SAIF Corporation and Snyder Roofing and Sheet Metal, Inc. (SAIF) petition for review of an order of the Workers’ Compensation Board (Board), contending that the evidence on which the Board relied in affirming an award to claimant of permanent partial disability (PPD) was legally insufficient to satisfy claimant’s burden of proof. Claimant cross-petitions, seeking review of the Board’s classification of his disability. We reverse the award of PPD and therefore do not reach claimant’s assignment of error regarding the classification of his disability.
Claimant was compensably injured in August 1995, suffering injuries to his face and head as a result of a fall.1 Following treatment of his physical injuries, claimant complained of, among other problems, headaches, moodiness, dizziness, and memory problems.
In June 1996, claimant was examined by a psychiatrist, Dr. Bellville. Bellville noted that claimant had just been released from a 140-day court-mandated alcohol treatment program and that he was still consuming alcohol. Bellville suspected that claimant had a history of alcohol abuse and mild to moderate depression; he was uncertain of the extent to which claimant’s “intellectual complaints” were due to “pre-injury, longstanding emotional or personality issues” or alcohol, rather than to his injury. He recommended that claimant receive neuropsychological testing to rule out mental disorders related to his injury.
Also in June 1996, claimant was examined by a neurologist, Dr. Bell. Bell noted claimant’s apparent mental deficits, stated that it was “difficult” to determine the extent to which they were due to his injury, and recommended further testing, including a neuropsychological evaluation and an EEG.
In December 1997, Dr. Binder conducted a neuropsychological evaluation of claimant.2 Binder noted claimant’s history of alcohol abuse, including court-mandated *494treatment. Claimant reported that, four days before the examination, he had consumed six to 10 beers. Claimant denied any previous history of psychiatric problems or treatment. Based on claimant’s performance on the neuropsychological tests administered to him, which generally was worse than that of persons with “well-documented” brain dysfunction and “more serious traumatic brain injuries,” Binder concluded that the results of the tests were invalid, due to lack of effort and “poor motivation” on claimant’s part. Binder declined to estimate the likely outcome from claimant’s injury because, according to the doctor, it was a type of injury “from which full recovery sometimes occurs and sometimes does not occur.” Binder stated: “When people do not fully recover, the residual is mild or minimal in terms of Oregon Workers’ Compensation definitions of disability attributable to brain injury.” He opined that it was “likely” that claimant was an “alcohol abuser” and that he suffered from antisocial personality disorder. Binder concluded that he was unable to determine whether a neuropsychological condition related to claimant’s injury “continues to exist”; that he did not know if claimant “has any permanent impairment”; and that claimant’s “probable impairment is nil, minimal, or mild.” Claimant’s treating physician, Dr. Potter, concurred in Binder’s findings.
In February 1998, SAIF closed claimant’s claim with an award of temporary disability but no award of permanent disability. Claimant requested reconsideration by the Department of Consumer and Business Services (DCBS). DCBS appointed three medical arbiters including, as pertinent here, an ear, nose, and throat specialist and a neurosurgeon, and submitted medical information and written questions to each arbiter, as appropriate to the arbiter’s medical speciality. See OAR 436-030-0155(1); OAR 436-030-0165(3).
In his May 1998 report, the ear, nose, and throat specialist, Dr. Holden, described claimant’s nasal fracture and other physical injuries, noted his residual nasal deformity, and noted that the latter may have been the result of claimant’s noncompliance and interference with his treatment “along with alcohol abuse.” Holden noted that, although claimant’s fractures had healed, he had “persistent symptoms of headache, dizziness, imbalance and problems *495with short term memory and cognition along with psychosocial problems of depression and inadequacy * * Holden reported that claimant was a “chronic alcoholic with moods of depression.” He concluded:
“My sense is that [claimant] will remain dependent upon society for his care and support and I do not see any likely alternative.
“I am also concerned about his brain/possible inner ear injury * * *, short term memory and cognition problems, drug abuse of alcohol and aspirin and social inadequacy.
* * Since I suspect that [claimant] has a post-traumatic inner ear concussion syndrome it would be advisable to have him tested * * * to further clarify his functional capacity mentally and physically.”
Dr. Williams, the neurosurgeon, took a medical history of claimant’s injury and conducted a neurological examination. In his May 1998 report, Williams stated:
“It is my opinion that the patient is ADL [activities of daily living] assisted, being limited with cognitive impairment and psychological impairment.
“It is my opinion that the patient in the head/brain injury impairment section functions as a Rancho-Los-Amigos Scale of Class III. He is alert and oriented. His behavior is appropriate for the most part. He does have impaired judgment and mild memory deficit. His language is mildly affected. He has noted personality changes. He does describe sleep disorders. I would not consider the worker safe to operate industrial machinery.”
On reconsideration, DCBS determined that, based on claimant’s invalid neuropsychological test results, no impairment rating was due. DCBS modified the award of temporary disability and affirmed the award of no permanent disability. It did not consider Holden’s and Williams’s reports because those reports had not been completed as of the time of the order on reconsideration.
Claimant requested a hearing. The record was supplemented with, as pertinent here, Holden’s and Williams’s reports. In the Conclusions and Opinion section of his order *496after hearing, the administrative law judge (ALJ) stated, in part:
“There is no question that claimant had preinjury psychosocial problems, including alcohol abuse. However, the evidence preponderates in favor of finding that claimant’s ability to function was diminished post-injury.
“The most that the attending physician could say, concurring in the report of the neuropsychologist, was that the etiology of claimant’s then current condition could not be determined because of the invalidity of the test results. Although arbiter Holden noted claimant’s difficulties in dealing with life prior to the injury, he concluded that claimant would be dependent on society for his support and care after the injury, and suspected an inner ear concussion syndrome as a result of the injury. Dr. Williams based his opinion of claimant’s impairment on the post-injury condition of claimant, noting the post-injury changes. Presumably, Dr. Williams had reviewed claimant’s medical records, and he did not attribute claimant’s condition to anything other than the compensable injury. The neuropsychologist indicated that claimant’s symptoms could be consistent with his head injury.”
Citing Kim E. Danboise, 47 Van Natta 2163, on recons 47 Van Natta 2281 (1995), aff'd sub nom SAIF v. Danboise, 147 Or App 550, 937 P2d 127, rev den 325 Or 438 (1997), the ALJ concluded that claimant had met his burden of proving that his post-injury condition was the result of his compensable injury. The ALJ awarded claimant 74 percent unscheduled PPD. The Board affirmed the ALJ’s award, specifically adopting the ALJ’s reasoning and conclusion that Williams’s opinion established claimant’s entitlement to the disability award.
In its assignment of error, SAIF argues that the Board erred as a matter of law in concluding that the medical opinions in this case satisfy claimant’s burden of proving that his permanent impairment is due to the compensable injury. According to SAIF, Williams merely measured the extent of claimant’s impairment; neither he nor any other medical expert determined the cause of claimant’s impairment nor expressly found that it was consistent with the compensable *497injury. SAIF urges that, because this record “abundantly discloses other possible sources of claimant’s impairment,” the record legally is inadequate to permit a factual inference that the compensable injury caused claimant’s permanent impairment.
Claimant responds that, under this court’s opinion in SAIF v. Danboise, a claimant can meet his or her burden of establishing impairment due to a compensable injury by presenting the report of a medical arbiter that (1) contains impairment findings that are consistent with the compensable injury, and (2) does not attribute the impairment to causes other than the compensable injury. Claimant argues that the Board’s application of the described test was legally correct and that, as a factual matter, substantial evidence in the record supports the Board’s order.
We begin with a brief review of the principles relevant to the burden of proof in this context. The burden of proving that an injury or occupational disease is compensable, and of proving the nature and extent of any disability resulting therefrom, is on the claimant. ORS 656.266. The claimant must meet his or her burden by a preponderance of the medical evidence. See, e.g., Barnett v. SAIF, 122 Or App 279, 282, 857 P2d 228 (1993). OAR 436-035-0005(10)3 is instructive in regard to the type and quality of medical evidence that is required to meet the applicable burden of proof. It defines “preponderance of medical evidence” as follows:
“ ‘Preponderance of medical evidence’ or ‘opinion’ does not necessarily mean the opinion supported by the greater number of documents or greater number of concurrences; rather it means the more probative and more reliable medical opinion based upon factors including, but not limited to, one or more of the following:
“(a) The most accurate history;
“(b) The most objective findings;
“(c) Sound medical principles; or
“(d) Clear and concise reasoning.”
*498See also ORS 656.005(19) (“ ‘Objective findings’ in support of medical evidence are verifiable indications of injury * * * that may include, but are not limited to, range of motion, atrophy, muscle strength and palpable muscle spasm. ‘Objective findings’ does not include physical findings or subjective responses to physical examinations that are not reproducible, measurable or observable.”). Those provisions indicate that, in order for a workers’ compensation claimant to meet the applicable burden of proof, the evidence put forth by the claimant ordinarily must be composed of express findings, opinions, and reasoning of the treating physicians or medical arbiters. See SAIF v. Gaffke, 152 Or App 367, 954 P2d 179 (1998) (in that case, “the question of causation is a complex one, requiring expert medical opinion”); Barnett, 122 Or App at 282 (particularly where a claimant’s injuries are complex in nature, expert medical evidence is necessary to meet the burden of proof).
With those evidentiary principles in mind, we turn to the issue raised in this case. The starting point is our decision in Danboise, upon which both parties rely. In Danboise, the medical arbiters found that the claimant had cervical impairment but did not expressly attribute the impairment to the compensable injury. The Board affirmed the award of PPD, stating in part:
“We agree that claimant has the burden of establishing that his cervical impairment is due to his compensable injury. Claimant may, however, meet that burden by presenting a treating physician’s or medical arbiter’s report that: (1) contains impairment findings that are consistent with [his or her] compensable injury; and (2) does not attribute those findings to causes other than the compensable injury.”
Danboise, 147 Or App at 552 (quoting Board’s order; internal quotation marks omitted; emphasis omitted). The employer appealed, contending that the Board applied an incorrect standard of proof. This court determined that
“the Board’s explanation of the evidence that claimant was required to produce was correct. As the Board found, the record in this case identifies no noncompensable factors that may have contributed to claimant’s impairment. The Board is correct that, when the record discloses no other possible source of impairment, medical evidence that rates the *499impairment and describes it as ‘consistent with’ the compensable injury supports a finding that the impairment is due to the compensable injury.”
Id. at 553 (emphasis added). We concluded that the findings made by the Board met that standard and that they were supported by substantial evidence in the record. Id.
As those two quoted portions of our opinion reveal, there is a subtle but significant difference between the Board’s and this court’s articulation of how the claimant in that case satisfied her burden to prove permanent impairment in the absence of a medical opinion making express findings that the impairment was due to a compensable injury. The Board’s order in Danboise stated that claimant satisfied her burden with evidence (1) that the physician or arbiter made impairment findings consistent with the compensable injury, and (2) that the same physician or arbiter did not attribute those findings to any other cause. In affirming the Board, our emphasis was somewhat different. We agreed that, in rating the impairment, the medical evidence expressly described the impairment as “consistent with” the compensable injury. Furthermore, we considered it significant that nothing in the record disclosed any other possible source for the impairment, not just that the medical experts failed expressly to attribute the impairment to other possible causes.
Properly understood, Danboise is an application of the principles that apply to factual inferences in administrative cases. Evidence to support a finding of fact ordinarily can include reasonable inferences derived from facts expressly adduced, as long as the facts on which the inference is based are themselves supported by substantial evidence in the record and as long as there is a “basis in reason” for connecting the inference to the facts from which it is derived. See City of Roseburg v. Roseburg City Firefighters, 292 Or 266, 271, 639 P2d 90 (1981) (so explaining). In Danboise, we concluded that the factfinder could infer causation from two “facts” in combination: (1) a medical opinion expressly describing a claimant’s permanent impairment as consistent with the compensable injury, and (2) the lack of any evidence that the impairment could be due to other possible causes. The logic *500underlying the inference was: claimant has an impairment consistent with the compensable injury; there is no evidence of any other cause for the impairment; therefore, the impairment is due to the compensable injury.4
Here, as SAIF points out, the record contains extensive evidence that claimant’s impairment may be due to other possible causes. Indeed, those other possible causes were the reason that several of the medical experts expressly declined to attribute claimant’s impairment to the compensable condition. Thus, this case is distinguishable from Danboise. Distinguishing Danboise does not end the inquiry, however. The fact that we agreed that an inference of causation could be made on the record in Danboise does not establish that, on this different set of facts, the inference cannot be drawn.
In this case, the Board expressly adopted “the ALJ’s reasoning and conclusion that the opinion of Dr. Williams * * * establishes that claimant was entitled to an unscheduled permanent disability award.” The ALJ had inferred causation from Williams’s arbiter’s report because Williams “presumably” had reviewed claimant’s medical records and nevertheless did not attribute claimant’s condition to anything other than the compensable injury. In essence, Williams’s silence on causation was the “fact” from which the Board inferred that he formed a medical opinion that claimant’s impairment was, to a medical probability, attributable to the compensable injury.
We agree with SAIF that the limited facts on which the Board relied will not bear the weight of the inference that the Board drew. As SAIF correctly points out, the record here *501contains significant medical evidence of at least one noncompensable factor that may have caused claimant’s impairment — namely, claimant’s alcohol abuse. The record suggests other noncompensable factors as well — i.e., that claimant was suffering from or has a history of depression and possibly other mental disorders that may be responsible for his impairment, and that the tests of the extent of claimant’s impairment were influenced by his lack of effort and poor motivation. Moreover, no medical expert offered a direct opinion that claimant’s impairment is attributable to the compensable condition. Nor did any medical expert even offer an express opinion that the impairment “is consistent with” the compensable condition.5 Thus, the facts from which the Board inferred causation reduce to these two: (1) Williams was aware of possible noncompensable causes for claimant’s condition; (2) Williams did not say that claimant’s impairment was attributable to those other possible causes. From those two facts alone, the Board inferred that Williams must have concluded that the compensable conditions, rather than the noncompensable ones, caused the impairment.
The Board’s reasoning appears to be: if Williams knew about other possible causes, and if he did not say that claimant’s impairment was attributable to the other causes, then he must have concluded that claimant’s impairment was not attributable to those causes. That reasoning simply is faulty. No principle in logic declares that if several things are possible, and none is selected as being probable, then a *502specific one of the possible things must be more probable than the others.
It may be that the Board considered some additional unarticulated fact or presumption in drawing an inference of causation in this case. For example, the Board’s inference might have a greater “basis in reason” if it could be presumed that a medical arbiter’s findings of impairment always relate only to the compensable condition. In prior cases, however, the Board has rejected that presumption. See Julie A. Widby, 46 Van Natta 1065 (1994) (there is no requirement that a medical arbiter report only impairment findings that are due to the compensable injury); see generally Kim E. Danboise, 47 Van Natta 2281-82 (discussing cases). In all events, however, our review is limited to the Board’s stated rationale. As the Supreme Court has explained:
“Sometimes a rational nexus between an evidenced fact and an inference drawn from it is obvious from common experience (e.g., we may infer from the fact of a wet street that it recently rained). In other cases, however, and particularly in cases involving expertise, the reasoning is not obvious (e.g., we may infer from present meteorological conditions that it will snow tomorrow). In such an inference, we will not assume the existence of a rationale. Rather, we look to the order to state the rational basis of the agency’s inference. The explanation need not be complex, but it should be sufficient to demonstrate the existence of a rational basis and to allow for judicial review.”
Roseburg City Firefighters, 292 Or at 271-72 (footnote omitted). If the Board had additional facts or presumptions in mind when it drew the inference that it did, the Board did not tell us what those were. As to the explanation that the Board gave, the Board’s reasoning for inferring causation in this case is insufficient.6
*503In sum, claimant presented no direct medical testimony to establish that his impairment is due to the compensable injury. Nevertheless, the Board inferred such causation from Williams’s arbiter’s report. The Board’s stated rationale, however, does not provide a sufficient basis in reason for any such inference. Consequently, we reverse the Board’s disability award and remand for reconsideration.
On petition, reversed and remanded for reconsideration; affirmed on cross-petition.
Among other injuries, claimant suffered a skull fracture, a mild closed-head injury described as a “small epidural hematoma at the anterior left temple region,” and nasal and facial fractures.
The 18-month delay in conducting a neuropsychological examination apparently was due in part to claimant’s incarceration for 10 months during 1997.
The rule was promulgated by the Workers’ Compensation Division of DCBS.
As a syllogism, the logic is flawed: the conclusion that the impairment is due to the compensable injury does not invariably flow from the factual premises — i.e., that the impairment is consistent with the compensable injury and that there is no evidence of other possible causes. The conclusion is not compelled because the mere absence of evidence of any other cause does not affirmatively establish that there is no other cause. As a result, the conclusion to be inferred — i.e., that the impairment is due to the compensable injury — at most is a permissible one. Consistent with that observation, Danboise does not hold that the inference of causation must be drawn, only that it may be.
The ALJ expressly relied on Binder’s arbiter’s report, characterizing it as containing a conclusion that claimant’s impairment “could be consistent” with the compensable injury. The Board, however, declined to adopt the ALJ’s reliance on Binder’s report. The contents of that report suggest why. Rather than observe that claimant’s impairment was or could be consistent with the compensable injury, Binder observed only that claimant’s injury was such that “full recovery sometimes occurs and sometimes does not.” That is an equivocal statement at best. Moreover, it cannot be viewed in isolation from the rest of Binder’s reported medical opinion. Binder found that his examination was invalid and explained why (e.g., that certain test scores were not consistent with the injury claimant had suffered; that certain of the neurological test results were inconsistent in ways that suggested they were invalid; that claimant was poorly motivated). The invalid testing, together with the existence of other potential causes for claimant’s impairment, caused Binder expressly and directly to state that he could not make a diagnosis regarding the presence or absence of residual consequences due to claimant’s injury. Binder ended his report stating: “I do not know if I claimantl has any permanent impairment because of the invalid test results.”
If the concurring opinion’s basic point is that experts generally need not express themselves with particular word choices, we agree, at least as a general proposition. See Freightliner Corp. v. Arnold, 142 Or App 98, 105, 919 P2d 1192 (1996) (“expert’s testimony need not be ignored merely because it fails to include ‘magic words’ ”). That does not mean, however, that the Board’s factfinding role extends to supplying a medical opinion when the substance of the opinion is significantly in doubt because of the expert’s failure to articulate it. See SAIF v. Calder, 157 Or App 224, 228, 969 P2d 1050 (1998) (“The Board is not an agency with specialized medical expertise entitled to take official notice of technical facts within its specialized knowledge.”); Rolfe v. Psychiatric Security Review Board, 53 Or App *503941, 951, 633 P2d 846, rev den 292 Or 334 (1981) (“It is one thing, however, to say that an agency may employ its experience and expertise to evaluate and understand evidence and quite another to allow it to use its special knowledge as a substitute for evidence presented at a hearing.”). The fact remains that, in the workers’ compensation area, the legislature expressly requires compensability and extent determinations to be made based on preponderant medical evidence. To meet that standard, a medical opinion must be expressed by the medical expert, even if less-than-artfully, rather than divined by the factfinder. See generally Uris v. State Compensation Department, 247 Or 420,424,427 P2d 753 (1967) (endorsing the “settled rule” that where a worker’s injuries are of such character as to require skilled and professional persons to determine the cause and extent thereof, the question is one of science and must necessarily be determined by testimony of skilled, professional persons). We endorse no more liberal standard of proof or liberal view of the Board’s factfinding role in these cases.