Petitioner Linn-Benton-Lincoln ESD (district) petitions for judicial review of an order of the Employment Relations Board (ERB) certifying the results of a self-determination election by a group of early childhood interventionists. The district assigns error to ERB’s order including the interventionists in the academically licensed employees’ bargaining unit. The district contends that ERB erroneously interpreted ORS 243.650(1) and mistakenly relied on its own previous decisions in concluding that the interventionists are eligible for membership in the “academically licensed” unit. The question for decision in this case is whether the interventionists are “academically licensed,” as the phrase is used in ORS 243.650. That statute provides, in part:
“(1) ‘Appropriate bargaining unit’ means the unit designated by the Employment Relations Board or voluntarily recognized by the public employers to be appropriate for collective bargaining. However, an appropriate bargaining unit cannot include both academically licensed and unlicensed or nonacademically licensed school employees. Academically licensed units may include but are not limited to teachers, nurses, counselors, therapists, psychologists, child development specialists and similar positions.”
For the reasons that follow, we affirm ERB’s order.
ERB’s material findings of fact are essentially undisputed. The district provides educational and support services to school districts in Linn, Benton, Lincoln, and Coos counties. Respondent Linn-Benton-Lincoln Education Association/OEA/NEA (union) represents a bargaining unit of approximately 62 employees who work in the district’s special education division (the academically licensed unit). The interventionists provide special education services to infants and young children in the district. OAR 581-015-0900. Interventionists must have earned, at minimum, a bachelor’s degree or have had comparable advanced training in early childhood education, special education, or a related field. OAR 581-015-1100(3). They are not required to hold an occupational or professional license from a competent authority in order to engage in their vocation. The district considers the interventionists to be “classified” employees. The district *561employs classified personnel in 36 different positions, none of which requires a professional license. Seven classified positions, including the position of interventionist, require a bachelor’s degree.
In April 1996, the union filed a unit clarification petition seeking to add the interventionists to the academically licensed unit. After a hearing, the administrative law judge (ALJ) issued a proposed order to include the interventionists in the unit. The district filed objections with ERB, which reversed the AU’s order. The union filed a petition for reconsideration, contending that ERB erroneously interpreted ORS 243.650(1) by concluding that interventionists are “unlicensed” and therefore ineligible to join with “academically licensed” employees in a single bargaining unit. On April 3, 1997, ERB issued a new order in which it decided that the interventionists are “academically licensed” and thus appropriately included in the unit.
ERB explained its decision on reconsideration as follows:
“[The union] rightly points out that this Board, in Mid-Valley Bargaining Council [v. Greater Albany School Dist., Case No. C-17-81, 6 PECBR 4766 (1981)], coined the term ‘academically licensed’ to describe a community of interest factor to be considered in cases concerning school district employees. For that purpose, this Board could have used a different term — such as ‘academically trained’ — to describe that community of interest factor. The focus of the factor, then, is on ‘academic;’ and this Board did not intend the term licensed,’ in the context of cases concerning school district bargaining units, to be strictly construed according to its dictionary definition. This circumstance is explained in the Mid-Valley Bargaining Council decision itself where we stated:
“ ‘* * * We specify that the certificate, license, degree, or the equivalent must be “academic,” as opposed to solely occupational, in order to preclude inclusion in the unit of persons who are required to possess a nonacademic license, such as a driver or chauffeur * * * or journeyman plumber * * *.’ 6 PECBR at 4780 * * *.
*562“According to the job description for interventionists, applicants must possess a ‘[b]achelor [sic] degree or comparable advanced training in the area of special education or early childhood education or related field.’ The training required for the position therefore is ‘academic.’ The degree or certificate that certifies that the holder has completed such academic training is a ‘license’ for purposes of the community of interest factor that this Board denominates ‘academically licensed.’ We also conclude that the position of interventionist is of like character to those occupations— other than ‘teacher’ — listed in the statute as examples of academically licensed employees. The enumerated positions, as they are employed by school districts, share the trait we recognized (concerning school nurses and psycho-metrists) in Mid-Valley Bargaining Council: their duties ‘complement those of the teachers in that they all are devoted to enhancing the educational development of students.’ 6 PECBR at 4778. The duties of the interventionists, basically intended to assist very young children in their development to help prepare them for school, clearly are complementary to those of the teachers to whom the children one day will be assigned.
“In light of the above analysis, we now hold that we erred in our original Order when we concluded that interventionists do not share with other members of [the union’s] bargaining unit the community of interest factor of ‘academically licensed.’ ” (Footnotes omitted, emphasis in original.)
Finally, ERB held that the interventionists do not have a clearly distinct community of interest from the other employees in the academically licensed unit and accordingly must be included in it. As a result, ERB ordered a self-determination election. In June 1997, the elections coordinator certified the election results, which favored union representation. We review ERB’s April 3 order because it is a foundation for the order certifying the union. ORS 183.482(8).
ERB is charged by statute with the responsibility of designating an appropriate bargaining unit for school employees. ORS 243.682(1).1 If the interventionists are “academically licensed,” as the phrase is used in ORS 243.650(1), *563they may be joined in an “appropriate bargaining unit” with other academically licensed school employees. If not, they may only be included within an appropriate bargaining unit consisting of other “unlicensed” or “nonacademically licensed” employees.
In this case, we review ERB’s interpretation of statutory terms in the context of an order in a contested case. The terms “academically licensed,” “nonacademically licensed,” and “unlicensed” are not defined by statute. They are inexact terms in that “the legislature has expressed [their] meaning completely, but that meaning remains to be spelled out in [ERB’s] rule[s] or order[s]. An inexact term gives the agency interpretive but not legislative responsibility.” England v. Thunderbird, 315 Or 633, 638, 848 P2d 100 (1993). On review of ERB’s interpretation of inexact statutory terms, our role is to determine whether the agency erroneously interpreted the meaning of those terms in its attempt to discern and apply the legislature’s intent. Id. If legislative intent is clear from the text and context of the statute, then we end our inquiry. If the intent is unclear, then we move to the second level of analysis and examine the legislative history underlying the statute. Finally, if the intent of the legislature remains uncertain, then we resort to general maxims of statutory construction. PGE v. Bureau of Labor and Industries, 317 Or 606, 611-12, 859 P2d 1143 (1993); Carroll v. Boise Cascade Corp., 138 Or App 610, 614, 910 P2d 1111 (1996).
We begin our analysis with an examination of the text and context of the term “academically licensed.” Neither the phrase nor its words are defined by statute. Webster’s Third New Int’l Dictionary, 1304 (unabridged ed 1993), defines “licensed” as “having a license,” and “license” as “a right or permission granted in accordance with law by a competent authority to engage in some business or occupation * * * which but for such license would be unlawful.” Although the interventionists need not hold a formal license from a competent licensing authority, ERB concluded that the plain and natural meaning of the word “licensed” is not controlling *564because of textual and contextual evidence that the legislature intended a broader meaning.
Textually, ERB noted the express inclusion of child development specialists by example in ORS 243.650(1) as “academically licensed” employees. Like interventionists, child development specialists need not hold a professional or occupational license. ERB apparently adopted the union’s argument that, because child development specialists need not be licensed in a conventional sense yet are academically licensed under the statute, interventionists may be likewise included because they fulfill a similar role in the education of children. Although the comparison provides a possible clue to the legislature’s intent, we agree with the district that it is not highly persuasive. OAR 581-023-0050(5), which specifies the qualifications for child development specialists, provides, in part:
“Individuals contracted as Child Development Specialists shall hold a Personnel Services Certificate, a Personnel Services Certificate of Accomplishment as issued by Teacher Standards and Practices Commission or be approved to serve in that capacity by the State Board of Education. * * * [Requirement of a master’s degree is recommended * * (Emphasis added.)
Child development specialists must either hold a professional certificate or receive an alternative approval from the State Board of Education in order to practice. Those professional requirements are closely identifiable with the plain and ordinary meaning of the word “license.”2
However, the listed examples of academically licensed positions are not exclusive. ERB decided that the interventionists “is of like character” to those listed because all of them “enhanc[e] the educational development of students.” The problem with that conclusion is that it is not evident from a first step PGE analysis.3 While the interventionists are academically trained, and undoubtedly share a *565community of interest with employees in the listed job positions, the statute does not unambiguously support ERB’s conclusion that the interventionists’ academic training makes them academically licensed.
The district, on the other hand, steadfastly argues that the text resolves the case in its favor because the dictionary definition of the word “licensed” is controlling. We disagree. The term we are called on to construe is a phrase consisting of two words: “academically” and “licensed.” We may not focus on the meaning of one of its component words to the exclusion of the other. The plain, natural, and ordinary meaning of the word “academic” is “belonging to, or associated with an academy or school, esp. of higher learning * * *: formed by school training or associations * * *: based on formal study at an institution of learning, esp. of higher learning.” Webster’s at 9. Institutions of learning do not award licenses to their graduating students: they award certificates, diplomas, degrees, and the like. In other words, they train and educate their students, rather than grant them legal authority to engage in a profession.
The ambiguity of the phrase arises from the conjunction of its constituent words. Because of the differences in emphasis between the words “academically” and “licensed,” they simply do not coalesce to signify any single, obvious meaning. The joinder of those words suggests at least two alternative, plausible meanings. The first is that the persons described must be both academically trained and must also be licensed by a competent legal authority. The second is that the persons described must be entitled to engage in a particular occupation or profession by reason of academic training. *566Each construction requires some clarification or enhancement of the language in the phrase in order to prevail. In order to arrive at the first suggested meaning, the district must add the word “trained” to the word “academically,” then add the word “and” between “trained” and “licensed.” On the other hand, if the second meaning is what the legislature intended to express, it coined a term for academic training that, in and of itself, qualifies a person for work. That intent could also have been more clearly expressed. It is the need to relate the words to each other in order to evoke a logical meaning for the phrase that makes its meaning ambiguous at the first level of analysis.4 Both of the meanings that we have suggested plausibly, although not decisively, capture the legislature’s intent. Accordingly, we conclude that a textual analysis of the statute does not fully resolve the question.
ERB based its decision largely on the contextual prong of the first level analysis prescribed in PGE. ERB concluded that its decision in the Mid-Valley Bargaining Council case is part of the context of the 1995 amendment to ORS 243.650(1), which introduced the term “academically licensed” into the determination of appropriate bargaining units. Neither party has cited any case that directly holds that agency decisions either may, or may not, be treated as context for statutory interpretation. However, when
“administrative expressions of policy are offered as context, courts must be cautious not to make policy in the guise of interpretation, or to allow agencies * * * to achieve through a court’s interpretation policy objectives that the enactment as promulgated was not meant or failed to embody.” DLCD *567v. Jackson County, 151 Or App 210, 218, 948 P2d 731 (1997), rev den 327 Or 620 (1998) (emphasis in original).
That caution applies to the tender of the Mid-Valley Bargaining Council decision as statutory context. The case never actually employed the precise term used by the legislature but instead used a series of other phrases such as “academic certificate, license, degree, or the equivalent,” 6 PECBR at 4772, 4780; “credentialed employe[e]s,” id. at 4776-77; “professional certification,” id. at 4777; “certification or educational attainment,” id.; “certificate, license, degree, or the equivalent must be ‘academic,’ as opposed to solely occupational,” id. at 4780; “certificated personnel,” id. at 4781 n 2. The Mid-Valley Bargaining Council case addressed whether “certificated personnel” such as teachers, nurses, psychologists, and similar positions were appropriately grouped together in a single bargaining unit. 6 PECBR at 4766-67. ERB considered the question under ORS 243.682(1), which directed it to consider “such factors as community of interest, wages, hours and other working conditions of the employees involved, the history of collective bargaining, and the desires of the employees.” At the time that case was decided, there was no other legislative guidance as to what groupings of employees might constitute an “appropriate” bargaining unit. ERB concluded that a broad unit was appropriate, encompassing all jobs “which, as a condition of employment, require an academic certificate, license, degree, or the equivalent * * * ” Id. at 4780. In 1995, ORS 243.650(1) was amended to include the “academically licensed” requirement, but it did not include any language about certificates, degrees, or equivalencies.
We agree that it was not implausible for ERB to conclude that the term “academically licensed” is shorthand for its terminology in Mid-Valley Bargaining Council. However, the lack of precise symmetry in terminology between the statutory text and ERB’s decision makes ERB’s conclusion less certain. It is unclear, for example, why the legislature did not use the broader terminology employed by ERB, including academic certification, degrees, and equivalents, if it intended employees such as the interventionists to join a single, professional bargaining unit. The omission may or may not be significant. For that reason, ERB’s decision in *568Mid-Valley Bargaining Council does not shine a bright contextual light on the legislature’s adoption of the inexact term “academically licensed” some 14 years later. On the other hand, the district has provided little assistance in the analysis of statutory context because of its mistaken view that the statute’s text is decisive. Accordingly, we conclude that the meaning of the questioned terms cannot be determined from a first level analysis under PGE.
Neither the parties nor ERB moved to the second level, consideration of legislative history, to inform the inquiry into legislative intent.5 Unfortunately, the legislative history does not clearly reveal the meaning intended by the term “academically licensed.” However, it does suggest that the general purpose behind adopting ORS 243.650 was to require educators and others holding professional positions to bargain separately from those employees who perform other occupational tasks.
The original version of the bill stated, “An appropriate bargaining unit cannot contain both teachers and school employees who do not have teaching licenses.” Draft, SB 750, February 27, 1995. A representative of the Oregon School Board Association, testifying in favor of that version, indicated an understanding that the measure would separate teachers from classified employees such as transportation workers:
“Bargaining for teachers and classified employees requires addressing the specific needs of employees in each bargaining unit. As school districts are becoming more heavily regulated through state and federal standards and mandates, there are an increasing number of different issues facing these two groups. This makes it increasingly difficult to support the Employment Relations Board’s conclusion [in Welches Education Association/OEA/NEA v. Welches *569School District No. 13, 12 PECBR 12/304 (July/August 1990)] that working conditions of both employee groups are not disparate. For example, with classified staff, there are requirements that are specific to that employee group, including overtime requirements, fingerprinting requirements, drug and alcohol testing for transportation employees, vacations, etc.” Testimony, House Labor Committee, SB 750, April 21, 1995, Ex F (statement of Chris Dudly, Oregon School Board Association).
In contrast, the Oregon Education Association proposed that the bill’s language be amended to include “noncer-tified professional staff’ in the teachers’ unit. Testimony, Senate Labor and Government Operations Committee, March 2,1995, Ex D (statement of John Danielson). In a public hearing before the House Committee on Labor, Danielson expressed concern that the original version
“le[ft] some people with no place to go. For example, there are districts where they have sociologists, where they have counselors, they have other professional people [who] do not have teaching licenses, nor are they required to have teaching licenses because of their nature in the system and their work with children and their work with other people in the district, they are more naturally aligned with the professional unit than they would be with the classified unit.” Tape recording, House Committee on Labor, SB 750, April 21, 1995, Tape 92, Side B (statement of John Danielson, Oregon Education Association).
Shortly after that testimony, the bill was amended to substitute the term “academically licensed” for the original language. SB 750, May 2, 1995.6
Although the foregoing testimonial history is informative, we are reluctant to draw decisive inferences concerning legislative intent from that history for several reasons. First of all, the statements related to earlier versions of SB 750, rather than to its final enacted form. Moreover, the statements were made by witnesses and are not direct expressions of legislative intent. See, e.g., State v. Guzek, 322 Or 245, 261, 906 P2d 272 (1995). Finally, any inference drawn from the *570nature of revisions to the bill after those statements were made is, at best, permissive.
Because the meaning of the statute remains uncertain even after examining the text, context, and legislative history of ORS 243.650, we “resort to general maxims of statutory construction to aid in resolving the remaining uncertainty.” POE, 317 Or at 612. One of those maxims holds that a court should attempt to construe the language of a statute in a manner consistent with its purpose. Bartz v. State of Oregon, 314 Or 353, 358, 839 P2d 217 (1992); Welliver Welding Works v. Farmen, 133 Or App 203, 210, 890 P2d 429 (1995). We conclude that following that maxim is the most rational path out of this particular thicket of statutory construction under the teaching of PGE.7
While legislative history is not dispositive, it does provide some guidance when viewed in light of the maxim that we are to construe in accordance with the general statutory purpose.8 The timing sequence between witness Danielson’s expression of concern that professional staff should be aligned with teachers in a single bargaining unit and the subsequent amendments to SB 750, does help tip the scales in favor of ERB’s construction of the statute. The interventionists are professional, academically trained staff and, as such, are similarly situated to the other nonteaching professionals that Danielson felt could fall between the cracks under the former language of the bill.
When we consider all relevant evidence of the legislature’s purpose, including the evolution of the statutory language during the 1995 legislative session, the similarity *571between the statutory terms and those used by ERB in the Mid-Valley Bargaining Council decision, and a comparison of the nature of the interventionists’ position with those positions expressly included by statute in the academically licensed classification, it seems consonant with that purpose to place the interventionists with other professionals in the academically licensed category. The interventionists’ position requires academic training and involves a direct professional service to students. Including them in the academically licensed unit is in keeping with what appears to have been the legislature’s general purpose in distinguishing between academically licensed and nonacademically licensed employees.9 Although our analysis takes a different route, we therefore conclude that ERB correctly interpreted ORS 243.650 to include the interventionists in the academically licensed unit.
Affirmed.
ORS 243.682(1) provides, in part:
“[I]n making its determination [ERB] shall consider such factors as community of interest, wages, hours and other working conditions of the employees *563involved, the history of collective bargaining, and the desires of the employees. The board may determine a unit to be the appropriate unit in a particular case even though some other unit might also be appropriate.”
While it may be tempting to conclude that the examples of academically licensed positions listed in the statute defeats ERB’s interpretation of the phrase based on the principle of ejusdem generis, we are wary of the risks inherent in assuming too much from a short list of qualified positions.
The district concedes that the phrase “and similar positions” is a form of express noncompleted legislation that ERB is given delegated authority to complete. England, 315 Or at 637. While ERB has delegated authority to determine *565what positions are similar to those listed in the statute, its discretion is limited by the meaning of the term “academically licensed.” Springfield Education Assn. v. School Dist., 290 Or 217, 224-25, 621 P2d 547 (1980). Although the examples listed in the statute furnish evidence of the term’s meaning, we may not simply defer to ERB’s apparent conclusion that the interventionists occupy a similar position. ERB’s authority is not based on a broad delegation of rulemaking authority resulting from a general, indirect specification by the legislature. If that were the case, then ERB’s exercise of its delegated authority would normally be controlling and PGE’s methodology would not apply. Meltebeke v. Bureau of Labor and Industries, 322 Or 132, 142 n 12, 903 P2d 351 (1995). However, here, as in PGE, we are called on to decide whether the specific, although in this case inexact, terms of a statute conflict with an agency’s interpretation of that statute. Therefore, PGE’s methodology governs our review.
The foregoing observation best explains the essential disagreement we have with the dissent’s view of this case. The dissent’s predominant focus on the word “licensed” leads it to conclude that the statute is not ambiguous, a conclusion that does not withstand scrutiny if the questioned term is construed as a phrase, rather than as two unrelated words. With respect, the dissent never squarely confronts the changes that must be made to the phrase in order to require it to read as the dissent would have it.
Nor do we mean to say, as the dissent suggests, that the word “licensed” has no meaning as used in the statute. However, we do agree with ERB that, modified by the adverb “academically,” “licensed” could, as part of a phrase, plausibly refer to academic training that, in and of itself, qualifies a person for designated employment.
In a footnote to its ruling and order on petition for reconsideration, ERB made clear that its analysis did not reach the examination of legislative history:
“[The district] asserts that it is ‘pure speculation’ that the legislature considered the Mid-Valley Bargaining Council case in drafting the amendments to ORS 243.650(1). * * * [The district’s] assertion ignores the circumstances that the case is part of the ‘context’ of the statute and that the legislature employed a term, ‘academically licensed,’ that-was used in the case and is not in general use as a term of art in other contexts.”
On June 1, 1995, the Conference Committee further amended the bill to include the examples of “academically licensed” employees contained in the final version. SB 750 amendments, Conference Committee, June 1,1995.
We are mindful of the criticisms that flow from a judicial search for legislative purpose at the third level of analysis under PGE. Jack L. Landau, Statutory Construction in Oregon, 32 Willamette L Rev 1, 57-66 (1996). However, we conclude that no other nontextual canon of construction informs our analysis in any meaningful way. See, e.g., ORS 174.030 (natural rights); Westwood Homeowners’ Assn, Inc. v. Lane County, 318 Or 146, 160, 864 P2d 350 (1993), adhered to as modified on recons 318 Or 327 (1994) (statutes should be construed to avoid unconstitutionality).
It is appropriate to allude to legislative history at the third level of the PGE analysis. See Windsor Ins. Co. v. Judd, 321 Or 379, 387-88, 898 P2d 761 (1995). Although the legislative history cited here is not clear enough, standing alone, to allow us to render a decision at the second level, in combination with Welliver Welding Works’s maxim of statutory construction, we find its guidance useful.
The dissent’s criticism of the relative strength of individual elements of our analysis at the second and third levels under PGE makes few points that we do not acknowledge. Nonetheless, because the indicia of legislative intent are not plain and overwhelming, of necessity, our resolution of the case relies on the cumulative weight of the available evidence of that intent at each stage of the analysis. The dissent’s proposed resolution of the case, because it mistakenly comes at the first level under PGE, does not fully sort out the evidence so as to arrive at the most probable construction of the legislature’s intent.