A colleague of mine has an old crappy computer. I am thinking about searching for some money to buy him a newer one with two large screens. This got me thinking: what does the evidence say regarding number and size of screens and work performance? Anecdotally, I suspect dual screens makes me a lot more productive, about 30%.

Search methods

I used Google Scholar. I restricted the publication date to >= 2005 because before this year, dual screens were pretty rare. Searching for the literature is not so easy because terms like screen, monitor and display all have other meanings. But after some searching, it appears that a commonly used term for this kind of research is display configuration. The following papers were skimmed:

There are more, but I have other things to do.

Review

Hill et al, 2007

Summary:

Ninety-six university and non university personnel participated in a comparison of single 20- inch, traditional aspect monitor (ST), dual 20-inch, traditional aspect monitors (DT), and single 24-inch, widescreen aspect monitor (SW) display configurations. Respondents edited spreadsheets, and text documents in a simulation of office work, using each of the display arrays. Performance measures, including task time, editing time, number of edits completed, and number of errors made as well as usability measures evaluating effectiveness, comfort, learning ease, time to productivity, quickness of recovery from mistakes, ease of task tracking, ability to maintain task focus, ease of movement among sources, ease of task management, management of windows, and overall approval were combined into an global evaluation of productivity. The single screen was the least productive display configuration overall and the widescreen the most.

There were many measures and the effect sizes varied quite a bit, but in general, larger screens were better and dual screens were better. The study did not have a dual large screen condition for some reason, so this cell is left empty in their table. This is annoying because this cell allows one to spot interaction effects if they are present. The research design was a repeated measurements 3×2 setup: 3 display configurations, 2 kinds of tasks. There was a lot of noise in their analyses due to the small sample size. In general, however, dual smaller screens was about as good as one larger screen. They do not present d values for the comparisons, and they are spread out in a hundred tables thru the report. They do report means and standard errors, so one can compute the standard deviation using the sample size, and then compute d values. This task is left to the reader (i.e. never gets done unless Gwern does it!). I did spend a bit of time entering some of the raw values into a spreadsheet, so whoever wants to build on this analysis can do so.

Kang and Stasko, 2008

Abstract:

It is becoming increasingly common to see computers with two or even three monitors being used today. People seem to like having more display space available, and intuition tells us that the added space should be beneficial to work. Little research has been done to examine the effects and potential utility of multiple monitors for work on everyday tasks with common applications, however. We compared how people completed a trip planning task that involved different applications and included interjected interruptions when they worked on a computer with one monitor as compared to a computer with two monitors. Results showed that participants who used the computer with two monitors performed the task set faster and with less workload, and they also expressed a subjective preference for the multiple monitor computer.

N=28, university students and employees. They removed a single person who never used the second monitor when he had the chance. Within subject/repeated measurements design. Setting:

In the Singlemon setting, participants used a computer running MS Windows XP with one 17” LCD monitor at 1024×768 resolution. In the Multimon setting, two monitors identical to the one used in the Singlemon setting were used. They were positioned side-byside at an angle of 160° as shown in Figure 1.

The analyses are standard ANOVA stuff. They are not very good because they analyze the two sessions independently, effectively making it into a double single measurement study with N=13-14 per group. Then they try a lot of more fancy interaction analysis with their totally underpowered study. Their figures claim lots of p<alpha effects, but the p values are not given and low p values seem unlikely based on the reported means and standard deviations. In general, this study is not very convincing.

Bi and Balakrishnan, 2008

Abstract:

With the ever increasing amount of digital information, users desire more screen real estate to process their daily computing work, and might well benefit from using a wallsize large high-resolution display instead of a desktop one. Unfortunately, we know very little about users’ behaviors when using such a display for daily computing. We present a week-long study that investigates large display use in a personal desktop computing context by comparing it with single and dual desktop monitor use. Results show users’ unanimous preference for using a large display: it facilitates multi-window and rich information tasks, enhances users’ awareness of peripheral applications, and offers a more “immersive” experience. Further, the data reveals distinct usage patterns in partitioning screen real estate and managing windows on a large display. Detailed analysis of these results provides insights into designing interaction techniques and window management systems more suited to a large display.

N=8. Design:

Our goal is to investigate how a person might use a large high-resolution display for daily work. In particular, we aim to compare large display use with traditional personal computing environments (i.e., single- or dual-monitors). Rather than conduct a controlled experiment to examine individual aspects in isolation, we carried out a diary study in a more realistic context, allowing us to explore usage in a broad range of computing activities over a five day period.

I didn’t read further.

Stegman et al, 2011

Abstract:

Several research studies have been published on user opinion and productivity of using dual monitor systems. These studies found that users typically enjoy using multiple monitors, but none found a strong increase in performance and productivity. Other researchers have focused on improving multiple monitor usability, but often without any statistical framework. This study compared single and dual monitor productivity measures: task time, cursor movement, and number of window switches. Additionally, window management styles (WMS) were studied in order to help designers understand user behavior better. WMS were broken into two categories, toggler and resizer, and then compared to the WMS created by Kang and Stasko (2008). The results of the research showed a significant difference between the number of open applications and a significant difference between single and dual monitors for the number of window switches. The only significant difference between the toggler and resizer WMS was the number of window switches, which was an interaction between the styles and the tasks.

N=36, “mostly from the School of Industrial Engineering at University of Oklahoma“. It seems to be repeated measurement design. The tasks were difficult and required switching between multiple windows. Despite this, they found no evidence for superior performance of dual screens.

Owens et al, 2012

Abstract:

Previous studies have found that using multiple monitors increases productivity, but there are also documented drawbacks to increased monitor count and/or size. In this study participants completed tasks requiring them to use multiple windows and applications simultaneously on four different monitor configurations: single 17” monitor, dual 17” monitors, single 22” monitor, and dual 22” monitors. Results indicated a performance benefit for dual monitor usage, regardless of monitor size. Participants preferred using dual 22″ monitors and least preferred a single 17″ monitor.

N=60. “Participants consisted of individuals who typically work in an office and conduct standard office tasks on a computer using Microsoft (MS) Windows/Office applications”. Design:

This study evaluated four screen configurations: a single 17” monitor, a single 22” monitor, dual 17” monitors, and dual 22” monitors. Each configuration was set up as its own workstation, with comparable standard keyboards and mice. The order of presentation of the workstation was counterbalanced across participants. Each participant interacted with each of the configurations by working through four sets of typical office tasks. Each set of tasks utilized 8 windows from various software applications, including: MS Office (Word, PowerPoint, Excel), MS Outlook, Adobe Acrobat Reader, an Internet browser, and Windows Explorer.

So, repeated measures with 4 conditions in different orders.

Results:

  • “Efficiency was measured by dividing task time by success (% correct). This efficiency measure reflects how quickly and accurately participants completed the task. … A one-way repeated-measures ANOVA was performed for the configurations. Results indicated no differences in efficiency across all four configurations, p > .05. Participants were equally efficient when completing tasks across all four configurations.

I.e. negative.

General conclusions

  • There are very few studies on this obviously important topic.
  • The studies that are there are not well reported and hard to incorporate into a formal meta-analysis.
  • The studies generally found no to little support for superior work performance using dual vs. single screens.
  • The studies generally found lots of report of users preferring two screens.

So, maybe dual screens don’t increase performance much, but they make users think they do, so people buy them and employees ask their employers for them (‘the selfish screen’ hypothesis). I’m a bit skeptical of this conclusion, but the fairly limited evidence points in that direction.

It bothers me that scientists spend years studying things like ego depletion, still manage to fuck that up, and don’t study things with obvious, day-to-day real-world applications (“should I buy dual screen setups for my employees?”, “should I buy dual screens so I can work more efficiently from home?”).