Expert report review

Computer simulation in the opposing accident reconstruction: what does a reviewer check before it reaches the jury?

A computer simulation in a reconstruction report is a calculation, and its result depends on the inputs and on whether the program is accepted for the use made of it. A Washington appellate court accepted PC-Crash for vehicle speed in State v. Phillips (2004), and in State v. Sipin (2005) the same court rejected its occupant model. A review checks the program and version, each input, the match with the evidence and the disclosure.

An engineer from behind watches a blurred crash simulation, a small model car on the desk

What is the difference between a simulation and an animation?

A simulation computes how vehicles or occupants move from physical laws and the inputs the expert enters. An animation draws a sequence the expert has already decided on. Courts apply a reliability test to simulations offered as proof.

Reconstruction software can do two different things. In a simulation the program takes masses, dimensions, tire and road properties, initial speeds and driver inputs, applies equations of motion and impact, and calculates where each vehicle goes. The output is a result of the calculation. In an animation the expert or a graphics artist moves models along paths that the expert's opinion already fixed. The output is a picture of the opinion.

Division One of the Washington Court of Appeals drew the legal line in State v. Sipin, 130 Wn. App. 403, 123 P.3d 862, filed November 21, 2005. That opinion replaced an earlier opinion in the same case, which the court withdrew on reconsideration. It speaks of computer-generated models or simulations, as opposed to animations, offered as substantive proof, and it requires simulations used as substantive evidence to come from programs generally accepted by the appropriate scientific community. The trial court in Sipin had ruled that the State's expert could rely on the results of his PC-CRASH simulation as the basis for his opinion.

A reviewer begins by sorting every graphic exhibit in the opposing report into one of the two groups. The same video file can contain both: frames rendered from computed positions and frames filled in by hand between them. The sorting decides which test the court applies and which questions counsel asks.

What happened in State v. Sipin?

The State used PC-Crash 6.2 and its multi-body occupant option to show who was driving a BMW that crashed. The court held the simulation inadmissible for lack of general acceptance for that use and ordered a new trial.

On March 6, 2000, Sipin's BMW Z-3 convertible crashed with Sipin and David Taylor aboard, and Taylor died. The State charged Sipin with vehicular homicide as the driver. Sipin testified that Taylor had been driving, because Sipin had gout that day and could not operate a manual transmission, and his daughter testified that Taylor was driving when the two men left. No neighbor saw who was driving.

The State's accident reconstructionist, Ronald Heusser, used version 6.2 of PC-CRASH and its multi-body occupant option to simulate how the two men moved inside and out of the car. He concluded that the passenger was thrown into the tree and the driver was thrown out behind the vehicle, which placed Sipin in the driver's seat. After trial, defense experts said the program had not been validated for predicting the movement of occupants inside a vehicle in a crash with multiple impacts.

In the opinion filed November 21, 2005, Judge Becker wrote for Division One, with Judge Baker concurring; Judge Kennedy, who wrote the earlier opinion, had died on September 16, 2005. The court held that the simulation evidence was inadmissible under Frye and that the error was not harmless, because the outcome of the trial might have been different without it. The conviction was reversed and the case remanded for a new trial with a new Frye hearing. The court added that it could not countenance the State's use of such evidence here or in future cases until the relevant scientific community reached a consensus on that use.

What test did the court apply to the simulation?

Three conditions: the computer was functioning properly, the input and underlying equations were sufficiently complete and accurate and disclosed to the opposing party, and the program is generally accepted by the appropriate community of scientists for the purposes at issue.

Washington applies the Frye test of general acceptance to novel scientific evidence. The Sipin opinion set out what that test means for a simulation. The proponent must show that the computer is functioning properly; that the input and underlying equations are sufficiently complete and accurate, and were disclosed to the opposing party so that it could challenge them; and that the program is generally accepted by the appropriate community of scientists to be valid for the purposes at issue in the case.

Each condition maps onto a part of a review. Proper functioning concerns the software installation, the version and any settings that change how the program computes. Complete and accurate input concerns every number the expert entered and its source. Disclosure concerns what the other side received: a printed summary, or the project files that let another engineer rerun the calculation. General acceptance concerns the specific module and the specific use, the point on which Sipin turned.

The condition on disclosure has a practical edge. Inputs that were never produced cannot be challenged, and under the conditions listed in Sipin disclosure is part of what the proponent has to show.

Why did the same program pass in one case and fail in another?

Acceptance attaches to a use, not to a program name. PC-Crash was accepted for vehicle movement in a single-impact crash in State v. Phillips (2004) and rejected for occupant movement in a multi-impact crash in State v. Sipin (2005).

State v. Phillips, 123 Wn. App. 761, decided by Division Two on October 5, 2004, arose from a crash on September 29, 2001, in which a car left a Kitsap County road and struck a telephone pole, killing the passenger. The State's expert used PC-Crash to estimate a speed of about 80 mph in a 40 mph zone. The trial court held a Frye hearing, and the court of appeals found that the program was generally accepted in the accident reconstruction community and that challenges to how the software was used went to weight, not admissibility. The conviction was affirmed.

The Sipin opinion distinguished Phillips. In Sipin the State used the multi-body option to predict the movement of occupants inside the vehicle in a crash with more than one impact, and the court found that use not generally accepted. A record that supports the program for one calculation did not carry over to another calculation done with a different module.

For a reviewer the question is therefore never whether a program is accepted in general. It is which module computed which result, for what kind of collision, and what validation work exists for that combination: staged crash tests compared with simulations, published studies, the developer's documentation of the model's limits. A report that names only the program, without the module and the version, leaves the court unable to answer the question on which Sipin turned.

How do federal courts test a simulation?

Under Rule 702 and Daubert. The proponent must establish that the opinion rests on sufficient facts or data and on reliable methods reliably applied; testing, peer review, error rate, standards and acceptance inform the court's assessment.

Federal courts do not use Frye. In Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, decided June 28, 1993, the Supreme Court held that the Frye general acceptance test was superseded by the adoption of the Federal Rules of Evidence. Justice Blackmun's opinion listed considerations for scientific validity: whether the theory or technique can be and has been tested, whether it has been subjected to peer review and publication, its known or potential rate of error, the existence and maintenance of standards controlling its operation, and its general acceptance in the relevant scientific community.

Rule 702, as amended in 2023, requires the proponent to show the court that it is more likely than not that the expert's knowledge will help the trier of fact, that the testimony is based on sufficient facts or data, that it is the product of reliable principles and methods, and that the opinion applies those methods reliably to the facts of the case.

For a simulation the Daubert questions become concrete. Has this module been tested against staged crashes of this type? Are the validation results published? What error did those tests find in speeds, positions or occupant paths? Did the expert follow the developer's procedures? The 2023 committee note adds that some challenges go to weight rather than admissibility; a review separates defects in the method, which belong in a motion, from disputes over individual inputs, which belong in cross-examination.

Which inputs does the reviewer check?

Every number the expert entered: vehicle masses with occupants and cargo, dimensions and center of gravity, tire and road friction, impact point and geometry, initial speeds and headings, braking and steering inputs, and occupant data where occupants are modeled.

The reviewer builds a table of inputs from the project file or, if only a printout was produced, from the report. Each row gets the value the expert used, the source the expert cited and the value the case evidence supports. Vehicle data often come from the program's database. The reviewer checks that the database entry matches the model year, body style and equipment of the actual vehicle, and that the mass includes occupants and cargo on the day of the crash.

Road friction is a frequent source of disagreement. A value measured at the scene with a test vehicle or drag sled under similar conditions differs from a value taken from a published range. The reviewer records which was used and whether the surface, the moisture and the tire condition on the day of the crash were accounted for. The same applies to coefficients that describe the impact itself, such as restitution and stiffness.

Initial conditions and driver inputs come next. Speeds, headings and the timing of braking or steering are inputs in many simulations, not outputs. Where the event data recorder supplies speed and brake status, the reviewer compares the expert's inputs with the download. Where the expert entered a speed to make the simulation match the scene, the speed is a fitted value, and the report should say so.

For occupant models the inputs extend to body size and mass, seating position, restraint use and the properties of interior surfaces. Sipin turned on this kind of model, and the reviewer checks whether the program's documentation supports it for the crash at hand.

How does the reviewer test whether the simulation matches the physical evidence?

By comparing simulated tire marks, impact and rest positions and crush with the measured ones, and by asking how many inputs the expert adjusted to get the match. A match reached by tuning several inputs proves less.

A simulation earns weight when its results agree with evidence it was not built from. The reviewer overlays the simulated paths on the scene diagram, compares the simulated rest positions with the measured ones, and compares the simulated contact areas and crush with photographs of the vehicles. Each discrepancy is recorded with its size.

Then the reviewer asks how the agreement was obtained. If the expert ran the simulation once with measured inputs and it reproduced the scene, that is evidence for the model. If the expert adjusted speed, friction, impact point and restitution across many runs until the rest positions matched, several combinations of inputs may fit the same evidence, and the report has to explain why the chosen combination is the right one.

The reviewer also checks evidence the simulation did not reproduce. Gouges, debris fields, yaw marks and secondary impacts that the model omits are listed. A model that matches rest positions but not the tire marks leading to them has not reproduced the crash.

Why does a single simulation run prove less than it seems?

Because inputs carry uncertainty. A sensitivity analysis varies speed, friction and other inputs across their plausible ranges and shows how far the result moves. A report without one presents a point where the evidence supports a range.

Each measured or estimated input has a margin. Friction from a published range, a speed estimated from crush, an impact angle read from photographs: each can move within limits the evidence permits. A sensitivity analysis reruns the simulation at the edges of those limits and records how much the conclusion changes.

Some conclusions survive the exercise. If the vehicle exceeds the speed limit in every plausible run, the range supports the opinion. Others do not. If a modest change in friction moves the point of impact into the other lane, the opinion that depends on the lane is weaker than a single run suggests.

The reviewer either finds the sensitivity work in the expert's file or asks for it. Where the project files were produced, the reviewer's engineer can run the variations independently and report the range. That range is often a central item in the review, because it states in numbers what the opposing opinion can and cannot support.

What should counsel request from the other side?

The native project files, the program name, version and modules, every run including those not relied on, input and output reports, the source of each input, and the rendered video with the computed data behind each frame.

Sipin made disclosure of input and equations part of the foundation for a simulation. A printed summary page lets the other side read the inputs but not test them. The native project file lets another engineer open the same model in the same program and version, inspect each value, and rerun it.

The request lists the program and version, the modules and libraries used, the vehicle database entries, all simulation runs in the file including runs the expert did not rely on, the input and output reports for each, the scene diagram or point cloud imported into the model, and any photographs or measurements used to set the inputs. For animations, it asks for the rendered video, the frame rate, the camera settings and the link between each frame and the computed data.

Runs the expert did not rely on are part of the record of the work. If earlier runs with measured inputs did not match the scene and later runs with adjusted inputs did, the sequence is part of how the opinion was formed. The reviewer prepares deposition questions on each change between runs.

When does an animation become a problem?

When it shows more than the calculation supports, or when the jury could take it as proof rather than illustration. Sipin treats simulations offered as substantive proof separately from animations.

An animation shows a single version of events with smooth motion, clear lighting and a chosen viewpoint. Details such as a driver's head turning or a pedestrian's posture may appear in the animation although no calculation produced them. The reviewer marks each element of the animation as computed, taken from evidence or added for illustration.

Viewpoint and timing matter as well. An animation viewed from above shows hazards a driver could not see from the seat. An animation played at a different speed changes the impression of how much time the driver had. The reviewer checks the frame rate and the time base and compares the viewpoint with the driver's eye position where visibility is an issue.

Sipin shows how one court handled the line: it applied Frye to simulations offered as substantive proof and set animations apart. A review gives counsel the material to ask for the same separation: a limiting instruction, labels on illustrative frames, or exclusion of frames that no calculation supports.

What does IAE do when the opposing report relies on a computer simulation?

An engineer from the IAE registry reviews the opposing reconstruction, including its tests and simulations, and lists what counsel should measure and download at an inspection. The review is issued in up to 5 business days, with a 48-hour rush.

The IAE accident reconstruction review is for counsel facing the other side's reconstructionist in an injury or property damage case in a US court. The engineer asks whether the opposing reconstructionist applied the intellectual rigor that practice in the field requires: inputs, reaction times, calculations, tests and simulations, and the limits of engineering opinion. Each point is tied to a page of the opposing report and to a rule, a ruling or a published method.

The client sends the opposing report and the case materials and describes the task. The secretariat confirms receipt the same day. Within one business day IAE confirms the format, cost, deadline and the specialist from its registry, and the client approves the candidate and the CV before work starts. The signed PDF is issued in up to 5 business days; cases with tight deadlines have a 48-hour rush.

Every specialist is a member of the public IAE registry and passed the council's attestation under the published Criteria Matrix. The Member ID or QR code on the document opens the specialist's category, specialization, attestation date and record status. The specialist confirms in the document that he or she has no interest in the outcome of the case. A review is evidence that the court weighs with the other materials; neither IAE nor the reviewer guarantees the outcome, and IAE is not a law firm.

Sources

Read also

All articles

Expert report review

Ready to discuss your case?

Send the materials: the secretariat will confirm the format, cost, deadline and the specialist within one business day.

Expertise request

Received

Cost and timeline estimate

WorkReview of an opinion
TimelineUp to 5 business days
ExpertIAE registry member