1 Introduction

1.1 Motivation

Driving is a process inducing high visual and cognitive load. Consequently, the driver experiences high workload on visual perception, cognitive information processing as well as manual responses (Recarte and Nunes 2003). In order to perform well, additional effort is needed to deal with the task load. This exceptional effort is even more crucial while driving at night time or twilight in urban areas as the driver needs to continually readapt and adjust to the items on the road and inside the car. Driver attention depending on the frequency and duration of glances away from the road scene (Jahn et al. 2005). The aim of developing such systems must to create more efficient in-vehicle displays that draw even less attention away from the primary driving task, decreasing driver workload and providing a safer driving experience as a consequence.

The National Highway Traffic Safety Administration Report gives a good overview of which consequences a lack of concentration might bring (Neale et al. 2005). The report states that driver inattention is associated with 78% of traffic collisions and 65% of near collisions. One of the reasons is assumed as “driving related inattention to the forward roadway and non-specific eye glances”. Designing in-vehicle interfaces and cockpit displays that do not distract is advantageous.

1.2 Mesopic Vision in Urban Areas

Driving at night time is more risky than driving during the day. Although only 25% of the overall traffic takes place during hours of darkness, the number of accidents is equal compared to the day time traffic (Rumar 2002). In the dark hours, road lighting and in-vehicle lights play an important role in visual performance during night driving. One major goal when designing in-vehicle information systems is decrease distraction associated with the device. The information displayed on an instrument cluster should be acquired quickly and safely without distracting too much from the primary driving task. Studies on visual performance while driving show that stimuli presented under different lighting levels are interpreted differently. Specifically, the error rate and reaction time increases with a decreasing luminance level. (Alferdinck 2006).

For some brands, the first in-vehicle color concept for the instrument cluster to be used for night-time driving were developed based on the light and luminance used in sea or underground vehicles, such as submarines. In such environments, persons were completely deprived of external light sources and operated in the scotopic visual range (Boyce 2006). The color chosen for those instrument cluster concepts was red as to not disrupt dark adaptation (Purkinje effect). At this lower levels the color red (700 nm) is not affected, because the sensitivity of the human eye shifts toward the blue end of the color spectrum as can be seen in Fig. 1. If the dark adapted eye senses only light of wavelength at the right end of the spectrum like red, then the rods of the eye won’t become saturated and stay adapted to the dark, because they are not sensitive to long-wavelengths.

Fig. 1.
figure 1

The CIE spectral luminous efficiency functions for photopic vision, V(k) and V10(k), and scotopic vision V’(k) compared with an example of a tentative spectral mesopic function for a typical mesopic light level (based on Alferdinck 2006, p. 265)

Full size image

Ordinarily, driving in rural areas at night time is classified as scotopic. However, an urban setting at night or twilight consists of many signals, signs, other vehicles and external street lamps that serve as light sources in addition to instrument cluster lights (Stockman and Sharpe 2006; Viikari 2008). Furthermore, other internal light sources such as navigation or entertainment systems also contribute to the overall lighting. These multiple light sources can activate mesopic vision in the driver. Since previous studies have shown that urban driving does not operate under scotopic vision, the possibility of a new instrument cluster lighting concept, other than red, becomes available.

In this study, our main objective was to compare task performance between two interior color concepts (old and new) in terms of interpretability and differentiability, attention, readability under mesoscopic lighting conditions. Specific to these terms, we defined the following research questions:

  • Interpretability and differentiability. How accurately can the information be read from each display? How efficiently can targets be recognized and identified?

  • Attention. How quickly can the information be read from the display?

  • Readability. Can the information presented in each cockpit be seen with the same efficacy? What is the detection threshold for information presented on the display?

In this paper, only the terms specified in the first two points are addressed. Both subjective (e.g., questionnaire-based; already published in (Götze et al. 2013) and objective (e.g., performance-based) data was collected in this experiment. The participants performed two separate experiments evaluating the aforementioned criteria. In this paper, only the first experiment is presented. The subjective data using a modified Post-Study Usability Questionnaire (Lewis 2002) was published earlier (Götze et al. 2013) and, the second experiment was published in Götze et al. (2014).

2 Method

2.1 Mesopic Vision Conditions

The aforementioned mesopic vision condition is one of the three light levels defined for the human vision. The visual system operates over a wide range of luminance from about 10–6 up to 106 cd/m2. Photopic vision is the luminance level defined as higher than 3–10 cd/m2 and is usually obtained at day time. Starting with twilight or even at night with streetlamps, moonlight or stars, the mesopic vision condition is active.

Depending on the literature the mesopic vision range is defined from 0.001 up to 3-10 cd/m2 (Boyce and Rea 1987; Dacheux and Raviola 2000). It is acknowledged that the upper luminance limit of this range cannot be precisely defined (Viikari et al. 2005; Plainis et al. 2005).

All three vision conditions have different characteristics: Photopic vision allows color vision and a good detail of discrimination, mesopic vision is associated with diminished color vision and less detail discrimination while scotopic vision is associated with no color vision and poor detail discrimination (Boyce 2006).

2.2 Instrument Cluster and Colors Used

In this study two different cockpit color concepts of an instrument cluster were compared in order to determine if a new color concept (e.g., white) is not worse than the currently used red color concept as can be found in many mid-range cars. Figure 2 shows the design of the concept as well as the size.

Fig. 2.
figure 2

The cockpit concept used in this study with its measurements

Full size image

The whole cockpit width including two small (fuel and water temperature) and two large rings (speedometer and tachometer) is 314 mm with a height of 117 mm. The diameter of the large rings was 96.4 mm each. All measurements were equal to the actual size of those in instrument clusters used in current vehicles.

The baseline color red had the wavelength of 603.2 nm. The luminance level of the cockpit was tailored for mesopic vision conditions and set to 7.8 cd/m2 while the white display had 11.0 cd/m2.

2.3 Framework Conditions of the Study

The study took place in an experimental room under mesopic lighting conditions in a range from 0.01 cd/m2 up to a maximum of 1 cd/m2 to maintain visual function within a mesopic range.

The experiment included two separate experiments, each consisting of two experimental parts assessing the white and the red color concept. Both experiments and experimental parts were fully randomized among all participants. The second part was described and published earlier as mentioned before (Götze et al. 2014). For this experiment a setting resembling a real vehicle was established. The participants sat in a car seat and had to perform two task simultaneously (see Fig. 3).

Fig. 3.
figure 3

Experimental setting of the first experiment. Participants sat in the car seat and performed two tasks simultaneously. On the lower screen a discrimination task was presented and on the upper screen, subjects observed a continuous tracking task.

Full size image

Participants had to perform two tasks. The main task presented on the larger upper screen was a driving-like continuous tracking task (CTT) demanding continuous attention and control in order to model a real driving situation and to ensure the participants were not constantly observing the cockpit display where the secondary task was shown. The aim of this secondary task was to observe the cockpit display shown on the smaller screen (Fig. 3) and perform a choice reaction time (CRT) task.

The CTT screen used was a Samsung 24’ computer screen (contrast 20000:1, reaction time 5 ms). The distance between the CTT screen and the cockpit screen was 60 cm. The cockpit screen had a presentation angle of 24’ and was placed on a 66 cm table in a distance of 70 cm from the participant’s eyes. All distances and presentation angles matched a real car setup.

In order to respond to the CTT, a joystick was placed on an arm rest to the right of the participants. In addition, a keypad to the left was used to respond to the CRT task.

2.4 Continuous Tracking Task (CTT)

A CTT is a visual-manual task, which requires continuous control activity of the participant (Eichinger, 2011). The task in this type of CTT was to control the position of a vertically and horizontally moving cross-hair toward a central point using a joystick. The movement of this target cross can be described as a first order instability. With no influence of the participant on the cross-hair movements, the cross-hair divergently approaches the display edges. In order to bring the target back to the center of the screen, participants have to make a corrective joystick input.

2.5 Discriminating Warning and Information Signals

The aim of the discrimination task was to observe the cockpit display and respond to small arrows in different colors (green, yellow, and red) presented in the upper part of the cockpit (Fig. 4). The chosen colors corresponded to regular colors in a car used for warnings (yellow), danger (red) and informative signals (green). The wavelength of the specific colors presented was as following: red 609.4 nm, yellow 578.8 nm and green 554.4 nm.

Fig. 4.
figure 4

Cockpit display with a left arrow shown in the upper mid. Participants had to respond to this arrow in time using the keypad.

Full size image

Participants had to respond to the direction of the presented arrows using the keypad. Only left and right arrows were used. In addition, three different positions of the arrows were shown to prevent guessing of the participants. In summary, a total of 12 different options per cockpit color could appear (2 directions [left/right] * 3 colors [red/yellow/green] * 3 positions [left/mid/right]).

The following performance errors were possible: Miss (participant did not respond in time), erroneous response (participant responded with the incorrect direction) and false alarm (participant responded with no arrow shown). If the participant discriminated and indicated the correct direction within the permitted timeframe, the response was a hit.

One experimental block (with either red or white display concept) contained of 108 trials and lasted 10 min. To decrease the predictability of the arrows even more, three inter-stimulus intervals were used: 3000, 5000 and 7000 ms. The signal presentation time was 1000 ms (as per Jahn et al. 2005; Merat and Jamson 2008). The task was to be as fast and as accurate as possible.

2.6 Procedure

After each volunteer’s arrival at the laboratory, information about the study was provided. All had to perform a visual acuity test using the Landolt C as well as a color vision test to check for visual impairments. Participants then completed a demographic questionnaire before initiating a dark adaptation period, which lasted at least 20 min (Lamb and Pugh 2004). Participants then started to perform the main experimental tasks. Before each experimental part, a training (54 trials) was carried out in order to counteract learning effects. After each part, participants answered the Post-Study Usability Questionnaire adapted to our study (Götze et al. 2013).

After the first experimental part, which contained two blocks (2 × 108 trials) of a particular display color concept, a five-minute block performing only the CTT without a discrimination task was performed. This established a baseline level of the CTT performance.

2.7 Data Acquisition

The cockpit display presentations of both experiments were prepared and executed with E-Prime 2.0 (Psychology Software Tools, Inc.). The data of the CRT experiment was acquired for both tasks separately (CRT task and CTT) using two computers. The results and performance time of the CTT was saved in a MySQL database. The reaction times and error rates of the CRT task were recorded with E-Prime. Questionnaires after each experimental part were performed orally by the supervisor since the mesopic vision of the participants did not allow otherwise.

2.8 Participants

Thirty healthy male volunteers participated in this study. The sample was limited to only male participants due to findings indicating possible gender effects on visual task performance (see Der and Deary 2006). All volunteers were aged between 24 and 54 years, with a mean age of 43 years. None of the participants reported to suffer from any visual or motoric impairments. The visual acuity and color vision tests were also successfully completed by all participants. The driving experience varied between 15 and 36 years with an average of 25 years. The driving distance per year ranged from 8000 to 70000 km with a mean of 25000 km.

3 Results

Statistical power of the analysis is assessed by two types of errors, α-error and β-error. In this study, in order to increase the power of the analysis, the α-error was established on the level of α = 25%, simultaneously decreasing the β-error to β = 75% (Bortz 2005).

3.1 CTT Performance

The Continuous Tracking Task was performed simultaneous to the CRT task in order resemble a real driving, multitask situation. The performance of the CTT was assessed as a standard error of the regression (RMSE) for each participant. Additionally, the mean global performance level and the global standard deviation (SD) of all participants were calculated. Participants with a result lower than mean ± 2.0 SD in at least two blocks were excluded as they did not fulfill the criterion of carrying out both tasks simultaneously. Three participants were excluded in this step (Table 1).

Table 1. Mean RMSE and SD of the CTT in particular experimental blocks are presented. Twenty-seven of 30 participants performed the CTT in the range of M ± 2 SD of the global performance of all participants.
Full size table

3.2 CRT Task – Error Rate

The error rate for each participant was calculated in order to examine the difference in global accuracy between the white and red cockpit color concept. The global accuracy is defined as the number of correct responses. The total number of trials was 432 (two blocks of each color with 108 trials) (Table 2).

Table 2. Error rate with SD and global accuracy of the CRT task. Twenty-seven of 30 participants performed the participants.
Full size table

A paired-samples two tailed t-test was used to examine differences in error rates between the two lighting concepts. There was no significant difference in global error rate between the white (M = 0.0174, SD = 0.0113) and red (M = 0.0167, SD = 0.0114) display found, t(26) = 0.5273, p = 0.6024, see Fig. 5.

Fig. 5.
figure 5

Choice reaction time task. Mean global error rate for the white (M = 0.0174, SD = 0.0113) and red (M = 0.0167, SD = 0.0114) cockpit display. No significant difference between the two color concepts was found.

Full size image

3.3 CRT Task – Reaction Time

For the reaction time analysis only hits were used. Furthermore, trials shorter than 200 ms and higher than 1500 ms were excluded as outlines. Additionally, for each participant the mean reaction time was calculated and trials in a range M ± 2.5 SDs were excluded (287 trials) (Baayen and Milin 2010) (Table 3).

Table 3. Reaction times of the CRT task. Only hits were used. Trials shorter 200 ms and higher 1500 ms were excluded.
Full size table

The mean global reaction time for the two lighting color concepts across all participants was calculated. A paired-sampled t-test was executed to examine any difference in global reaction time between the two colors. There was no significant difference found between white (M = 716.50 ms, SD = 89.26) and red (M = 717.10 ms, SD = 84.55) cockpit display, t(26) = 0.077, p = 0.939, see Fig. 6.

Fig. 6.
figure 6

Choice reaction time task. Mean global reaction time for the white (M = 716.50 ms, SD = 89.26) and red (M = 717.10 ms, SD = 84.55) cockpit display. No significant difference between the two was found.

Full size image

3.4 CRT Task – Reaction Time for Different Arrow Colors

To examine the effect of color concept on reaction time to the three different arrow colors (red, yellow, green), a two-way repeated measures ANOVA was performed for the two cockpit color concepts. Additionally, paired-samples t-tests were performed to examine any difference between the arrow colors in particular.

The arrow color had a significant effect on the reaction time (independent of the lighting color concept): F(2,52) = 46.166, p < 0.001 (Table 4).

Table 4. Reaction times of the CRT task for a specific arrow color. Only hits were taken into account. Trials shorter 200 ms and higher 1500 ms were excluded.
Full size table

Participants responded to yellow arrows (M = 701.70 ms, SD = 85.72) significantly faster than to both other colors green (M = 720.20 ms, SD = 8801) and red (M = 731.40 ms, SD = 88.37), t green(53) = 5.8302, p < .001, t red(53) = 8.7454, p < .001, see Fig. 7. Furthermore, green arrows were recognized and responded to significantly faster than red ones, t(53) = 3.4190, p < 0.001.

Fig. 7.
figure 7

CRT task. mean reaction time of the different arrow colors presented on the display. Merged data of both color concepts.

Full size image

A significant interaction between the cockpit color and the arrow colors was found, F(2,52) = 5.469, p = 0.007, see Fig. 8. No significant difference was found between the green arrows presented on the white (M = 718.4 ms, SD = 92.12) and red (M = 722.10, SD = 85.43) color concept, t(26) = 0.3994, p = 0.6929. There was a significant difference found between the red arrows presented on the white (M = 737.3 ms, SD = 92.89) and red (M = 725.5, SD = 84.96) display, t(26) = 1.6085, p = 0.1198. No significant difference was found between the yellow arrows presented on white (M = 697.6 ms, SD = 87.40) and red (M = 705.8 ms, SD = 85.40), t(26) = 0.9287, p = 0.3616, see Table 5.

Fig. 8.
figure 8

CRT task. mean reaction time of the different arrow colors (plotted on the x-axis) presented on the display. Separated data of both color concepts.

Full size image
Table 5. Reaction times of the CRT task for a specific arrow color separated by color concept. Only hits were taken into account. Trials shorter 200 ms and higher 1500 ms were excluded.
Full size table

3.5 CRT Task – Reaction Time for Different Arrow Positions

A two-way repeated measures ANOVA was performed to examine the effect of an arrow position and the cockpit color concept on reaction time. Moreover, paired-samples t-tests were executed to examine the difference between arrow positions in particular (Table 6).

Table 6. Reaction times of the CRT task for a specific arrow position separated by color concept. Only hits were taken into account. Trials shorter 200 ms and higher 1500 ms were excluded.
Full size table

The arrow position had a significant effect on the mean reaction time F(2,52) = 8.201, p = 0.001. There was no significant interaction found between the lighting color concept and the arrow position F(2,52) = 0.250, p = 0.780, see Fig. 9.

Fig. 9.
figure 9

CRT task. mean reaction time of the different arrow colors (plotted on the x-axis) presented on the display. Separated data of both color concepts, white (gray line) and red (black line).

Full size image

4 Discussion

The main aim of the current study was to compare two colored cockpit concepts (white and red) in terms of interpretability, differentiability and attention under mesopic conditions (relevant for driving in urban areas at night or twilight). A Choice Reaction Time task was used to facilitate a comparison between the accuracy and speed of recognizing a specific signal in the cockpit display while doing a driving-like task. The aim was to investigate whether the two color concepts yielded the same results for the accuracy and reaction time using different signal colors (red, yellow, and green) and positions (left, center, and right) with a CRT task. No significant difference for the accuracy was found. Furthermore, no significant difference for the reaction time depending on the position and direction was found. For the colors, no significant difference between the cockpit concepts were found for yellow and green; exclusively the red color showed a significantly faster reaction time with the red color display compared to the white one. Taking these results into account as well as the results of the previously published parts of this study, no difference between the two concepts can be established. The interpretability, differentiability, readability and attention of and towards information under mesopic vision, based on the performance and data recorded in this study, is independent of the lighting color concept. Possible future directions may involve investigating possible effects that may occur when drivers’ vision has not yet adapted to the environment and or is interrupted in some way.

5 Conclusion

Summing up, the study shows that a white color lighting concept for in-vehicle cockpits, as tested, has neither objective and nor subjective disadvantages over a red color concept under mesopic vision. These results can help in-vehicle display designers or car manufactures with new possibilities for lighting concepts in urban areas