Cavazos complains of the sentence imposed upon him based upon irrelevant extraneous hearsay data. We vacate and remand for resentencing.

Cavazos pled guilty to possession with intent to distribute approximately 10 ounces of heroin in violation of 21 U.S.C.A. § 841(a)(1). At the time of sentence the court was informed that Cavazos had never before been involved in criminal activity and had no prior record. In response the court drew a hypothetical situation of a bank robber who inferentially had robbed banks before, but who was apprehended for the first time when 'he blows the brains of the president out with a gun that he held up to his head while he robbed him. Now, that's the first time he has a prior arrest record. That's the first time he had been apprehended.'

In further colloquy with the court, counsel for Cavazos again referred to the lack of a previous record and asked the court how may prisoners had stood before it without any record. The court responded,

Most of them, most of the ones that have appeared before me in the drug traffic have had no prior arrest record. But you've got to remember too that computer statistics show from the Bureau of Prisons and the FBI that an average of six unsuccessful transactions, sometimes even conspiracies have been concluded before the person was ever apprehended. So, you can't always go by the fact that the person has not been apprehended. Sometimes people, according to the statistics of the FBI and the Bureau of Prisons show that a person gets by with ten, fifteen, twenty, sometimes as many as 200 narcotics sales. Narcotic sales are done undercover. They are a little more difficult to apprehend than some of our most heinous crimes such as murder, rape and things of that kind. But Mr. Miller, when you take into account the number of lives that are destroyed by a person who engaged in this type of elicit (sic) traffic, you can't very well say that he doesn't have a record now. It's a bad, bad record. * * *

The court then made it clear that he never granted probation in a case of this kind and imposed a maximum sentence of 15 years imprisonment followed by a special parole term of 15 years.

We, of course, do not review the severity of a sentence imposed within the statutory limits. But we have appellate responsibility to give 'careful scrutiny of the judicial process by which the particular punishment was determined.' United States v. Hartford, 5 Cir. 1974, 489 F.2d 652, 654.

The bank robber misanalogy drawn by the court together with the non-record, irrelevant, hearsay computer statistics upon which the court apparently relied may well have led the court to impute to Cavazos anywhere from 'six successful transactions' to having gotten by 'with ten, fifteen, twenty, (maybe) as many as 200 narcotic sales' which led the court to conclude that 'It's a bad, bad record.' On the contrary, it is patent from the record that Cavazos had never been arrested or convicted of any criminal activity before the occasion in question.

In United States v. Espinoza, 5 Cir. 1973, 481 F.2d 553, it was represented to the court that the appellant had never been convicted of a felony. The court nevertheless remarked that whether or not he had actually been convicted appellant's record was bad. The court denied appellant's request to submit evidence to rebut the court's statement. Relying on Townsend v. Burke, 1948, 334 U.S. 736, 68 S. Ct. 1252, 92 L. Ed. 1690 in which the Supreme Court said that a sentence based on assumptions concerning the prisoner's record which are materially untrue is inconsistent with due process of law, this Court opined,

that despite the broad discretion left to the trial judge in assessing background information for sentencing purposes, . . . a defendant retains the right not to be sentenced on the basis of invalid premises.

Id. at 555.

The use of statistics, even if calculated from a sound base and placed in proper context, neither of which was shown to have been done here, is an invalid premise on which to impose sentence. This approach injects hypothetical extraneous considerations into the sentencing process and contradicts the judicially approved policy of individualizing sentences that are tailored to fit the offender. Williams v. New York, 1949, 337 U.S. 241, 248, 69 S. Ct. 1079, 1083, 93 L. Ed. 1337, 1342.

Nor do we condone a mechanical sentencing policy that eliminates probation from consideration in every case in the category here involved. We iterate,

We affirm our disapproval of statements by a trial judge reflecting a fixed sentencing policy based on the category of crime rather than on the individualized record of the defendant.

Hartford, supra, at 655, quoting from United States v. Baker, 2 Cir. 1973, 487 F.2d 360.

VACATED AND REMANDED FOR RE-SENTENCING.