Ludwig Benner, Jr., P.E.
V. P., Events Analysis, Inc.
12101 Toreador Lane
Oakton, VA 22124
John R. Saams, P.E.
Dir, Fire Protection Services
Events Analysis, Inc.
Alexandria, VA. 22031

Presented by Saams at the
Miami, FL, November 14, 1990

© 1990 by Events Analysis, Inc.



Fire investigations are one important source of new knowledge for the Fire Engineering profession. The purpose of this paper is to alert fire protection engineers to new investigation systems that can add substantially to the profession's knowledge base, and to stimulate a more vigorous demand for their use. Study of current practices discloses major shortcomings in their conceptual base. The paper briefly describes varying perceptions of a “fire” and resultant problems with the scope, objectives, methodologies, outputs and uses of fire investigations. It proposes adaptation of new systems-based advances in investigation technology to fire investigations. By demanding better investigations, Fire Protection Engineers can accelerate the rate of growth and improve the quality of new knowledge about fire phenomena on which their practice rests.

I. What is a fire?

What is a “fire?” Before addressing the topic of fire investigation, it is imperative to examine briefly the various perceptions of the fire phenomenon that will be investigated. A precedent for this examination is found in accident investigation research, where perceptions of the accident phenomenon were found to be a dominant influences on the scope, objectives, methodologies, outputs and uses of investigations. [1]

Fire investigation sources indicate that several differing perceptions of the fire phenomenon influence fire investigations. Those perceptions include the widely used “fire triangle” or “fire tetrahedron” models of fire elements; the idea of a three-phased fire model consisting of incipient, free burning and smoldering phases; [2] and the more sophisticated computerized fire progression models developed by fire researchers [3]. Elsewhere, especially in the media, fires are viewed in a narrower context of a single event, as demonstrated by the frequent reporting that “the cause of the fire” had not been established. In the legal arena accidental phenomena are typically treated as a sequential chain of events [4].

Based on numerous arguments for doing so (presented elsewhere [5]) we have applied the general systems model and underlying concepts which leads us to think about a fire as a process. More specifically, we view a process, consisting of a set of sequential and concurrent causally-linked interactions over time, between certain forms of thermal energy and other system elements, that produce harmful outcomes.

These perceptions are important, because they form the conceptual framework driving most current fire-related work, including investigations as well as research, design, testing and installation tasks.


Based on a comparison with our experience in accident and fire investigations, traditional fire investigation practices pose major problems. These problems range from the scoping of the fire to be investigated to the uses of the investigation work products. Let's look at these problems, and then look at how they have been resolved. [6]

1. Scoping investigations.

By scope, we mean the range of events and circumstances covered by an investigation. For investigators, determination of the scope of a fire establishes what they will investigate, and the breadth of the new knowledge they produce. How does an investigator define the scope of a specific fire to be investigated? What guidance is given to an investigator for determining when the fire began, and when it ended? How far back in time should the investigator look for information about the fire?

The significance of these questions is the impact of the investigator's answers have on everything that follows. The investigator's concepts of the fire phenomenon affect the scoping and all other aspects of an investigation. If the concept is flawed, the investigation will be flawed. If the perception of the fire phenomenon driving an investigation is the fire triangle, the investigator will search for the fuel, heat and oxygen that produced the fire, and probably how they came together. If the work is driven by the fire phases model, the investigator will want to find out about all three phases and focus on what happened during these phases. If the process concept prevails, the investigator will want to address what happened between the first aberrant event and the ultimate harm produced by the fire process, and the effects of fixes over the life cycle of the facility - a much broader scope.

2. Objectives:

Why investigate a fire? Traditional reasons are to develop an opinion about the fire origin [7]. A second typical reason is to determine the "cause of the fire" to prevent future fires [8]. Some “investigations” are performed to fill blanks on forms [9], or use blank forms to keep notes [10]. Other investigators investigate to determine losses and subrogation options for claims purposes [11]. The objective of many local fire investigations is to comply with the law requiring investigations or even the determination of cause. Some investigations are conducted to satisfy political or social or legal demands [12]. Some investigations are conducted to determine violations of the law or regulations [13]. Some are even conducted to determine if present codes and standards are adequate.

Thoughtful analysis of these objectives suggests that almost all demand opinions of the investigator. None of these objectives address directly the addition of new knowledge on which persons with fire safety responsibilities must act in the future. Opinions - such as "cause" - are breeding grounds for controversy

What should be the objective of a fire investigation ? What theoretical basis, - that is what concepts and principles - should govern the setting of fire investigation objectives? How should the investigation objectives be framed so success or failure of an investigation can be measured?

3. Fire investigation methodology selection

A large number of investigation tasks or techniques - investigation tools [14] - are available in the literature, but no formalized investigation system or comprehensive fire investigation methodology could be identified. The apparent approach observed most frequently in reference documents is the adaptation of engineering -related techniques or “common sense” to investigations. Said another way, if you are an engineer with system knowledge, your investigation capabilities are assumed!

This overlooks the many differences in the knowledge base, skills and thought process required by system experts vs fire investigators. For example, the investigative knowledge base requires knowledge of investigation systems and technology to bring order to unstructured and often fragmentary observations. Investigation skills focus on tools for data gathering and analysis of ambiguous events, while systems skills are focused on empirical knowns, efficiency, cost, schedules, and related needs to produce desired outcomes.

A good test for anyone who has performed a fire investigation: What methodological choices do you think exist today? What methodology did you use, and where has that methodology been researched, critiqued and formalized? Does it produce replicable results? Does it provide a technically sound method for resolution of conflicts, rather than the conventional "weight of evidence" approach? If your fire were investigated independently by someone else, how replicable would the results be?

4. Fire investigation Outputs:

Three types of investigation outputs have been observed. The primary output is a narrative description of what happened, with conclusions and, sometimes, recommendations. The main problem with narrative descriptions of anything, especially fires where many things are occurring simultaneously, is the linear limitation of the written or spoke word. This form demands that the reader mentally reconstruct the process being described, a very difficult task even with perfectly flowing narrative descriptions of fires. Further, this form of presentation is very prone to oversights and omissions [15].

A second prevalent output is fire investigation forms [16]. The main problem with forms of any kind is that the data requested is usually a prejudgment of the forms designer, incomplete, imprecise, and forcing conclusions about the incident being reported. A second defect is that the questions posed on the form usually reflect what the form designer(s) believe is important about a fire or mishap. As one who has experienced the problems force-fitting observed mishap data into blanks specified by a form, users of forms should be strongly warned that the fit is usually not good. Even quantified data like the selection of the time of a fire or mishap calls for a conclusion by the investigator about when the phenomenon began, with low probabilities of replication. What do you enter if the incident lasted 10 hours or 2 days, as when hazardous materials are involved?

A third type seen occasionally is the schematic or alternatively a mathematical model of a fire scenario. The main problem with this type of output is that mathematical models are largely linear representations, at best, or "snapshots" or time slices at a particular point in time. With graphic models, the elements of the model are often abstractly or ambiguously worded, because the modeler lacks a formalized discipline for the construction of each model element. Another problem is the scope of such models, discussed earlier.

The difficulty with most narrative and form outputs is that they do not lend themselves to prediction and control actions during activities in progress, long before the next fire losses occur. A second consequence is that the outputs do not provide a measure of the effectiveness of the fire protection engineering efforts, or support a methodology for measuring that effectiveness proactively,.

5. Uses of investigation outputs

Most uses of fire investigations are, in fact, based on a “fly-fix-fly” philosophy, or retrospective. These retrospective uses range from statistical analyses of past fires to claims process and punitive compliance actions. Some investigation outputs may be used to verify predictions and in engineering work or code modifications, but the actions were still retrospective in that the serious loss had to occur before a change was introduced. We observed no evidence that investigations were used to monitor performance and drive changes before significant losses occurred except through new codes, standards or rulemaking actions and subsequent compliance efforts.

6. Consequences of fire investigation problems

Having identified these problems, why should they concern anyone? Because they are practical questions that each fire investigator must answer each time an investigation is started. Since there are not prescribed answers, all investigators attempt to answer these questions in their own way. In the absence of sound research-based guidance, the answers reflect the best judgment of the individuals, influenced by whatever their life experience has taught them. Thus they do get answered. However, the consequences of this action are of concern. One inevitable consequence of the lack of a widely accepted formal system of investigation (methodology) when coupled with current insistence on opinions of "cause," is controversy! Other observed consequences include
  • scope of the data reported is generally inadequate, unnecessarily limiting choices for remedial action because the phenomenon is inadequately understood
  • the quality of the outputs and reported data is not rigorously tested, resulting in hasty and unsubstantiated judgments like "human error" or "poor judgment" or "negligence"
  • replicability of outputs is impossible, resulting in controversy because the investigative process, testing and reporting criteria are left to the discretion of individuals applying different ideas and methods
  • users of the data are led to misdirected, delayed and inefficient countermeasures of all kinds including improper design, construction, training, monitoring and future investigations.

These adverse effects are evidenced by continuing fire losses, the relatively small number of options for controlling fire phenomena, the lack of a science-based measure of success for investigators, litigation involving fires, and in the difficulty in defining fire risks in a way that permits real-time rather than post-fire monitoring of and verification of the control of those risks.

In summary, the consequences are that we wait too long and then try to fix the wrong things after a fire.

What would be better?


Concepts developed for accident investigation have been found to have direct application to fire investigations. Relevant innovations for fire investigators are suggested below.

1. Scope.

By thinking of a fire as a process , investigators and fire protection engineers are encouraged to seek out aspects of the phenomenon and interactions that otherwise escape attention. Note that this concept encompasses both accidental and willful fires. A preferred investigation system should be adaptable to both types of origins, and should support investigations before or after fires occur. The "process" perception seems to be generally accepted, at least implicitly, by most fire protection engineers, because of their need to do predictive analyses and make proactive decisions. Fire models seem to reflect the process view to a degree, for example.

First, think of fire as a process [17], focusing on events that lead to subsequent changes in people or things, serially and in parallel, from the beginning to the end of the process. From an engineering perspective, this provides a basis for defining a beginning and end of the fire being investigated, and for scoping the investigation. The fire process can be said to begin when someone or something introduces a change into a stable static or stable dynamic state, producing cascading changes with the fire-related harmful outcome(s). The fire ends with last successive harmful event produced by the fire process, and linked to the beginning event. The harmful events includes intangible losses - e.g., loss of credibility, good will, public confidence - as well as smoke, ashes and burn victims. The scope of the investigation must address that entire process if the full range of action options to control losses are to be identified and evaluated to find the optimum actions.

2. Objectives

Adaptation of concepts for scientific investigation (research) is helpful in setting fire investigation objectives. The fundamental objectives of scientific inquiry are the understanding, prediction and control of phenomena - or new knowledge. The scientific investigation process starts with a phenomenon that is not understood, and uses observations, measurement, testing and logical reasoning to develop an ordered description of or hypothesis about a phenomenon. To gain new understanding and control, the investigator needs to develop new knowledge, order it and test it for validity. To be of value, the new knowledge must be in a usable form by any party needing to make the predictions or achieve control.

The logic for applying objectives of scientific inquiry to fire investigation is compelling. Understanding, prediction and control describe, in readily understandable terms, why fire investigations should be performed. Development of new knowledge by investigating a fire is similarly readily understood, from the perspective of any user. To be usable, the new knowledge about a fire must be in a form that fire protection engineers and others can use. Ideally, a complete description of a fire should serve all users for all purposes.

This can be achieved if the investigator produces a valid, tested and comprehensive description of the fire, free of guesses and opinions about fire cause, and readily acknowledging remaining uncertainties. A concrete description of the fire process interactions that produced the final outcome, from beginning to end, could then be used to draw any conclusions users of investigators' work products might wish to draw, for their own purposes. With a complete understanding and description of the fire events and changes which sustained it from beginning to end, the need for controls and the identification of specific actions is a short next step.

From this, it can be seen that the objectives of fire investigations should be the development of a valid description of what happened during the fire (new knowledge), in a form that can be used by persons responsible for the proactive control of fire losses. Note that this objective rejects "origin and cause" objectives, with good reason. Settling for the origin is too narrow a scope, and making a judgment on the "cause" of an entire process is an unacceptable oversimplification of the complex fire loss process. Seeking a cause misdirects the investigative effort, misleads lay persons into an overly simplistic view of fire control needs, is an unnecessary exercise in abstraction which contributes no useful technical knowledge about the specific phenomenon, precludes many options for more effective control of fire processes and risk, is too often based on choices or checklists reflecting last year's problems and experience, introduces an unnecessary source of controversy, and is of no practical predictive value to fire protection engineers and most other users. Hopefully, this tradition will not choke off adoption of sorely needed changes in fire investigation objectives.

3. Methodologies:

Two major mishap investigation systems have been developed and formalized. Both are applicable to fire investigations. The systems are the STEP [18] and MORT [19] investigation systems. Both are process-oriented, using energy flow concepts, mishap models, and flow charting with time lines. Both focus on discovering and defining multiple cause/effect relationships rather than causes or causal factors. STEP uses specially formatted event sets, linking tests and sequential logic extensively to organize data and aid timely problem discovery. MORT uses fault trees and deductive logic, along with a MORT generic check list extensively. Both use adaptations of change analysis concepts, though in different ways. Neither requires the use of forms as a data collection vehicle. Both take a broad view of the scope of the mishap phenomenon, and are aimed at dispelling mystery and facilitating improvement in performance.

Of the two, STEP-based methods are the most rigorous for the investigator in that validation tests of the individual flow chart building blocks and events pairs, and of the completeness of the mishap description are prescribed. STEP is applied to many functions requiring new knowledge. Of the two, STEP has the stronger procedures for developing options to control future risks. Probably the most useful element of both is the flow charted data displays.

Little comparative research of the two systems has been reported. In one study [20] the multilinear events sequencing model was judged "the best investigative model currently available." STEP and MORT training are readily available, and can be practiced frequently before or after fires occur.

4. Outputs:

Instead of supporting win/lose, designation of blame or confrontational activities, the outputs from STEP and to a substantial degree MORT investigations are judgmentally neutral, in that they provide objective timed descriptions of interactons, rather than subjective opinions. STEP particularly aids in resolution of differences. as when considering hypotheticals. The STEP worksheet format, for example, provides almost instantaneous feedback on the viability of a hypothesis during an investigation. Both accommodate estimates of inherent or residual risks [21], and provide work products for monitoring future performance.

The timed, sequenced graphic outputs quickly communicate event scenarios with minimal effort by users. They may be checked easily for quality control purposes [22].

Both STEP and MORT can produce non-linear process descriptions, with minimal judgment calls.

STEP provides a format for discovering new control options, and evaluating them in terms of the degree of improvement that would be expected if each were implemented. The outputs also permit evaluation of any alternatives anyone can dream up, while disciplining unsupportable ideas.

If you can't flow chart it, you don't understand it.


Both the STEP and MORT systems provide a basis for PROACTIVE actions - in terms of engineering design, task design, process control and monitoring, engineering feedback, and risk acceptance decision making. For fire protection engineers, the value of a flow chart to help determine system needs and fire "process intervention points" - and the loss-reduction capability of each - can not be overstated. By using flow charted scenarios, it is an easy task to monitor current and future activities to discover changed fire risks, for example. Flow charts also provide a superior "corporate memory" for all purposes.

IV. Conclusions

The foregoing discussion of problems with current fire investigation practices, and the opportunities offered by recent advances in investigation systems and technology suggests the following actions in the fire investigation field.
  1. 1. Promote the widespread adoption of the perception of a fire as a process, and develop conventions for determining the beginning and end of the fire process for various investigation purposes.
  2. Develop a restatement of fire investigation objectives, incorporating in the objectives the need for new knowledge, in a form capable of supporting concrete actions by persons with a role in fire risk management.
  3. Adapt appropriate new investigation system concepts, principles and techniques to the fire investigation process, to encourage the development of complete non-judgmental descriptions of the process, identify more options for control of risks, and provide a basis for assessment of the relative effectiveness of alternative control options.
  4. Encourage managers to demand improved fire investigation performance and outputs available through use of the new investigative systems and technologies.
If Fire Protection Engineers, in their roles as managers and technical professionals, demand these changes in the fire investigation field, they s can accelerate the rate of growth and improve the quality of new knowledge about the fire phenomenon on which their practice is based.


  1. Benner, L., “5 Accident Perceptions”, Professional Safety, February 1982. p 21
  2. Roblee, C. L. and McKechnie, A. J., “ THE INVESTIGATION OF FIRES”, (1981) Prentice Hall, Englewood Cliffs, NJ
  3. C. Perroni in Fire Journal, July-August 1989, p 27.
  4. Prosser, W.L., Wade, J. W. and Schwartz, V. F., CASES AND MATERIALS ON TORTS, (1976) The Foundation Press, Inc., Mineola, NY, p 345
  5. Hendrick, K.M. and Benner, L., INVESTIGATING ACCIDENTS WITH STEP (1986). Marcel Dekker, New York. p 17-22.
  6. National Fire Academy, FIRE ARSON DETECTION, STUDENT MANUAL, UNIT 2 (1983) Federal Emergency Management Agency, Emmetsburg, MD p 1
  7. Carroll, J. J., FIRE AND ARSON INVESTIGATION, (1979), Charles C. Thomas, Springfield, IL p 83.
  8. NITS FIRS system.
  9. NFPA 906M
  10. NFPA 906M-10
  11. Report of the Dupont Plaza or MGM fires
  12. Michigan v Tyler et al, Supreme Court of the United States, May 31, 1978, No. 76-1608 for an insightful discussion relevant to this objective.
  13. Perroni, C., _________ Fire Journal, July/August 1989, p 25

  14. NITS FIRS reports, for example.
  15. Drysdale, D., AN INTRODUCTION TO FIRE DYNAMICS (1985), John Wiley & Sons, NY p 1
  16. U. S. Dept. of Energy MORT ACCIDENT INVESTIGATION MANUAL , SSDC 27, Second Edition, (1985), System Safety Development Center, Idaho Falls, ID
  17. Harvey, M. D., MODELS FOR ACCIDENT INVESTIGATION, (1985) Alberta Workers' Health, Safety and Compensation, Edmonton, Alberta, page v


[1] Benner, L., “5 Accident Perceptions”, Professional Safety, February 1982. p 21

[2] Roblee, C. L. and McKechnie, A. J., “ THE INVESTIGATION OF FIRES ”, (1981) Prentice Hall, Englewood Cliffs, NJ

[3] See discussion of models by C. Perroni in Fire Journal, July-August 1989, p 27.

[4]. See Prosser, W.L., Wade, J. W. and Schwartz, V. F., CASES AND MATERIALS ON TORTS, (1976) The Foundation Press, Inc., Mineola, NY, p 345

[5] The arguments are developed extensively in Hendrick, K.M. and Benner, L., INVESTIGATING ACCIDENTS WITH STEP (1986). Marcel Dekker, New York. p 17-22.

[6] Fire investigation herein is intended to embrace all investigations of the fire phenomenon, including arson and research activities.

[7] National Fire Academy, FIRE ARSON DETECTION, STUDENT MANUAL, U NIT 2 (1983) Federal Emergency Management Agency, Emmetsburg, MD p 1

[8] Carroll, J. J., FIRE AND ARSON INVESTIGATION, (1979), Charles C. Thomas, Springfield, IL p 83. ("The fire investigator makes an important contribution to society each time he makes an investigation and a positive determination as to the cause of a fire.")

[9] NITS FIRS system.

[10] NFPA 906M

[11] NFPA 906M-10

[12] cite a report of the Dupont Plaza or MGM fires

[13] See Michigan v Tyler et al, Supreme Court of the United States, May 31, 1978, No. 76-1608 for an insightful discussion relevant to this objective.

[14] Perroni, C., _________ Fire Journal, July/August 1989, p 25

[15] When narrative reports are flowcharted using newer methods, significant gaps in the description or explanation of an incident have become evident in every case where this was done. In one extreme case, the narrative describing a fatal accident did include a single event during the accident; despite this deficiency, a penalty was levied based on alleged deficiencies and violations demonstrated by the "accident."

[16] NITS FIRS reports, for example.

[17] Drysdale, D., AN INTRODUCTION TO FIRE DYNAMICS (1985), John Wiley & Sons, NY p 1

[18] U. S. Dept. of Energy MORT ACCIDENT INVESTIGATION MANUAL , SSDC 27, Second Edition, (1985), System Safety Development Center, Idaho Falls, ID

[19] Hendrick, K.H. and Benner, L., INVESTIGATING ACCIDENTS WITH STEP, (1987) Marcel Dekker, NY

[20] Harvey, M. D., MODELS FOR ACCIDENT INVESTIGATION, (1985) Alberta Workers' Health, Safety and Compensation, Edmonton, Alberta, page v

[21] STEP uses 20 element estimated risk assessment code matrices to indicate risks, while MORT uses a subjective "less than adequate" indicator.

[22] When is the last time you saw or used specifications for fire investigation work product quality controls?