Proceeding pursuant to CPLR article 78 to review so much of a determination as, after a hearing, dismissed the petitioner from her position as a tenured teacher upon a finding that she had been guilty of conduct unbecoming a teacher and of insubordination. Petition granted, determination annulled insofar as reviewed, on the law, with costs, the charges of conduct unbecoming a teacher and insubordination are dismissed, respondent is directed to reinstate petitioner to her classroom duties, with back pay, less the amount of compensation earned in any other employment or occupation and any unemployment benefits she may have received, and the proceeding is remitted to Special Term for a determination, after a hearing, of whether back pay should be further reduced in accordance with the guidelines set forth in our decision in Matter of Short v Nassau County Civ. Serv. Comm. (59 AD2d 157). The petitioner, a tenured teacher who had served in the Great Neck public schools for some 18 years, was charged with, inter alia, conduct unbecoming a teacher and insubordination. The formal charges arose out of allegations that she had prepared two of her classes for the vocabulary portion of the Stanford Achievement Test by revealing to them many of the words that were to appear on the examination. It was alleged that the petitioner had engaged in such misconduct with her fifth grade class in 1977 and with her sixth grade class the following year. The petitioner denied the charges and demanded a hearing pursuant to section 3020-a of the Education Law. At the conclusion of the hearing, which consumed over 2,500 pages of transcript, the majority of the three-member panel sustained the charges of conduct unbecoming a teacher and insubordination and the petitioner was dismissed from her position. She subsequently instituted the proceeding at bar. In our view, the determination was not supported by substantial evidence and, accordingly, must be annulled insofar as reviewed. The evidence against the petitioner was entirely circumstantial, consisting for the most part of a disputed statistical analysis of the performance of her students. The inquiry began when, following the 1978 Stanford Achievement Test, five of the petitioner’s sixth grade pupils reported to their parents that several of the words of the examination were familiar to them because the petitioner had reviewed them during class vocabulary drills. The parents in turn notified the principal of the petitioner’s school, and the question was raised as to whether, in the course of her vocabulary drills, the petitioner had purposefully revealed and taught specific words which she knew would appear on the examination. When confronted with this suggestion, the petitioner *569denied any wrongdoing. She asserted that her vocabulary drills had been based entirely upon a set of cards which she had composed personally. The cards contained 999 words which petitioner had gathered over the years from various books read in her classes. As part of the drills, which began some four to six weeks prior to the examination, pupils in the petitioner’s class were given lists of words for study. The parents who first brought the matter to the principal’s attention produced five lists which were said to be part of those given to their children in petitioner’s class. After further inquiry by the principal, formal charges were filed against the petitioner. At the hearing which followed, the respondent relied essentially on a statistical approach. It was established that, in 1978, in sharp contrast to other sixth grade students in the school, those in the petitioner’s sixth grade class had scored 26 percentile points higher on the test than they had the year before. Similarly, it was shown that the petitioner’s 1977 fifth grade class had enjoyed an improvement of over 20 percentile points but had then shown a corresponding decrease the next year. In further support of the charges, the respondent called Dr. Carl Winston, the District Co-ordinator of Research and Evaluation for the Great Neck public schools. He testified that there were 285 so-called “key words” on the vocabulary portion of the Stanford test. These were the words which the students were either asked to match with words of similar meaning or were offered as choices in the matching process. Of the 285 key words, 118, or 41.4%, appeared on the petitioner’s cards. Dr. Winston testified that he had examined other standard achievement tests and had found that, of the key words used on those tests, only about 2% or 3% appeared on the petitioner’s cards. From this disparity, Dr. Winston concluded that the petitioner must have obtained her words from the Stanford Achievement Test. Moreover, Dr. Winston found that a high percentage of the words appearing on three of the lists provided by the parents also appeared on the Stanford Achievement Test while a very low percentage appeared on other standard achievement tests. To refute the respondent’s charges, the petitioner produced two expert witnesses in the field of educational research, statistics and testing. These witnesses, who testified in petitioner’s behalf without compensation, did not find the fluctuation in the students’ performance indicative of intervention in the testing process. The substance of the witnesses’ testimony was that student scores could dramatically increase where the teacher employs rote instruction and extensive drills, encourages students to do much individual reading and exposes the students to literature in class. In this regard, the petitioner testified without contradiction that, in addition to the extensive vocabulary drills her students received in class, all but three of the students had read at least 40 books during the school year and all had been exposed to some 100 books during the same period. As to a subsequent decline in performance, there was expert testimony that discontinuance of such teaching methods could well result in a significant drop in student scores. Further, the petitioner’s experts were of the view that comparisons between the Standford Achievement Test and other standard tests were of little value. The Stanford test draws on so-called “content words” taken from books dealing with areas such as science and social studies. Other tests rely exclusively on basic reading textbooks. Thus, if the petitioner’s vocabulary drills drew upon “content” material, it was to be expected that a greater percentage would appear on the Stanford test than on the others. The petitioner’s experts also took strong exception to the statistical analysis presented by Dr. Winston. He had laid heavy emphasis on the fact that 41.4% of the key words on the test appeared on the petitioner’s cards. The petitioner’s experts, however, found little significance in that statistic. More important, according to them, was that, of the 999 words appearing on the petitioner’s cards, only about 11.8% were on the test. *570Significantly, of the nearly 4,000 words in the various glossaries used by the petitioner in compiling her cards, a similar percentage, 10.31%, appeared on an older addition of the Stanford Achievement Test. Finally, there was no evidence that the petitioner had seen the 1978 test before it was administered. In this regard, the respondent was compelled to rely on the fact that, in 1973, the year prior to the initial utilization of the Stanford test in the Great Neck public schools, a packet containing the test materials for fifth and sixth graders was reviewed by all teachers. The petitioner, who was then teaching a fourth grade class, would have also had access to the packet. As to the specific test in question, however, the respondent’s own witness testified that the petitioner had picked up the examination on the very morning it was given. In 300 Gramatan Ave. Assoc. v State Div. of Human Rights (45 NY2d 176, 180-181), the Court of Appeals defined substantial evidence as follows: “The concept of substantial evidence, a term of art as related to administrative decision making, is rather easily verbalized but, when put to use in respect to a particular determination, frequently causes difficulty and disagreement, as witnessed here by the divergence at the Appellate Division (see Matter of Stork Rest. v Boland, 282 NY 256, 274, supra; see, also, Matter of Paulsen [Catherwood], 27 AD2d 493, 495). It is related to the charge or controversy and involves a weighing of the quality and quantity of the proof (Matter of Di Nardo v Monaghan, 282 App Div 5, 7; McCormick, Evidence [2d ed], § 352); it means such relevant proof as a reasonable mind may accept as adequate to support a conclusion or ultimate fact (New York State Labor Relations Bd. v Shattuck Co., 260 App Div 315, 317). Essential attributes are relevance and a probative character (Edison Co. v Labor Bd., 305 US 197, 229; Matter of Ralph v Board of Estimate of City of N. Y., 306 NY 447, 454). Marked by its substance — its solid nature and ability to inspire confidence, substantial evidence does not rise from bare surmise, conjecture, speculation or rumor (cf. Matter of Milea v Easy Appliance Div., Murray Corp., 29 AD2d 730). More than seeming or imaginary, it is less than a preponderance of the evidence, overwhelming evidence or evidence beyond a reasonable doubt (see People ex rel. Consolidated Water Co. v Maltbie, 275 NY 357, 370; Matter of Erin Wine & Liq. Store v O’Connell, 283 App Div 443, 446, affd 307 NY 768; Matter of Paulsen [Catherwood], 27 AD2d 493, 495, supra; cf. State v Randecker, 79 Wn 2d 512, 517-518; District of Columbia v Heman Ward, Inc., 261 A2d 836, 840 [DC]).” In our view, the compelling nature of the independent expert testimony presented by the petitioner, coupled with her demonstrated lack of prior access to the 1978 examination, lead inexorably to the conclusion that the determination with respect to that test was not supported by substantial evidence. And since the only evidence offered in relation to the 1977 test was the fluctuation of class scores, those charges must also fail. The petitioner offered substantial expert testimony to the effect that such fluctuations, in the circumstances at bar, were without significance. As to petitioner’s contention that it was improper to permit Dr. Winston, an employee of the Great Neck public schools and one of respondent’s witnesses, to remain in the hearing room in order to assist the district in the presentation of its case, we agree. The balance at the hearing was unfairly tipped in favor of the district in that it had the “technical” assistance of its key witness for the entire trial, while the petitioner’s experts were not present in the hearing room, except when they gave their own testimony. Moflen, P.J., Hopkins, Damiani and Titone, JJ., concur.