimulation and null trials using a 255-length m-sequence [61] with m = 2 and minimal stimulus onset asynchrony (SOA) of 2 seconds (resulting in SOAs ranging from 2-26 seconds). For each bimodal trial, participants see two checkerboards presented to the left and right of a central fixation (34 milliseconds duration) and simultaneously hear a 300 milliseconds binaural tone at one of three frequencies (300, 600, or 1200 Hz, equal numbers of trials pseudorandomly ordered). For unimodal trials, participants either only hear a tone or see the checkerboards. For each trial, participants respond by pressing a button with their right index finger if they hear or see any stimuli. Scanning parameters for this task are the same as in the Resting state scan.

Field maps

Differences in the magnetic susceptibility of head tissues, bone and air lead to inhomogeneities in the magnetic field in the scanner, which particularly affect EPI MRI. To measure the field inhomogeneities, an SPGR gradient-echo sequence with the same parameters as the Resting state and Sensorimotor tasks are acquired, but with two TEs (5.19 milliseconds and 7.65 milliseconds). The phase difference between the two TEs can be used to calculate field maps in order to unwarp image distortions caused by field inhomogeneities. Acquisition time is 54 seconds.

Resting state

To assess intrinsic aspects of neural connectivity, we gather eyes-closed resting state data from participants for approximately 8 minutes and 40 seconds, with the first 20 seconds discarded, comparable to the Stage 2 fMRI Resting state data.

Sensorimotor task

To assess basic sensorimotor neural responses, MEG data are recorded while participants perform a simple audio/visual sensorimotor task identical to that in the Stage 2 MRI session. Following this, there is an additional passive stage in which 120 trials of unimodal stimuli are presented every 1 second, half with auditory tones at one of three frequencies (300, 600, or 1200 Hz) presented for 300 milliseconds, and half with a checkerboard patterns presented for 34 milliseconds, and no requirement for the participant to make motor responses. The purpose of this passive stage was to estimate auditory and visual evoked responses independently, to help separate auditory and visual responses during the main sensorimotor task.