Stride time and stride length variability are both parameters that are related to the control of the rhythmic stepping mechanism [8]. In motor control in general, high variability is related to major attention involvement [9], whereas low variability reflects automatic processes that require minimal attention [9,10]. Performing a motor task while walking, such as carrying a cup [7] has been related to an increased variability of stride time, but not stride length. Verbal fluency task is a frequently used attention-demanding task in dual-task paradigm [1]. In contrast to attention-demanding motor tasks and to their older counterparts, young adults showed no significant change in stride-to-stride variability while performing a verbal fluency task [3,5,6]. Recently, Beauchet et al. [11] reported that, compared to a verbal fluency task, backward counting out loud from 50 significantly increased the coefficient of variation (CV) of stride time in a group of older adults aged 75 years and older who had a broad range of cognitive function abilities (e.g., some had mild dementia). The authors suggested that these findings could be explained by a possible age-related difficulty in the ability to appropriately allocate attention between both tasks due to a major competitive interaction with executive function while dual tasking. Little information is available about the impact of backward counting on stride-to-stride variability in healthy young adults. The only published study using this mental arithmetic task in a small group of healthy young adults showed that the CV of stride length while backward counting did not change compared with walking alone [12]. No data are available about the impact of backward counting on stride time variability.

Previous studies have shown that stride time variability increases when stride velocity decreases [13-16]. Because stride velocity often decreases under dual-task condition [1,12], dual-task related increase in CV of stride time could be provoked either by stride velocity decrease, the attention-demanding task, or both. The understanding of the role of stride velocity, as a potential confounder in the relationship between stride time variability and the involvement of attention in gait control is important. In contrast to stride time variability, variability of stride length in young adults remained low across different gait speeds while walking alone [17]. Furthermore, no significant stride length changes appeared under dual-task condition [3,5,6]. Such results suggest a constant stereotype pattern for stride length regulation, independent of gait speed.

Changes in gait patterns due to the simultaneous performance of a walking-associated task have been reported previously among healthy young adults and interpreted as interference related to competing demands for attention resources involved in both tasks [1]. Both dual-task-related performance changes in gait and backward counting found in our study support this statement. However, unlike previous results obtained in older adults that showed major dual-task related gait changes [1,11], only minor changes in gait parameters were found in our sample of healthy young adults. Several interpretations of these results are possible.

The explanation of dual-task interference is usually based on the assumption that attention resources are limited [20]. According to this theoretical approach, dual-task interference will only occur if the available central resource capacity is exceeded, provoking a performance decrease in one or both tasks. Therefore, interference suggests an overload of the central resources associated with an inability to appropriately adapt allocation of attention between two simultaneously performed tasks. The manner in which attention is divided between two tasks in dual-task paradigm mainly depends on both the priority given (or not) to one task and the attentional load of each task [1,20-22]. In our study, subjects were asked to combine both walking and backward counting without prioritizing either one of the tasks, creating a condition in which attention is divided. Both tasks used in our dual-task paradigm are relatively easy and do not require major attention. Backward counting out loud from 50 is a simple mental arithmetic task requiring low attention involvement in healthy young University students. Therefore, the total attentional load mobilized to simultaneously perform both tasks could not overload the available central resources, and thus only provoked little interference with minor gait changes. All significant dual-task related gait parameter changes in our study were relatively small. Gait speed decreased from 130 cm·s^-1 to 123 cm·s^-1 and the CV of stride time increased from 1.8 to 2.1%. In addition, although decrease of stride velocity while backward counting was related to increase of stride time, change in stride-to-stride variability for stride time was not associated with the attentional component of backward counting.

Previous studies have shown that dual-task related gait changes also depend on the type of measured stride parameters [1,7,8]. A change in single support time while performing a walking-associated attention-demanding task has been shown in healthy young adults [2], suggesting that young adults devote attentional resources to balance control during single-limb support. Few studies have explored the effect of a walking-associated task on the rhythmic stepping mechanism in young adults [4,6,7]. Our findings showed no significant effects of backward counting on means and CV of stride length. Furthermore, dual-task related changes in stride time could be explained by a decrease in stride velocity and variability between subjects, but apparently not on attentional components related to backward counting. Such results suggest that, in contrast to stride velocity, the control of the rhythmic stepping mechanism requires only minimal attention. Only two studies using a motor task as attention-demanding task while walking have shown significant modifications in stride time variability of young adults. Grabiner et al [7] found an increase in stride time variability while simultaneously carrying an 8-ounce cup placed in a saucer while walking. Ebersbach et al. [4] reported a significant decrease in stride time when walking with a rhythmic finger tapping task, interpreted as a magnet effect, a term used to describe the tendency of biological oscillators to attract each other. However, both studies did not examine the role of walking speed as a potential confounder in the relationship between stride time variability and the involvement of attention in gait rhythmicity control.

Most studies exploring dual-task related gait changes have focused on mean values of stride parameters [1], whereas stride-to-stride variability is considered as a sensitive marker for gait control [9,23,24]. Among the temporal gait parameters, stride time reflects the walking rhythm, and is therefore taken as an index of the rhythmic stepping mechanism control [9]. In older people, there is increasing evidence that stride time variability may be related to executive function. Recently, Hausdorff et al. [25] showed an association between high CV of stride time and a relative decline in executive function among healthy older adults, and Sheridan et al. [26] reported a similar relationship between high CV of stride time and impaired executive function in demented older adults. Furthermore, Beauchet et al. [11] recently reported a specific increase of CV of stride time in a group of older adults with a range of cognitive function abilities while backward counting, but not with a verbal fluency task. Whereas verbal fluency mainly relies on semantic memory [27], counting backward essentially depends on the working memory [28] and is therefore more directly related to executive functions. Thus, the dual-task-related increase in CV of stride time while counting backward could be related to competitive interaction with executive function.

The findings of the present study demonstrate that the dual-task related increase in mean value and CV of stride time was apparently related to stride velocity and subjects' effect, but not independently to attentional interference. Although the effect of stride velocity on variability is complex [29], similar positive correlations between increase in stride time variability and decrease of stride velocity have been reported previously [13-16]. Thus, it seems that in young adults the control of gait rhythmicity, stride velocity variability, stride length variability and likely stride time variability, is an automated process that demands little or no attention.

Interestingly, the decrease in the stride velocity during dual tasking was related to an increase in stride time but not to changes in stride length. This result confirms previous findings, which suggested that stride length is not affected by dual tasking, despite changes in gait speed and the performance of attention-demanding tasks [3,5,6,17]. Our subjects decreased stride velocity only by increasing their stride time, without modifying their stride length. This increase in stride time has been related to an increase in the double-support phase [1,2], which may serve to reduce attentional demands during the swing phase and lower the risk of a loss of balance under dual-task. Therefore, the change in the gait pattern during dual task might represent a strategy aimed at maintaining an optimal index of movement consistency in term of energy costs, attentional demand, and efficiency of gait control. The isolated increase in stride time under dual tasking may be explained by two interpretations. First, stride length and stride time could depend on different cerebral control areas. Second, stride time could be more sensitive to interference than stride length.

In our sample of young University students dual tasking had a greater effect on the performance of backward counting than it did on gait velocity. This result could be interpreted as an implicit strategy of the participants, in this specific dual-task situation, to rather give priority to gait safety than to arithmetic task performance. A similar strategy has been showed in older adults [30].

A possible methodological limitation of the present study might be related to the number of strides required to obtain a representative and suitable measure of stride-to-stride variability. Analyzing steady-state walking over 15 meters, the number of steps collected in our study was around 20, whereas Owings et al. claimed that accurate estimation of step kinematics variability required at least 400 steps [31]. Another question that calls for future study is how other, more difficult "dual-tasks" might affect the variability of gait in healthy young adults.