1. Introduction
Working memory (WM) is the ability to efficiently maintain and
manipulate temporarily available information, and provides a crucial
functional backbone for complex behaviours, such as memory, learning,
and problem solving (Baddeley, 2012; Diamond, 2013; D’Esposito and
Postle, 2015). Motivated by the centrality of WM throughout human
cognition, as well as the remarkable vulnerability of WM to ageing, most
theories or hypotheses on cognitive ageing have implicated WM as one
fundamental cause of more general age-related declines in cognitive
performance (Park et al., 2002; Reuter-Lorenz and Sylvester, 2005; Craik
and Salthouse, 2011; Anderson and Craik, 2017). Therefore, illustrating
the neural mechanism of WM ageing would help untangle the cognitive
ageing process and develop strategies to protect and promote cognition
in old age.
Over the past decades, studies using positron emission tomography (PET)
and functional magnetic resonance imaging (fMRI) have shown that WM
relies heavily on the prefrontal cortex (especially its dorsolateral
part) and regions of the parietal lobes (e.g., superior parietal lobule
and inferior parietal lobule) (Curtis and D’Esposito, 2003; Wager and
Smith, 2003; Bledowski et al., 2009; Rottschy et al., 2012), and
numerous studies have examined their alterations in ageing, linking
extended patterns of neural activity and functional connectivity changes
to WM impairment in old age (Spreng et al., 2010; Grady, 2012). When
compared to their younger counterparts, older individuals have shown
both decreased and increased WM-related activity (Rypma and D’Esposito,
2000; Grady et al., 2008; Carp et al., 2010; Piefke et al., 2012;
Toepper et al., 2014), and the interpretation of these age-related
changes has been a long-standing issue in the field. While decreased
activity is usually considered to be a reflection of neurocognitive
decline, those increases (i.e., hyperactivity or over-recruitment) have
been viewed as signs of either compensation or dedifferentiation
(Cabeza, 2002; Reuter-Lorenz, 2002; Reuter-Lorenz and Cappell, 2008),
based on the correlations between brain activities and task performance.
In addition, recent evidence from functional connectivity investigations
on WM ageing, including those using task or resting-state fMRI, has
revealed diverse results in older individuals as well, ranging from
weakened long-range functional correlations (mainly between frontal and
posterior regions) (Podell et al., 2012; Heinzel et al., 2017; Tsvetanov
et al., 2018), to increased functional covariations between adjacent
cortical regions during tasks (Cook et al., 2007; Sambataro et al.,
2012).
Taken together, these findings highlight the important roles for both
neural activity and functional connectivity in the WM ageing process,
consistent with the idea that the two functional measurements provide
complementary information on the brain physiology underlying cognition
(Friston, 2011; Turk-Browne, 2013; Misic and Sporns, 2016). However, to
the best of our knowledge, limited studies have applied the two methods
in concert to better understand cognitive functions or their ageing
(Murphy et al., 2016; Tsvetanov et al., 2018). This lack of integration
might be partly because most connectivity studies have been conducted
based on resting-state fMRI data, which lacks cognitive measurements
during data acquisition, and for methods to examine connectivity in task
fMRI data, such as psychophysiological interactions (PPI) (Friston et
al., 1997) and dynamic causal models (DCM) (Friston et al., 2003),
task-evoked activations have been incorporated into the modelling of
time series (Friston, 2011), thus without clear distinctions between
brain activity and connectivity. Hence, we proposed that the distinction
and joint evaluation of brain activity and connectivity would improve
the understanding of how the brain was functionally shaped for higher
cognitive function and how cognitive ageing occurred in the brain.
We therefore applied a recently introduced approach, background
connectivity, to process task-induced blood oxygen level-dependent
(BOLD) signals. The background connectivity could be thought of as an
extension of resting connectivity, where one does not need to assume a
stable default state, and the logic of background connectivity
emphasizes that the BOLD signals during performing cognitive tasks carry
two components of task-related information, one is the task stimuli
reaction (task-evoked activity) and the other one is the current
cognitive status maintenance (state-related activity) (Al-Aidroos et
al., 2012; Norman-Haignere et al., 2012; Turk-Browne, 2013; Duncan et
al., 2014); therefore, the connectivity among regions could be assessed
independently by regressing out the task-evoked activity. In the present
cross-sectional research, we aimed to characterize those two functional
components with a classical visual N-back WM task and to then examine
their contributions to age-related changes in WM.