Serial and parallel processing in multitasking: Concepts and the impact of interindividual differences on task and stage levels. [original]

PARALLEL PROCESSING IN MULTITASKING

Serial and Parallel Processing in Multitasking: Concepts and the Impact of 2

Interindividual Differences on Task and Stage Levels 3

Jovita Brüning 1*, Valentin Koob2*, Dietrich Manzey1, and Markus Janczyk2 5

1 Department of Psychology and Ergonomics, Technische Universitaet Berlin 6

2 Department of Psychology, University of Bremen 7

Jovita Brüning

https://orcid.org/0000-0002-9554-4179

Valentin Koob

https://orcid.org/0000-0003-0186-7186

Dietrich Manzey

https://orcid.org/0000-0002-6977-7024

Markus Janczyk

https://orcid.org/0000-0002-9958-3220

Competing Interests: The authors have no competing interests to declare. 12

Raw data and stimulus material are available via the Open Science Framework at 13

https://osf.io/tw8qx/?view_only=f5e4bec9428844e9b0434894b1d3a4c9 14

Author Note 15

* Shared first-authorship. The order of the first authors is alphabetical. 16

This research was supported by a grant awarded to Jovita Brüning within the Priority Program 17

1772 on “Human performance under multiple cognitive task requirements: From basic mechanisms to 18

optimized task scheduling” funded by the Deutsche Forschungsgemeinschaft (DFG) and by grant 19

JA2307/6-1 funded by the DFG and awarded to Markus Janczyk. This study was not preregistered. 20

Correspondence concerning this article should be addressed to Jovita Brüning, Work, Engineering and 21

Organizational Psychology, Technische Universitaet Berlin, Marchstrasse 12, D-10587 Berlin, 22

Germany. E-mail: jovita.bruening@tu-berlin.de. 23

PARALLEL PROCESSING IN MULTITASKING

Abstract 1

In multitasking research, a central question revolves around whether humans can process 2

tasks in parallel. What 'in parallel' refers to, however, differs between research perspectives 3

and experimental approaches. From a task-level perspective, parallel processing can be 4

conceived as to whether complete tasks are processed in an overlapping manner and how this 5

impacts task performance. In contrast, a large body of literature solely focuses on the central 6

stage of response-selection and whether it can run in parallel with other processing stages, an 7

approach we refer to as the stage-level perspective. Importantly, although each perspective 8

addresses related topics and highlights inter-individual differences, they evolved through 9

independent lines of research. In two experiments, we have taken a first step to investigate if 10

individuals' tendencies for an overlapping versus serial processing mode on the task level are 11

related to vulnerabilities for task interference on the stage level. Individual preferences for 12

either task processing mode were assessed in the task switching with preview (TSWP) 13

paradigm. Individuals’ vulnerability for task interference was assessed with the backward 14

crosstalk effect (BCE) in a classical dual-task. Our results suggest that individuals who prefer 15

overlapping relative to serial task processing at the task level are less vulnerable to task 16

interference during response selection, indicated by a smaller BCE. This difference, however, 17

only emerged in the second experiment with an increased sample size and with task-stimuli 18

that facilitate a bottom-up separation of tasks in the dual-task. 19

Keywords 20

multitasking; backward crosstalk; task switching; interindividual differences 21

PARALLEL PROCESSING IN MULTITASKING

Public Significance Statement 1

There is a considerable range from basic to more applied research in multitasking. However, 2

few studies have tried to link the different research perspectives and experimental approaches 3

employed in basic multitasking research, and even fewer have addressed their relation to more 4

applied multitasking. The present study addresses this gap using an interindividual differences 5

approach. We show that individuals who process multiple tasks in a parallel way in a less 6

restrictive paradigm tend to be less vulnerable to task interference in a paradigm derived from 7

basic multitasking research. Thereby, the study contributes to the literature by describing a 8

relation between multitasking effects found on the level of specific processing stages and on 9

the task level. 10

PARALLEL PROCESSING IN MULTITASKING

Serial and Parallel Processing in Multitasking: Concepts and the Impact of Inter-1

individual Differences on Task and Stage Levels. 2

Introduction 3

The idea that multitasking has detrimental effects on performance became a dominant 4

claim in recent decades. Whether you chat with friends or browse the public media, the 5

consensus is that multitasking doesn't actually work and is, thus, more harmful than useful 6

(e.g., MacMillan, 2016; Rosen, 2008; Schindler, 2015). Arguments favoring this claim 7

emerge from a still ongoing debate: Can we really process and perform tasks “in parallel”, or 8

is the introspective impression of parallel processing just the result of efficient task-9

scheduling, but essentially serial processing? At least, it is relatively clear that people often (if 10

not always) engage in multitasking, that is, they perform two or more tasks with some 11

temporal overlap. 12

An attempt to answer these questions must first consider what ‘parallel processing’ 13

actually means. Although this seems rather simple at first glance, it turns out that the concept 14

is used differently across research perspectives. One perspective on parallel processing relates 15

to two-process theories of information processing. Examples are the distinction of automatic 16

and controlled processes (Shiffrin & Schneider, 1977), volitional and automatic control of 17

action (Norman & Shallice, 1986), reflective and impulsive systems (Strack & Deutsch, 18

2004), or fast and slow thinking (Kahneman, 2011). Common to these theories is the 19

assumption that information processing either runs automatically, that is, via direct stimulus-20

response translations with no or only little voluntary control, or in a more controlled way, 21

demanding deliberate allocation of attention (Norman & Shallice, 1986). In this context, it is 22

often noted that only automatic processes can in principle run in parallel with other processes 23

without necessarily affecting them, whereas controlled processing is essentially serial 24

(Shiffrin & Schneider, 1977). This does not exclude the occurrence of interference effects 25

between automatic and controlled processes, though. The latter can arise, for example, if 26

PARALLEL PROCESSING IN MULTITASKING

automatic processing leads to an automatic attention capture (Shiffrin & Schneider, 1977), or 1

to outcome conflicts with a controlled process running in parallel (Navon & Miller, 1987). A 2

broader perspective on parallel processing, one that is also adopted here, is the one proposed 3

by resource models of information processing (Kahneman, 1973; Navon & Gopher, 1979; 4

Wickens, 1984, 2002). According to these models, the human information processing system 5

has limited processing resources that can be shared by different cognitive processes. This 6

implies that all sorts of cognitive processes can, in principle, run at the same time (i.e., ‘in 7

parallel’). However, this parallel processing may come along with interference effects due to 8

the competition for resources in addition to possible interference related to the kinds of 9

crosstalk effects mentioned above (see Schacherer & Hazeltine, 2021, for a recent discussion). 10

Another noteworthy distinction between research perspectives is how fine-grained 11

(i.e., at which ‘level’) they address the issue of parallel processing: First, parallel processing 12

can be considered on the level of specific information processing stages, and second, it can be 13

considered on the level of entire tasks. The stage level is addressed by a large set of literature 14

on basic mechanisms of multitasking in cognitive psychology (see, e.g., Koch et al., 2018; 15

Pashler, 2000). Here, the question to what extent tasks are processed in parallel is addressed 16

separately for the different processing stages from input to output with a main focus on the 17

central stage of response selection. In contrast, human factors and human performance 18

research often take a more holistic perspective on parallel processing on the task level. The 19

main question addressed by this perspective is to what extent whole tasks can be performed 20

concurrently by assessing concurrent task performance and comparing it to single-task 21

performance (see, e.g., Navon & Gopher, 1979; Norman & Bobrow, 1976; Wickens et al., 22

1981, 2013). Whether options of parallel processing on the stage level are actually used to 23

optimize concurrent performance on the task level is thus far unclear, though. In the 24

following, we will point out similarities and differences between these two perspectives of 25

parallel processing on the stage and the task level. We will also address the topic of inter-26

Loading more pages...