Cerebellar contributions to working memory
Introduction
It has been long established that the cerebellum is involved in the control of movement (Holmes, 1939, Glickstein, 1993). This is partly predicated on the finding that in non-human primates, the heaviest projections to the cerebellum arise in the cortical motor system (Brodal, 1978, Glickstein et al., 1985). However, more recent evidence suggests that cerebellar cortical circuits also process information from areas not directly related to motor control (including areas of the prefrontal cortex; Schmahmann and Pandya, 1997). In particular, such projections are best characterized from Walker’s area 46 (Hoover and Strick, 1999, Kelly and Strick, 2003, Middleton and Strick, 2001, Walker, 1940). The connections of these cortical areas with their cerebellar counterparts form independent modular loops. The motor cortex (area 4) is reciprocally connected with lobules HV, HVI, HVIIB and HVIII of the cerebellar cortex. Prefrontal area 46 (Hoover and Strick, 1999, Kelly and Strick, 2003, Middleton and Strick, 2001) is reciprocally connected to vermal and hemispheral parts of lobule VII. Evidence from diffusion imaging in humans and non-human primates suggests that the prefrontal loop has selectively grown during the course of evolution (Ramnani et al., 2006). While the anatomy of this pathway is now well defined, the processing of prefrontal inputs within connected cerebellar cortical territories is still poorly understood (Ramnani, 2006).
Evidence from neuropsychology has demonstrated that some cerebellar syndromes are associated with cognitive deficits traditionally linked to frontal lobe dysfunction (Drepper et al., 1999, Schmahmann and Sherman, 1998). In particular, some have suggested that cerebellar lesions result in specific impairments in processes related to articulatory planning (Silveri et al., 1998, Zettin et al., 1997). However, it is difficult to draw precise anatomical conclusions on the basis of heterogenous clinical populations. On the basis of this evidence alone, no inferences about local information processing in the cerebellar cortex can be made because lesions in this location have profound distal effects in connected frontal lobe areas (crossed cortico-cerebellar diaschisis; von Monakow, 1914). In contrast, neuroimaging evidence can investigate information processing in precise anatomical locations, in healthy populations. Such studies have provided clear evidence of cerebellar activity that can be explained purely in terms of cognitive demands (Chen and Desmond, 2005, Desmond et al., 1997, Desmond et al., 1998, Desmond and Fiez, 1998, Desmond, 2001).
Recent theoretical accounts of cortico-cerebellar information processing suggest that cerebellar cortical circuits acquire forward models of cerebral cortical information processing that facilitate the automatic execution of those processes, whether in motor or cognitive domains (Ramnani, 2006). The modular organization of the cortico-cerebellar system suggests that forward models in each of these domains must be acquired in distinct areas of the cerebellar cortex. Control theoretic accounts suggest that highly practiced execution of actions engage cerebellar circuits (Ito, 2005, Kawato and Wolpert, 1998, Miall and Wolpert, 1996, Wolpert and Kawato, 1998), probably cerebellar components of the motor loop. Similarly, it was predicted that familiar and routine cognitive operations should activate cerebellar cortical components of the prefrontal loop (Ramnani, 2006). Here, we test this hypothesis by manipulating working memory load within a task that requires speech motor control.
The Paced Serial Addition Task (PASAT; Gronwall, 1977) is well known for imposing high cognitive load. This task involves auditory presentation of single digit numbers in pseudo-random order. On presentation of every number (n), that number must be added to the preceding number heard in the sequence (n − 1). Thus, when each number is presented, the previously heard number (n − 1) must be remembered to complete the addition, but the number preceding that (n − 2) must be excluded from the calculation (see Fig. 1). This task imposes a number of specific cognitive demands that include the operation of verbal working memory, the phonological loop, speech production, addition and inhibition of cumulative total. Our control task was designed to reduce the cognitive load while holding constant the sensory and motor demands of the task. In comparing the two conditions, we aimed to localize activity evoked by increased cognitive load in the experimental condition (which was over-learned and therefore routine) in relation to the control condition. We predicted that this comparison would reveal activations in the human homologue of macaque prefrontal area 46 (human areas 9/46 of Petrides and Pandya, 1999)—a finding which would be consistent with other reports (Cohen et al., 1997, Curtis and D’Esposito, 2003, Passingham and Sakai, 2004, Schumacher et al., 1996, Smith and Jonides, 1998, Smith and Jonides, 1999). Importantly, we also predicted such activations in connected cerebellar cortical components of the ‘prefrontal’ loop (lobule VII, including Crus I and Crus II). Only the sensory and motor demands of the task were common to both conditions. Thus, a feature of our analysis strategy was to seek validation of our methods by localizing activity related to both experimental and control conditions (a ‘conjunction’ analysis; Friston et al., 1999). Such activity was expected in the auditory areas of the superior temporal gyrus and the ventral areas of precentral cortex containing representations of orofacial musculature. Furthermore, we predicted that this conjunction would reveal activation of the cerebellar cortical components of the ‘motor’ loop (lobules IV, V and VI).
The requirement to produce an overt verbal response during working memory tasks is relatively demanding compared with other working memory tasks. Hence, our task required the execution of speech movements during fMRI. ‘Sparse sampling’ is commonly employed to overcome the confounding effects of speech-related head motion on MRI data (Abrahams et al., 2003, Amaro et al., 2002, Gracco et al., 2005, Hall et al., 2005). We present a novel variant of this technique in which a period of scanner silence was regularly introduced to enable participants to produce overt verbal response in the absence of EPI scanning. This took advantage of the slow time-constant of the BOLD response, in that it could be sampled after cessation of the motor response. The silent period also allowed us to record and score verbal responses on every stimulus in order to ensure appropriate level of task performance (to our knowledge, other studies employing the PASAT have not used this specific combination of methods).
Section snippets
Participants
15 Right-handed volunteers (aged 18–29; 9 females) participated in the study after giving written and informed consent. They confirmed that they had no known neurological or psychiatric history. The study was conducted in accordance with the Royal Holloway University of London (RHUL) MRI Rules of Operation that incorporate rules set out by the Medical Devices Agency. This study received approval from the RHUL Psychology Department Ethics Committee.
Experimental design
The aim of the study was to test whether the
Behavioral results
In the first part of the preparation phase (traditional PASAT) all participants performed to a satisfactory level (82% correct; SD, ± 7.8 across group), and were therefore allowed to progress to the following stages. Previous studies have repeatedly demonstrated that the performance of healthy volunteers on the PASAT under nearly identical psychophysical conditions is similar to those of our study (range 78% to 84%; Audoin et al., 2003, Audoin et al., 2005a, Audoin et al., 2005b, Forn et al.,
Discussion
The aim of our study was to test the anatomically specific hypothesis that cerebellar cortical zones, which are interconnected with the prefrontal cortex, would be activated by the cognitive demands of our experimental task. In line with our hypothesis, our experimental task (ADD) activated vermal territories of the cerebellar cortex that lay within lobule VII. This area is known to be interconnected with the prefrontal cortex in non-human primates (Kelly and Strick, 2003). These areas of the
Acknowledgments
This was supported by an unrestricted educational grant by Bayer Schering Pharma AG to D.W.L. and a grant from the Royal Society to N.R. We would like to thank Joshua Balsters who assisted with scripting and data acquisition.
References (101)
- et al.
Does the cerebellum contribute to cognitive aspects of speech production? A functional magnetic resonance imaging (fMRI) study in humans
Neuroscience Letters
(1998) - et al.
Modeling geometric deformations in EPI time series
NeuroImage
(2001) - et al.
Experimental designs and processing strategies for fMRI studies involving overt verbal responses
NeuroImage
(2004) - et al.
Temporal dynamics of cerebro-cerebellar network recruitment during a cognitive task
Neuropsychologia
(2005) - et al.
Persistent activity in the prefrontal cortex during working memory
Trends in Cognitive Sciences
(2003) - et al.
Neuroimaging studies of the cerebellum: language, learning and memory
Trends in Cognitive Sciences
(1998) - et al.
Dissociation of frontal and cerebellar activity in a cognitive task: evidence for a distinction between selection and search
NeuroImage
(1998) - et al.
A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data
NeuroImage
(2005) - et al.
Cortical Reorganization During PASAT Task in MS Patients With Preserved Working Memory Functions
NeuroImage
(2006) - et al.
Analysis of fMRI time-series revisited
NeuroImage
(1995)