• STC exposure. Research has indicated that crash risk is significantly elevated within the first
six months of licensure. Due to insurance considerations and the need for a driver to possess
rudimentary driving skills, the current work involved teens who had a minimum of six
months of driving experience. It is possible that the effects of the STC observed in the
current work may be different (e.g., accentuated) for teens within the first six months of
licensure. Certainly, if the STC can guide teens towards positive driving behaviors, results
may show a greater positive effect during STC exposure and potentially stronger carry-over
effects after the STC is deactivated. It would be beneficial to conduct a study similar to that
presented here but with teen drivers within the first six months of licensure. Furthermore,
the positive results obtained for the STC from the present research effort would likely carryover
to older teen and adult novice drivers. The degree of benefit for these additional groups
is not known and would require additional research.
• Duration of treatment exposure. Several of the dependent variables exhibited significant
findings and trends throughout the data acquisition stages that were indicative of both
positive and negative behavioral adaptation through continued STC exposure. However, the
duration of STC exposure was limited to six weeks due to project timing and financial
considerations. It would be informative to extend the period of STC exposure to document
more fully the longer-term effects of STC use.
• Duration of transfer exposure. Similar to the duration of treatment exposure an extended
period of transfer exposure would provide additional information regarding the effects of
STC deactivation on driving behaviors. In particular, does the STC facilitate improved
driving behaviors beyond the two-week no-STC timeframe employed in the current work?
At what point might any improvements observed in the transfer stage asymptote and how
long before the effects diminish?
• Participant sample size. The sample size of the STC field evaluation was limited, with 15 teen
drivers at each of two locations (N=30). While preliminary power analyses indicated this
sample size would be sufficient, a study of this magnitude and potential real-world impact on
teen driver safety would benefit from an increased number of participants to determine if the
positive trends in the datasets would be statistically significant. A larger study would more
reliably determine what behaviors are likely to be affected by the STC and could more
precisely quantify the degree of behavioral change. One of the findings from the current
work is that several of the follow up statistical interactions, main effects, and paired
comparisons were not significant despite marked differences between the means (which
were often also paired with low standard error). Many of these did not reach significance due
to the extent of the Bonferonni correction employed to control for the increased number of
comparisons. Increasing the participant sample size might allow these trends to reach
significance thus allowing researchers to have greater confidence in the direction of the data
trends.
• Control group. The addition of a control group in future research is recommended. The
control group would be a between subjects factor that would monitor drivers in parallel with
the treatment group. A control group could be further matched based on gender and driver
53
experience. The control group would then allow for a comparison between actual treatment
effects compared to those individuals without the STC.
• Sophistication and performance of the implementation. In order to implement STC
capabilities in participants’ personal vehicles for the field evaluation, it was necessary to
devise portable, inexpensive “aftermarket” instrumentation that would not damage, mar, or
otherwise interfere with the vehicle. Actual OEM implementations would be expected to
have much better integration, accuracy, and reliability. Limitations to the tested system in
terms of reliability, nuisance alarms, intrusiveness, and esthetics may have influenced system
effectiveness or user attitudes.
• This project had the objective of demonstrating a prototype STC. The features and
functions incorporated into the system were based on consideration of the literature on
teenage drivers and their associated crash characteristics. In implementing these functions,
design features and parameters, such as thresholds, feedback display timing and
characteristics, and adaptive algorithms, the research team had to make selections based on
existing literature, expert judgment, or informal piloting. Implementation was also based on
cost considerations for a non-intrusive experimental platform introduced into a participant’s
vehicle. The project did not have the much more extensive resources that would be required
to empirically determine the optimal design choices for every function. Rather, bestjudgment
choices were used to provide a proof-of-concept. At this point, it may now be
useful to conduct further study to optimize system design, which may enhance the findings
and trends observed in the present study.
• While the results of the current work did indicate some positive benefits of STC use, we
encourage readers to interpret the results with caution due to the limitations associated with
the current work. Further large-scale research is required that examines increased exposure
to an STC and comparisons to a control group. Furthermore, expanding the participant pool
to other states will likely encourage exposure to varied driving behaviors of teens around the
country that may in turn reflect different seat belt use rates, speeding behaviors and the
overall utility of the STC.
• Finally, as with many research efforts, there is a potential confound with respect to
participant engagement in the study. In particular, participants were aware that data
collection equipment was installed in their vehicle and they may have biased their behaviors
simply due to the presence of the equipment regardless of whether or not it provided
feedback. As with most behavioral studies we assume this confound was omnipresent and
likely influenced all drivers similarly in the current work. It was beyond the scope of the
current work to examine the extent of the confound’s influence.
• Previous research (Dingus et al., 2006) of longer-term monitoring of driver behavior
indicates that drivers seem to forget they are being monitored early in the study period.
Relative to the current work an examination of driver behavior between the treatment
groups indicated that those drivers in the baseline condition did not significantly modify
their behavior either at the beginning of the study period or at the end, suggesting that the
presence of the system alone (i.e., not providing feedback during the baseline periods) had
little effect on driver behavior.