Measurement of difference thresholds for vehicle acceleration
Experiment 5: Absolute difference threshold for mid-range accelerations
Since Experiment 3 established that the driving simulator did not have absolute validity for perception of acceleration differences (because of inconsistencies between acceleration cues in different modalities, and the inadequacy of vestibular cues), a further experiment was carried out to measure the absolute difference threshold for mid-range accelerations, using the same experimental design but with a real car on a test track.
Method
Experimental design
Each participant completed 16 pairs of experimental drives, in which VAO was paired four times with each of VAO, VA3, VA6 or VA9. The order of presentation was opposite for two o f the four to control for order effects. The order o f pairs was counterbalanced for each participant. The dependent variable was again frequency o f correct pair-wise judgments.
Confidence in judgments, and self-reports o f cues used
Participants were asked to report their level of confidence in their pair-wise judgments, in the same way as in Experiment 4. At the end of the experiment they completed a questionnaire in which they rated the importance in their judgments of a range of perceptual cues, again using a six point scale. The items in this questionnaire differed somewhat from those in Experiment 4, because of differences in the cues available in the real car/test track perceptual environment:
• Distance to visual reference point • Time to reach a visual reference point • Passing roadside objects
• Passing road markings • Loudness of engine noise • Pitch of engine noise • Monitoring the Rev Counter • Monitoring the Speedometer • Feeling pushed into the seat • Tilt of the vehicle
Test vehicle
The test vehicle was a 2007 Mk5 Golf TSi with a 1.41itre spark-ignition (gasoline) engine (Figure 6-11). In order to achieve acceleration levels approximately consistent with those in Studies 3 and 4, the engine output was reduced via the engine control unit (ECU) to 90bhp (from its standard 122bhp). The vehicle was fuelled with Shell V-Power (99 RON24).
Different levels o f YAmax were achieved using a controller provided by Revo Technik, connected to the car’s On-Board Diagnostics (OBD) port, which communicated with the ECU to increase engine output as shown in Table 6-9.
Data for velocity and position were recorded on each drive using a Racelogic VBOX Micro mounted inside the vehicle, equipped with GPS to locate the position of the vehicle to within ±5m, and recording data at a sampling frequency of 10Hz. Mean acceleration rates for the four engine output conditions are also shown in Table 6-9. The rate for condition VA3 was higher than expected, indicating that the controller did not function as intended in this condition. Nevertheless, using the mean acceleration rate, this condition still provided a useful data point.
Figure 6-11. Experim ent 5 test vehicle (2007 Mk5 Golf 1.4 TSi, gasoline fuelled)
VA condition Nominal engine output (bhp) Actual mean acceleration (m s'2) S tandard deviation of actual acceleration (m s'2) VAO 90.0 1.02 0.11 VA3 92.7 1.05 0.11 VA6 95.4 1.06 0.09 VA9 98.1 1.08 0.10
Table 6-9. Experim ent 5: test vehicle engine output for each VAmax condition
Test track and route
The test route used the TRL Large Loop test track, driven in a clockwise direction (Figure 6-12). The Large Loop was 2.25km in length and similar in design to a single side of a motorway,
including gantries, crash barrier and signage. There were two main sections to each drive: the acceleration zone where participants accelerated the vehicle from 30mph to 55mph in 4th gear, and the question zone, where participants were required to choose the drive on which the car had accelerated the fastest. The acceleration zone was level and enclosed by trees which eliminated the possibility of trackside distractions. The preceding long curve gave ample time for participants to reach and maintain 30mph.
The acceleration zone consisted of four main sections (Figure 6-13):
1. 30mph zone, where participants were required to achieve a speed of 30mph and maintain 4th gear in preparation for the start of the acceleration event
2. Starting point for accelerations, indicated by a change in surface type, and by traffic cones. 3. Target speed zone, where most drives reached the target speed of 55mph.
4. Deceleration zone, where drivers began to slow down for the bend ahead
Question zone
Acceleration zone
Zone where 50mph+ was typically reached Deceleration zone Start point of acceleration 30mph zone
Figure 6-13. Acceleration zone
Procedure
Each participant was accompanied by a TRL researcher throughout the track session. The researcher initially drove the test vehicle to the question zone, where the participant was briefed, then drove to the acceleration zone and demonstrated how to complete the acceleration. Participants were briefed to depress the accelerator pedal fully when accelerating, and keep it fully depressed until the dashboard speedometer indicated they had exceeded 55mph (thus ensuring that the pedal remained fully depressed as the speed reached 50mph). The researcher and participant then exchanged places in the question zone and the participant began the first of two practice drives (further practice drives were made if a participant failed to complete the first two correctly).
On each experimental drive, participants attained and maintained a speed of 30inph in 4th gear in the 30mph zone. On reaching the starting point they fully depressed the accelerator pedal, keeping it fully depressed until the speed had exceeded 55mph, at which point they braked to decelerate. Participants then drove round the loop to the question zone. The researcher then adjusted the VAmax condition using the controller (out of sight of the participant), and, after each drive pair, asked the
participant to select on which drive the car had accelerated the fastest. For the first two drive pairs, participants stopped in the question zone, but for later pairs they were allowed to choose to give their responses while on the move (slowly) through the question zone, and all chose to do so. Completion o f all 16 pairs took approximately 1.5 hours, and participants took a 10-minute break after 8 pairs.
On completion of all pairs, the participant and researcher again exchanged places, and the researcher drove the test vehicle off the test track.
Analysis
The ratios of correct to incorrect responses for each VAmax condition (using the mean actual acceleration rates) were fitted to a cumulative normal distribution using the Probit regression procedure with maximum likelihood estimation.
Participants
49 participants completed the experiment (all different from those in Experiments 3 and 4). 24 participants were male, 25 female and had a full UK driving licence. Again the sample was intentionally recruited from relatively experienced drivers (mean number of years of driving experience 21.0 ± 13.2; mean age 40.3 ± 14.5 years) to ensure that participants were able to attend to the perception task rather than requiring cognitive resources for the driving control task.
Results
Perception o f acceleration differences
Figure 6-14 shows the main results. The ratios of correct to incorrect responses were clearly larger than those measured in Studies 3 and 4 for the equivalent speed range in the driving simulator (Figures 6.6 and 6.8), and all were statistically significant.
Table 6-10 summarises the Probit regression model fitted to the data, and the difference thresholds calculated from the model.
100 90
u. 80
(U to§
7 0 CL (/> CL) 60 O 50 <u HO <2 40 8 30 cu 20 10 0 Correct ■ Incorrect 3 4 7A ctual acceleration d iffe ren ce (%)
Figure 6-14. Experim ent 5: Perception of acceleration difference, 4th gear, 30 to 50 mph, 2007 Mk5 Golf 1.4 TSi, gasoline fuelled, test track
Condition Intercept P 5C2(2) P D L .7 5 relative to DL75 (ms'2) standard stimulus VAO (% ) DLC relative to standard stimulus VAO (% ) 30-50mph acceleration 0.011 0.026 0.606 0.739 0.08 7.66 11.34
Table 6-10. Experim ent 5: Regression model and difference thresholds, 30mph - 50 mph acceleration
Vehicle acceleration
condition pair Confidence rating (SD)
VAO-VAO 6.19(2.25)
VA3-VA0 6.45 (2.14)
VA6-VA0 6.34 (2.14)
VA9-VA0 6.46 (2.06)
Table 6-11. Experiment 5: Participants’ ratings of confidence in th eir pair-wise judgm ents
Confidence in pair-wise judgments
Table 6-11 shows participants’ mean ratings of confidence in their pair-wise judgments for each VAmax condition. Ratings were similar to those for the equivalent speed range in Experiment 4 (if anything, slightly lower). The small differences between VAmax conditions were not significant.
Self-reports o f cues used in pair-wise judgments
Table 6-12 shows the means o f participants’ ratings o f how much different perceptual cues influenced their pair-wise judgments about the drive in which the acceleration was fastest. The most highly rated cues were distance and time to a visual reference point, and monitoring the speedometer. Neither loudness nor pitch o f the engine noise, nor feeling pushed into the seat, was rated particularly highly.
Perceptual Cue Mean rating (SD): how much cue
influenced pair-wise decision
Distance to visual reference point 3.57(1.27)
Time to reach a visual reference point 3.18(1.35)
Passing roadside objects 2.24(1.44)
Passing road markings 1.75 (1.23)
Pitch of engine noise 2.06(1.28)
Loudness of engine noise 1.96(1.28)
Monitoring the Rev Counter 1.45(1.46)
Monitoring the Speedometer 3.47(1.30)
Feeling pushed into the seat 2.08(1.40)
Tilt o f the vehicle 1.02(1.03)
6 point scale with verbal anchors fo r each point (1 = None, 2 = Very little, 3 = Little, 4 = Moderate, 5 - Much, 6 - Very Much).
Table 6-12. Participants’ ratings of how far different perceptual cues influenced their pair wise judgments about on which drive the acceleration was fastest
Discussion
Perception o f acceleration differences
The most striking aspect of Experiment 5 is that the difference threshold measured for mid-range accelerations (0.08 ms'2) was around half the values measured in the driving simulator, thus confirming my earlier conclusion that the driving simulator did not have absolute validity for perception of acceleration. This value was also around half the values measured by MacNeilage et al. (2007), supporting the suspicion I expressed in Chapter Four that difference thresholds in actual driving might be lower because o f the much richer mix o f cues available. As I noted in Chapter Four, the difference in manufacturers’ claimed 0-60mph acceleration times between engine size increments tends to be in the region of 12-15% (e.g. between 1.41itre and 1.61itre engine versions of the same model). It seems reasonable to expect that manufacturers have learned to establish engine size increments that provide acceleration benefits exceeding the difference threshold. 12-15% differences just exceed my measured value for DLa, meaning that drivers would be able to perceive them on most occasions, at least in a careful paired comparison, if not necessarily in naturalistic driving.
Confidence in pair-wise judgments
Confidence ratings again did not increase significantly with increasing VAmax difference, even though the frequency of correct responses was higher than in Experiment 4. This supports the inference drawn from Experiment 4 that judgments were being made via non-conscious perceptual processes.
Perceptual cues used, as reported by participants
Among the most important cues that participants reported using were distance and time to a visual reference point. These cues bear some similarity to the time to lead vehicle cue in Experiment 4 (although visual reference points available to Experiment 5 participants were all stationary, rather than moving as was the lead car in Experiment 4). Perhaps rather surprisingly, neither o f the auditory cues, (loudness or pitch o f engine sound) was rated highly. Neither was the feeling o f
being pushed back into the seat, a cue that was missing from the driving simulator studies. Yet the frequency of correct judgments was significantly higher in Experiment 5 than in Experiments 3 and 4. This again suggests that participants were not consciously aware of the cues that were influencing their judgments.