Sir John Beaumont:
This is an appeal by special leave from the judgment and order of the High Court of Judicature at Patna dated 19th September 1947 dismissing an appeal against the judgment and order of the Court of the Sessions Judge of Ganjam-Puri dated 23rd July 1947 whereby the appellant was convicted of the offence of murder under S. 302/34, Penal Code, and sentenced to death. At the conclusion of the arguments their Lordships announced that they would humbly advise His Majesty that the appeal be allowed and would state their reasons later. This they now proceed to do.
 It is not in dispute that on 11th October 1946 one Kalia Behara was brutally murdered at a place between Berhampur, where he lived and carried on business as a jutka driver, and Golantra, to which he was driving with passengers in his jutka. It is unnecessary to refer to the details of the murder; though it may be noted that the motive attributed to the appellant was that he is a relation of accrued 1 and 2 who are said to have been on terms of enmity with the deceased, but both of whom were acquitted of the murder. Eight persons were charged with the offence and tried by the Sessions Judge of Ganjam-Puri. The learned Judge convicted six of the accused including the appellant who was accused 7 and one Trinath, who was accused 5. The six convicted persons appealed to the High Court at Patna. Two of the appeals were allowed, but the other appeals, including these of the appellant and Trinath, were dismissed. The only question which arises on this appeal is whether there was evidence upon which the appellant could be properly convicted.
 The evidence against the appellant consisted of, (a) the evidence of Kholli Behara who had taken part in the murder and had become an approver; (b) the confession of Trinath recorded under S. 164, Criminal P. C., which implicated both himself and the appellant in the murder, but which was retracted in the Sessions Court; and (c) the recovery of a loin cloth identified as the one which the deceased was wearing when he was assaulted, and a khantibadi, or instrument for cutting grass, in circumstances alleged to implicate the appellant.
4.The law in India relating to the evidence of accomplices stands thus: Even before the passing of the Indian Evidence Act, 1872, it had been held by a Full Bench of the High Court of Calcutta in R. v. Elahee Buksh, 5 W. R. Cr. 80 : (Beng. L. R. Sup. vol. 459), that the law relating to accomplice evidence was the same in India as in England. Then came the Indian Evidence Act which by S.133 enacts that:
"An accomplice shall be a competent witness against an accused person; and a conviction is not illegal merely because it proceeds upon the uncorroborated testimony of an accomplice."
Illustration (B) to S. 114, Evidence Act, however, provides that:
"The Court may presume that an accomplice is unworthy of credit unless he is corroborated in material particulars."
Reading these two enactments, together the Courts in India have held that whilst it is not illegal to act upon the uncorroborated evidence of an accomplice it is a rule of prudence so universally followed as to amount almost to a rule of law that it is unsafe to act upon the evidence of an accomplice unless it is corroborated in material respects so as to implicate the accused; and further that the evidence of one accomplice cannot be used to corroborate the evidence of another accomplice. The law in India, therefore, is substantially the same on the subject as the law in England, though the rule of prudence may be said to be based upon the interpretation placed by the Courts on the phrase "corroborated in material particulars" in Illustration B to S. 114.
5. The approver in the present case was a man aged about twenty. He was arrested on 12th October 1946, the day after the offence, and on 14th October was sent by the police to a Magistrate, who was called as a witness at the trial, and who on 15th October recorded a statement of the approver under S. 164, Criminal P.C. In this statement the approver described the murder, and alleged in effect that he and Trinath had been engaged by the appellant to assist in the murder, which they did; that in the struggle the cloth of the murdered man became untied and the appellant threw it over a bush. On 17th February 1947 the approver having been tendered a pardon, gave evidence before the Committing Magistrate. His evidence followed the general lines of his statement made under S. 164 but added some further details. In particular he said that the appellant gave him a khantibadi which he, the approver, subsequently gave to Trinath who handed it over at the appellant's house. The approver also alleged that the appellant gave to each of them, himself and Trinath, a sum of Rs. 25, presumably as remuneration for the part they had taken in the murder. On 8th July 1947, the approver gave evidence in the Sessions Court. His evidence was that he knew nothing about the murder, and he denied all the facts to which he had deposed before the Committing Magistrate. He affirmed that his evidence before the Committing Magistrate was the result of beating and tutoring by the police, and he denied that he had made any statement at all before a Magistrate under S. 164, Criminal P. C. Thereupon the Sessions Judge brought the evidence of the approver given before the Committing Magistrate upon record under S. 288, Criminal P. C., the effect being to make the evidence given before the Committing Magistrate evidence in the case for all purposes. Both the learned Sessions Judge, and the learned Judges of the High Court in appeal, preferred the evidence given by the approver before the Committing Magistrate to his evidence given in the Sessions Court. Some discussion took place in the High Court as to whether under S.157, Evidence Act the Court could use the statement made by the approver under S. 164, Criminal P. C. Section 157 is in these terms :
"In order to corroborate the testimony of a witness, any former statement made by such witness relating to the same fact at or about the time when the fact took place, or before any authority legally competent to investigate the fact, may be proved."
A statement made under S. 164, Criminal P. C., can never be used as substantive evidence of the facts stated, but it can be used to support or challenge evidence given in Court by the person who made the statement. The statement made by the approver under S. 164 plainly does not amount to the corroboration in material particulars which the Courts require in relation to the evidence of an accomplice. An accomplice cannot corroborate himself; tainted evidence does not lose its taint by repetition. But in considering whether the evidence of the approver given before the Committing Magistrate was to be preferred to that which he gave in the Sessions Court, the Court was entitled to have regard to the fact that very soon after the occurrence he had made a statement in the same sense as the evidence which he gave before the Committing Magistrate.
 Although the learned Judges of the High Court accepted the evidence of the approver given before the Committing Magistrate, they appreciated that it would be unsafe to act upon such evidence unless it were corroborated in the manner required by the rule of prudence. Apart from the suspicion which always attaches to the evidence of an accomplice it would plainly be unsafe, as the Judges of the High Court recognised, to rely implicitly on the evidence of a man who had deposed on oath to two different stories. The learned Judges stated the principle upon which they proposed to act in these terms: "It is of the utmost importance, however, that such evidence (i. e., the evidence of an approver admissible under S. 288, Criminal P. C.), before it can be acted upon as against other accused persons, should be corroborated by independent evidence in material particulars. The corroboration must be not only with regard to the occurrence but also as against each of the accused sought to be implicated in the crime."
 Acting upon this principle the learned Judges proceeded to consider, in the case of each accused, whether the evidence of the approver was sufficiently corroborated against him. They were not prepared to accept the confession of Trinath, who was also an accomplice, as sufficient corroboration of the evidence of the approver ; and in the case of Mata Simma, who was accused No. 6, against whom there was only the evidence of the approver and Trinath, they allowed the appeal; but as against the appellant, they found sufficient independent corroboration in the discovery of the deceased's cloth and the production of the khantibadi. With regard to the cloth, as already noticed, the evidence of the approver was that the appellant threw the cloth over a hedge, and it was proved at the trial that the cloth was found in the place pointed out by the approver. This fact no doubt was of value as supporting the credibility of the approver's story, but there is nothing beyond the statement of the approver to connect the appellant with the cloth. It was not found at the appellant's house, or in any place under his control, and the statement made by the approver that it was the appellant who threw the cloth to the place where it was found is of no more, and no less, value than his statement that the appellant took an active part in the murder. Sir Valentine Holmes for the crown admitted that he could not rely on the finding of the cloth as a piece of corroborative evidence implicating the appellant. With regard to the khantibadi the evidence of the approver was that the appellant gave it to him, that he passed it on to Trinath, and Trinath handed it over at the house of the appellant, and the police recovered a khantibadi from the house of the appellant where it was produced, somewhat unwillingly, by the appellant's mother. No blood stains were found on the khantibadi so produced, and there is nothing on the record from which it can be inferred that it was that khantibadi which was employed in the murder. Had a khantibadi been the sort of instrument which would be unlikely to be found in the house of an agriculturist it would no doubt be a striking coincidence that a khantibadi was handed over by Trinath at the appellant's house and one was subsequently found there. But the evidence is that a khantibadi is an instrument commonly possessed by agriculturists, and there was nothing strange in finding one at the house of the appellant. The High Court attached some importance to the unwillingness of the mother of the appellant to produce the khantibadi, but such unwillingness is in accord with the unco-operative attitude which agriculturists, and particularly female members of the family, usually display towards police investigations. In their Lordships' view neither the finding of the piece of cloth nor the production of the khantibadi tends to implicate the accused in the crime, nor affords such corroboration of the evidence of the approver as the rule of prudence requires.
 Sir Valentine Holmes did not rely strongly upon these pieces of alleged corroboration. He concentrated his argument mainly on the contention that the High Court was wrong in not accepting the confession of Trinath as sufficient corroboration of the evidence of the approver. This involves consideration of the position of the confession of a co-accused under Indian law. Section 30, Evidence Act, enacts :
"When more persons than one are being tried jointly for the same offence, and a confession made by one of such persons affecting himself and some other of such persons is proved, the Court may take into consideration such confession as against such other person as well as against the person who makes such confession."
9. This section was introduced for the first time in the Evidence Act of 1872, and marks a departure from the Common Law of England. It will be noticed that the section applies to confessions, and not to statements which do not admit the guilt of the confessing party. In the present case the Courts in India appreciated this, and ruled out statements made by certain of the accused which were self-exculpatory in character. The statement of Trinath was, however, a confession. Section 30 seems to be based on the view that an admission by an accused person of his own guilt affords some sort of sanction in support of the truth of his confession against others as well as himself. But a confession of a co-accused is obviously evidence of a very weak type. It does not indeed come within the definition of "evidence" contained in S. 3, Evidence Act. It is not required to be given on oath, nor in the presence of the accused, and it cannot be tested by cross-examination. It is a much weaker type of evidence than the evidence of an approver which is not subject to any of those infirmities. Section 30, however, provides that the Court may take the confession into consideration and thereby, no doubt, makes it evidence on which the Court may act; but the section does not say that the confession is to amount to proof. Clearly there must be other evidence. The confession is only one element in the consideration of all the facts proved in the case; it can be put into the scale and weighed with the other evidence. Their Lordships think that the view which has prevailed in most of the High Courts in India, namely that the confession of a co-accused can be used only in support of other evidence and cannot be made the foundation of a conviction, is correct. Sir Valentine Holmes puts his case in this way. He relies on one of the examples given by the Evidence Act of the sort of facts to which the Court should have regard in applying illustration B to S. 114, Evidence Act. The example is in these terms:
"A crime is committed by several persons. A, B and C, three of the criminals, are captured on the spot and kept apart from each other. Each gives an account of the crime implicating D, and the accounts corroborate each other in such a manner as to render previous concert highly improbable."
Sir Valentine contends that Trinath's confession was made independently of that of the approver, that neither he nor the approver had any reason for falsely implicating the appellant, and that the confession does afford sufficient corroboration to justify acceptance of the evidence of the approver, even if it does not amount to corroboration in material particulars within illustration (B) of S. 114. The evidence on record, however, does not support this argument. The confession of Trinath is a very short one and gives only the bare outline of the story. It discloses nothing which the police had not been able to ascertain from the approver, and affords no intrinsic evidence of its truth. It was, as already noted, retracted in the Sessions Court Retraction of a confession by an accused is a common phenomenon in India. The weight to be attached to it must depend upon whether the Court thinks that it was induced by the consideration that the confession was untrue, or by realization that it had failed to secure the benefits the hope of which inspired it. Their Lordships will assume that the confession of Trinath was not weakened by its retraction. Even so, the approver and Trinath were, according to both their statements, working together on the day of the murder when they were summoned by the appellant to take part in the crime; they were arrested on the following day so they had had a day, in which they must have appreciated that they were under suspicion, in which to arrange their story. After their arrest they were for some two days in police custody before they were sent up together, and with other accused, to the Magistrate to have their confessions recorded, and there is no evidence that they were kept apart during this period. In that state of the evidence, it is impossible to say that the approver and Trinath were kept apart from each other, and that their previous concert was highly improbable. Sir Valentine Holmes has relied strongly on the case In re B. K. Rajagopal, ILR (1944) Mad 308 : (air (31) 1944 Mad 117 : 45 Cr. LJ 373 FB), in which the Court founded a conviction upon the evidence of an accomplice supported only by the confession of a co-accused. Their Lordships whilst not doubting that such a conviction is justified in law under S.133, Evidence Act, and whilst appreciating that the coincidence of a number of confessions of co-accused all implicating the particular accused given independently, and without an opportunity of previous concert, might be entitled to great weight, would nevertheless observe that Courts should be slow to depart from the rule of prudence, based on long experience, which requires some independent evidence implicating the particular accused. The danger of acting upon accomplice evidence is not merely that the accomplice is on his own admission a man of bad character who took part in the offence and afterwards to save himself betrayed his former associates, and who has placed himself in a position in which he can hardly fail to have a strong bias in favour of the prosecution; the real danger is that he is telling a story which in its general outline is true, and it is easy for him to work into the story matter which is untrue. He may implicate ten people in an offence, and the story may be true in all its details as to eight of them, but untrue as to the other two, whose names have been introduced because they are enemies of the approver. This tendency to include the innocent with the guilty is peculiarly prevalent in India as Judges have noted on innumerable occasions, and it is very difficult for the Court to guard against the danger. An Indian-villager is seldom in a position to produce cogent evidence of alibi. If he is charged with having taken part in a crime on a particular night when he was in fact asleep in his hut, or guarding his crops, he can only rely, as a rule, on the evidence of his wife, members of his family, or friends to support his story, and their evidence is interested and not likely to carry weight. The only real safeguard against the risk of condemning the innocent with the guilty lies in insisting upon independent evidence which in some measure implicates each accused. This aspect of the matter was well expressed by Sir George Rankin in Ambikacharan Roy v. Emperor, 35 C.W.N. 1270 at p. 1274 : (AIR (18) 1931 Cal. 697: 33 Cr. LJ 19 SB.).
 In the present case their Lordships are in complete agreement with the Judges of the High Court in declining to act upon the evidence of the approver supported only by the confession of Trinath. These two persons appear to have been nothing but hired assassins. They had ample opportunity of preparing their statements in concert, and in addition, the approver has sworn to two contradictory stories, and Trinath has denied that his confession was true. It is true that no motive is shown for their falsely implicating the appellant, but motive is often a matter of conjecture. It may be that these two men thought it advisable to say falsely that they were acting on the instigation of another rather than on their own initiative, or they may have had reasons of their own for wishing to conceal the name of the real instigator.
 For the above reasons their Lordships are of opinion that the conviction of the appellant cannot stand.
Conviction set aside.