If we allocate undivided attention to a task, its execution will often be more successful as compared to situations when our attention is distracted by a concurrent task. Thus, it is everyday experience that paying attention to the visual environment is affected by the concurrent execution of a motor task. Consider driving a car whilst repeatedly pressing the buttons of your car stereo device in search of your favourite radio program or CD track. In such a situation, your monitoring of the traffic events outside will likely be rendered less efficient compared to a condition when you are focussed on the visual task alone. Empirical data corroborate this view. For example, Mioni et al. (2016
) found temporal discrimination thresholds in the visual but not the auditory modality to be elevated by performing a concurrent finger tapping task in young healthy subjects. Similarly, Fuller and Jahanshahi (1999
) reported that, in patients with schizophrenia, the performance of a task requiring visual-selective attention declined during concurrent finger tapping. These data suggest that even relatively simple motor tasks can significantly affect the efficiency of visual processing.
One approach to understand the performance decline typically observed under dual-task conditions, when two continuous tasks have to be executed simultaneously, is a resource sharing account (see Tombu & Jolicoeur, 2004
, for an overview). This framework assumes that two tasks can be performed in parallel, but that the amount of processing capacity is strictly limited. Due to the limited resources, the available processing capacity has to be shared between the two tasks, rendering task processing of both tasks less efficient. The decrease of processing efficiency under dual-task conditions, compared to the processing of each single task in isolation, is observed as the dual-task cost. Several versions of the resource sharing model have been proposed. Kahneman’s (1973
) original proposal suggested a more or less undifferentiated pool of mental resources that can be allocated to different task demands. Navon (1984
), and Wickens (2002
) assumed multiple resources that can be shared across tasks, giving rise to dual-task costs whenever two or more task processes or stages draw from the same specific resource. A special case are central capacity sharing models (Navon & Miller, 2002
; Tombu & Jolicoeur, 2004
) which accept the idea of multiple task stages, but assume resource sharing at central processing stages only. These models consider the structural bottleneck account of dual-task costs, with its implication of serial task processing at central stages (Pashler, 1994
), as a special case of capacity sharing, when task 1 and task 2 get all of the available capacity, respectively, in serial succession. A model that encompasses aspects of both the structural bottleneck, and of the resource sharing account, has been proposed by Logan and Gordon (2001
) in their ‘executive control of the theory of visual attention’ (ECTVA) model.
The ‘theory of visual attention’ (TVA) introduced by Bundesen (1990
; see also Bundesen, Vangkilde, & Petersen, 2015
, for a recent update) is a framework well suited for assessing how the efficiency of visual information uptake is affected by a concurrent motor task. TVA conceptualizes visual processing capacity as a set of attentional parameters. These parameters can be estimated, on the individual level, by modelling a subject’s performance in a simple psychophysical task, i.e., whole report of briefly presented letter arrays. In short, TVA assumes that visual information uptake is accomplished across two processing waves. During the first, unselective wave, evidence values are computed during a massive parallel processing of the visual input, where objects from the display are matched to long-term memory representations. In the second, selective wave of processing, the available attentional capacity is distributed across the objects in the visual field, and weighted according to the evidence values. All objects compete with each other in a race towards visual short-term memory (VSTM) which has a limited storage capacity of about four elements in healthy, young, participants. Objects receiving more attentional weight race with a faster speed and gain higher probability to be encoded into VSTM. Encoded objects are selected and available for further processing in the cognitive system. Thus, in TVA, the efficiency of visual information uptake is represented by three parameters reflecting the perceptual threshold (parameter t
0), the rate of visual processing (parameter C), and the storage capacity of VSTM (parameter K). These parameters reflect origin, slope, and asymptote, respectively, of the exponential growth function by which the individual whole report performance is modelled according to the equations provided by TVA (see Kyllingsbæk, 2006
; Habekost, 2015
, for a tutorial overview).
Based on TVA, it is possible to individually describe attentional parameters representing the efficiency of visual information processing. Compared to ‘classical’ response time based measures, a number of important advantages arise with respect to the analysis of dual-task effects. It is not only possible to assess the effects induced by a concurrent motor task on visual information uptake by quantifying, for each individual participant, whether and to what degree changes of the perceptual threshold, rate of information uptake, and VSTM storage capacity are invoked. In addition, TVA-based analysis also allows for a comparison between single- and dual-task conditions according to qualitative aspects related to task processing. In TVA, it is assumed that the parameter visual processing speed C
and VSTM storage capacity K
are indexing processes that are relatively constant, within a given individual, across comparable stimulus and task conditions. Indeed, they were interpreted as having a latent trait character (e.g., Finke et al., 2012
). However, it might be possible that, when measured in a dual-task scenario, these parameters reflect variable performance from moment to-moment, traded off in a time-sharing manner. In other words, in the dual-task condition, participants might start and stop the entire task process in which the TVA parameter estimates are embedded, depending on whether or not the participants were paying attention to the visual task. Then, the C
estimate, for example, rather than reflecting a constant rate of information uptake across the dual-task condition, might be an average of actual C
and a non-operating task (where C
could possibly even equal 0). Therefore, two statistical analyses were run to explore whether, in dual-task conditions, the TVA parameter estimates actually reflect a relatively constant performance that can be validly modelled using the TVA-fitting process, or rather provide an overall average across very low versus optimal performance. First, goodness of fit measures were obtained for each participant that reflect the degree to which variance in the empirical performance in the different whole report conditions can be predicted by the individual TVA parameter estimates. Second, the variability of the individual parameter estimates under single- and dual-task conditions was assessed by a bootstrapping procedure (Efron & Tibshirani, 1993
) to investigate the possibility of a broader distribution of the estimates under dual-task conditions.
Effects of a concurrent visual task have been recently assessed within a TVA-based framework by Poth, Petersen, Bundesen, & Schneider, (2014
). These authors found a reduction of visual processing speed, but no effects on the perceptual threshold and the storage capacity of the VSTM. Our study combines—to the best of our knowledge, for the first time—the TVA approach with a continuous motor task in a dual-task procedure. We assessed whether visual processing speed is also affected under a concurrent non-visual task, and whether VSTM storage capacity would be affected as well. As part of this special issue, this attempt can offer new insight into how visual processing is affected by performance of a concurrent motor task. It also offers novel possibilities to assess qualitative differences between single- and dual-task conditions.
In this study, we combined a concurrent motor task, in the form of a repetitive finger tapping, with a visual task assessing the efficiency of visual information uptake. Based on TVA (Bundesen, 1990
), parameter estimates were derived, both under single- and dual-task conditions, that reflected distinct components of visual processing capacity; that is, the perceptual threshold, the speed of visual processing, and the storage capacity of VSTM. Additionally, goodness-of-fit values were obtained for each condition to check whether parameters were validly estimated under both single- and dual-task conditions. Moreover, by applying a bootstrapping procedure, quantitative estimates of the reliability of the parameter estimates in each condition were obtained to test for possibly increased fluctuation of visual attentional performance in dual-task compared to single-task conditions.
Our results showed that concurrent tapping affected visual processing in a significant way. Both the speed of visual processing, and VSTM storage capacity declined under dual-task- compared to single-task conditions. In contrast, the perceptual threshold remained unaffected. These results suggest that a concurrent motor task taps attentional aspects of visual-processing capacity. Participants seem to process information at a lower rate and also to store less pieces of information in VSTM, but are not less sensitive for stimulus registration at minimal exposure durations.
The effect on processing capacity is remarkable when considering the fact that the tapping task was performed on a very high level, with more than 96% accuracy, under both single- and dual-task conditions. Obviously, then, tapping was not a very demanding task and subjects were readily able to keep motor performance in the dual-task condition on a level comparable to the single-task condition. Nevertheless, this rather easy task with only a minor cognitive demand was sufficient to significantly reduce efficiency of visual information uptake in participants at middle to higher age.
The analysis of goodness-of-fit values for the single- and dual-task conditions indicated that a very high variance of the empirical data was explained by the TVA parameter model estimates in both conditions. Moreover, bootstrapping analyses of the parameter estimates showed that the robustness of these estimates was comparable between single- and dual-task conditions. These results clearly do not suggest that the dual-task condition created a higher trial-to-trial variability in the way the participants approached the task. Instead, they support the assumption of the TVA-based fitting that relatively constant parameters underlie whole report performance of a given individual—also across the entire duration of the dual task.
These data are appealing for two reasons. First, they suggest that performing a concurrent motor task relies on attentional resources that are necessary for visual information uptake. Second, they are compatible with a capacity sharing account of motor-cognitive dual-tasking and justify the assumption that both tasks share a common central resource. Given the very short, near-threshold, exposure durations that are most critical for estimating visual processing speed C, these results would be difficult to reconcile with an attention switching account. Contrary to the prediction made by a switching account of dual tasking, there was no evidence of a time-based trade-off in processing the visual task under dual-task conditions, such that participants would switch between a state of paying attention (with a “normal” processing rate at the level of the single task), and a state of not paying attention to the display (with a rate of processing approaching 0). Such behaviour would be reflected in both a violation of the TVA model, giving rise to a decline in the goodness-of-fit, and in an increase of the variability of the bootstrapping estimates. Our analyses showed that this was not the case.
Of course, time-sharing accounts cannot be completely ruled out on the basis of our present findings. After all, there are lots of ways for costs of more difficult or higher-demand central processing to influence the time course of other processes (e.g., costs of switching between monitoring different tasks relative to task difficulty). Therefore, additional studies with experimental settings tailored to investigate this issue in more detail would be required. For example, combining TVA-based whole report with a “classical” psychological refractory period (PRP) paradigm could allow for more fine-grained temporal distinctions.
Our results also render another explanation for our data rather unlikely, namely that participants visually monitored the tapping device in the dual-task condition. The consistency with respect to both model fitting and bootstrapping estimates across single- and dual-task conditions speaks against such an assumption. Arguably, as participants would need to shift not only eye fixation but also turn their heads towards the tapping device, this should result in a marked change of visual threshold estimates (whereby trials with low-exposure duration in particular would be affected) and in reduced parameter robustness in general. Taken together, the high comparability between single- and dual-task conditions with respect to goodness-of-fit and bootstrapping estimates is in line with a resource sharing account predicting qualitatively similar but quantitatively less efficient visual processing in the dual compared to the single task.
Within the framework of TVA, parameter C
reflects the amount of attentional capacity that can be allocated to the processing of objects in the visual field (Bundesen, 1990
; Bundesen et al., 2015
). Accordingly, a reduction of C
would indicate that the amount of attentional capacity is decreased by the presence of a concurrent motor task. A plausible explanation would be that the motor task receives attentional weighting which leaves less attentional capacity available for visual processing. In other words, the concurrent motor task acts as sort of a distractor receiving attentional capacity. Two conclusions can be drawn from such an assumption. First, the decrease of processing speed assessed by the whole report task can be regarded as a quantification of the amount of attentional capacity that is used by the concurrent motor task. Second, due to the non-visual nature of the motor task, this suggests that central attention rather than visual attentional capacity is shared between the concurrent tasks. That is, the attentional capacity as conceptualized by TVA reflects, at least to some degree, central attentional resources instead of purely visual processing capacity. This has already been suggested by clinical studies in which processing speed has been associated with global cognitive ability (Bublak et al., 2011
), or with a non-visual task reflecting central attentional capacity (Kluckow, Rehbein, Schwab, Witte, & Bublak, 2016
). Note, however, that this is the first study to suggest a relationship between TVA-based visual processing speed and central attentional capacity in healthy subjects. While Poth et al. (2014
) also found a reduction of processing speed under the influence of a concurrent visual task, this interference could be interpreted as a competition of visual attentional resources. Nevertheless, it must also be noted that both tasks involve a spatial component insofar as the TVA task utilises six stimuli spread out across the visual field, whilst the tapping task relies on the learning of a sequence which is spatially organised. Thus, it is also possible that rather than drawing on a general central attentional capacity, the tasks more specifically tap into a form of spatial attention. However, it is not possible to distinguish the degree to which the attentional changes found in this paper are reflective of either spatial attention or a more general attentional capacity.
parameter reflects VSTM storage capacity in TVA, which represents object categorisations that are available for further processing. Essentially, and in accordance with the ECTVA framework of Logan and Gordon (2001
), this is a stage of response selection, which results in naming of the letters in the case of whole report. In the presence of a concurrent motor task, response selection is made more demanding by the fact that not only do letters have to be named, but also that finger movements need to be selected. Here, executive control is necessary, and our results suggest that this stage is also characterized by resource sharing. A possible explanation could be that when more representations have to be maintained in parallel in a passive store such as VSTM, the reliability of these representations is reduced, owing to decay or interference (see e.g., Jonides et al., 2008
), and response selection is rendered more difficult.
A limitation of our study is that our investigation involved subjects of middle to higher age. Therefore, the results need first to be replicated in younger subjects, before their applicability can be reliably evaluated. However, our results can provide a first step towards a deeper understanding why motor-cognitive dual-task effects seem to be especially pronounced under concurrent visual processing demands in the elderly (Boisgontier et al., 2013
). Furthermore, they set a valuable framework for neuropsychological studies in patients with lesions in brain regions relevant for cognitive-motor functions, which are currently underway.