Cross and Fisher 1977 is a landmark and oft referenced study. The site truewheelers.org has this listed as the 1978 AAA Report: “This report describes the methods and results of the classic 1977 Cross/Fisher car/bike collision study”.
Here is the landing page to the complete volume I, DOT-HS-803-315 300 page .pdf file from the national Transp Library, which lists the problem types, and results.
The numbers/results seem to agree between the truewheelers html version and the HS803315 document.
Note that for whatever reason, vol 2 and vol 3 (DOT-HS-803-316 and 317) .pdf’s are not at the NTL website (volume III used to be here but that link died years ago).
In the study, the authors categorized the 36 types of crashes (Type 1 through 36) into 7 groups (Group A through Group G.).
The Cross/Fisher Class – Types, Percentage injuries/fatalities
| Accident type | Cross/Fisher Study: (injuries/fatalities) |
| Class A: Bicycle ride-out from driveway, alley and other midblock locations | |
| Type 1: residential driveway ride-out | 5.7%/6.7% |
| Type 2: commercial driveway ride-out | 3.2/2.4 |
| Type 3: parallel direction driveway ride-out | 2.5/2.4 |
| Type 4: ride-out over shoulder or curb | 2.5/3.6 |
| Total Class A: | 13.9%/15.1% |
| Class B: Bicycle ride-out at controlled intersection | |
| Type 5: stop sign or yield sign | 10.2%/7.8% |
| Type 6: signal phase change; cyclist caught in intersection | 3.1/0.6 |
| Type 7: ride-out at signal: multiple threat | 2.0/2.4 |
| Total Class B: | 17.0%/12.0% |
| Class C: Motorist turn/merge/drivethrough/driveout | |
| Type 8: Motorist driveout from commercial driveway/alley | 5.3%/0% |
| Type 9: Motorist failure to yield at stop or yield sign | 10.2/1.2 |
| Type 10: Motorist failure to yield at signal | 1.9/0 |
| Type 11: Motorist backing from driveway | 0.8/0 |
| Type 12: Motorist didn’t even slow for sign or signal | 0.5/1.2 |
| Total Class C: | 18.7%/2.4% |
| Class D: Motorist overtaking/overtaking threat | |
| Type 13: Motorist overtaking, cyclist not seen | 4.0%/24.6% |
| Type 14: Motorist overtaking/out of control | 0.7/4.2 |
| Type 15: Motorist overtaking/counter active evasive action | 1.7/2.4 |
| Type 16: Motorist overtaking/misjudged space required to pass | 2.0/1.8 |
| Type 17: Motorist overtaking/cyclist’s path obstructed | 2.0/0.6 |
| Motorist Overtaking: Type unknown |
0.1/4.2 |
| Total class D: | 10.5%/37.8% |
| Class E: Bicyclist unexpected turn/swerve | |
| Type 18: Bicyclist unexpected left turn; parallel paths; same direction | 8.4%/8.4% |
| Type 19: Bicyclist unexpected left turn; parallel paths; opposite direction | 3.2/3.0 |
| Type 20: Bicyclist unexpected swerve left; parallel paths; same direction | 1.5/3.6 |
| Type 21: Wrong-way bicyclist turns right; parallel paths | 1.1/1.2 |
| Total Class E: | 14.2%/16.2 |
| Class F: Motorist unexpected turn |
|
| Type 22: Motorist unexpected left turn; parallel paths; same direction | 1.3%/0.6% |
| Type 23: Motorist unexpected left turn; parallel paths; opposite direction | 7.6/0.0 |
| Type 24: Motorist unexpected right turn; parallel paths | 5.6/1.8 |
| Total Class F: | 14.5%/2.4% |
| Class G: Other |
|
| Type 25: Vehicles collide at uncontrolled intersection; orthogonal paths | 2.8%/0.6 |
| Type 26: Vehicles collide head on, wrong way bicyclist | 3.6/2.4 |
| Type 27: Bicyclist overtaking | 0.9/0.6 |
| Type 28: Head-on; wrong way motorist | 0.8/1.8 |
| Type 29: Parking lot | 0.8/0.8 |
| Type 30: Head-on; counteractive evasive action | 0.1/0 |
| Type 31: Bicyclist cuts corner when turning left | 0/0.6 |
| Type 32: Bicyclist swings wide when turning right | 0.3/0 |
| Type 33: Motorist cuts corner when turning left | 0.4/0 |
| Type 34: Motorist swings wide when turning right | 0.1/0 |
| Type 35: Motorist driveout from on-street parking | 0.3/0 |
| Type 36: Weird | 0/7.2 |
| Total Class G: | 11.2%/13.8% |
Notes: The study covered 166 fatalities, and 753 non-fatal injuries.
71% of the type 13 were in the dark, vs. only 30% of fatalities in general.
Tan study, type 13: Rural roads were overrepresented almost 2:1 (62 to 38%), Darkness overrepresented, High (60-70kph) and Very high speed (80+kph) roads overrepresented, 2 lane roads overrepresented.
Related link: The first Cross study.
Apparently you can order printed copies from ntis.gov DOT-HS-4-00982 Volume 1, Volume 2 and Volume 3.
Other Studies
The city of Toronto has a comprehensive 2003 study, and here is a comparative report from 1982 and Misoula, MT.
Listening to Bike Lanes, Hiles delves into the overtaking (Group D, especially type 13) crash types.
FHWA Pedestrian and Bicycle Related Research Reports Lists a bunch of stuff, including references to FHWA-RD-95-163, Pedestrian and Bicycle Crash Types of the Early 1990s [full 192 page scanned .pdf or a summary page that erroneously says it is out of print and unavailable]. This report has a section on “Comparison to Earlier Results by Cross and Fisher”. See comment below about mis-categorization of, e.g., right-hooks as “parallel paths”. The Crash-Type Manual for Bicyclists (FHWA-RD-96-104, “by Carol Tan Esse”) also published 1996 (and appears to be all the same stuff / extracted from Pedestrian and Bicycle Crash Types of the Early 1990s mentioned above) It follows the general Cross & Fisher types, then adds a few others ( Here is [from archive.org] [ original link now dead] a handy page that took the dozens of individual pdfs and collated them into one table). This same pub number is apparently synonymous with “Hunter, W.W., J.C. Stutts, and W.E. Pein,Bicycle Crash Types: A 1990’s[sic]Informational Guide, Publication No. FHWA-RD-96-104, FHWA, Washington, DC, April 1997.” in the Federal Highway Administration University Course on Bicycle and Pedestrian Transportation, lesson 3.
Also see PBCAT; I think they use same graphics for crash types.
Wachel and Lewiston, Risk Factors for Bicycle-Motor Vehicle Collisions at Intersections is one of the few studies that includes exposure data.
Lists of links to many other studies: industrializedcyclist.com, bikexprt.com, bicyclinglife.com/Library,
The FHWA Bicycle Road Safety Audit (May 2012 FHWA-SA-12-018) has a handy Table 1. where they list 4 result-sets according to the gross grouping of Crossing vs. Non-crossing paths, along with the more specific breakdowns — they list 4 datasets: Bicycle-Motor Vehicle Crash Types from FHWA’s Six-State Study (which turns out to be the “…of the early ’90s” referred to above), North Carolina Rural / Urban, and Orlando, Florida Metropolitan Planning Organization (MPO).
NCHRP
(the National Cooperative Highway Research Program; A forum for coordinated and collaborative research somehow related to the National Academies)…
NCHRP 17-18(3) Reducing Collisions Involving Bicycles, types of problems uses the usual crash-type pictograms, same as used in other FHWA docs. Seems to be from the mid-2000s.
I recently became the proud owner of a copy of Forester’s classic Bicycle Transportation, 2nd edition, 1994. In it he discusses the Cross and Fisher data at great length and detail.
In table 5-8 and 5-10; he re-splits the crash types into sub-types which illuminate fault basis to a great extent. The splits are c/s/w/t for: correct road position, sidewalk, wrong side of road, and cyclist swerve (why is that one t? t=turn?)
He states, without any further explanation as to how these seemingly precise (down to the ten-thousandths!):
“The statistical error in sample stratification is that rural and urban car-bike collisions were sampled by different plans but are grouped as if they had been sampled by one plan. As a result rural collisions are overstated by a factor of 1.454 while urban collisions are understated by a factor of 0.9663”
I can’t quite see how the numbers to jive up, so to take an example of crash type 5 which according to the web-version of Cross and Fisher equals 10.2% of the 761 non-fatals, with the text elaboration saying that only 4% of this type were at rural intersections.
So, in table 5-8 & 10 of Bicycle Transportation, Type 5 appears broken into three incarnations: 5c/urban = 9.3%, 5w/urban = 2.6%, and 5c/rural = 2.0%
There is a fairly long and detailed discussion of how Cross-Fisher categorized Rural and Urban…
Below is a long excerpt from Volume I (big .pdf!) of the original study.
page 118-120…
THE ACCIDENT LOCATION
The accident location is described below in terms of the urban-rural designation of the area, the predominant use of the land in the vicinity of the accident site, and the characteristics of the roadway the operators were traveling just prior to the accident
URBAN-RURAL DESIGNATION
Traditionally, the location of all types of traffic accidents is designated as urban or rural by the investigating officer or by another representative of the law enforcement agency. Law enforcement agencies most commonly differentiate urban and rural areas in terms of either the incorporation status of the area or the number of inhabitants who reside within a built-up area. As a consequence, urban accidents may be defined as those which occur within the political boundary of an incorporated area or those which occur within communities inhabited by more than some prescribed number of persons (sometimes 2,500 or more and sometimes 5,000 or more). Rural accidents are those that are not designated as urban.
Clearly, it is not possible to draw valid inferences about the characteristics of an area knowing only that it was designated urban or rural by a representative of an enforcement agency. During a preliminary examination of the accidents in the study sample, it was noted that many of the accidents that were officially designated as rural, in fact, occurred in densely populated residential communities located in the unincorporated fringe of a large population center. Although such areas were unincorporated and therefore officially rural, the characteristics of the areas were urban in every important respect. A smaller, but significant, number of cases were noted in which accidents officially designated as urban
occurred in areas that were truly rural in character. Therefore, with the classification criteria that are used by law enforcement agencies, it is altogether possible that an accident designated as rural may have occurred in an area that is truly urban in character, and vice versa.
It is for this reason that all the accidents in the study sample were reclassified using more meaningful classification criteria. For purposes of defining accident causation, the most important differences between
urban and rural areas are the posted speed limit and the motorists’ expectations about encountering bicyclists in the area. Other differences that may be important include the presence of sidewalks, the presence of street lighting, roadway surface type, and the type and surface condition of the roadway shoulder. For the non-fatal sample, accidents were usually classified as rural if they occurred in an area where a) the posted speed limit was 45 miles per hour or more; b) there were no curbs or sidewalks adjacent
to the roadway; c) street lights were not present at intersections; and d) at least 50% of the area within one-half mile radius of the accident site was open…
p.122:
At the present time, neither this study nor other studies reported in the literature provide the information needed to formulate a highly reliable estimate of the proportion of bicycle/motor-vehicle accidents that occur in areas that are truly urban or truly rural.
> Hit from behind tends to have a higher fatality rate but right hooks are no joke, especially when the vehicle is a truck or a bus.
The question should be about which type of crash has the higher risk for bicyclists, rear-end or right hook, not which is more severe IF it happens. The risk of being killed or seriously injured on a particular ride in a particular crash type is essentially the probability of the crash type times the probability of being killed or seriously injured given you are involved in a crash of that type:
Risk = (probability of crash type i) x (severity of injury|crash type i)
Right hooks may have a lower likelihood of being killed or seriously injured IF you are involved in one, but are much more likely than being rear-ended so the risk is greater. The problem is that the latest study on crash types Pedestrian and Bicyclist Crash Types of the Early 1990’s
categorized right hook crashes as “parallel paths,” which is BS. Wayne Pein participated in the preparation of that report but his experience was so negative that he subsequently left the UNC HSRC.
Regardless, most people look only at the consequences of a particular crash type (severity of injury|crash type i) instead of the risk.
Bob Shanteau
—
Alas, Bob is correct of course; page 93 of the document linked above:
Crashes where the bicycle and motor vehicle were on parallel paths accounted for more than 35 percent of the crashes. These distributed into the following categories:
Motorist turned or merged into path of bicyclist 12.1 percent
Bicyclist turned or merged into path of motorist 7.3 percent
Operator on wrong side of street 2.8 percent
Motorist overtaking bicyclist 8.6 percent