Thirty patients with CFS were recruited from a tertiary referral clinic associated with the University of New South Wales in Sydney, Australia, which provides a graded-activity oriented cognitive-behavioural therapy program. All patients had been diagnosed according to the international diagnostic criteria for CFS 
. Forty control subjects were recruited by advertisement from the local community. The groups were matched for age, sex, body mass index (BMI), hours of weekly exercise, and education.
Exclusion criteria for the study were: pregnancy, primary sleep disorder, endocrine, neurological, autoimmune or cardiovascular disease, and any major depressive, psychotic or substance abuse disorder. Medications known to affect autonomic functioning including beta-blockers, benzodiazepines, corticosteroids, and any other centrally-active drugs were also exclusionary.
The Human Research Ethics Committee of the University of NSW approved this research (HREC Approval No: 08085). The study was conducted in accordance with the principles expressed in the Declaration of Helsinki. All participants gave written informed consent before taking part.
Testing was carried out at a comfortable ambient temperature (23±3°C). Participants were asked to abstain from caffeine, alcohol and exercise for 12 hours prior to testing. Participants were connected to physiological sensors and sat in a semi-reclined relaxed position for 5 minutes during baseline recordings. Participants then rated their current sensations of ‘physical’ and ‘mental’ fatigue and perceived energy levels on a series of 1–10 Likert scales (1
none at all, to 10
absolutely the most possible) before and after cognitive testing. Presentation of the first two tests was counter-balanced to control for possible order or carry-over effects on performance, including fatigue and learning. However, the Stroop task was always presented last as previous findings by our group found that this cognitive task generally provided greater autonomic activation than the other cognitive tasks used in this study. At the completion of the Stroop task, participants were asked to relax until HR had returned to baseline.
Questionnaires were used to obtain medical history and demographic information and to assess health behaviours, symptoms, and functional impairment. Education was assessed as years of education completed. The number of hours/week of exercise sufficient to cause a noticeable increase in HR and breathing (e.g. brisk walking) was recorded.
The 34-item Somatic and Psychological Health Report (SPHERE) 
assessed a wide range of physical and psychological symptoms; with an empirically-derived, validated subscale (the SOMA) identifying the key clinical features of prolonged fatigue states. The Kessler 10 (K10) produced a global measure of “psychological distress” based on questions about the level of anxiety and depressive symptoms 
. The Brief Disability Questionnaire (BDQ) 
assessed functional impairment, specifically the ‘days in bed’ and ‘days out of role’ were quantified days over the past month during which the respondent stayed in bed or was unable to carry out usual daily activities due to illness. The Pittsburgh Sleep Quality Index (PSQI) 
recorded the quality and pattern of sleep. The short-form McGill Pain Questionnaire provided information on the sensory, affective and evaluative dimensions of pain experiences 
Cognitive Performance Tasks
Computerised versions of three standard cognitive tests, the Digit Symbol test, Spatial Working Memory task and the Stroop-Colour-Word test were used 
. These tests permit assessment of cognitive functioning across the modalities previously found to be affected by CFS (including sustained attention, working memory, and response flexibility) and produce indices of response speed, as well as performance accuracy. Additionally these modalities have been found to be affected by variation in HRV and thus the chosen tests will permit an evaluation of an association between typical cognitive difficulties in CFS and altered HRV. To obtain an accurate assessment of response speed, all tests were self-paced, requiring a response before the next trial appeared on the screen. In the digit symbol test, the digits 0 to 9 were presented each with a different symbol beneath it. One of the symbols from the set was displayed and participants were required to key in the corresponding digit. As soon as an answer was keyed in another symbol appeared. Participants were asked to complete as many of these trials as possible within a 3 minute period. The total number of answers, percentage correct and average reaction time (RT) was recorded.
The spatial working memory task consisted of a 9 square grid in which 3 squares flashed in sequence. The participant was asked to recreate this sequence by mouse-clicking the squares in the same order as the original sequence. If correct, the sequence re-appeared, extended by one additional square. The sequence continued to grow in length until the participant could no longer retrace it from memory and made an error. Following this, a new 3 square sequence began immediately. This test consisted of 6 such sequences. The maximum path length, average error position and average reaction time were recorded.
The Stroop Colour Word test presented a randomised sequence of tasks consisting of an upper case word (either NAME or COLOUR) and a lower case word red, green or blue coloured either, red, green or blue. If the upper word was ‘NAME’ the participant had to press the button corresponding to the name of the word below it. If the word ‘COLOUR’ appeared, the subject had to press the button corresponding to the colour of the text below. Participants were asked to answer as many items as possible within a 5 minute period. The percentage correct and average reaction times were recorded.
Physiological Measures and Data Preparation
HR was monitored via standard three-lead ECG chest electrodes and respiration was monitored via a strain gauge transducer (Pneumotrace, UFI, USA) around the chest. All sensors were connected to an ML880 16 channel Powerlab using LabChart 7 software for recording and analysis (ADInstruments, Bella Vista, Australia). Physiological measures were recorded for the duration of the testing process.
At the completion of testing, the raw data was analysed in LabChart 7. HR and respiratory rate were calculated from raw ECG and respiratory recordings. Mean HR and respiratory rate were sampled in 1 minute blocks throughout the baseline, testing and recovery periods.
The root mean square of successive differences of R-R intervals (rMSSD) – a time domain measure of HRV was calculated using the HRV module (ADInstruments, Bella Vista, Australia). This measure has been shown to be an accurate index of vagal/respiratory sinus arrthymia modulation of heart rate 
. HRV analysis was performed from the recorded ECG for a stable 3 minute period for baseline, digit symbol and Stroop recordings, and a stable 2 minute period for the spatial working memory and recovery periods.
It is well documented that following a stressor, the vagus plays a major role in restoring HR to baseline values 
. The moment of HR recovery was defined by the point at which instantaneous HR [calculated in beats per minute (bpm)] had returned to baseline values for at least 5 seconds.
Analyses were performed using PASW Statistics for Windows version 18 (SPSS Inc., Chicago, IL, USA). Normality of variables was ascertained by graphical methods (normal Q-Q plots). We used t-tests for simple group comparisons of participant characteristics (age, level of education, activity levels, BMI, current symptoms and functional disability); and for analysis of between-group differences of baseline values of physiological parameters and in rated fatigue and energy levels; the Chi-square test was used to assess independence of categorical data such as sex. General linear model (GLM) repeated measures ANOVA served to assess differences in repeated measures, between-group differences, and interactions between group and repeated effects for HR, HRV, as well as changes in rated fatigue, energy and effort in relation to performing the cognitive tasks. Cohen’s d or partial eta squared (ηp2) were calculated to indicate size of effect. Kaplan-Meier survival analysis was used to assess differences in time-to-recovery to resting HR after the cognitive stressor. Correlational (Pearson) analyses tested bivariate associations between current symptoms or HRV and cognitive performance parameters. To further evaluate the relative importance of HRV and other variables with documented potential to affect cognition (i.e. current severity of fatigue and other somatic symptoms, and psychological distress) to outcomes in cognitive performance multiple regression modelling using the enter method was used.