Method and analyses
This research was conducted as a quasi-experiment and combined elements of field observation with formal experimental design (Lehman, 1991). This approach allowed for conventional ‘pre- and post-intervention’ assessment to be conducted so that any effects of the PRIMEs on riding behaviour could be determined against the baseline measures. This approach has been used in similar research of this kind (Fildes et al, 2005; Mackie and Scott, 2015; Hirsch et al, 2017, 2018; Stedmon, McKenzie, Langham, McKechnie, Perry and Wilson, 2021, 2022, 2023; Winklebauer et al, 2021).
Participants
This research relied on an opportunistic sample of motorcyclists. Across all the trial sites 32,213 motorcycles were observed and from these 9,919 lead motorcycles were analysed in more detail.
Apparatus
Data were captured at each site using small and inconspicuous weatherproof video cameras typically attached to roadside posts or trees (Figure 5).
The cameras captured 1080p video at 60Hz for time periods of at least 20hrs, stored in 512Gb microSD cards. Power-packs were used to collect continuous data through the data collection periods.
At each site, three cameras were installed facing: towards the rider, behind the rider and perpendicular to the rider a short distance ahead of the last PRIME road marking (Figure 6).
The road markings were installed using 3MTM StamarkTM High Performance 100 mm wide permanent tape. This material was chosen because it provided increased visibility, grip and safety, even in the wet. It had also been used in previous research (Bricelj, Merkun, Brumec and Hudej, 2016). The material also offered high levels of adhesion to the road surface and provided a permanent marking that would not be disturbed by other vehicles (i.e. general traffic and heavy goods vehicles).
Design
For the research, funded by the Road Safety Trust, a specific focus on left-hand bends was proposed. Other funding provided opportunities to conducting initial investigations of right-hand bends.
The independent variable in this research was the PRIME road markings which had two levels: Baseline (without PRIMEs installed) and PRIME (with PRIMEs installed).
Baseline and PRIME data were collected on a number of occasions, as specified below:
- Baseline 1 and 2 – two separate weekends before PRIMEs were installed
- PRIME 1 – the weekend after PRIMEs were initially installed
- PRIME 2 – six or eight weeks after the PRIME 1 data collection
- PRIME 3 and 4 – to investigate the nature of sustained behaviour effects
- PRIME 5 and 6 – to investigate the nature of long-term behaviour effects
A range of dependent variables were identified to capture data about the potential influence of PRIMEs on rider behaviour (Table two).
| Dependent variable | Measure (units) | Apparatus | Reason |
|---|---|---|---|
| Speed | Miles per hour (mean, standard deviation, mode, 85th %tile) | Side facing camera to measure speed between two points (i.e. 10m apart) | To assess any changes in speed due to PRIMEs |
| Position | Lateral lane position at the final PRIME and at the apex (mean, standard deviation, median, mode) | Forward-facing camera to measure lateral position on approach to the bend. Rear-facing camera to measure position at the apex | To assess any changes in lateral lane position due to PRIMEs |
| Braking | Brake light illumination (count) | Rear-facing camera to capture brake light illumination | To assess any changes in braking on bends due to PRIMEs |
| Motorcyclists | Pillion, lead, group riders (count) | From video data collected for other measures | To identify rider characteristics |
| Use of PRIMEs | Use of the final PRIME (count) | From video data collected for other measures | To identify how many riders used PRIMEs |
In addition, rider interviews were conducted using a scripted protocol during the 2021 and 2022 road trials. These took place at the Green Welly Stop at Tyndrum and Inveraray waterfront as they were both popular meeting points and ride-out destinations for motorcyclists. The findings are presented in the summary results.
Procedure
Prior to data collection, trial sites were assessed and where necessary upgraded with various measures such as: resurfacing, line repainting, new crash barriers, vegetation removal, vehicle restraint systems (VRS), and motorcycle friendly ‘bikeguard’ installations. Each location therefore provided a setting where the road environment had been brought up to the best possible standard, prior to investigating rider behaviour. Comparison sites were selected that were of a similar standard and did not require engineering works. This meant that any extraneous variables were controlled as much as possible so that they would not otherwise influence rider behaviour (e.g. poor road surface, obscured views, potholes, poor safety provisions).
The weather during the trials was generally good. Contingency weekends were planned in case poor weather may have affected the numbers of motorcyclists out on the roads and integrity of the data. If light rain showers occurred the data were generally included for analyses. Only if the showers were heavy and caused wheel spray or if other vehicles had their windscreen wipers operated was the data excluded from analyses.
Data were captured during the typical motorcycle season (i.e. May to September). Weekends were chosen for as this was generally when motorcyclists ride for leisure/social purposes.
Each weekend cameras were set up at every trial site and recorded all road traffic during Saturday and Sunday from 09:00 to 17:00. Power supplies were replenished through the weekend and cameras were collected on Sunday evenings.
Care was also taken to make sure that no changes to the sites were undertaken during the pilot trials (i.e. scheduled road works).
Ethics and risk assessment
An independent review of potential ethical issues was conducted by Dr Martin Langham who acted as an external auditor for the project. Approval was granted in accordance with general principles of the British Psychological Society and International protocols.
A risk assessment was also conducted in order to safeguard the research activities. Induction training was undertaken so that roadside safety protocols were adhered to and the correct PPE was worn at all times.
The design for the PRIME road markings and road sign went through a formal application process for authorisation of non-prescribed traffic signs (Road Traffic Regulations Act 1984: Sections 64 and 65). Approval was granted prior to the trials taking place. Following on from this, independent road safety audits were conducted in accordance with the Design Manual for Roads and Bridges (DMRB) to oversee the safe installation of PRIMEs at all trial sites.
For more details on the method and analyses, please refer to Stedmon, McKenzie, Langham, McKechnie, Perry and Wilson (2021, 2022).
Data analyses
Once the data had been processed, they were then analysed in a number of ways.
Initially Baseline 1 and Baseline 2 datasets were compared by conducting a T-Test (t) to identify any differences between them. Where any significant differences were observed, effect size was calculated using Cohen’s (ds) equation.
Where the Baseline 1 and Baseline 2 datasets were observed to be the same (i.e. there was no significant difference) they were combined into a single dataset (i.e. ‘Baseline’). Where any difference was observed, Baseline 1 and Baseline 2 were kept as separate datasets and compared individually with the PRIME 1 and PRIME 2 datasets.
Speed and lateral position data were analysed using one-way Analysis of Variance (ANOVA) techniques. Where any significant results were observed, effect size was calculated using a partial eta squared (η2) analysis. Post-hoc Bonferonni-Hoch analyses were conducted in order to determine where significant differences occurred between the datasets. Tests for effect size were conducted using Cohen’s (ds) calculations.
Braking behaviour and use of PRIMEs datasets were analysed using Chi Square (X2) tests. Where any significant results were observed, effect size was analysed using Cramér’s V (V) calculations. Further post-hoc analyses were performed by calculating standardised residuals in order to determine where significant differences occurred between the datasets.
Due to project restraints and as the data processing relied on specific and discrete manual counts, one researcher (Prof Stedmon) conducted the data processing and analyses. This researcher reviewed and re-checked data during the data processing activities. During the 2020 trials intra-rater reliability was assessed instead of inter-rater reliability (Stedmon, McKenzie, Langham, McKechnie, Perry and Wilson, 2021). This followed the process set out by Mackey and Gass (2005) where ratings were conducted at different time intervals (i.e. T1 and T2) and then analysed in the same way as inter-rater reliability. Cohen’s Kappa (k) calculations were conducted for samples of data for speed, lateral position and braking in the 2020 road trials. For speed and braking perfect matches were observed (k=1.0) due to the discrete nature of these data. For lateral position k=0.92 indicating a very high agreement and only minor differences in coding at the thresholds between the three lane positions.
For detailed statistical analyses for each site please refer to Stedmon, McKenzie, Langham, McKechnie, Perry and Wilson (2021, 2022, 2023).