Background
Ageing affects both brain volume and cognitive ability in non-demented older people. Brain volume declines with age: autopsy studies estimate around 2–3% per decade from around age 40 years [
1]; [
2], whereas neuroimaging studies estimate around 5% per decade [
3], with changes being non-uniform across brain structures [
4]. Changes in cognitive ability with age are generally described for two main components of intelligence: (a) crystallised abilities e.g. general knowledge, vocabulary, which are relatively well-preserved even into early dementia, and (b) fluid abilities, which tend to decline from early adulthood. Fluid-type abilities typically require abstract reasoning, particularly under time pressure, with new materials, where previous experience provides no advantage [
5].
Over the last century several studies investigated the relationship between brain size and intelligence [
2] in young adults. Before accurate measures using neuroimaging techniques were possible, head size was used as a proxy for brain size, and there was a small, but statistically significant, association between head size and intelligence: in adults a mean correlation of around 0.2 [
6]. Head size is closely related to brain size, but these cannot be seen as equivalent. The size of the skull vault reflects the maximum size of the brain, and is attained by around age six years [
7]. In older people, brain size (but not head size) decreases with ageing-related cerebral atrophy. Head size therefore reflects maximal, rather than current, brain size. In older people there is an association between head size and cognitive ability (r = .07 to .21) [
8,
9], but these studies did not account for prior cognitive ability. This is important because differences in cognitive ability in old age may be due to the stability of these differences from earlier in life, rather than a decline due to age. About 50% of the inter-individual variation in cognitive ability is stable from age 11 to almost 80 years [
5].
Neuroimaging allows an in-vivo, non-invasive measure of actual whole brain volume that is a current and much more accurate measure of actual brain size than proxy measures such as head size. A meta-analysis of
in vivo brain volume and intelligence reviewed 37 independent samples (n = 1,530), and found a correlation between brain volume and intelligence of 0.33 [
10]. Twenty-four of the studies were in adults (r = .41 for females, r = .38 for males, .33 for sexes combined), but the mean age was not reported. Therefore, in neuroimaging studies, there is a consistently documented moderate correlation between brain size and cognitive ability in young adulthood but few studies have investigated whether this relationship persists into older age. Clinical studies in older people often comment on atrophy seen on neuroimaging, and it may be assumed that that 'bigger is better': that individuals with significant atrophy will perform less well [
11]. However, the results of studies of brain size and cognition in older people do not confirm this. One review [
12] suggests that "When structure-cognition associations are found, they are not easily replicated and appear sensitive to the sample composition and choice of cognitive measures." (p. 736)
One study that investigated the relationship between brain size and cognition in older adults measured intracranial area (ICA) – an estimate of maximal brain size [
13] – and several regional brain volumes in 97 unmedicated healthy older men (mean age 67.8, SD 1.3 years). Regional brain volumes correlated with tests of premorbid and fluid intelligence and tests of visuospatial memory (r = .20 to .32) [
14]. The relationships between specific cognitive tests and regional brain volumes were best summarized by a significant positive relationship between the latent traits of a general brain size factor and a general cognitive factor (
g) (structural equation modelling, correlation = .42) and not by associations between individual tests and particular brain regions. Therefore, in healthy elderly men, the relationship between brain region volume and cognitive ability may be largely due to longstanding associations between general cognitive ability and overall brain size.
In a cross-sectional community survey in Australia [
15], MRI scans were performed on 446 individuals (52.2% male, aged 60–64 years) who performed cognitive tests in addition to giving details of education, health and well-being. There was a correlation between cognitive ability (both crystallised and fluid) and both whole brain volume (WBV) and intracranial volume (ICV, a measure of maximal brain size) for men (r = .15 to .18) but not women (r = .02 to .06). There was no association between cognitive ability and a measure of atrophy – WBV corrected for ICV ((WBV - ICV)/ICV) – for men or women (r = 0 to -.1).
Only one study to date has published data on the relationship between brain size and cognitive ability in older people aged over 75 where cerebral atrophy is more likely to have an influence. In 92 individuals aged 79 (57.6% male) from Aberdeen, UK, there was no statistically significant association between total intracranial volume and cognitive ability (η
2 = .009) [
16]. There is therefore a need for further studies, with information about prior and current cognitive ability and brain size, to investigate the relationship between cognitive ability and brain volume in both men and women over age 75.
In summary, ageing is associated with both generalised brain atrophy, and changes in cognitive ability – a relative preservation of crystallised-type and decline in fluid-type intelligence. Brain size is associated with cognitive ability in adulthood, and it has been suggested that the relationship between brain size and cognitive ability in old age is due to the persistence of this relationship from earlier life. Studies investigating the biological basis of ageing have often been cross-sectional, and have reported associations between regional brain volumes and specific cognitive domains in older people [
12]. These are often interpreted as meaning that ageing-related atrophy in a particular brain region is associated with deterioration of a specific cognitive ability [
17]. However, the association between brain structure and cognitive ability may originate earlier in life and remain stable over the lifespan [
14]. Moreover, this association may be between overall brain volume and general cognitive ability, rather than between regional brain volumes and specific cognitive domains [
14,
18]. Studies of the structural basis of cognitive ageing therefore need to consider both maximal and current brain volume; both prior (typically using a valid estimate) and current cognitive abilities; and both specific and general cognitive abilities.
We therefore investigated the relationship between brain size and cognitive ability in community-dwelling older people (aged 75–81). We hypothesised that: (a) current brain volume would be associated with current cognitive ability, (b) this association would be attenuated when corrected for maximal (prior) brain volume and (c) the association would be due to the relationship between general cognitive ability and overall brain size, not specific cognitive domains and brain regions.
Discussion
In 107 community-dwelling people aged 75 – 81 years WBV correlated positively with tests of current, 'fluid'-type [
10] intelligence. Correlations between regional brain volumes and specific tests became non-significant when adjusted for WBV or ICA. The association between WBV and general cognitive ability was almost wholly attenuated by correcting for ICA. Therefore, associations between brain regions and specific cognitive abilities in (healthy) older age may be largely due to the underlying life-long association between overall brain size and general cognitive ability [
14,
18].
This replicates the results in a previous study of healthy older men [
14]. Our study includes more subjects, both men and women, with a narrow, older, age range [
26]. We used minimal exclusion criteria, and therefore our participants should be more typical of the general population than studies selecting healthy participants (e.g. [
14]). The prevalence of common diagnoses such as hypertension and diabetes show this group to be comparable to other older volunteer populations [
4]. One study of individuals aged over 75 [
27] did not find an association between cognition and prior brain size, which may be due to different amounts of white matter damage, or different sample characteristics (e.g. more male participants). Although results from a volunteer population with a limited age range may not be generalisable to the elderly community as a whole, this methodology has important benefits. Since the main correlate with cognitive function is normally age, studying a cohort with a narrow age range allows the relationship between brain MRI data and individual differences in cognitive ability to be investigated [
26].
Using volunteers means that there may be selection bias effects we are unaware of, but these are unlikely to produce inflated associations within the cohort. Also, although all subjects scored > 24 on the MMSE, some may be in subclinical stages of dementia, which could affect both brain volumes and cognitive function.
Previous studies have suggested there may be a sex difference in any association between brain size and cognitive ability, with a stronger association for men [
15]. In this study including both men and women an association between brain size and cognition persisted when correcting for sex statistically. Our study was not powered to look for sex differences, but emphasises the importance of consideration of sex when investigating brain volumes and cognitive ability.
The concept of 'cerebral reserve' has been introduced to account for differences in cognitive ability in individuals with identical amounts of brain ageing [
28]. Cerebral reserve has been characterized by both passive and active components: in passive cerebral reserve, the size of the brain is often used as a proxy for the brain's capacity to withstand the damage accrued by ageing; in active or 'cognitive' cerebral reserve, the brain actively tries to compensate for brain damage using preexisting cognitive processing techniques, or recruiting compensatory approaches. Education or occupation are often used as proxies for cognitive reserve [
16]. Cerebral and cognitive reserve are not mutually exclusive concepts, and it is likely that both are involved in protecting the ageing brain from damage accrued with time [
28]. The underlying neural mechanisms are suggested to be a) neural reserve, where preexisting neural networks have greater capacity (the passive model) or efficiency, and b) neural compensation, where pathology is compensated for by the recruitment of alternative networks [
29]. The finding that cognitive ability in older age correlates with larger brain volume is consistent with the passive model of cerebral reserve [
28], and the attenuation of this relationship by education is consistent with the active model [
16]. Educational attainment is an important consideration in cohort studies of cognitive ability [
30]. It is closely related to childhood cognitive ability (rho = .41 in this study). There is, however, concern in using educational attainment in this cohort born in 1921 – 1926: their educational progress will have been affected by the onset of the Second World War in 1939 (reflected in the median of nine years of formal education). This influence is likely to differ between men and women, and those born in different years, therefore the effect of education in this cohort should be treated with caution, and cannot be generalized to subjects born in other years.
In this cohort maximal brain size (ICA) but not global atrophy was associated with general cognitive ability aged 75–80. The lack of an association between brain atrophy – whether measured by regression or ratio method – [
31] and either cognitive ability in older age or cognitive change is interesting. Christensen et al. (2007) also found a correlation between cognitive ability and maximal brain size, but not atrophy, for men but not women [
15]. It is possible that atrophy of specific brain regions is associated with decline of specific cognitive abilities [
32]. In this study, only logical memory was associated with global atrophy, and in the opposite direction to that expected: i.e. more atrophy correlated with better memory performance. This is unlikely to be a type I error, as similar trends are seen across all the brain regions (uncorrected for ICA), and are stronger in regions that are likely to be associated with memory (e.g. temporal lobe). A negative association between hippocampal volume and memory performance (r ~ 0.25) has been reported in children, adolescents and young adults, whereas the association in older adults is extremely variable [
11]. The underlying mechanism for this is unclear, but may due to inadequate neural pruning causing memory to be mediated less effectively in younger people [
33]. In older adults, the variability in results may be due to statistical methods of normalizing for age and head size [
11,
12]. Studies with a narrow age range, such as this one, largely eliminate the influence of age. It is also important that results are presented both raw and adjusted for head (or prior brain) size [
11]. Correction for head size is a controversial area [
11,
12], and we suggest that it should be seen not only as a methodological or statistical issue, but that corrections may in fact obscure the very relationships they are attempting to investigate.
Despite the intuitive appeal of a 'bigger is better'[
11] association between brain volume and cognitive performance, structure-cognition associations, where they are found, are not easy to replicate, and are sensitive to the composition of the sample tested, and the cognitive tests used [
12]. Differences in white matter volume [
18] or grey matter density [
34] may be more important than overall volume.
The association between cognitive ability in the eighth decade and maximal brain size – fixed by around age six, and possibly determined by postnatal growth – [
35] suggests that research into the biological basis of age-related cognitive change should take a lifecourse perspective [
30]. Influences occurring in early life, or even before birth, can affect cognitive ability many decades later [
36]. However, it should be noted that the effect size of maximal brain size on cognition is small: ICA accounts for only around 7% of the variance in
g.
In this healthy cohort, where there is unlikely to be high degrees of atrophy, ICA (as a proxy for WBV) is likely to be close to the actual WBV in earlier life. The association between brain volume and cognitive ability may be different in pathological cognitive impairment. For example, in dementia, correcting for prior brain size may still reveal associations between profound atrophy and impaired cognition [
37], and indeed atrophy of specific brain regions such as hippocampus and entorhinal cortex may predict those likely to develop Alzheimer's disease [
38]. However, in normal cognitive ageing changes are likely to be more subtle [
12]. Cross-sectional studies such as this one, performing multiple comparisons, cannot be conclusive about the strength of associations or the direction of causation. Longitudinal studies over many years are required to accurately establish the relationship between brain structure and function, including cognitive change. Future studies should examine these interrelationships in healthy and cognitively impaired older individuals to establish if there is a point at which the correlation between regional atrophy and specific cognitive tests does become truly significant.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SS recruited the participants, performed the cognitive tests, performed the statistical analysis and wrote the first draft of the manuscript. JW established the imaging facility and set up the imaging protocol. CR performed the volumetric analyses. ID, JS and JW conceived of, designed, and supervised the study. CR, ID, JS and JW revised the manuscript critically for important intellectual content, and all authors read and approved the final manuscript.