dissent and vote to grant the petition, annul the determinations and dismiss the charges against petitioners, with the following memorandum: The petitioners herein, who are two collection agents employed by the respondent Transit Authority, were charged with "miscon*909duct and/or incompetence” resulting from a violation of four rules of the respondent authority. The specification stated that on June 9, 1976 petitioners "improperly handled revenue” at the 81 Street Station of the "A” line of the IND division of the New York City subway system. After a hearing, the hearing examiner sustained the charges and the respondent dismissed petitioners from their positions. The instant proceeding then ensued. In confirming the respondent’s determinations, the majority concludes that the inference of guilt drawn by the hearing officer and adopted by the respondent, was one reasonably drawn from the evidence which outweighed the inference of innocence, and constituted the requisite quantum of substantial evidence. We disagree. In our view, the respondent did not satisfy its burden of proving the charges by substantial evidence. The essential facts adduced at the hearing are not in dispute. The record indicates that the respondent had become concerned about discrepancies in revenue received from high turnstiles (enclosures which permit entry to the subway when an attendant is not on duty). A procedure was instituted on the morning of June 9, 1976 to check on the high turnstile at 81 Street and Central Park West in Manhattan, which had showed persistent discrepancies. Three Transit Authority employees (an accountant, auditor and safe inspector) met with an accountant from Price Waterhouse & Co. at the selected turnstile. With the four men present, all of the coins were removed from the safe, into which the turnstile automatically deposits them, and placed in a bag. The bag was taken to the token booth where two of the four men making up the inspection team counted the coins and were observed by the other two men. The actual count of the items taken from the safe of the high turnstile, which was numbered N-45 H-l, was conducted in Booth N-45, using a counting machine at about 9:00 a.m. In conducting the count the first time, all of the coins were poured onto a 12-inch tray. At the beginning of this count, for a period of several minutes, all of the coins were fed into the counting machine without any initial check by visual inspection and without manual separation of slugs. The machine began to jam as some coins were rejected by it. At that point, the two inspectors who were conducting ■the count began to visually examine the coins and manually separate the slugs from the tokens; they continued to feed the tokens into the counting machine. The total number of slugs found was 20. The remaining coins which passed through the counting machine was 2,290, which made a grand total of 2,310 coins. The 20 slugs were stacked separately near the counting machine. A second count was taken by feeding the 2,290 coins back into the counting machine, and the figure of 2,290 was again found. The coins were then put back in the original bag, returned to the high turnstile, placed in the safe and the safe was locked. The inspection team went upstairs and waited until the revenue collectors, pursuant to their prescribed duties, came to collect the bag. The revenue collectors’ job was to remove the contents of the bag, place the contents in another bag, and deposit that bag in a safe located in Token Booth N-44, which was approximately one block away. The token collector at Booth N-45 testified that he saw the inspection team return the bag to the high turnstile after their initial two counts and lock the safe of the high turnstile. He further testified that from that point on no one approached the turnstile except the two revenue collectors who came about an hour later. The token collector saw the two revenue collectors unlock the high turnstile and begin their job, but saw nothing unusual or incriminating. However, after a few seconds, his attention was diverted by numerous members of the public, who approached his window asking for tokens, and he lost sight of the two revenue collectors. The two revenue *910collectors performed their job completely unobserved by the inspection team (since the team was upstairs), made their "drop” at Booth N-44 and, sometime after 10:00 a.m., the inspection team went to the safe in Booth N-44 and took out all the items. The bag was tagged on the outside to show which turnstile it had come from and was sealed. Inside there was a yellow paper identifying one of the revenue collectors as the petitioner Coker. The inspection team conducted two more counts in Booth N-44 using a different counting machine than was used earlier and, as a result of these counts, found 37 slugs and 2,273 tokens, for a combined total of 2,310 coins, which was exactly the same as the combined total in the earlier counts in Booth N-45. At the hearing, the petitioners denied having removed any tokens from the bag and replacing them with slugs between the pickup at the turnstile and the drop in the safe in Booth N-44. They testified that after removing the contents of the bag from the high turnstile, and in accordance with required procedure, they immediately secured the bag with the proper seal and placed it in the safe. The hearing officer rejected petitioners’ testimony and their arguments and found that "the preponderance of the evidence * * * establishes the charges”. In his report sustaining the charges, the hearing officer stated, among other things, the following: (1) "after the 'drop’ the count in Booth N44 was made in the same fashion” as in Booth N-45; (2) "The Authority * * * presented circumstantial evidence that tokens were removed from the contents of the High Turnstile and. replaced by slugs during a period when only the Respondents [petitioners herein] had access to such contents”; (3) "the testimony of the disinterested Authority witnesses, including that of the outside accountant, is credible with respect to the validity of the two counts made”; and (4) "The evidence that slugs on occasion pass through the system is outweighed by the care and supervision exercised in the particular counts about which there was testimony.” It is clear that the two elements of the respondent’s case which led to the inference of guilt drawn by the hearing examiner were (1) the discrepancy in the slugs discovered between the counts in Booth N-45, and the subsequent counts in Booth N-44 and (2) the exclusivity of possession by petitioners of the bag between the counts in Booth N-45 and the counts in Booth N-44. The inference of guilt based on the existence of these two elements would be a rational and logical one if the evidence also established that (1) each of the counting machines did not usually accept slugs and (2) that the manual separation of slugs from valid tokens after their visual identification was a reasonably reliable and accurate method of removing all slugs from a mass of coins also containing tokens. In drawing the inference of guilt, it is readily apparent that the hearing officer was not persuaded "that the first counting machine failed twice to reject slugs, that [the two inspectors making the count] failed to observe that slugs had entered the counting machine on the first count [that the inspectors observing the count] failed to observe that such slugs entered the counting machine and that the second counting machine rejected slugs which had been accepted by the first such machine”. However, our reading of the record discloses substantial and critical items of proof which, in our view, support findings contrary to those made by the hearing examiner and which create serious doubts as to his evaluation of all of the proof. On the basis of this proof, we are dissuaded from joining in confirming determinations based upon circumstantial evidence from which an inference of guilt based more on suspicion than substantial evidence is drawn. The initial flaw in the hearing officer’s findings is found in his statement "that the count in Booth N-44 [after the drop by petitioners] was made in the same fashion” as in Booth N-45. The *911proof in the record does not support this finding. The first count in Booth N-44 was conducted in the following manner: (1) the coins were visually observed and the 37 slugs were manually separated from the other coins, stacked, and put to the side; and (2) the remaining coins, 2,273, were passed through the counting machine. On the second count in Booth N-44, the slugs remained separated and the 2,273 other coins were again fed into the machine. The first count in Booth N-45, during which 20 slugs were uncovered, was significantly different in that the coins were initially fed into the counting machine for several minutes without any attempt at visual observation and manual separation of slugs from valid tokens. After an undisclosed number of coins passed through the machine, the machine jammed. It was only then that the inspectors began to visually observe and manually separate the slugs from the coins which had not as yet been fed into the machine. On this count,. conducted as above described, 20 slugs were discovered. On the second count, conducted in Booth N-45, these 20 slugs remained separated and the 2,290 remaining coins were fed into the machine. In addition to this proof, which reflects a significant difference in the conduct of the first counts in each booth, the record also contains substantial proof with respect to the acceptance and rejection of slugs by the counting machines which is inconsistent with the hearing officer’s finding "that slugs [only] on occasion pass through the system” (emphasis supplied). The evidence also reveals the presence of serious constraints against accurate visual identification of slugs and valid tokens and the reliable manual separation and removal of slugs which considerably undermines the validity which the hearing officer attached to the two counts made. The proof to which we refer, consists of the following: (1) The admission by the token collector at Booth N-45 that the counting machine in that booth, which was used for the first count, takes some slugs and rejects others, with the result that slugs are often counted as tokens; (2) The concession of one member of the inspection team that, with respect to the counts in Booth N-45, some of the slugs could have «gone through the machine; (3) The testimony of another token collector that the counting machines used by the respondent accepted slugs; (4) The uncontradicted testimony that some slugs so resemble valid tokens they would escape visual inspection when poured into the tray prior to being fed into the counting machine; (5) The uncontradicted testimony that slugs had eluded visual observation as well as the machines at the authority’s central headquarters at Jay Street, Brooklyn, and that slugs were returned in bags from Jay Street to the individual token booths; and (6) Evidence that the lighting conditions in these booths were not conducive to accurate visual identification of slugs and valid tokens. The record does not contain any clear testimony or other evidence which discloses how many of the coins had passed through the machine on the first count in Booth N-45 either before the manual separation of slugs from the other coins was commenced or after the separation of the 20 slugs was completed. The proof is rather vague as to the actual rejection of slugs by the machine during the initial feeding process in Booth N-45 and nowhere in the record is the number of slugs rejected by the machine, if any, in this initial process mentioned or hinted at. As to the jamming of the machine which prompted the inspection team to halt the feeding of the coins until the removal of the 20 slugs was completed, the proof does not establish that the jamming was caused by the presence of slugs rather than by the more likely cause of the concentration of coins. There is no proof that any of the 2,290 coins fed into the machine during the second count in Booth N-45 after the 20 slugs were separated from the mass of coins, or that any of the *9122,273 remaining after the separation of the 37 slugs on both counts in Booth N-44, were rejected by the respective machines. The hearing officer’s finding that "slugs on occasion” pass through the system cannot be reconciled with the admissions and concessions of the authority’s own witnesses as to the frequency with which slugs passed through counting machines. These admissions and concessions, as well as the absence of proof that any coin fed into the machine after the manual separation of slugs had been completed was rejected and the frequent return of slugs from headquarters to booths, more clearly establishes that slugs were routinely accepted by counting machines than that they passed through "on occasion”. The inference that these unrejected coins were valid tokens is not a reasonable one considering the unreliability of the manual separation procedures. The additional proof as to the lighting conditions and the inspection team’s limited experience in counting, confirm the conclusion that the manual separation of slugs from the valid tokens following visual identification was not conducted with the care and accuracy found by the hearing officer and was not the reliable and accurate method of checking which he evaluated it to he. On the basis of the total proof, it is more logical and reasonable to infer, as to the 17 additional slugs in Booth N-44 after the drop, that all or some of these slugs had passed through the machine in Booth N-45 during the first stage of the feeding process and that the remainder of them eluded visual identification, escaped manual separation and thereafter passed through the machines unrejected. This inference, which supports innocence, is a more logical and reasonable one than the inference drawn by the hearing officer which supports guilt, to wit, that these additional slugs had been placed in the bag by the revenue collectors after they removed an identical number of valid tokens after the pickup at the high turnstile and before the deposit of the coins in the Booth N-44 safe. The fact that the total number of coins counted in both booths was exactly the same, as well as the fact that one of the revenue collectors was not regularly assigned to this route, also militates against the inference of guilt and bolsters = the more logical and reasonable inference of innocence. The authority’s own misgivings with this checking system and the quality of this proof are betrayed by disguising a clear-cut charge of stealing in such general, vague and cosmetic terms as misconduct, incompetence and mishandling of revenues. The utilization of administrative jargon to state such a clear-cut accusation is indicative of the authority’s lack of confidence in its case against these employees. The authority, on the basis of the evidence presented, has not satisfied its burden of proving guilt either by the preponderance of evidence as found by the hearing examiner or by the substantial evidence standard more frequently used in the determination of appeals from administrative determinations. A case against these employees cannot be made out merely by choosing an inference of guilt based on mere suspicion rather than the very reasonable inference of innocence. In expanding on the test of substantial evidence, the Court of Appeals, in Matter of Stork Rest. v Boland (282 NY 256, 273-274), stated: "A finding is supported by the evidence only when the evidence is so substantial that from it an inference of the existence of the fact found may be drawn reasonably. A mere scintilla of evidence sufficient to justify a suspicion is not sufficient to support a finding upon which legal rights and obligations are based. That requires 'such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.’ ” (See, also, Matter of Francese v Waterfront Comm. of N.Y. Harbor, 56 AD2d 535.) By its conclusion that the circumstantial evidence here rises to the level of a preponderance of the evidence, or substantial evidence, the majority de*913prives these employees of their livelihood on the basis of a record which would not support a conviction for a criminal charge of stealing or a recovery of $8.50 in a civil action. Clearly, no one can argue that the circumstantial evidence adduced at this hearing would satisfy the burden of proving guilt beyond a reasonable doubt in a criminal proceeding. Nor would it satisfy the burden of proof in a civil suit to recover $8.50. In a civil suit where the evidence is circumstantial, as here, the inference sought to be drawn must be the only one that can be fairly and reasonably drawn from the facts and any other explanation must be clearly and reasonably excluded (Galbraith v Busch, 267 NY 230, 233-235). As a result, the majority permits the imposition of a more drastic penalty, recommended by a hearing officer selected and compensated by the accuser, than is legally permissible or possible on the same facts and evidence in either a criminal proceeding or a civil lawsuit before a Judge with or without a jury. Finally, it should be noted that among the five cases decided under Matter of Pell v Board of Educ. (34 NY2d 222, 238), was the case of Matter of Best v Ronan, where the court upheld a charge of " 'nickeling’ ” or mishandling of funds belonging to the New York City Transit Authority. However, that case is clearly distinguishable from the case at bar, since petitioner’s activities therein were observed for a period of more than 18 days. In the case at bar, petitioners’ actions during the actual few moments that they were in possession of the bag were never observed by the inspection team or any other official or employee of the respondent authority. We also note that the petitioners — the two collection agents — did not know each other and that this was their first joint assignment. To assume that under such circumstances they immediately conspired to and did steal flies in the face of common sense. Accordingly, we dissent and vote to grant the petition and annul the respondent’s determinations.