In the contest for district delegates to the Con-
stitutional Convention from the Thirty-third Senate District in Bronx County at the general election of November 8, 1966, a close vote separates appellant-respondent Rice from respondent-appellant Santangelo. The election of the other two delegates is not in dispute.
The recanvass by the city Board of Elections show a difference of eight votes. Rice was shown by this recanvass to have received 45,281; Santangelo, 45,289.
This court has heretofore decided in an appeal on constitutional grounds that the Supreme Court has jurisdiction to review the canvass and recanvass of votes (Matter of Rice v. Power, 19 N Y 2d 106).
In a proceeding pursuant to subdivision 5 of section 330 of the Election Law, the court at Special Term, reviewing the recanvass, determined that 10 votes which had been added to the vote for Rice by the Board of Elections on the basis of examining a voting machine should be deducted from his count.
The court also determined, however, that the Board of Elections in recanvassing had made a numerical mistake of 10 votes adverse to Rice in adding his vote. Therefore, the numerical result of Rice’s vote as the court found it was the same as the vote found on recanvass, 45,281.
The court also credited to Santangelo two votes on paper ballots which had not been credited by the board on recanvass. The court’s totals, therefore, were Rice, 45,281; Santangelo, 45,291. The Appellate Division has affirmed these findings.
The essential legal issues as they reach this court are, therefore, whether the 10 votes for Rice shown on the machine on recanvass should be credited to him, or, as the court at Special Term found, deducted; and whether the two ballot votes for Santangelo should be credited to him as the court found.
If Bice is not entitled to the 10 votes in dispute on the machine, Santangelo will have a majority of eight votes, even if he is not *477credited with the paper ballot votes. If Bice is entitled to the 10 votes and Santangelo is not entitled to the two votes, Bice will have a majority of two. If Bice is entitled to the 10 votes and Santangelo to the two votes, the result will be a tie.
The focus of the controversy is a voting machine (No. 83867), one of two machines used in the 54th Election District of the 83rd Assembly District. Bice’s counsel regards this machine as “ crucial to the outcome of the election ”. The inspectors in the election district, examining this machine after the polls had closed, entered on their tally sheet 16 votes for Bice on Bow D (Conservative party). On recanvass the Board of Elections entered in their tally sheet 26 votes on this line for Bice.
It is undisputed that when this machine was examined by the Beferee in this present section 330 proceeding, it showed exactly what the recanvass tally of the Board of Elections showed, 26 votes for Bice.
Based on the recollections of witnesses present when the counters of the machine were examined on election night and the inspectors made entry in their tally of 16 votes, the Beferee found that the vote which Bice was entitled to receive was 16 rather than 26 votes and the court at Special Term confirmed this finding.
The statutory responsibility to recanvass the vote shown on any voting machine is expressly placed by law in the city Board of Elections (Election Law, § 274, subd. 1). The board, says the statute, “ shall recanvass the vote cast thereon ”,
Upon such recanvass, “If * * * it shall be found that the original canvass of the returns has been incorrectly made from any machine ”, the board shall make an appropriate statement of its recanvass (§ 274, subd. 2). The recanvass so made “ shall thereupon supersede the returns filed by the inspectors of election ”. It is thus the plain purpose of the statute to require that where there is a difference between the inspectors at the election district and the Board of Elections as to what is shown by the voting machine, the superseding canvass by the Board of Elections must prevail.
A difference between the original election district canvass and the recanvass is, of course, subject to judicial review in a proceeding under section 330 of the Election Law. The Supreme Court is given summary jurisdiction to determine a controversy *478arising from the canvass of returns by a city Board of Canvassers and may direct a recanvass or correction of the board’s determination (§ 330, subd. 5).
But if the court is to change the result of the board’s recanvass, which the statute expressly provides shall supersede that of the election inspectors, it must do so on a record which will show reliably that the board on recanvass has been mistaken in its result. The proof in this case does not reach that level. The voting counters on machines are purposefully designed mechanically to remain unchanged until they are unlocked and released.
There can be no doubt whatever that, when the recanvass of the machine in question was made by the board, Bow D showed 26 votes for Bice. This is what the Beferee saw when he re-examined the machine some time after the board had recanvassed it.
Several witnesses, including the inspectors of election of both political parties, testified, consistently with the entry made on their tally sheet, that 16 was the number they saw on the machine on election night; and this was the number purportedly entered in his tally by the policeman on duty.
But if this machine counter was unchanged at the time it was recanvassed, these earlier entries and recollections necessarily must be incorrect.
The rest of the numbers on the machine when recanvassed corresponded with the tally of the inspectors of election. It has not been shown how, and no one has offered any reasonable explanation, a single counter, out of a great many on the machine, could have been changed between the original canvass and the recanvass.
Conceivably, the people who examined the machine and made the tally might have read “ 1 ” for 44 2 ”. No special reason for personally remembering this counter among a large number, long before it was known the result might be affected by it, is shown.
The inspectors testified to recollections which agreed with their tally. It is for the correction of just such tally errors, which experience shows occur from time to time routinely in canvassing, that the statute authorizes the official recanvass and directs that it ‘4 shall thereupon supersede ’ ’ the original canvass.
*479The normal legal presumption attaches to the reliability of a superseding canvass over the original one. This presumption might be overcome by evidence .showing some reasonable mechanical probability of a change in the machine between the canvass and the recanvass. That has not been shown here and the court is required, on such a record as this, to accept the superseding canvass of the Board of Elections. (See, generally, Matter of Carson, 164 Misc. 945, affd. 254 App. Div. 801; Matter of Pettit, 246 App. Div. 895; Matter of Creedon, 264 N. Y. 40.)
When the machine was examined by the Referee in March, 1967, several months after the recanvass, it was found there was no 1 ‘ lockout ’ ’ device attached to Row D, Line 23, where Rice’s name appeared as a Conservative candidate. This device, also referred to in the record as an “ endorsement strap ” or “ compensator ”, is designed to prevent voting for the same candidate more than once in situations where a voter has more than one choice of candidates for the same office, here three.
But the absence of this device, assuming it had not been installed before the election, could offer no possible means of changing the number of votes for Rice on this line from 16 to 26 between the canvass and the recanvass.
It would, on the contrary, have been a mechanical impossibility to have pulled the lever down 10 times after the polls had closed and before the recanvass by the Board of Elections because, if there had been any further voting use of the machine, it would have been reflected both in the “ protective counter ” and the separate ‘ ‘ public counter ’ ’ of the machines which are designed to guard against any such change.
The public counter of the machine, as shown by the original canvass of the inspectors on election night; by the recanvass made by the Board of Elections; and by the Referee’s later examination in March, 1967, all indicated exactly the same number: 306; and the protective counter, similarly, .showed the same number: 00375. If a lever had been pulled down for Rice or anyone else after the polls closed, these counter numbers would have reflected such additional use of the machine.
Further, the “ lockout ” device does not serve the function of preventing a voter from pulling down a specific lever more than once to record an additional vote at that place. This, it is obvious, would be impossible on any voting machine anywhere *480because the machine will not record a vote until the voter moves the general machine handle to close it; and then it will record only those votes with the lever then pulled down. The lockout, therefore, is solely to prevent a voter from voting for the same candidate on more than one line.
That the lockout devices were actually on this machine when used at the election is suggested by the general proof showing the care taken in the preparation of the machine. Besides this, the 26 votes for Bice on the Conservative line, shown in the recanvass, were not higher, as they would probably be if absence of lockout played any part in the election itself, but lower than other candidates on the 'Conservative line.
It is thus quite impossible for the presence or absence of lockout to have had any effect on a change in vote after the election was over since such a change could be effected only by moving the general machine handle to permit voting levers to be pulled down and this, in turn, could not be done without reflecting itself in the protective counter and the public counter. The absolute identity and consistency of these counters at the polling place, at the recanvass and at the judicial examination demonstrate such a post-election change could not occur.
Therefore, the “lockout” device has no relevancy to any possible change in the vote of Bice on Bow D, Line 23. Indeed, Santangelo’s argument before the Beferee on this aspect of the proof was that no votes should be counted for Bice on this line because of the absence of the “lockout”; an argument not based on a machine change after election, but on a failure of protection during election itself. This argument was rejected by the Beferee.
Nor did the Beferee base his decision on a “ change ” in the counter on Bow D, Line 23, on the absence of a “lockout”. The Beferee noted that it “ was not established at the hearings ” that the change was the “ direct result of the absence of the protective device ”. He regarded as “ speculative ” the possibility that “ moving the machine to storage ” might have caused the change. It is not only speculative; its mechanical impossibility is obvious.
He concluded “some unknown agency” of a kind not suggested in any of the proof had “caused the counter here involved ” to change from 16 to 26. He said this had occurred *481u after the canvass and recanvass had been completed ” bnt he probably meant between the canvass and the recanvass, since number 26 was seen on the recanvass.
There is no proof that at the time of recanvass by the Board of Elections the machine had no “ lockout ” on Bow D, Line 23. On the contrary, the board’s recanvass records do not show any mechanical irregularity on the machine and the absence of the lockout device was shown months later when the Beferee examined the machine and at a time when the process of stripping down some of the machines had begun—a subject which neither side explored. No possible mechanical means has thus been .shown by which the change in this single counter could have been effected between the canvass and the recanvass.
This result adds 10 votes to the votes credited to Bice, which would give him two votes more than Santangelo; but this leads to a consideration of the further question whether Santangelo should be credited, as the Special Term and the Appellate Division found, with two votes cast for him on paper ballots not accorded him by the recanvass.
The statute authorizing recanvass does not distinguish between the result shown on machines and the result shown on paper ballots. There is a good discussion of the power of Boards of Election to recanvass paper ballots in Matter of O’Shaughnessy v. Monroe County Bd. of Elections (15 A D 2d 183, opinion by Halpern, J.).
But there is, of course, a difference in stability and durability between a voting machine and a paper ballot. This difference must be taken into account in a judicial review under subdivision 5 of section 330 of the Election Law on the reliability of the count on recanvass.
Where the question is whether a ballot marked in a particular way should or should not be counted, the resolution of the legal problem by the recanvassing board and its review by the court may be undertaken with considerable assurance.
But ballots are fragile things. Unlike voting machines, they may be lost, misplaced, or voided by marks or mutilation. A judicial review to determine under the opening paragraph of section 330 of the Election Law “ any question of law or fact arising ” in respect of the recanvass of the result depending on paper ballots, differing from the original canvass, would suggest *482a broader scope of judicial inquiry in reviewing the recanvass than would be appropriate for machines.
This would especially be so where the inspectors have counted a ballot on election night and it is missing and not found on recanvass. The Board of Elections cannot, of course, in carrying out its statutory function, count on recanvass what it does not find. But under a proper record, the court may be called on to determine in this and in similar discrepant situations whether the original canvass is reliable enough to stand nevertheless.
For very long periods before the general use of the voting machines, the legislative policy in New York, and its judicial implementation in dealing with paper ballots, suggests a heavy reliance on the original election district canvass. The ballots were not normally recanvassed except ballots marked protested, void or wholly blank by the inspectors and these were especially preserved for later judicial examination.
There was neither an official recanvass of the general count of the ballots, nor any judicial power summarily to pass on them and make a determination different from that of the inspectors. The recanvass by the Board of Elections was based on an examination of the tally sheets or returns of the inspectors.
If the inspectors had been mistaken in their actual count, the plenary judicial remedy to correct the result lay in quo warranto and not in a summary proceeding under the Election Law. This extraordinary and plenary action still lies and searches the underlying accuracy of election returns.
The history of the development of and the limitations upon official recanvass and judicial review when elections were largely held with paper ballots is to be seen in a number of cases as the statutory and judicial policy emerged.
Before the reforms effected by the changes in the Election Law of 1896, it was the practice, as Bartlett, J., noted in Matter of Stewart (155 N. Y. 545, 549 [1898]) to destroy both the ballots and the memoranda of count by the inspectors immediately after the election, and the inspectors’ statement of the count became the official and controlling evidence of the result.
The statute of 1896 required, among other things, the preservation of the contemporaneous tally made by the inspectors from which they made their return. The court held that where there *483was a difference shown in count between the tally sheets and the returns of the inspectors, and where there was no challenge to the accuracy of the tally sheets themselves, these latter, “ the contemporaneous, self-proving record of the canvass ” (supra, p. 553) would control and the Board of County and City Canvassers could be compelled by mandamus to act accordingly. The scope of the controversy is more fully explored in the opinion of Ingraham, J., at the Appellate Division (24 App. Div. 201).
But as the judicial construction of the Election Law developed after Stewart (supra), it became clear that the contemporaneous tallies and consistent returns of inspectors as to the count made by them were not open to later change either by recanvassing boards or by the court in a summary proceeding, except as to those ballots marked void, protested or wholly blank which were separately sealed and subject to re-examination of various kinds.
This direction of decision is to be seen in three main cases: People ex rel. Brink v. Way (179 N. Y. 174 [1904]); Matter of Hearst v. Woelper (183 N. Y. 274 [1905]); and People ex rel. March v. Beam (188 N. Y. 266 [1907]). It is “very clear”, Chief Judge Parker wrote in Brink (p. 180), “ that the legislature does not intend to permit the court to order a recount of the votes in the box”. In March, Judge Bartlett observed (p. 271) that it had become “ settled by the decisions of this court that a recount is not authorized by the Election Law ”. See, also, People ex rel. Brown v. Freisch (215 N. Y. 356) which dealt with the limitations on the review of the protested, void and blank ballots. As to ballots other than those, it was again noted (p. 369) that “ [j]ur is diction is not granted to a direct recount or recanvass of such ballots.”
The present statute, as it has been seen, permits the recanvass by the city Board of Canvassers of paper ballots as well as of machines. Paper ballots, now limited to absentee and military ballots, play a relatively small part in a modern election. But in laying down guidelines for the judicial review of a recanvass of the count of paper ballots, the historical experience with election procedures followed when paper ballots were universally used may be helpful. Unless there is some substantial attack on the integrity of the contemporaneous tallies and returns of the inspectors of election whose duty it is to examine and count the ballots, they should be accepted by the court.
*484The fullest judicial inquiry may he indicated to determine the question of integrity of the tallies and returns, including the absence of any ballot on the recanvass or its physical condition or marking. In .such an inquiry records dealing with the number and type of ballots transmitted to the election district may throw important light on the accuracy of the inspectors’ original canvass. Where there is consistency between the records and the inspectors’ canvass, the court may find a reliable ground for accepting the original canvass.
The difference in stability and reliability between paper ballots which, without any attribution of wrong, may be misplaced or lost, and a voting machine counter, seems to suggest a pragmatic differential in judicial approach.
The contemporaneous inspectors’ tally discussed in Stewart {supra) may become the most reliable remaining basis for judicial determination as, conversely in dealing with machines, the counter seen on recanvass and judicial review seems to be the most reliable basis for determination.
The records of the Board of Elections show that two military ballots were returned by the voters and sent by the Board of Elections to the 79th Election District of the 85th Assembly District. The returns of the inspectors for this election district showed two military ballots were counted for various candidates, including Santangelo.
The recanvass showed two military ballots were received from the 79th District of the 85th Assembly District cast for San-tangelo ; but, since one of the ballots belonged to the 79th District of the 83rd Assembly District, it was not credited to Santangelo by the board.
It is not .suggested that this ballot should have been counted for him in the district in dispute. It differs in its vote for several offices from the original canvass of military ballots in the district in dispute.
It is suggested, rather, that the ballot belonging to the 83rd Assembly District had been misplaced at the time of recanvass and the original military ballot counted for Santangelo in the original canvass for the 79th District of the 85th Assembly District had also been misplaced.
Thus, the records show reliably two military ballots were received for and sent to the 79th Election District of the 85th *485Assembly District and the original canvass shows two military ballots were counted for Santangelo.
A finding based on a similar situation was made and affirmed relating to the 26th Election District of the 80th Assembly District. Two ballots, one absentee and one military, were credited to Santangelo in this district. Only the absentee ballot was seen on recanvass and the military ballot was not credited. But there is documentary proof that a military ballot was counted for other candidates and proof by the inspectors that there was one military and one absentee ballot in the canvass. In these circumstances, the court at Special Term and the Appellate Division upheld the accuracy of the canvass.
These factual determinations supporting the accuracy of the original canvass of paper ballots in the two election districts are sufficiently supported by the record. The 10 additional votes for Bice on the basis of the recanvass and the acceptance of the conclusions at Special Term as to the two paper ballots cast for Santangelo lead to a tie in the result of the election.
The order of the Appellate Division should be modified to direct that the return on recanvass be corrected to show a tie between petitioner-appellant Bice and respondent Santangelo and, as modified, affirmed, without costs.