All the plaintiffs in this diversity action, except for Vernon McFadden who is the widower and personal representative of Addie Harris McFadden, have the unusual status of being successful popular musicians. They brought suit against Gusto Records, Inc. and G.M.L., Inc. in this Tennessee diversity action seeking royalties from the use of master recordings of their songs. Essentially, this is a breach of contract action. The case was tried without a jury and the district court awarded the plaintiffs a total of $843,209.89 plus prejudgment interest for failure to pay royalties due. Gusto and G.M.L. have appealed this judgment. The defendants do not dispute they are liable; they merely contend the amount owed is far less than the amount awarded by the district court. For the reasons which follow, we affirm.
As the district court noted, the musicians in this suit have achieved a high degree of success in “popular music.” B.J. Thomas is most famous for his recording of the theme song for the movie “Butch Cassidy and the Sundance Kid,” “Rain Drops Keep Failin’ on My Head.” Burt Bacharach and Hal Davis wrote the song originally for Bob Dylan to record. Thomas also had a number one hit with “Another Somebody Done Somebody Wrong Song.” Among his *397accolades are eleven gold records, two platinum albums, and five Grammy awards.
Shirley Owens Alston, Doris Coley Jackson, Beverly Lee and Addie Harris McFadden, members of a singing group known as the Shirelles, had their first hit, “I Met Him on a Sunday,” in 1958 while they were still attending high school in New Jersey. One year later the group was successful with “Dedicated to the One I Love.” Among their accolades are twelve top ten hits, including two number ones: “Soldier Boy” and “Will You Still Love Me Tomorrow,” a song recently repopularized by its use in the movie “Dirty Dancing.”
Gene Pitney’s most successful song, “Only Love Can Break a Heart,” reached number two on the popular charts in 1962. Pitney was especially successful singing movie theme songs, “A Town Without Pity” and “(The Man Who Shot) Liberty Valance,” and writing songs for other popular stars, including “Rubber Ball” for Bobby Vee, “Hello Mary Lou” for Ricky Nelson, and “He’s a Rebel” for the Cry-stalls.
As the district court aptly noted, record companies transferred the master recordings of the plaintiffs’ songs through “mesne conveyances,” until G.M.L. purchased them in the mid-1980’s. G.M.L. owned the masters, while Gusto sold copies of them for the retail trade. Mr. Gayron “Moe” Lytle is the president, sole shareholder, and sole director of Gusto and G.M.L.
Gusto and G.M.L. concede that an owner of the master recordings incurs any royalty obligations that arise during its ownership. Defendants make this concession after having profited throughout most of the 1980’s from sales and licensure of the plaintiffs’ masters without much regard to this obligation. In fact, this stance of neglecting the plaintiffs’ rights was consistent with most record companies who owned the plaintiffs’ master recordings through much of the 1970’s and throughout the 1980’s.
The district court found after a five day trial that Pitney should recover $187,762.44 plus interest, that Thomas should recover $177,299.77 plus interest, and that each member of the Shirelles should recover $119,537.07 plus interest. In reaching these amounts the district court accepted, with minor exceptions, the plaintiffs’ expert’s testimony on custom and usage in the music industry. Most importantly, the court accepted his testimony that, absent a contractual provision to the contrary, the musician receives half of the fees received from licensing the masters to unaffiliated third parties. The trial court’s decision was also based in large part upon the unseemly record keeping practices of Gusto, which forever prevent an exact determination of royalties earned by the plaintiffs. Gusto and G.M.L. timely filed this appeal.
The defendants assert the trial court erred on numerous grounds. First, they argue that clear language in the contracts of B.J. Thomas and the Shirelles prevents the consideration of evidence concerning industry custom and practice in determining royalties from domestic licensing. Thomas and the Shirelles entered into contracts with Scepter Records, Inc. in 1968 and 1961, respectively. The provisions of the two contracts in question are identical, except as to the royalty rate, and provide the following:
4. For the rights herein granted and the service to be rendered by you we shall pay you as royalty a sum equal to [4% for the Shirelles] [5% for Thomas] of the net retail list price in the United States of America based on 90% of all double-faced records manufactured and sold by us and paid for, on both faces of which are embodied only the selections recorded hereunder; and one-half of the respective amounts of such royalties of 90% of all records manufactured and sold by us and paid for on only one face of which are embodied only the selections recorded hereunder. In the case of phonograph records and other copies manufactured and sold in foreign countries by any subsidiary, affiliate, licensee or nominees to whom we have supplied a copy or duplicate of a master or matrix or tape of any such recordings, we will pay you one-half of the United States of America royalty rate out of all net li*398cense fees paid and received by us for phonograph records and other copies so manufactured and sold.
8. All recordings and all records and reproductions made therefrom together with the performances embodied therein, shall be entirely our property, free of any claims whatsoever by you or any person deriving any rights or interest from you. Without limitation of the foregoing, we shall have the right to make phonograph records, tape recordings or other reproductions of the performances embodied in such recordings by any method now or hereafter known, and to sell and deal in the same under any trade mark or trade names or labels designated by us, or we may at our election refrain therefrom.
Gusto and G.M.L. assert that paragraph four is unambiguous with regard to royalties from domestic licensing, arguing that the “obvious implication” of silence on the issue is that the parties agreed there would be no such royalties. We do not find that implication so obvious.
Both parties agree that New York law should be applied to resolve this issue. The resolution of any ambiguity in a written contract is to be determined by the court as a matter of law. Schuler-Haas Electric Co. v. Aetna Casualty & Surety Co., 40 N.Y.2d 883, 389 N.Y.S.2d 348, 357 N.E.2d 1003 (1976). Not only are these contracts silent on many issues, e.g. domestic licensing royalties, many of the provisions which the contracts do set forth are perfectly obscure. Paragraph eight of both contracts, for example, if read independently, could be interpreted as a provision in which Thomas and the Shirelles signed away any and all rights. Indeed, the defendants assert this is what the paragraph means with respect to royalties from domestic licensing. Without an integration clause or more evidence in the documents supporting this harsh interpretation, we cannot accept this reading of the contracts. Because of the ambiguities in these contracts, we believe the district court properly looked beyond the written contract to determine the true intentions of the parties.
New York cases have consistently held that custom may be used to clarify ambiguities or to “fill gaps” in an agreement. E.g., Tannenbaum v. Zeller, 552 F.2d 402, 414 (2d Cir.), cert. denied, 434 U.S. 934, 98 S.Ct. 421, 54 L.Ed.2d 293 (1977); Franklin Research & Dev. Corp. v. Swift Elec. Supply Co., 340 F.2d 439, 443 n. 2 (2d Cir.1964). Gusto asserts these cases are inap-posite because custom and practice cannot be used to contradict the express terms of the contracts. See, e.g., Western Union Tel. Co. v. American Communications Ass’n, 299 N.Y. 177, 86 N.E.2d 162 (1949). The problem with this argument is that the contracts in question are silent with respect to royalties from domestic licensing; there are no express terms in the contracts explaining the parties’ intentions on this issue. Paragraph 4 sets forth the royalty rate on the record company’s own sales and on foreign sales by “any subsidiary, affiliate, licensee or nominee[]_” Paragraph 8 merely establishes the record company’s title free of any claim to ownership by the artists. Because no provision in the contracts of B.J. Thomas or the Shirelles pertains to royalties from domestic licensing income which could be contradicted by incorporating into the contracts the custom in the music industry with respect to that issue, the district court properly accepted the fifty percent royalty rate as the rate intended by the parties.
Gusto and G.M.L. next argue that the district court incorrectly calculated the royalty rate for foreign license income. They assert that the royalty rate designated for foreign sales by “subsidiaries], affiliate[s], licensee[s], or nominees” set forth in paragraph four dictates a much lower rate than the fifty percent rate used by the district court. The clause in question provides a royalty rate based upon recordings manufactured and sold outside the United States “by any subsidiary, affiliate, licensee, or nominees to whom [the master’s owner has] supplied a copy or duplicate of a master....” The defendants rely upon *399the use of the term licensee, arguing that it should apply to all licensing arrangements the owner of the masters engages in.
These arguments fail for three reasons. First, as the record reflects, many licensing contracts include fixed fee payments to the owner of the masters. Attempting to compute an artist’s royalty payment in such an instance with the formula set forth in paragraph four does not make sense. One of the three variables used in the formula is based upon the “retail list price” of the record sold, another is based upon the volume of records sold. Neither of these variables would be helpful in computing a royalty when the payment involved is a fixed fee for use of a master, unless the licensee is using the master recording to copy for the retail music record industry and the licensee gives the owner of the masters, the licensor, data on the volume of records sold. Reviewing this record, we believe such a combination would be unlikely except when the licensee has some business relationship with the owner of the masters. This takes us to our second point.
The contract and the term “licensee” can be reasonably interpreted, without creating any internal inconsistencies or illogical conclusions, when we view the clause in question in its entirety. The term “licensee” is found in the paragraph where the formula for royalties from retail sales is set forth. It is the third word in a series of four; we read “licensee” as being of a similar kind to the other words in the series: “subsidiary,” “affiliate,” and “nominee,” e.g., Board of Educ. v. Barni, 66 A.D.2d 340, 345, 412 N.Y.S.2d 908 (N.Y.App.Div.1979), rev’d on other grounds, 49 N.Y.2d 311, 425 N.Y.S.2d 554, 401 N.E.2d 912 (1980); Fishman v. Islip, 20 Misc.2d 180, 181, 189 N.Y.S.2d 979 (N.Y.Sup.Ct.1959); the term “licensee” is used in paragraph four to describe an entity having some business relationship connecting it with the owner of the masters, not an independent entity acting for its own profit.
G.M.L. and Gusto provide just such an example of connected entities. Owned and operated by the same individual, G.M.L. owns the master recordings, while Gusto is in the business of manufacturing and marketing copies of those recordings. By reading paragraph four in its entirety, we see that it was designed to include the royalty rate for this type of relationship and that the parties did not set forth a royalty rate for the foreign licensing income at issue.
The final problem with the defendants’ argument is the fact that the artists received royalties at a rate of fifty percent on foreign licensing income in the past. The prior construction of a contract by the parties before the contract became the subject of this controversy is a prominent factor in determining the meaning of the contract. See, e.g., Croce v. Kurnit, 737 F.2d 229, 235 (2d Cir.1984). That neither Gusto nor G.M.L. were parties to the contract at the time the fifty percent royalty rate payments were made is of little assistance to their argument. As Gusto and G.M.L. concede, their purchase of the master recordings created an obligation to pay the royalties to the artists on those masters. Looking to the construction of the ambiguous contracts by prior owners of the masters helps in determining the obligation. Together these three factors, the results from applying the royalty formula set forth to typical licensing payments; the entire structure of paragraph four; and the past construction of the contract, convince us that the district court properly accepted fifty percent as the royalty rate for the foreign license income at issue in the Thomas and Shirelles contracts.
Gusto and G.M.L. next argue that the district court erred in awarding Gene Pitney any license income. They assert, as they did in the contracts with B.J. Thomas and the Shirelles, that any payment of licensing income is in conflict with the clear language of Gene Pitney’s contract. The paragraph in question is number four in Gene Pitney’s contract and is similar to paragraph four in the contracts of Thomas and the Shirelles; it states:
4. The Company agrees (a) during the first year of the agreement to pay the Artist a royalty of three (3%) percent, and during the second year of this agree*400ment to pay the Artist a royalty of four (4%) percent of the suggested retail list price of ninety (90%) percent of all records embodying on both sides thereof the compositions performed by the Artist and recorded hereunder manufactured and actually sold by the company or any subsidiary, affiliate or licensee to whom the Company has supplied a copy or duplicate of a master, matrix or embodiment of such composition or compositions and (b) to pay the Artist one-half (V2) the preceding amounts, respectively, for all records embodying such composition or compositions on only one side thereof manufactured and actually sold during each of such years, within the territorial limitations of the United States of America.
Gusto and G.M.L. again argue the use of “licensee” should be interpreted to include all licensing arrangements and again we reject that argument. For the same reasons the Thomas and Shirelles contracts do not address the foreign license income at issue in this case, we believe the fourth paragraph in Gene Pitney’s contract does not address the payment of licensing income, except when the licensor has a business relationship connecting it with the owner of the masters. Accordingly, we believe the district court correctly applied the fifty percent rate to the licensing income at issue in Gene Pitney’s contract.
Gusto and G.M.L. also argue the district court erred in finding them liable for royalties incurred by the prior owners of the masters. Gusto and G.M.L. mistakenly rely upon New York law in asserting that they should be liable only for royalty obligations arising after the date they purchased the masters. The determination of whether Gusto and G.M.L. are liable for royalties owed by the prior owner of the masters, Koala Records, is not determined by looking to the contracts between the plaintiffs and the original owners. Rather, we must look to the agreements between the defendants and Koala. The issue is: what did Koala sell to Gusto and G.M.L.? Did their sale of the master recordings include the transfer of royalty obligations incurred but not yet paid? To resolve this issue we are directed by the “most significant relationship” test, see Restatement (Second) of Conflict of Laws § 188 (1971), to the laws of Tennessee (Both G.M.L. and Koala are Tennessee based companies); however, neither party presented precedent from Tennessee on this issue, nor could we discover any in our research. We believe Tennessee would probably follow the general rule that a corporation purchasing the assets of another does not assume liabilities unless one of three exceptions is met: an express or implied assumption as to liability, consolidation, merger, or other similar connection between the two companies, or fraud in the transfer between the companies.
The objective facts set forth in the record support the conclusion that there was an assumption in this case that G.M.L. and Gusto would accept Koala’s liability. Among other things, the defendants claimed the rights to royalties earned before the date of sale, March 6, 1984, and the right to charge expenses of prior owners against the plaintiffs’ royalties. The implication of such broad rights is that Koala’s sale included much more than the mere sale of the master recordings. Indeed, viewed in the light most favorable to the plaintiff, the defendants’ own expert explicitly testified that Koala’s sale and Gusto’s purchase included Gusto’s obligation to assume the payment of accrued but unpaid royalties. We do not believe the district court committed a clear error by finding Gusto and G.M.L. liable for royalties owed by the prior owner of the masters.
Lastly, Gusto and G.M.L. argue that the district court incorrectly accepted the plaintiffs’ expert’s thirty-three percent incremental increase in the damage estimate. A trial court’s determination of damages is to be set aside only upon the determination that it is clearly erroneous. Smith v. Manausa, 535 F.2d 353, 354 (6th Cir.1976). Gusto and G.M.L. characterize the increase in damages as “arbitrary,” “unfair,” “guess work,” and “the purest form of speculation.” They ask this court *401to strike this increase in the damage award.
Not surprisingly, Gusto and G.M.L. all but ignore the factual findings of the district court, which set forth the reasons for the necessity of an incremental increase in the damages:
In the first place, the record-keeping practices of Gusto Records are (to use the kindest possible expression) unorthodox. By its own admission, Gusto keeps no general or subsidiary ledgers, no sales records for individual items, no perpetual inventory, and indeed no sort of record which could document the relation of any income to a particular master or artist.
The only thing the defendants had to offer were certain items which they denominated “royalty pay cards.” As the name implies, these documents are small pieces of paper purporting to show the sales of products and the royalties due the artists thereunder. This form of documentation appears to be peculiar to Gusto Records; at least, the testimony at trial indicated that no one had ever seen the like. If such is the case, then all concerned can be thankful, for a more woefully inadequate method of recording royalty obligations is hard to imagine. Calculation of the artists’ royalties is supposedly the cards’ raison d’etre. Yet the format of the cards is such that there is no place on them to perform the computation.
But this is not the sole (or even the chief) difficulty. One of the biggest problems with this whole case is that no one seems to know how much Gusto product is out there in the marketplace. The defendants have effectively admitted that there exists no reliable record of the “universe” (as it is called) of Gusto products. At Gusto, even catalogs were apparently kept only on a sporadic basis. There was uncontradicted testimony that there is in the marketplace a large amount of Gusto product for which no corresponding “pay cards” have been prepared. Moreover, because of Gusto’s grossly inadequate inventory records, the data on the “pay cards” cannot be supported by the kind of documentation upon which a reputable accountant would normally insist. In fact, a form of “record-keeping” which is compiled independently of all other documents can never be checked for accuracy at all. At very least, it creates a potential conflict of interest in the company; in the wrong hands, it can easily become an instrument of fraud and deceit. In short, this Court finds that the so-called “royalty payment cards” are a wholly inadequate index of the amount due to these artists.
An equally serious question concerns certain records which do exist, but are being withheld from this Court by the actions of the defendants. The bulk of Gusto & G.M.L.’s actual financial records are apparently in the possession of the accounting firm of Noel & Hundman in St. Louis, Missouri. Not surprisingly, the plaintiffs sought to depose the members of this firm, and they caused a subpoena duces tecum to be issued directing the firm to produce the records at the deposition. But on June 30, 1988, Noel & Hundman, at the insistence of the defendants, moved in the United States District Court for the Eastern District of Missouri to quash the subpoena. On July 19, 1988, The Honorable William L. Hungate, District Judge, issued an order finding that the information sought was privileged under Missouri’s accountant-client statute, Mo.Rev.Stat. § 326.151 (1986), and granting the motion to quash the subpoena. As a result, none of the members of the firm were deposed, and none testified at the trial. The financial records in their possession remain a mystery to all others, including this Court.
These actions by the defendants effectively put an end to any hope of obtaining a complete and accurate determination of just what it is that the defendants owe the plaintiffs.
In a case like this one, where damages are uncertain and where expert testimony on highly specialized subjects must be offered, a defendant with something to hide will not do anything which could lighten the plaintiff’s burden of proof, or *402give himself a burden of his own. But by allowing negative inferences to pile up, the defendants in the present case have done just that. We often hear of people butting their heads against a brick wall, but until now this Court had never heard of anyone building a brick wall expressly for that purpose.
In an action for royalties, it is a well established rule in New York that a “plaintiff need only demonstrate a ‘stable foundation for a reasonable estimate of royalties ....’” Contemporary Mission, Inc. v. Famous Music Corp., 557 F.2d 918, 926 (2d Cir.1977) (quoting Freund v. Washington Square Press, Inc., 34 N.Y.2d 379, 383, 357 N.Y.S.2d 857, 314 N.E.2d 419 (1974)). “The party who has caused the loss cannot insist on theoretical perfection.” Contemporary Mission, Inc., 557 F.2d at 928 (quoting Entis v. Atlantic Wire & Cable Corp., 335 F.2d 759, 763 (2d Cir.1964)).
The evidence in the record shows the plaintiffs’ expert had a “stable foundation” upon which to make estimations. Because of the unseemly record keeping of Gusto, the precise calculation of damages owed to the plaintiffs are impossible to determine. Gusto’s records, or more appropriately Gusto’s lack of records, prevents an exact determination of the universe of products and licenses. In order to make up for this deficiency, the plaintiffs’ expert considered many factors, including sales of comparable products for which data is available, customs of the industry, sales and leasing figures from possessors of the masters owned by G.M.L., and the limited records of Gusto and G.M.L. We cannot say that the district court’s reliance upon such testimony for the computation of damages, especially in light of the actions of Gusto and G.M.L., amounts to clear error.
For the foregoing reasons, we affirm the judgment of the district court.