Plaintiff1 appeals a judgment dismissing a writ of habeas corpus. He asserts that the Board of Parole and Post-Prison Supervision (the Board) failed to follow the applicable statute and administrative rules in extending his parole release date for two years. We reverse.
Many of the arguments that plaintiff raises parallel those that we discussed in Weidner v. Armenakis, 154 Or App 12, 959 P2d 623 (1998), withdrawn by order July 13, 1998, reasoning reaffirmed and readopted in Merrill v. Johnson, 155 Or App 295, 964 P2d 284, rev den 328 Or 40 (1998),2 and involve the Board’s authority to extend a parole release date under ORS 144.125(3) (1991). That statute requires the existence of a severe emotional disturbance constituting a danger to the health or safety of the community before the Board may extend the release date. In Weidner, we concluded that ORS 144.125(3) (1991) did not require that there be a psychiatric or psychological diagnosis of that condition. Rather, we held that it was the Board’s responsibility to decide whether a prisoner had such a condition and that in doing so it could rely on all the evidence in the record; it was not limited to considering only the psychiatric or psychological report.
In this case, plaintiff relies on an administrative rule that we did not consider in Weidner, because it was not cited or argued to us. That rule shows that, at least during the period that the rule was in effect, the Board believed that a psychiatric or psychological diagnosis that met the statutory requirements was essential to the Board’s authority to extend the release date. Thus, the broader construction that we gave the statute in Weidner, to which we adhere in this opinion, is not controlling for that period. Because plaintiff committed his offenses when the rule was in effect, it continues to apply to the Board’s consideration of his case. The *263Board, however, followed the approach that we approved in Weidner, with the result that it has not decided whether the psychologist’s diagnosis meets the statutory standard. The trial court on remand should give the Board an opportunity to make that decision.
Plaintiff has been imprisoned since early 1989 for offenses that he committed in late 1988. Shortly after his imprisonment, the Board applied the matrix system that was then in effect and set a parole release date of October 30, 1997. On March 7,1997, a psychologist interviewed plaintiff; the psychologist prepared a report a few days later. On May 7, the Board issued an order extending plaintiff’s parole release date to October 30,1999. It gave the following explanation of its decision:
“The [B]oard, based on all the information it is considering at this hearing finds that the doctor’s diagnosis coupled with all the information it is considering, does result in a • finding of a present severe emotional disturbance that constitutes a danger to the health or the safety of the community. The [Bjoard has considered this matter under the laws in effect at the time of the commitment offense.”
The first issue is whether the Board could extend plaintiffs release date without first determining that the psychologist had diagnosed plaintiff with a severe emotional disturbance that constitutes a danger to the health or safety of the community. That issue arises under ORS 144.125(3) (1991), which provided at the time of plaintiff’s offenses:
“If a psychiatric or psychological diagnosis of present severe emotional disturbance such as to constitute a danger to the health or safety of the community has been made with respect to the prisoner, the board may order the postponement of the scheduled parole release until a specified future date.”
In Weidner, we held that the statute referred to a legal, rather than medical, standard; that an express diagnosis that met all of the statutory criteria was therefore not necessary; and that the Board was entitled to use any information before it in determining whether the prisoner’s condition satisfied the statutory criteria. 154 Or App at 17. We rejected the dissent’s argument that the statute required the Board to *264determine from the psychiatric or psychological report alone whether there was a diagnosis that gave the Board the authority to extend the release date. 154 Or App at 21-22 (Warren, J., dissenting).
In this case, plaintiff has cited OAR 255-60-006 (1988) in support of the position that we rejected in Weidner, an administrative rule that we did not consider in that case. That rule shows that, at the time of plaintiffs offenses, the Board understood its role in very much the way that the dissent in Weidner suggested. OAR 255-60-006 (1988) covered a number of issues related to a prisoner’s release on parole. The relevant portions are subsections (7) and (8):
“(7) The Board may order a psychiatric/psychological report anytime prior to release. If the record indicates that a psychiatric or psychological condition of severe emotional disturbance, such as to constitute a danger to the health or safety of the community, is present, the Board may consider deferring parole release until a specified future date.
“(8) If the evaluation does not make a finding of severe emotional disturbance such as to constitute a danger to the health or safety of the community, the Board shall affirm the parole release date and set parole conditions.”
(Emphasis added.) Plaintiff also points out that in 1983 the Board’s chair stated that the Board had no basis to refuse to release another prisoner, because all three psychological reports indicated that he was not emotionally disturbed or dangerous. The Board’s attorney’s legal opinion was that its hands were tied. See O’Bremski v. Maass, 905 F2d 281, 285 (9th Cir 1990).
The Board has authority to adopt rules applicable to parole. ORS 144.050. OAR 255-60-006(8) was such a rule, and the Board adopted it within its delegated authority. The rule expressly required the Board to affirm a parole release date unless the evaluation made the finding that ORS 144.125(3) contemplates.3 It is, of course, axiomatic that an *265agency must follow its own rules. Moore v. OSP, 16 Or App 536, 519 P2d 389 (1974); see also Aetna Casualty & Surety Co. v. Sue A. Blanton, D.C., 139 Or App 283, 287, 911 P2d 363 (1996). Even if an agency is not required to adopt a rule, once it has done so it must follow what it adopted. Harsh Investment Corp. v. State Housing Division, 88 Or App 151, 744 P2d 588 (1987), rev den 305 Or 273 (1988). The rule may limit what an agency would otherwise be able to do. “An agency which is vested with discretion by statute may limit its own discretion in its regulations.” Wyers v. Dressler, 42 Or App 799, 807, 601 P2d 1268 (1979), rev den 288 Or 527 (1980), overruled on other grounds by Mendieta v. Division of State Lands, 148 Or App 586, 941 P2d 582 (1997), rev dismissed 328 Or 331 (1999). Thus, the question is not, as it was in Weidner, what the statute requires; rather, the question is what the rule requires.* **4
As the majority and dissenting opinions in Weidner show, it is possible to read ORS 144.125(3) in more than one way. In OAR 255-60-006(8), the Board gave it a meaning that was closer to that of the Weidner dissent than of the Weidner majority. So long as that rule was in effect, it required the Board to treat the statute as requiring a formal finding in the psychiatric or psychological evaluation as a prerequisite to the Board’s authority to extend a prisoner’s parole release date. That requirement imposes a greater limit on the Board’s authority to extend a release date than does allowing *266the Board to make its own determination, based on all the evidence in the record, of the prisoner’s condition. The rule, thus, is more favorable to a prisoner than is the view of the statute that we adopted in Weidner. Because the rule was in effect at the time of plaintiffs offenses and was more favorable to plaintiff than Weidner held ORS 144.125(3) to be,5 it would violate the ex post facto provisions to apply ORS 144.125(3), rather than OAR 255-60-006(8), to plaintiff. See Williams v. Board of Parole, 112 Or App 108, 112, 828 P2d 465, rev dismissed 313 Or 300 (1992) (rules made pursuant to delegated authority are subject to ex post facto analysis). Because the Board did not determine whether the psychological report made a finding that satisfied the statutory conditions, it erred in extending plaintiffs parole release date.6
We next consider whether the psychologist’s report would permit the Board to find that the psychologist diagnosed plaintiff as having a severe emotional disturbance that constitutes a danger to the health or safety of the community.7 If it would not, then under the rule the Board did not have the authority to extend plaintiffs parole release date and the trial court on remand should direct plaintiffs release. If it would, then the trial court on remand should direct plaintiffs release unless the Board, under the relevant legal authority including OAR 255-60-006(8) (1988), determines to extend plaintiffs parole release date. See Meadows v. Schiedler, 143 Or App 213, 221, 924 P2d 314 (1996) (similar conditions on remand when Board improperly applied amended statute).
Because of the importance of the psychological report under the rule, we summarize it at some length. The *267psychologist first noted that plaintiffs last disciplinary citation came in 1990, for having THC in his system. He then summarized plaintiffs description of his offenses, noting that plaintiff admitted the conduct that constituted the crimes but asserted that he acted in self-defense. Plaintiffs affect was appropriate to mildly anxious; there was no evidence of hallucinations, delusions, or other psychotic thought processes. His judgment appeared adequate, although his insight was slightly limited.
The psychologist then described the results of standardized tests, which suggested someone who had “difficulty incorporating the values and standards of society. Such individuals would tend to be rebellious toward authority figures and often in conflict with them.” They would be somewhat impulsive, striving for immediate gratification, acting without considering the consequences, with poor judgment and considerable risk-taking. Such people tend to be insensitive toward others, with elements of hostility and aggressiveness in their makeup.
From the testing and his interview, the psychologist concluded that plaintiff had normal needs for closeness but did not anticipate positive interactions with others. “Mr. Peek tends to intermingle feelings with thinking during problem solving and decision making activities.” “He tends to modulate or control emotional discharges much less than other people. He would probably call attention to himself because of the intensity with which he expresses feelings.” “He has a very negative, angry attitude toward the environment.” His capacity for control and tolerance for stress will usually be adequate, but a great deal of stress would have greater impact on his thinking. He tends to misinterpret his perceptions and to act inappropriately, in part because of excessive negativism and anger. The psychologist did not diagnose an Axis I (clinical disorder) condition; on Axis II (personality disorder), the diagnosis was 301.9, Mixed Personality Disorder with Antisocial and Narcissistic Features. The psychologist concluded that “[o]ver the past five years, [plaintiff] seems to have made determined efforts to rehabilitate himself. However, in my clinical judgment, I should still consider him to be a danger to the community.” The danger *268would be lessened if he were to complete violence prevention and intensive substance abuse programs.
Determining whether the psychologist’s diagnosis satisfies the standard of the statute and the rule is not a simple task. As we suggested in Weidner, the statute creates a legal, not a medical standard, so applying a medical diagnosis to that legal standard necessarily requires some translation from one system to the other. The legal determination is whether a “prisoner’s emotional disorder was (a) present, (b) severe, and (c) one that made the prisoner a ‘danger to the health and safety of the community.’ A ‘diagnosis’ of a mental disorder, standing alone, will not always provide the necessary information * * 154 Or App at 18. In his diagnosis, the psychologist relied on the fourth edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM-IV).8 As we noted in Weidner, the DSM-IV emphasizes that a clinical diagnosis of a mental disorder is not sufficient to establish that an individual’s condition meets a particular legal standard. “ ‘[T]he fact that an individual’s presentation meets the criteria for a DSM-IV diagnosis does not carry any necessary implication regarding the individual’s degree of control over the behaviors that may be associated with the disorder.’ ” 154 Or App at 19 (quoting DSM-IV) (emphasis in Weidner). That is, a diagnosis does not in itself necessarily indicate whether the condition is severe or whether it is the condition that creates whatever danger the person presents.
Under DSM-IV, 301.9, Personality Disorder Not Otherwise Specified, is a diagnosis of
“disorders of personality functioning that do not meet criteria for any specific Personality Disorder. An example is the presence of features of more than one specific Personality Disorder that do not meet the full criteria for any one Personality Disorder (‘mixed personality’), but that *269together cause clinically significant distress or impairment in one or more important areas of functioning (e.g., social or occupational).”
The psychologist diagnosed a personality disorder that, has both narcissistic and antisocial features. DSM-IV points out that many highly successful individuals display narcissistic personality traits. Only when the traits are inflexible, maladaptive, and persisting and cause significant impairment or distress would they constitute Narcissistic Personality Disorder. DSM-IV at 661. Similarly, it points out that one must distinguish Antisocial Personality Disorder from criminal behavior that is not accompanied by personality features that are characteristic of the disorder. Again, only when the traits are inflexible or otherwise significantly problematic do they constitute a disorder. DSM-IV at 649.
Reading the psychological evaluation as a whole, thus, there appear to be questions whether the diagnosis is of a severe disorder and whether the disorder causes plaintiff to be dangerous. The evaluation could permit the Board to find that the diagnosis satisfies those conditions, but that conclusion is by no means compelled. It is, thus, something that the trial court on remand should permit the Board to consider. We therefore reverse and remand to permit it to give the Board that opportunity.
Reversed and remanded with instructions to direct petitioner’s release unless the Board, applying OAR 255-60-006 (1988), determines to extend petitioner’s parole release date.
As has become our style in habeas corpus cases, in this opinion we will refer to the party seeking the writ as “plaintiff,” although it would be more accurate to describe him as “petitioner.”
Although we withdrew our opinion in Weidner after we learned that the case had become moot, for convenience we will refer to it in this opinion, although technically Merrill is the precedential case.
The dissent suggests that “the evaluation” in OAR 255-60-006(8) (1988) refers to the Board’s evaluation. There is no textual foundation for that conclusion. Nothing else in the rule refers to a Board evaluation. On the other hand, the *265reference to “the” evaluation follows immediately after a reference to a psychiatric or psychological report and most logically refers to it — indeed, there is nothing else in the rule to which “the evaluation” can refer. The use of “the” suggests that the rule refers to a specific evaluation that it has already identified; only the psychiatric or psychological report would qualify. The writer of the rule simply used two different terms to refer to the same thing.
That is not, as the dissent suggests, an issue of delegation. The Board has not delegated anything. It retains its discretion under ORS 144.125(3) to determine whether to extend the parole release date for a person who satisfies the criteria of the statute and the rule. All that the Board has done is to establish the type of evidence that is necessary to satisfy the statutory standard. The Board, of course, determines whether the evidence meets that standard. In that respect, the rule is similar to the requirement of a writing to prove the existence of certain agreements, see ORS 41.580, or the requirement of evidence corroborating a confession or the testimony of an accomplice. See ORS 136.425; ORS 136.440. It is even closer to the rule that expert testimony is usually necessary to prove negligence in professional malpractice cases. See, e.g., Griffith v. Blatt, 158 Or App 204, 211, 973 P2d 385 (1999); Docken v. Ciba-Geigy, 101 Or App 252, 790 P2d 45, rev den. 310 Or 195 (1990).
The Board amended these rules effective April 5, 1990. We do not consider the meaning or effect of the amendments.
Plaintiff also relies on rules of the Corrections Division involving the transfer of inmates to and from the Mental Health Division as establishing the criteria that a Board finding must meet. Those rules are not rules of the Board and are otherwise irrelevant to a decision to release a prisoner on parole. We do not consider them further.
Because of its failure to recognize the effect of OAR 255-06-060 (1989), the Board did not focus on whether the psychologist’s report satisfied the criteria of the rule. Instead, it drew its own conclusions about plaintiffs condition. The dissent would rely on those conclusions to sustain the Board’s order. They do not, however, resolve the issue that the rule requires the Board to resolve.
Plaintiff argues that the diagnosis must be based on DSM-III, a predecessor to DSM-IV, because that was the version that existed when the legislature adopted ORS 144.125(3). However, as we indicated in Weidner, the legislature did not adopt the medical terminology and thought of a particular period but, rather, created a legal standard. Whether a medical diagnosis satisfies that legal standard does not depend on whether the diagnosis uses the terminology that was current when the legislature created the standard.