"Hypotheses are free. You don't need "evidence" to form a hypothesis - you only need, at a minimum, a problem statement. The more hypothesis that are formed, the better the investigation will be. Some hypotheses will be supported by data (which can then be called "evidence" that supports the data) and some will be specifically refuted by data (which can then be referred to as "evidence" refuting the hypothesis.) In the end, a final hypothesis will be the one that is best supported by the data (evidence.)"

I would agree that CONSIDERING hypotheses is free of the need for data and evidence. DATA is INFORMATION that can be documented and verified. EVIDENCE is DATA that has been assessed as to its relevance and reliability in a specific context. All EVIDENCE is based on DATA, but not all DATA is EVIDENCE. CONSIDERING hypotheses and the FORMULATION of hypotheses are BOTH part of the reliable application of the SM to the investigation of fire and explosion incidents. The FORMULATION, as opposed to CONSIDERATION of hypotheses does require evidence.

"This may be subtle but may go to the "theory-ladenness" of observations that philosophers (like Duhem, Kuhn and Feyerabend) discussed. Data gets meaning when evaluated against a hypothesis - otherwise it is just raw, meaningless data. You can collect all you want but until there is a hypothesis against which to evaluate the data; it may be noise or signal - it's sitting in a box waiting to be "assigned" value. The alarm going off at 1 am is indicative of an alarm going off at 1 am. If your hypothesis is that "a fire would activate the alarm system" then the data that the alarm activated supports the hypothesis. If the hypothesis is that "an alarm would activate near the time that the fire started (or at least when it achieved sufficient energy to meet the alarm activation parameters)" then the data that the alarm activated at around 1 am, evaluated against the hypothesis, provides support for the fire starting sometime just before 1 am."

"I don't know how anyone else's brain works but I cannot help but form hypotheses as soon as I get an assignment (or used to.) What do you think when you receive a call for an alarm activation in a high-rise apartment complex at 5 PM? Can you form the hypothesis that this is probably food on the stove? You have no "evidence" yet (except the minimum - which is all you need). As you collect more data the hypothesis will either be supported or refuted. Is it "valid" in science to form hypotheses without evidence? Did Mendeleyev have "evidence" of the existence of elements where there were gaps in his table? He predicted (hypothesized) that an element (he called eka-silicon) would fit in a gap in his table (if his theory of periodicity was correct.) He hypothesized that in 1869 and in 1886, "germanium" was discovered with the properties Mendeleyev had hypothesized. Paul Dirac hypothesized the existence of "anti-electrons" in 1928 - "positrons were discovered in 1932. Einstein's general theory predicted gravitation lensing which wasn't shown to exist until 1919. The hypothesis, in and of itself, is neither "valid" nor "invalid" simply because it is postulated, with or without "evidence". One could argue that the argument that a hypothesis must be based on "evidence" could be considered circular if one accepts the premise that "evidence is data given meaning by evaluation against a hypothesis; that is, when it supports or refutes a hypothesis." How can you form a hypothesis if it requires data that supports or refutes itself?"

The SM as applied to the investigation of fire and explosion incidents involves BOTH Inductive and Deductive Reasoning. Inductive reasoning can produce hypotheses. The same is true of deductive reasoning. There is a difference between the two that I believe aligns with your statements also. The difference is that Inductive Reasoning involves CONSIDERING hypotheses that might explain what could have happened. You respond to a fire and you have a report of a fire fighter trapped after falling through a floor. Your experience is that FFs fall through the first floor of residential houses with basements. You now CONSIDER a hypothesis that the origin could have been in the basement and the fire fighter made entry into the house from the front door on the first floor level. You have no data or evidence as of yet, but your going to start to collect data to see if there is evidence to FORMULATE an origin hypothesis of a fire starting in the basement. The FF could have fallen through the second floor and became trapped on the first floor.

However, the SM does not care about what MIGHT have happened, the application of the SM cares about the most reliable determination of what DID happen in this specific incident. Thus, moving through the steps of the application of the SM, there must be evidence of such to produce a reliable determination. Otherwise, it is a subjective determination. This is Deductive Reasoning, where you are now collecting data and determining if it is evidence for the FORMULATION of what did happen in this specific incident. Thus, both CONSIDERING hypotheses using Inductive Reasoning AND FORMULATING hypotheses with evidence using Deductive Reasoning are part of the reliable application of the SM to the investigation of fire and explosion incidents.

"Certainly, the number of potential hypotheses regarding a fire's origin or cause is very large but probably not infinite (although in practicality, the combinations of possible heat sources, fuels and circumstances that could bring them together is so large as to be, effectively, "infinite.") Fortunately, examination of the scene reduces the number of plausible ignition scenarios at a given, specific scene and they can be narrowed down from "infinite" to manageable."

It is infinite if you are not restricted by the need for evidence. "The fire was caused by little green men from Mars. Without the need for evidence of such, how is this not a hypothesis that can be developed? With no constraints, there are an infinite number of hypotheses that can be developed. "OK, Doug, little green men from Mars is an highly exaggerated example that is not realistic." Fair enough. But, WHY? Is it because it is an untestable hypothesis? Is it an untestable hypothesis because it was formulated WITHOUT evidence?

"But the more hypotheses that the investigator formulates, the better their investigation will be - because they will have considered more possibilities. I can form the hypothesis "I bet the heat source was electricity" before I even arrive at a scene. If I arrive at the scene and there is no electricity, my hypothesis fails to be supported. The formation of the hypothesis is not "invalid" because fires can be caused by electrical events based on my knowledge, training and experience - Just not this one. And, the hypothesis, although it failed provides some defense against "confirmation bias" or "expectation bias" in that "I formed the hypothesis that the ignition source could be electricity. I looked for a potential refutational data and found it in the fact that there was no electrical service to the property. I then moved on to form other hypotheses."

You are describing the application of BOTH Inductive and Deductive Reasoning in the application of the SM to fire investigations as discussed in 921. So the reliable application of the SM produces BOTH "hypotheses" that are CONSIDERED without any data or evidence and "valid hypothesizes" FORMULATED with evidence. In contrast, hypotheses FORMULATED without evidence are "invalid hypothesis"

"The fact that 921 refers to "invalid" hypotheses is partly because I failed to submit a proposal to remove it before the deadline and is one area where I would play the 921 is a Guide" card and justify my deviation from that paragraph."

NFPA 921 does a poor job of explaining how to reliably apply the SM to a specific set of data and evidence. I tried in the last edition to make changes, but my proposals were not accepted.

"It is not a "negative corpus" methodology if there is data to support or refute the hypothesis. It seems that you are now trying to force that terminology ("negative corpus") into hypothesis development where it does not belong."

I said that FORMULATING hypotheses WITHOUT data and evidence using DEDUCTIVE Reasoning is a "negative corpus" methodology. This stems from an unreliable application of the SM. Yet, the fire investigation community understands "negative corpus," as a practical matter, "if I rule out every other hypothesis, but one, then I can determine what happened." This is the Sherlock Holmes famously quoted methodology, which is unreliable. Yet, fundamentally, "negative corpus" is rooted in the unreliable application of the SM, where one applies Deductive Reasoning without evidence. So it does belong in hypothesis development.

Along those lines, the unreliable application of the SM involves the elimination of all hypotheses except one, to reach a determination. First, how does one reliably eliminate an infinite number of hypotheses? Secondly, how does one "rule out" a hypothesis that has not first been "ruled in?"

Douglas J. Carpenter, MScFPE, CFEI, PE, FSFPE
Vice President & Principal Engineer
Combustion Science & Engineering, Inc.
8940 Old Annapolis Road, Suite L
Columbia, MD 21045
(410) 884-3266
(410) 884-3267 (fax)
www.csefire.com