Can a Boxer Engine Reduce Leg Injuries Among Motorcyclists? Analysis of Injury Distributions in Crashes Involving Different Motorcycles Fitted with Antilock Brakes (ABS)

Objectives: Several studies have shown that motorcycle antilock braking systems (ABS) reduce crashes and injuries. However, it has been suggested that the improved stability provided by ABS would make upright crashes more frequent, thus changing the injury distributions among motorcyclists and increasing the risk of leg injuries. The overall motorcycle design can vary across different categories and manufacturers. For instance, some motorcycles are equipped with boxer-twin engines; that is, with protruding cylinder heads. A previous study based on a limited material has suggested that these could provide some leg protection; therefore, the aim of this research was to analyze injury distributions in crashes involving ABS-equipped motorcycles with boxer-twin engines compared to similar ABS-equipped motorcycles with other engine configurations. Methods: Swedish hospital and police records from 2003–2014 were used. Crashes involving ABS-equipped motorcycles with boxer-twin engines (n = 55) were compared with similar ABS-equipped motorcycles with other engines configurations (n = 127). The distributions of Abbreviated Injury Scale (AIS) 1+ and AIS 2+ were compared. Each subject's injury scores were also converted to the risk for permanent medical impairment (RPMI), which shows the risk of different levels of permanent medical impairment given the severity and location and of injuries. To compare injury severity, the mean RPMI 1+ and RPMI 10+ were analyzed for each body region and in overall for each group of motorcyclists. Results: It was found that AIS 1+, AIS 2+, and PMI 1+ leg injuries were reduced by approximately 50% among riders with boxer engines. These results were statistically significant. The number of injuries to the upper body did not increase; the mean RPMI to the head and upper body were similar across the 2 groups, suggesting that the severity of injuries did not increase either. Indications were found suggesting that the overall mean RPMI 1+ was lower among riders with boxer engines, although this result was not statistically significant. The mean values of the overall RPMI 10+ were similar. Conclusions: Boxer-twin engines were not originally developed to improve motorcycle crashworthiness. However, the present article indicates that these engines can reduce leg injuries among riders of motorcycles fitted with ABS. Though it is recommended that future research should look deeper into this particular aspect, the present findings suggest that the concept of integrated leg protection is indeed feasible and that further engineering efforts in this area are likely to yield significant savings in health losses among motorcyclists.


Introduction
Motorcycles provide limited injury protection in the case of a crash (DaCoTa 2012). Previous research has indicated that a crash (Otte 1998). Though other studies have reported that leg injuries account for approximately 50% of all Abbreviated Injury Scale (AIS) 2+ injuries among motorcyclists (MAIDS 2004;NHTSA 2008), the majority of the fatal injuries are to the head, even among riders with helmets (DaCoTa 2012;NHTSA 2008).
Two general mechanisms of leg injury in a motorcycle crash have been suggested: direct impacts to the leg (which would also occur if the leg leaves its space) and crushing of the leg between the motorcycle and the other vehicle or the road surface (Elliot et al. 2003;Otte 1994).
The majority of the original research in the area of leg protection for motorcyclists was conducted in the 1980s and 1990s Rizzi (DaCoTa 2012). A rather simple countermeasure to address leg injuries has been conventional crash bars, usually made of loops of steel tubes projecting to the side of the motorcycle (Rogers and Zellner 1998). Studies based on in-depth investigations of 133 real-life crashes showed no overall benefits; that is, the proportion of injured leg regions was nearly identical for motorcycles with and without crash bars (Ouellet et al. 1987). Though there was evidence suggesting that crash bars were sufficient to preserve the leg space in many crashes, it was argued that leg space preservation was not strongly related to serious leg injuries, mainly because the leg often did not remain in the leg space during the collision (Ouellet et al. 1987). Furthermore, frontal crash tests with conventional crash bars showed greater chest and head accelerations due to the rotation of the upper body (Noordzij et al. 2001;Rogers and Zellner 1998).
The same issue-that is, increased injury risks to the upper body-has been reported in a more advanced leg protector presented by the Transport Research Laboratory (TRL), which was based on external and internal knee energy absorbing regions, a rigid structure supporting those regions, and a fairing enclosure (Chinn et al. 1985). Several crash tests using different methods were performed independently by TRL and by the International Motorcycle Manufacturers Association, resulting in contradictory claims for the efficacy of the TRL leg protectors (Mohaymany and Eghbalian 2007). Further testing was then carried out using methods defined by ISO 13232, using a Hybrid III dummy fitted with frangible legs (Rogers and Zellner 1998). Overall, the crash tests showed a disadvantage for the TRL leg protectors: the risk for leg fractures was reduced, although head injury risks were increased (Rogers and Zellner 1998). As a matter of fact, the TRL leg protectors never saw real-life implementation.
However, other technologies have been introduced to reduce injuries among motorcyclists. One of these is motorcycle antilock braking systems (ABS), which several studies have reported to be effective in reducing real-life crashes and injuries (Highway Loss Data Institute 2013; Rizzi et al. 2015;Teoh 2013). Previous research has shown that ABS generally provides shorter stopping distances (Green 2006) for both experienced and novice riders (Vavryn and Winkelbauer 2004). Though Roll et al. (2009) suggested that ABS may increase riders' confidence when applying full brakes, other studies have pointed out that stability improvements per se may explain the large benefits of ABS (Gail et al. 2009;Kato et al. 1996;Rizzi et al. 2015;Teoh 2013). This aspect may have important implications. It has been suggested that the improved stability due to ABS would reduce sliding crashes (i.e., crashes in which the riders slide on the ground into the collision object), thus making upright crashes more common with ABS. This may change the distribution of injuries among riders of ABSequipped motorcycles . Though the material included in this study was limited, it was found that riders who crashed in an upright position were more likely to suffer leg injuries than prone ones involved in similar crashes.
The fitment rate of ABS in Europe will increase dramatically in the coming years due to the legislation voted by the European Parliament to make ABS mandatory on all new motorcycles with engine displacement over 125cc from 2016 (Eu- ropean Commission 2012). It was therefore argued by Dawkes and Löfdahl (2014) that the development of fairings that offer increased leg protection will be even more important in the future. However, the possibility that some motorcycle designs may inherently offer some degree of leg protection may have not been investigated thoroughly by previous research.
The overall motorcycle design can vary across different categories and manufacturers. For instance, some motorcycles have been equipped since the 1920s with a horizontally opposed flat-twin engine, which means that the cylinders are overhanging horizontally in front of the riders' legs. This engine configuration is also known as boxer-twin engine. Figure 1 shows an illustration of a motorcycle equipped with a boxer-twin engine (top) and a similar one with a singlecylinder engine (bottom).
A previous study (Hurt et al. 1981) collected in-depth data of 900 motorcycle crashes in the Los Angeles urban area using on-scene investigations during the period 1976-1977. The findings showed that leg injuries were less common among riders of motorcycles with boxer-twin engines, although this was based on a very limited number of cases (n = 11). To the author's knowledge, however, no further research based on real-life crashes has investigated this particular issue. Therefore, the aims of the present study were to • analyze the distributions of all injuries in crashes involving ABS-equipped motorcycles with boxer-twin engines compared to similar ABS-equipped motorcycles with other engine configurations; • compare the risk for impairing injuries in those crashes; • investigate whether leg injuries may be reduced in crashes involving ABS-equipped motorcycles fitted with boxer-twin engines.

Material
Swedish hospital records for the period 2003-2014 were acquired from the Swedish national accident database (STRADA). If a crash is also police-reported, it is normally recorded in STRADA with the same crash identification number as the hospital report, which means that hospital data can be automatically merged with police records to obtain vehicle information. Crashes involving motorcycles fitted with boxer-twin engines were identified and compared to similar ones fitted with other engines configurations. These motorcycles were included in the categories touring, standard, custom, sport touring, and on/off. VIN numbers were used to acquire data on motorcycle ABS for those models with optional ABS fitment. In total, 55 crashes involving ABS-equipped motorcycles with boxer engines were compared with 127 involving ABS-equipped motorcycles with other engine configurations. Table 1 gives a brief overview of the material included in the analysis. For both groups, with and without boxer engines, multivehicle crashes were the most common ones; in the majority of those crashes, the opponent vehicle was a passenger car (93 and 84% of cases, respectively). The number of singlevehicle crashes into road barriers was limited in both groups (1 and 2, respectively). Table A2 (see online supplement) shows a detailed list of the make/models. The analysis included only motorcycles with MY 1996 and onwards and at least a 700cc engine displacement. In total 12 models with boxer-twin engines were compared to 31 models with other engine configurations. All of them were ABS-equipped.

Risk of Permanent Medical Impairment
Motorcyclists' initial diagnoses were converted to the risk of permanent medical impairment (RPMI), which is an estimation of the risk of a patient to suffer from a certain level of medical impairment, based on the injuries diagnosed according to the AIS 2005 scale (Association for the Advancement of Automotive Medicine 2005). A medical impairment is considered permanent when no further improvements in physical and/or mental functions are expected with additional treatment and is independent of occupation, hobbies, etc. An injury is given a degree of medical impairment between 1 and 99%. The risk is derived from risk matrices for least 1% permanent medical impairment (RPMI 1+; see Table A1, online supplement) as well as at least 10% medical impairment (RPMI 10+), as presented in a previous empirical study (Malm et al. 2008). For instance, AIS 2 injuries at the lower extremities have a 50% risk of at least 1% permanent impairment (see Table A1). Though PMI 1+ injuries include all levels of impairments, PMI 10+ injuries generally result in persistent symptoms affecting activities on a daily basis (Rizzi et al. 2013).
The overall RPMI for each rider was calculated based on the product of the risks of not being injured, described by Gustavsson (1990); see Eq. (1), where n is the number of injuries sustained by each motorcyclist.
(1) This approach has been used in a number of previous studies to analyze hospital-reported injuries among car occupants (Stigson et al. 2011), pedestrians ), cyclists (Rizzi et al. 2013), as well as motorcyclists ).

Analysis of Injury Distributions
In order to ensure the comparability of the 2 groups, checks were made on the distributions of a number of factors that could possibly confound the results such as age, use of protective equipment, speed area, fitment of combined braking systems (CBS) or traction control systems (TCS), etc.
The number of impairing injuries (PMI 1+) was given by the accumulated risk for permanent medical impairment (RPMI 1+) for each body region. The distributions of AIS 1+, AIS 2+, and PMI 1+ injuries for all body regions were analyzed across both groups. This was done in order to investigate the possibility that leg injuries may be reduced in crashes with boxer-twin engines but also that other injuries to the upper body may be increased, as previously reported for conventional crash bars and the TRL prototype (Noordzij et al. 2001).
Due to the limited size of the material, the 10 body regions originally used in the RPMI matrices (see Table A1) were regrouped into 6 body regions, as shown in Table 2. The difference between the proportions of AIS and PMI injuries for each body region was tested using Fisher's exact test (Agresti 1992). In order to compare injury distributions in the upper body, the expected number of leg injuries among boxer riders was also calculated; that is, the number that would give equal percentages of leg injuries across the 2 groups. This was done by calculating x in Eq.

Rizzi
(2) for AIS 1+, AIS 2+, and PMI 1+ injuries: n expected leg injuries boxer n expected all injuries boxer = n actual leg injuries boxer + x n actual all injuries boxer + x = n actual leg injuries others n actual all injuries others . (2) The overall RPMI for each rider was calculated according to Eq.
(1). In order to compare the severity of the reported injuries, the mean values of RPMI 1+ and RPMI 10+ (mRPMI 1+ and mRPMI 10+) were also calculated, as previously done in other studies Strandroth et al. 2012). This was done for each body region separately, as well as for all motorcyclists, to calculate the overall injury severity in each group. The relative difference between the 2 groups' mean values of RPMI was calculated and tested by an independent 2 sample t test conducted for unequal sample sizes and variance. Finally, the location and AIS severity of leg injuries among the 2 groups were compared by calculating the risk for AIS 1+ and AIS 2+ injuries for specific leg portions (hip, femur, knee, tibia, ankle, foot). As an example, the risks for AIS 2+ knee injuries are shown below. Similar calculations were performed for AIS 1+ injuries and for the other leg portions mentioned above.

Results
The comparison between crashes involving boxer engines and others showed very similar distributions of crash types, rider age and gender, speed areas, road conditions, use of protective equipment, fitment of CBS and TCS, etc. These results are shown more in detail in Table A3 (see online supplement). Skin injuries (i.e., scratches and bruises) were the most common AIS 1+ injuries in both groups of crashes (see Table 3). No statistically significant differences were found for any body region other than for lower extremities (P = .001). Leg injuries accounted for 9% of all injuries among riders with boxer engines and 20% of injuries among other riders, thus giving a 57% reduction. The calculations based on Eq.
(2) showed that  the injury distributions across the other body regions would have been very similar if the 2 groups have had an equal share of AIS 1+ leg injuries. The most common AIS 2+ injuries for riders with boxer engines were at the upper extremities (31%), and injuries at the lower extremities were more common among other riders (36%). No statistically significant differences were found for any body region other than for lower extremities (P = .003). These injuries accounted for 17 and 36% of all AIS 2+ injuries among boxer riders and others, respectively, thus giving a 54% reduction. The calculations based on Eq.
(2) showed that the distribution of AIS 2+ injuries to the other body regions would have been similar if the share of AIS 2+ leg injuries had been the same.
Similar to AIS 2+ injuries, the most common PMI 1+ injuries for riders with boxer engines were at the upper extremities (33%), and injuries at the lower extremities were more common among other riders (50%). No statistically significant differences were found for any body region other than for lower extremities (P = .04). These injuries accounted for 26 and 50% of all PMI 1+ injuries among boxer riders and others, respectively, thus giving a 48% reduction. The calculations based on Eq.
(2) showed that the distribution of PMI 1+ injuries to the other body regions would have been similar if the 2 groups had had an equal share of PMI 1+ leg injuries.
No substantial difference was found in the mean values of RPMI 1+ and RPMI 10+ (mRPMI 1+; mRPMI 10+) across different body regions; see Table 4. Indications were found suggesting that the overall mean RPMI 1+ among riders with boxer engines was lower than those with other engine configurations, although this result was not statistically significant (P = .23). The mean values of the overall RPMI 10+ were similar (P = .94).
Further analysis of injuries at the lower extremities showed that no knee or foot injuries had been reported among riders with boxer engines; see Table 5. However, the most common AIS 1 leg injuries among other riders were toe fractures and ankle joint sprains, and the most common AIS 2 injuries were fractures of the fibula (ankle) and metatarsals (foot). Ten AIS 3 leg injuries were reported among other riders, of which 5 were femoral or distal femoral fractures. On the other hand, injuries at the lower leg were less common among riders with boxer engines, as only one open tibia fracture (AIS 3) was reported. In general, AIS 2 injuries accounted for 57 and 66% of leg injuries among riders with boxer engines and others, respectively.
Among riders with boxer engines, the largest reductions in AIS 1+ and AIS 2+ injury risks were found for the knee, tibia, and foot, whereas no reductions were found for AIS 1+ and AIS 2+ ankle injuries; see Table 5.

Results and Implications
Despite the limited material, the present study found some significant results and new insights were given. It was found that leg injuries among riders with boxer engines were reduced by approximately 50% (see Table 3). In particular, no knee or foot injuries were reported among riders with boxer engines (see Table 5), although the present material might have been too limited for a proper breakdown of leg injuries. It was also found that the number of upper body injuries did not increase among riders with boxer-twin engines; in fact, the injury distributions across the other body regions would have been very Rizzi similar if the 2 groups had had an equal percentage of leg injuries. Furthermore, the severity of the reported injuries (measured with the mean value of RPMI) to the head and upper body was similar. In addition, the severity of the reported leg injuries was similar across the 2 groups. As a consequence of these findings, the overall mean RPMI 1+ was lower among riders with boxer engines, although this result was not statistically significant. The mean values of the overall RPMI 10+ were similar; this seems logical because the majority of the reported leg injuries were AIS 2+ (see Table 5). As can be seen from Table A1, AIS 2 leg injuries have a 50% RPMI 1+ and a 3% RPMI 10+, which means that a 50% reduction of these injuries would have a greater overall impact on RPMI 1+ than on RPMI 10+. In the present study, police and hospital records were used, and the final injury outcome of the crashes was analyzed. In other words, the available material did not include the level of information needed to understand the possibly different injury mechanisms among riders with boxer engines and others. However, there are a few aspects of this research that need to be discussed.
Two general leg injury mechanisms in a motorcycle crash have been suggested: direct impacts and leg crushing (Elliot et al. 2003;Otte 1994). Previous studies have shown no overall benefits for conventional crash bars (Ouellet et al. 1987;Rogers and Zellner 1998). Ouellet et al. (1987) suggested that this was due to the failure to keep the leg within the protection space and that more rigid leg protection systems fitted to the motorcycle would not have any substantial benefits without addressing this issue. In addition, the potential of leg impacts on the protector itself may increase leg injury risks (Ouellet et al. 1987). Another study by the same authors (Hurt et al. 1981), though, suggested that boxer engines could reduce leg injuries by approximately 50%. Though this finding was based on 11 crashes, it should be noted that there are a few differences between the present study and Hurt et al. (1981). For instance, ABS was still in a prototype phase at that time, which means that no ABS motorcycles could be possibly included in that study. Nonetheless, the present article seems to confirm the findings of Hurt et al. (1981); that is, leg injuries can be reduced with boxer engines in injury crashes.
Though it was suggested by Hurt et al. (1981) that leg entrapment could be limited by an engine cylinder, this injury mechanism was not reported to be predominant by Ouellet et al. (1987), primarily because the leg often did not remain in the leg space during the collision. Based on the present study, though, it is unclear whether only injuries resulting from one mechanism were reduced or a combination of both. Though leg entrapment may not be the main injury mechanism, it may seem logical that this would be reduced by a solid structure such as a cylinder. With regard to direct impacts, it could be argued that the protruding cylinder heads may be closer to the leg than conventional crash bars. This could decrease the degree of movement of the leg during the crash, thus keeping it within the protected area. Though it cannot be excluded that some leg injuries could be due to cylinder itself, it should be noted that most of the injury reduction with boxer engines was concentrated in the knee and lower leg regions (see Table 5). This may seem logical because those areas are closer to the cylinder heads (see Figure 1) and therefore would be expected to benefit the most in frontal or side collisions. This aspect also suggests the causality between the reported reductions of leg injuries and the fitment of boxer engines on ABS-equipped motorcycles.
On the other hand, previous research has suggested that leg protectors, such as the TRL concept, may increase the head injury risk (Rogers and Zellner 1998). This important issue was not found in the present study, because the mean RPMI to the head and upper body was similar across the 2 groups. However, it should be kept in mind that the ISO 13232 crash tests represent 7 car-motorcycle crash configurations, of which 6 were at the same motorcycle speed (Berg et al. 1998). In the present study, all types of crashes (occurring at different speeds) were analyzed, including single-vehicle, which accounted for some 45% of the material. The issue of increased head velocity due to rotation of the upper body may be most relevant in collisions into the side of a car, which is a typical crash scenario at intersections. This crash type accounted for 21% of the present material (see Table A3). Therefore, caution may be needed when comparing these overall results based on injury crashes with specific crash test configurations. Though the mentioned aspects may partly explain the differences between the present findings and previous crash tests, at this stage it could be argued that the ISO 13232 impact constellations may need to be further discussed, as suggested by Berg et al. (1998).
By comparing injury distributions in crashes involving ABS-equipped motorcycles with boxer engines and other engine configurations, it was found that • AIS 1+, AIS 2+, and PMI 1+ leg injuries among riders with boxer engines were reduced by approximately 50%. • the largest reductions of AIS 1+ and AIS 2+ leg injury risks were found for the knee, tibia, and foot. • the number of injuries to the head and upper body did not increase among riders with boxer engines. The mean risk for impairing injuries (RPMI 1+; RPMI 10+) to the head and the upper body were similar across the 2 groups, suggesting that the severity of injuries to the head and upper body did not increase either. • overall, the present study indicated that leg injuries were reduced among riders with boxer-engines, although no injury mitigating effects were found. • though boxer-twin engines were not originally developed to provide leg protection to motorcyclists, the present findings suggest that the concept of integrated leg protection is indeed feasible and that further engineering efforts in this area are likely to yield significant savings in health losses among motorcyclists.

Limitations
The present study was based on a number of assumptions and limitations that need to be discussed. First of all, the available crash data were limited. ABS-equipped motorcycles with boxer-twin engines were compared with similar ones (also fitted with ABS) from the same manufacturer as well as from other ones. Checks on possibly confounding factors were made to ensure their comparability in terms of crash and injury risks (see Table A3). The distributions of crash type, speed area, rider age and gender, and use of helmets and other protective gear were in fact very similar across the 2 groups. However, the distributions of motorcycle type (i.e., touring, standard, on/off, sport touring) were not similar. On/off motorcycles (also known as dual-purpose) were overrepresented among motorcycles with boxer engines, due to the limited crash data involving large on/off machines with other engine configurations. Though this aspect could confound the results, it was argued that the riding position was similar across the included motorcycles. A further confounding factor could be that motorcycles with boxer-twin engines may be slightly more stable than others, due to their lower center of gravity. However, this aspect is not expected to confound the results to any large degree. In fact, it may suggest that the leg injury reductions may be underestimated in the present study, because improved stability would increase the risk of leg injuries in a crash, as suggested in a previous study .
A further limitation is that the original 10 body regions used in the RPMI matrices were grouped for analysis, due to the limited material. Though it could be argued that such grouping seemed to be made in a logical manner (see Table 2), it is clear that the analysis of injury distributions would have been more powerful with the original 10 body regions (see Table A1). Similarly, all crash types were analyzed together, because the material was too limited for a separate analysis of single-vehicle crashes and multivehicle crashes.
Finally, it should be noted that the risk matrices used to calculate RPMI were initially developed for passenger car occupants. It could be argued that different road users have different risks of sustaining certain injuries (say, leg injuries for motorcycles and passenger car occupants); however, when an injury is sustained, the risk of not fully recovering from it should be the same. Though there is reason to believe that a certain injury should have a certain risk of permanent medical impairment regardless of how it was acquired, further research should confirm this.

Future Research
The data collected for the present study included too few boxer motorcycles without ABS to be further analyzed (n = 11). It could be hypothesized that the benefits of protruding cylinder heads for leg protection could be greater for ABS-equipped motorcycles than for motorcycles without it. It has been suggested that ABS riders are more often in an upright position during the crash than those without it (Dawkes and Löfdahl 2014), which would logically expose the legs even more ). Thus, a greater reduction of leg injuries could be expected by fitting some kind of leg protection to a motorcycle with ABS. The same idea could be taken even further, because the benefits of improved motorcycle design may be greater when motorcycles are fitted with ABS. For instance, the compatibility issue between road barriers and upright motorcycles fitted with some kind of leg protection seems promising and should be further investigated. It could be even hypothesized that some of the benefits of motorcycle ABS in terms of injury reduction (Rizzi et al. 2015) could actually be due to the leg protection offered by boxer engines. Though these may be intriguing hypotheses, it was not possible to investigate them with the present material.
Based on the present findings, it is recommended that future research should look deeper into this issue. The analyzed boxer-twin engine was firstly produced in 1923 and still is installed on some of the top selling large-displacement motorcycles in Europe (European Association of Motorcycle Manufacturers 2013). This aspect should guarantee enough exposure to generate crash data to analyze. In the present study, police and hospital records were used. Though there may be other ways to collect detailed injury data (i.e., telephone interviews, questionnaires, etc.), it is clear that an analysis of police records alone would not be sufficient to carry out such research. The same material could also be suitable for an evaluation of integrated airbags, which have been available as optional equipment on a motorcycle model for a number of years (DaCoTa 2012).
The present results may seem somewhat surprising, because boxer-twin engines were not developed to provide leg protection to motorcyclists. The basic idea was (and still is) that, because these engines are air-cooled, the position of the cylinders would be more favorable for the cooling airstream. However, this may not be the first case of vehicle safety being improved as a result of coincidence, rather than focused engineering , and the location of the injury reductions associated with boxer engines seemed to be consistent with the orientation of the leg. If confirmed by future research, the findings of this article could have important implications for motorcycle safety. The present article does not recommend a broad implementation of boxer-twin engines on motorcycles as a solution to address leg injuries. In other words, the benefits of boxer-twin engines in terms of leg protection may be only an example of what could be achieved: the present results suggest that the concept of protecting motorcyclists' legs with vehicle technology is indeed feasible and therefore more focused engineering efforts should be put to address this issue. Other technologies could be also used, such as knee airbags. In other words, the concept of integrated rider protection could be taken to a new level, without drastically modifying the look and handling of a motorcycle. In addition, consumer acceptance may be higher today than in the past. For instance, in a 2010 Swedish survey, 80% of the interviewed motorcyclists stated that their next motorcycle will be fitted with ABS (Nordqvist and Gregersen 2010). This may not have been the case some 20 years ago.