Due to the high prevalence of cognitive dysfunctions of about 40–70% in multiple sclerosis (MS), the clinical assessment of cognitive dysfunctions is central to the characterization of this disease (Chiaravalloti and DeLuca, 2008; Rocca et al., 2015). Processing speed, memory and attention are the most frequently and earliest affected cognitive domains in MS (e.g., Litvan et al., 1988; Beatty et al., 1989; Demaree et al., 1999; Denney and Lynch, 2009; Amato et al., 2010; Denney et al., 2011; Langdon, 2011). Therefore, cognitive assessment in MS in clinical studies and daily practice usually focuses on decrements in these most prominent domains (Benedict et al., 2006; Amato et al., 2013; Rocca et al., 2015). Opposed to this, executive functions are less examined, possibly because executive deficits are less frequently reported (Rocca et al., 2015) or because testing of this domain is assumed to be more time consuming (Amato et al., 2013; Mückschel et al., 2016).

Executive functions refer to a family of mental processes needed, e.g., when you have to concentrate and pay attention as well as to coordinate different actions (Diamond, 2013). A large body of studies has reported impaired executive functions in MS (e.g., Rao et al., 1991; Foong et al., 1997, 1999; Guimarães and Sá, 2012; Amato et al., 2013; Preston et al., 2013; Mitolo et al., 2015). These functions are particularly relevant in everyday activities requiring the coordination of two or more tasks at the same time, such as driving a car, preparing a meal, or working at your desk in an office. Dual tasking deficits have been reported for many neurological conditions, e.g., closed-head injury (Stablum et al., 2000; Dell’Acqua et al., 2001, 2003), Huntington’s disease (Beste et al., 2012; Vaportzis et al., 2015) or Parkinson’s disease (Nieuwhof et al., 2017; Salazar et al., 2017). Such dual-tasking situations and other employment-related factors are increasingly recognized in the treatment of MS (Krause et al., 2013), and require the hierarchical organization and processing of several individual actions (Dippel and Beste, 2015). It has been shown that dual-tasking conditions, where different cognitive processes related to the selection of actions have to be monitored in parallel, are very sensitive to even slight alterations in the functioning of neuronal networks (Beste et al., 2013a). The efficient usage of widely distributed functional networks including frontal, subcortical, parietal, and primary sensory regions are essential for performance in dual-tasking (Dux et al., 2006; Marois et al., 2006; Szameitat et al., 2006; Stelzel et al., 2008; Chmielewski et al., 2014; Yildiz and Beste, 2014; Gohil et al., 2015, 2017; Stock et al., 2017). Since MS may be described as a disease affecting the human “connectome” (Griffa et al., 2013) and functional connectivity between brain areas is critically affected by microstructural lesions especially in white matter structures (Bonzano et al., 2009, 2011; Droby et al., 2016), dual-tasking processes may be of particular relevance and are very sensitive to detect early and subtle cognitive (executive) dysfunctions in MS. Additionally, it has been shown that dual-task performance is modulated by neurobiological systems (Schulz et al., 2012; Beste et al., 2013b; Yildiz et al., 2013, 2014; Stock et al., 2014) that are either directly or indirectly affected in MS, like the dopaminergic system (Pacheco et al., 2014; Dobryakova et al., 2015). All these facts suggest that dual-tasking abilities are sensitive to key pathophysiological processes in MS. Until now, research on dual-tasking in MS has mainly focused on relatively simple cognitive tasks that are performed during motor activity like walking (Hamilton et al., 2009; Holtzer et al., 2014a,b; Wajda and Sosnoff, 2015; Downer et al., 2016; Learmonth et al., 2017) or balancing (Monticone et al., 2014; Butchard-MacDonald et al., 2017). Studies on dual-tasking during walking consistently showed a slowing of gait in dependence of disease severity in MS (for review, see Leone et al., 2015). Only few studies focused on the effects of walking on the cognitive tasks, but the results are inconclusive (Leone et al., 2015). Hamilton et al. (2009) reported cognitive performance decrements only when task difficulty was elevated, whereas Allali et al. (2014) could not find any significant effect. Overall, since tests on dual tasking in MS have focused on cognition-motor interactions up to now (Wajda and Sosnoff, 2015) they may be more specific for motor aspects than probably assessing cognitive deficits. Therefore, current knowledge on dual-tasking in MS is biased by frequent deficits in motor functions. Additionally, these approaches do not allow to finetune the difficulty of the task, i.e., the amount of cognitive load applied. Established cognitive tests may be not demanding enough to detect subtle cognitive deficits, especially during early stages of MS. The brain is able to compensate for pathological changes in cognitive processes up to a certain point, until these compensatory mechanisms finally break down and manifest as clinically relevant deficits, which is known as cognitive reserve (Rocca et al., 2009; Stern, 2009; Barulli and Stern, 2013). The possibility to adjust the cognitive test to the patients’ performance level may therefore allow to examine the extent of dual tasking and response selection deficits in more detail.

To the best of our knowledge dual-tasking functions have never been examined in MS using procedures that allow a parametrical scaling of the difficulty of dual-tasking in order to detect subtle executive control deficits and that also have a well-established theoretical background regarding cognitive control processes. In the current study we investigated dual-task performance in MS patients using a standard psychological-refractory period (PRP) paradigm (Welford, 1952; Pashler, 1994). Changes in dual-tasking are assumed to provide evidence for early executive dysfunctions in MS. Here, two tasks are presented in close succession and participants are asked to respond as quickly as possible to each task. The term PRP was first used by Welford (1952) to describe the finding that responses (RT2) on the stimulus of the second task (S2) are slower or less accurate when this stimulus was presented shortly after another first stimulus (S1) signaling a different reaction (RT1) (= PRP effect). With increasing time (stimulus onset asynchrony, SOA) between the stimuli signaling different reactions, the PRP effects becomes smaller (Pashler, 1984, 1994; Wu and Liu, 2008). It is assumed that the PRP effect is caused by the postponement of S2 processing due to fully engaged capacity-limited mechanisms still processing S1 (Pashler, 1994; Dell’Acqua et al., 2001, 2003). These central mechanisms likely comprise response selection processes (Van Selst and Jolicoeur, 1997; Van Selst et al., 1999; Dell’Acqua et al., 2003). Therefore, the PRP effect may serve as an index of dual-task interferences on response selection processes. On a neuroanatomical level, it has been shown that the PRP effect is mediated via a widely distributed network involving the superior and middle frontal gyrus (Dux et al., 2006; Marois et al., 2006; Szameitat et al., 2006; Stelzel et al., 2008), as well as areas in the parietal cortex (Hesselmann et al., 2011). Since the PRP task hence depends on widely distributed functional networks, and these are strongly affected in MS (Griffa et al., 2013; Droby et al., 2016), the PRP task is likely to show dysfunctions in MS. However, since dual-tasking measures using the PRP are particularly demanding when the SOA between two stimuli requiring a response is small, we hypothesize that especially short SOA condition should reveal deficits in MS patients, compared to controls.

The results of the neuropsychological testing as well as demographical and clinical data are summarized in Table 1. The behavioral results for response speed and error rates for S1 and S2 responses are shown in Figure 2.

We examined dual-tasking performance in patients with MS using a PRP paradigm. The results clearly showed impaired dual tasking in MS. These deficits did not affect the speed of responding, but the ability to respond accurately.

The observed relatively small RT differences on responses to the S1 stimulus reflect a typical pattern observed in the PRP (Wu and Liu, 2008). In contrast to responses on the S2 stimulus, the first response is not subject to limitations in response selection capacities. The finding that MS patients did not respond slower on the S1 stimulus shows that differences between groups were not simply due to motor deficits in MS patients, but due to differences in cognitive response selection processes in MS. Moreover, the findings that differential effects between SOA length and group (MS vs. controls) were not evident for the reaction time data, but only for the accuracy data underline that motor response speed is not affecting the pattern of results. As expected, differences between groups were largest in the most difficult conditions upon responses on the S2 stimulus, where both tasks were presented with only a short gap in between (i.e., 16 ms and 133 ms). In these conditions, MS patients committed more errors than controls. In the other SOA conditions, no group differences were obtained. Based upon this it seems that MS patients have stronger limitations in response selection resources than controls. Until now this has not been shown for MS. However, not only responses on the visual (S2) stimulus were less accurate, but MS patients also committed more errors in response to the auditory (S1) stimulus. This effect, where responses on the S1 stimulus are affected by responses on the S2 stimulus, is known as backward crosstalk effect (Miller, 2006; Janczyk et al., 2014) and shows that action goals in dual-tasking are activated in parallel and not in a step-by-step fashion (Janczyk et al., 2014). Overall, the results show that MS is associated with more limited response selection capacities but does not affect the way how responses are selected. It has been shown that the classical PRP effect is mediated via a widely distributed network involving the superior and middle frontal gyrus (Dux et al., 2006; Marois et al., 2006; Szameitat et al., 2006; Stelzel et al., 2008), as well as areas in the parietal cortex (Hesselmann et al., 2011). Additionally, it has been shown that dual-task performance is modulated by neurobiological systems (Beste et al., 2013b; Yildiz et al., 2013; Stock et al., 2014) that are either directly or indirectly affected in MS, like for example the dopaminergic system (Pacheco et al., 2014; Dobryakova et al., 2015). Therefore, it can be assumed that the changes observed in MS patients are due to a combination of structural neuroanatomical and neurobiological factors.

Importantly, the test applied allows scaling the difficulty of dual-tasking by means of different SOAs. This test may therefore be suitable also for adaptive testing in patients with more severe disease symptoms. This test feature is currently lacking in efforts to use multi- or dual-tasking processes in MS. Standard neuropsychological tests, like the Stroop task and the PASAT may also be considered to measure some facets of response selection. However, these tasks do not test dual-task performance and hence the demands on response selection processes are lower. Current research on dual-tasking in MS focusses on simple cognitive tasks that are performed during motor activity (for a review, see Leone et al., 2015) and thus assess rather cognition-motor interference effects. Existing studies on dual-tasking during walking or balancing strongly operationalize this important executive control functions via pure motor processes that are known to be dysfunctional in MS. Therefore, current knowledge on dual-tasking in MS is biased by the frequent deficits in motor functions and cannot provide accurate insights into the cognitive deficits underlying dual-tasking difficulties per se. Due to the task-inherent scaling of the difficulty to perform dual tasks (via the SOAs) the test is most likely applicable to a wide variety of MS-patients; i.e., from the mild to moderate to even strongly affected patients.

Clearly, the moderate sample size as well as the heterogeneity of the MS patients included in this study is a limitation. However, the results are robust, as underlined by the strong effect sizes and the fact that the post hoc power analyses revealed a power of above 99%. The robustness of the PRP is further proved by the fact, that the inclusion of processing speed as a covariate did not change the main pattern of results. Information processing speed is one of the most affected cognitive domains in MS (e.g., Rao et al., 1989; Demaree et al., 1999). Since the PRP is assumed to be based on an information processing bottleneck (e.g., Pashler, 1984, 1994), processing speed differences may have an impact. Even though both groups differed significantly in processing speed the covariate analysis showed that the observed interaction effect of SOA and group was stable even after controlling for processing speed differences. The age differences between control group and patient group pose a further limitation of this study. Age effects on dual tasking are manifold and associated not only by changes on a neurobiological level but also in strategy and motivation (Bier et al., 2017). However, and most important, the covariate analysis showed that age had no impact on the observed effects in this study. In comparison to other studies of aging on dual tasking (e.g., Li et al., 2001; Verhaeghen et al., 2003; Verhaeghen, 2011; Beurskens and Bock, 2012; Chmielewski et al., 2014; Bier et al., 2017), the age differences as observed here were rather small. Other factors differing between controls and MS patients (e.g., BDI and MFIS) did also not bias the results of this study, as shown by the covariate analysis controlling for these factors. Future studies should incorporate neurophysiological and/or MRI data to examine the neuropathological mechanisms behind the observed deficits in more detail.

The study shows that, using the applied PRP paradigm, multi-tasking deficits could be detected in patients with early stages of MS. Because of its scalability, the PRP allows an adaptive testing of important cognitive processes. The PRP may readily complement neuropsychological assessments in clinical care as well as in research settings, as the PRP is easy to conduct and does not require expensive computer hardware or gait analysis devices. The PRP paradigm as applied in the current study may 1 day serve as a prototype for an improved standard diagnostic instrument to detect executive control and response selection deficits in MS. Future research should focus on further cognitive parameters that can complement the PRP in the detection of cognitive deficits in MS. The current findings hopefully help to establish a new, more sensitive cognitive diagnostic instrument to be included in the clinical care of MS patients that can be applied on different subpopulations independent of disease severity.

CB and TZ conceived the study, analyzed the data, and wrote the paper. MM conceived the study and analyzed the data. MP conceived the study and collected the data.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.