Stanley G. Curtis brings this appeal from a decision of the Tax Court, P-H Memo T.C., P 78,265 (1978), and the Commissioner of Internal Revenue (Commissioner) cross-appeals. At issue are the correctness and adequacy of various findings of fact made by the Tax Court in its ruling on the case.

This case originated in 1969 when the Commissioner selected Curtis' tax returns for the years 1960-1968 for audit. After some investigation, the Commissioner determined that Curtis' returns for this period understated his income. Because Curtis, who worked as a "courthouse lawyer" and who also engaged in stock and real estate transactions, had kept inadequate records and then had failed to explain his financial dealings to the Commissioner, the Commissioner undertook the task of redetermining his income by use of the "net worth" method. As explained in Holland v. United States, 348 U.S. 121, 75 S. Ct. 127, 99 L. Ed. 150 (1954), in a net worth case,

the Government, having concluded that the taxpayer's records are inadequate as a basis for determining income tax liability, attempts to establish an "opening net worth" or total net value of the taxpayer's assets at the beginning of a given year. It then proves increases in the taxpayer's net worth for each succeeding year during the period under examination and calculates the difference between the adjusted net values of the taxpayer's assets at the beginning and the end of each of the years involved. The taxpayer's nondeductible expenditures, including living expenses, are added to these increases, and if the resulting figure for any year is substantially greater than the taxable income reported by the taxpayer for that year, the Government claims that the excess represents unreported taxable income.

Id. 75 S.Ct. at 130.

In August 1972, the Commissioner determined deficiencies in Curtis' federal income tax returns for the years 1960-1968, through use of the net worth method. The Commissioner also sought to hold Curtis liable under I.R.C. § 6653(b)1 for additions to tax for 1960-1968.2 Curtis petitioned for review of the Commissioner's determination in the Tax Court. After trial, that court found that Curtis was liable for deficiencies and additions to tax in the following amounts:

                     Additions to tax
Year  Deficiency  under I.R.C. Sec. 6653(b)
----  ----------  ----------------------
1960     none              none
1961     none              none
1962     none              none
1963     none              none
1964  $ 6,359.00        $3,180.00
1965    2,145.00         1,073.00
1966    9,499.00         4,750.00
1967    2,512.00         1,256.00
1968   18,627.00         9,314.00

Both Curtis and the Commissioner appeal from the Tax Court's ruling. Curtis asserts that the Tax Court's finding that he did not possess a "cash hoard" at the beginning of the period at issue was clearly erroneous, and that the apparent increases in his net worth from 1960 to 1968 were in fact due to his use of his cash hoard. Thus, he claims, it was improper for the Commissioner to redetermine his taxable income by use of the "net worth" method, and, correspondingly, the Tax Court was in error in assessing deficiencies and additions to tax for the years 1964-1968. Alternatively, Curtis argues that the Tax Court's opinion did not contain findings adequate to support its deficiency determinations for these years. The Commissioner supports the Tax Court's rejection of Curtis' cash hoard claim and its findings of deficiencies and additions to tax for 1964-1968. The Commissioner cross-appeals, however, contending that the Tax Court's findings of no deficiencies for the years 1960-1963 were clearly erroneous.

We affirm the Tax Court's rejection of Curtis' cash hoard claim. But because the Tax Court made no subsidiary findings and offered no explanation to support its ultimate conclusions regarding Curtis' tax liability, or lack thereof, for 1960-1968, we vacate those findings and remand to the Tax Court for a fuller explication of its decision.

A. Standard of Review.

Since both Curtis and the Commissioner ask us to review factual determinations made by the Tax Court, we first set out the standards which guide us in this task. It is well settled that our review of the Tax Court's fact findings is to be guided by the principles set forth in Fed.R.Civ.P. 52(a). See, e.g., Snider v. Commissioner, 453 F.2d 188, 192 n.4 (5th Cir.1972); Blackstone Realty Co. v. Commissioner, 398 F.2d 991, 996 (5th Cir.1968); 5A Moore's Federal Practice P 52.03(5) (2d ed. 1980). For an exposition of these principles, we can hardly improve upon Judge Morgan's learned discussion in Golf City, Inc. v. Wilson Sporting Goods Co., Inc., 555 F.2d 426 (5th Cir.1977), where, in the course of remanding to the district court for an eludication of its finding of a conspiracy violative of § 1 of the Sherman Act, 15 U.S.C. § 1, he wrote:

Rule 52(a) mandates that in all civil actions tried without a jury "the court shall find the facts specially . . . ." One purpose of the rule is to aid the appellate court by affording it a clear understanding of the ground or basis of the decision of the trial court. Silberblatt, Inc. v. United States, 353 F.2d 545, 549 (5th Cir.1965). Another purpose of the rule, identified forthrightly by Judge Frank, is to engender care on the part of the trial judge in ascertaining the facts:

It is sometimes said that the requirement that the trial judge file findings of fact is for the convenience of the upper courts. While it does serve that end, it has a far more important purpose that of evoking care on the part of the trial judge in ascertaining the facts. For, as every judge knows, to set down in precise words the facts as he finds them is the best way to avoid carelessness in the discharge of that duty: Often a strong impression that, on the basis of the evidence, the facts are thus-and-so gives way when it comes to expressing that impression on paper. The trial court is the most important agency of the judicial branch of the government precisely because on it rests the responsibility of ascertaining the facts. When a federal trial judge sits without a jury, that responsibility is his. And it is not a light responsibility since, unless his findings are "clearly erroneous," no upper court may disturb them. To ascertain the facts is not a mechanical act. It is a difficult art, not a science. It involves skill and judgment. As fact-finding is a human undertaking, it can, of course, never be perfect and infallible. For that very reason every effort should be made to render it as adequate as it humanly can be.

United States v. Forness, 125 F.2d 928, 942-43 (2d Cir.), cert. denied, 316 U.S. 694, 62 S. Ct. 1293, 86 L. Ed. 1764 (1942) (footnotes omitted). See United States v. Merz, 376 U.S. 192, 199, 84 S. Ct. 639, 11 L. Ed. 2d 629 (1964) (more careful consideration of problem assured if finder of fact required to state not only end result, but also process by which result reached).

The "facts" that 52(a) admonishes must be found specially are not all of one kind. They lie along a conceptual line from those termed "subsidiary" to those labeled "ultimate." For example, in the instant case a subsidiary fact might be that appellant Wilson notified Golf City on a certain date that Wilson would not sell Golf City pro-line golf equipment. An ultimate fact might be that PGA and the manufacturers conspired to prevent Golf City from obtaining pro-line goods. Ultimate facts generally are derived from subsidiary facts through a cause-effect reasoning process. That is, when subsidiary facts A, B, and C are shown to exist, then, because of the fact-finder's knowledge of the way the world works, he is able to conclude that, more likely than not, ultimate fact X also exists. He infers X from A, B, and C.

Rule 52(a) does not require that the trial court set out findings on all the myriad factual questions that arise in a case. As we have said: "Courts need not indulge in exegetics, or parse or declaim every fact and each nuance and hypothesis." Gulf King Shrimp Co. v. Wirtz, 407 F.2d 508, 516 (5th Cir.1969). On the other hand, parsimony in setting out subsidiary facts may run afoul of the rule:

We have concluded that the findings of the district court do not provide a sufficiently definite predicate for proper appellate review. Many of the findings are couched in conclusory terms. Some were announced solely as ultimate findings without support which clearly reflected a choice between conflicting accounts of events or between alternate legal interpretations of those events. This court cannot be left to guess. The findings and conclusions we review must be expressed with sufficient particularity to allow us to determine rather than speculate that the law has been correctly applied.

Hydrospace-Challenger, Inc. v. Tracor/MAS, Inc., 520 F.2d 1030, 1034 (5th Cir.1975). See Lettsome v. United States, 434 F.2d 907, 909 (5th Cir.1970). Wright and Miller observe that "(i)deally findings of fact should be clear, specific, and complete, without nonrealistic and uninformative generality on the one hand, and on the other without an unnecessary and unhelpful recital of nonessential details of evidence." 9 Wright & Miller, Federal Practice and Procedure § 2579, at 711 (1971). Relating the degree-of-thoroughness question to the purposes of Rule 52(a), the findings of the trial court must be sufficiently detailed to give us a clear understanding of the analytical process by which ultimate findings were reached and to assure us that the trial court took care in ascertaining the facts.

Rule 52(a) also provides that "(f)indings of fact shall not be set aside unless clearly erroneous . . . ." Justice Reed's definition of "clearly erroneous" has proved durable over the years:

A finding is 'clearly erroneous' when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.

United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S. Ct. 525, 542, 92 L. Ed. 746 (1948). Findings that are sufficiently thorough to comply with Rule 52(a) must also survive the "clearly erroneous" crucible if so challenged on appeal.

Golf City, supra, 555 F.2d at 432-434 (footnotes omitted).

Judge Morgan thus identifies two types of attacks that may be levelled at a trial court's findings of facts, both of which have been raised in the present case: the findings may be challenged as inadequate to give a clear understanding of the process by which the court's ultimate conclusions were reached and thus inadequate to permit appellate review, or they may be challenged as clearly erroneous. We turn now to the challenges raised in this case.B. Curtis' Cash Cache Cashiered.

In the trial of this case, Curtis sought to account for his increases in net worth from 1960 to 1968 by showing that he possessed a "cash hoard," dating back to the mid-1940's, on which he began to draw only after 1960. He now claims that the trial court's finding that he did not possess a cache of cash is clearly erroneous. We cannot agree.

Curtis' version is simply stated. His wife's father, Amerigo Antonelli, was engaged in the early 1940's in the manufacture of fireworks. At about this time, Antonelli obtained a munitions contract from the federal government. All did not go well, however, and Antonelli was subsequently indicted and convicted for fraudulent dealings with the Government in the execution of the contract. Curtis' wife Alvera testified at trial that in 1946 shortly before her father went to jail her father gave her $135,000 in cash which she deposited in a cedar chest in her bedroom. Then in 1948, after her father's death, the cash hoard was, according to Alvera, transferred to her sister's house. Although Alvera had married her first husband, Sam Barco, Jr., in 1950, her testimony was that she never told Barco about the cash.3 Curtis testified that his wife's sister eventually gave him the money to hold for Alvera and that his use of this money accounted for his increases in net worth.

The Tax Court did not find this story credible. It pointed out that Curtis had offered no objective evidence to indicate that the $135,000 ever existed and that other evidence dictated a contrary conclusion. It noted, for example, that Amerigo Antonelli, the purported source of the cash, had filed a pauper's oath in order to prosecute an appeal of his criminal conviction and that Curtis had not mentioned the gift through the long course of the Commissioner's investigation of Curtis' tax returns. The Commissioner also points to other evidence supporting the Tax Court's finding. He emphasizes the fact that Alvera claimed never to have told her first husband of the cache's existence, despite the fact that much of their marriage was staged upon the shoals of financial difficulty. Further, the Commissioner points to various representations made in court by Alvera in her support action against her first husband indicating her extreme impecuniousness.

In these circumstances, the Tax Court's findings are adequate and are not clearly erroneous. The facts set out above create a picture not unlike that in Holland v. United States, 348 U.S. 121, 75 S. Ct. 127, 99 L. Ed. 150 (1954), in which the taxpayers sought to escape criminal liability for income tax evasion by proving the existence of a cash hoard to defeat the Government's reliance upon the net worth method to establish the crime. Although the Government did not introduce any direct evidence to dispute the taxpayer's claim, the Supreme Court held that the Government could properly rely "on the inference that anyone who had $104,000 in cash would not have undergone the hardship and privation endured by the" taxpayers during the period in which the alleged cash hoard lay unused. Id. 75 S.Ct. at 134. A similar inference is warranted by the evidence in this case. We hold that, after considering this inference and the other evidence cited by the Tax Court in support of its conclusion, the Tax Court's finding that Curtis did not possess a cash hoard is not clearly erroneous.

C. Deficient Deficiencies.

When it came to the point of finding the amounts of Curtis' liability for income tax deficiencies and additions to tax under § 6653(b), the Tax Court stated only its ultimate conclusions that Curtis had no additional liability for the years 1960 to 1963 and that he was liable for deficiencies and additions to tax for 1964 to 1968. See p. ---- supra. Curtis attacks the latter findings as inadequate since they do not explain how this result was reached or, in the alternative, as clearly erroneous. The Commissioner contends that the findings of no additional liability for the years 1960 to 1963 are clearly erroneous. We do not reach the question whether the Tax Court's findings for any of the years at issue are clearly erroneous; rather, we hold that the Tax Court's findings were inadequate to permit us fairly to review its ultimate conclusions and remand for an eludication of the Tax Court's reasoning.

In Holland, supra, the Supreme Court set out guidelines for the Government's use of the net worth method of proof. These guidelines were necessary, it thought, because use of the net worth method "involved something more than the ordinary use of circumstantial evidence . . . . (C)areful study indicates that it (the net worth method) is so fraught with danger for the innocent that the courts must closely scrutinize its use." Id., 75 S.Ct. at 130. In developing its guidelines, the Court stated that

an essential condition in cases of this type is the establishment, with reasonable certainty, of an opening net worth, to serve as a starting point from which to calculate future increases in the taxpayer's assets. The importance of accuracy in this figure is immediately apparent, as the correctness of the result depends entirely upon the inclusion in this sum of all assets on hand at the outset.

Id. at 134.

In light of these considerations and the principles set forth in Golf City, supra, we conclude that the Tax Court's findings as to the amounts of the deficiencies and additions to tax for which Curtis was liable for the period 1960 to 1968 are inadequate. First, the Tax Court made no finding of Curtis' opening net worth either in 1960, the date at which the Commissioner sought to establish it, or in 1964, the date after which the court determined deficiencies were proved. Further, the court neither explained why it rejected the Commissioner's proof of unreported taxable income for 1960 to 1963 nor why its findings of unreported taxable income for 1964 to 1968 varied from that which the Commissioner claims to have proved at trial.4 In all candor, we are at a loss to divine the method by which the Tax Court reached its conclusions. Our review of the record has discovered no evidentiary source for them,5 and, in the absence of any explanation for their derivation, we cannot review them. A bit of folk wisdom tells us that one cannot make a silk purse from a sow's ear. It follows, in as a fortiori a manner as is possible with folk wisdom, that cannot one construct a deficiency from thin air. We do not require that a lower court's ultimate conclusions come to us clothed in the robes of state or in any particular degree of sartorial splendor. In this case, however, the Tax Court's conclusions sport only the emperor's new clothes.

For these reasons, the appropriate disposition of this case is to vacate the Tax Court's findings of deficiencies and additions to tax and to remand to the Tax Court to enable that court to make the subsidiary findings necessary to render its ultimate conclusions comprehensible, or, if necessary, to modify its conclusions to conform to the evidence.6 The court need not take additional evidence if it can perform its task on the basis of the extant record. It is free, however, to take additional evidence if it is necessary. See, e.g., Golf City, supra, 555 F.2d at 435-436; United States v. Trout, 386 F.2d 216, 223-225 (5th Cir.1967); 9 Mertens, The Law of Federal Income Taxation § 50.93 (rev.perm.ed. 1977).

We briefly summarize our disposition of this appeal. First, we have affirmed the Tax Court's finding that Curtis did not possess a cash hoard. Second, we have vacated the Tax Court's findings as to the amount of the deficiencies and additions to tax for which Curtis is liable for the period 1960 to 1968 and remanded to that court so that it can either explain the process by which it reasoned from the evidence to its ultimate conclusions or modify its conclusions to conform to the evidence. Finally, we have considered Curtis' other assignments of error and find them to be without merit. Thus, on remand, the only task remaining for the Tax Court is that of rendering adequate findings to sustain whatever ultimate conclusions it reaches as to Curtis' liability.

Accordingly, the decision of the Tax Court is AFFIRMED IN PART; VACATED AND REMANDED IN PART.