Thiokol Corporation contracted to provide Star 48 motors to McDonnell Douglas Corporation for inclusion in McDonnell Douglas’ upper-stage Payload Assist Module, which is designed to propel satellites from the Space Shuttle to a geosynchronous orbit 22,000 nautical miles from Earth. In the contract, Thiokol warranted that the motors would be free of defects in material, labor, and manufacture and that the motors would comply with contract drawings and specifications. Following the failure of two satellites to reach their intended orbit, McDonnell Douglas brought this warranty action against Thiokol. The district court entered judgment against McDonnell Douglas after an eleven-day bench trial. We must decide whether Thiokol breached its warranties. We have jurisdiction, 28 U.S.C. § 1291 , and we affirm.

Background

In 1976, McDonnell Douglas Corporation (“McDonnell Douglas”) entered a contract with the National Aeronautics and Space Administration (“NASA”) to develop and market an upper-stage Payload Assist Module (“PAM”). The PAM propels satellites from the Space Shuttle to a geosynchronous orbit some 22,000 nautical miles from Earth. It consists of several components, including: (1) a cradle that links the satellite to the Space Shuttle; (2) a mechanism that deploys the satellite from the Shuttle’s cargo bay; (3) an upper-stage motor to move the satellite from the Shuttle’s 160 mile parking orbit into a transfer orbit; and (4) a smaller motor to take the satellite from the apogee of the transfer orbit to its intended orbit. In 1976 there were no upper-stage motors in production that suited McDonnell Douglas’ needs. To obtain such a motor, McDonnell Douglas completed Specification Control Drawing 1B98497 (the “Specification Control Drawing”) on August 18, 1976, which conceptualized a motor that would meet its needs.

In response to the Specification Control Drawing, Thiokol Corporation (“Thiokol”) developed a prehminary design for an upper-stage motor. Thiokol’s prehminary design incorporated a carbon-carbon exit cone, rather than the more common earbon-phenohc exit cone. The use of the hghter carbon-carbon exit cone was necessary in order to comply with the weight limitations set forth in the Specification Control Drawing. Thiok-ol sent its prehminary design to McDonnell Douglas for its review. Because of the “young maturity” of carbon-carbon technology, Thiokol also sent a separate prehminary design of the motor’s nozzle assembly, which indicated that Thiokol planned to use a carbon-carbon cone.

In 1978, Thiokol and McDonnell Douglas entered a contract for the development and qualification of an upper-stage motor. The motor became known as the Star 48 motor. The development and qualification contract was labeled the 7011 contract. This contract established a rigorous series of development and qualification tests that Thiokol’s proposed design had to meet before Thiokol was allowed to begin production.

Before Thiokol could continue with development of the motor under the contract, it had to obtain McDonnell Douglas’ approval of the design. The development phase thus called for a prehminary design review in which McDonnell Douglas and Thiokol engineers jointly reviewed the design of the Star 48 motor. McDonnell Douglas gave Thiokol its approval following the preliminary design review. The approval extended to the carbon-carbon exit cone.

After receiving McDonnell Douglas’ approval, Thiokol conducted five engine test-firings, all of which were witnessed by McDonnell Douglas engineers. McDonnell Douglas engineers were allowed to inspect the motors following the firings and received reports analyzing the results of the tests.

After a second design review, which focused on whether the motor design complied with the requirements of the Specification Control Drawing, McDonnell Douglas again gave Thiokol its approval. Six more motors were test-fired and a qualification program review was held. The purpose of this review was to obtain McDonnell Douglas’ approval of the motor, which was necessary to allow Thiokol to move into the production phase. McDonnell Douglas gave its approval, concluding that the motor met all of the performance and technical requirements of the Specification Control Drawing.

Following qualification of the motor, but before the parties agreed to a production contract, Thiokol test-fired another motor. This test resulted in a motor failure, and an investigation ensued. Although the cause of the failure was not determined with certainty, McDonnell Douglas issued a statement indicating that the most probable cause was a low density/low quality cone. The failure occurred even though the motor met all of the contract’s specification and acceptance requirements.

NASA and the United States Air Force conducted and independent investigation of the test failure. Significantly, this investigation concluded, “The state of knowledge about the material properties of earbon/car-bon involute exit cones is such that a meaningful margin of safety cannot be established for the Star 48 exit cone.” The NASA/Air Force investigation recommended that McDonnell Douglas expand its testing procedures to detect density variations in cones that otherwise met technical and acceptance standards. McDonnell Douglas thus knew that a test firing resulted in a failure, that NASA and the Air Force believed a meaningful margin of safety could not be established, and that its acceptance testing could not detect all density variations. Despite its knowledge, McDonnell Douglas did not adopt the NASA/Air Force recommendations.

McDonnell Douglas and Thiokol entered into a production contract (the “7047 contract”) in 1981. The contract required Thiokol to “maintain the system of producing hardware as established under [the 7011 contract].” The contract also incorporated Article 18 of McDonnell Douglas’ Terms and Conditions Guide. That article provides:

On February 3, 1984, the Space Shuttle Challenger used the PAM system to deploy WESTAR-VI, a private communications satellite. The WESTAR-VI satellite failed to reach its transfer orbit. An interply density variation, which is an area of reduced density between the layers of carbonized cloth that comprise the carbon-carbon exit cone, caused the Star 48 motor to fail. Three days later, the Challenger deployed the Palapa B-2 satellite, which also failed to reach its transfer orbit because of a motor failure caused by an interply density variation.

On February 7, McDonnell Douglas convened a Failure Investigation Committee to investigate the WESTAR-VI and Palapa B-2 failures (collectively the “failures”). The Committee issued its final report in September 1984, concluding that the most probable cause of the failures were interply density variations, but that the “exact cause of these exit cone failures could not be determined.” The Committee further concluded that the “[materials, components, and subassemblies of the nozzle assemblies were produced to the established standards and specifications and essentially satisfied all acceptance criteria applicable to the failed motors.” Additionally, the Committee found no evidence that the materials, labor, or manufacture of the WESTAR-VI and Palapa B-2 exit cones varied from contract specifications and requirements.

In March, 1985, McDonnell Douglas made a formal written warranty claim against Thiokol. It alleged that Thiokol breached the warranty provisions of the 7047 contract. McDonnell Douglas sought incidental damages of $10,926,000 (costs associated with the failure investigation) and consequential damages of $6,947,000 (the cost of replacing carbon-carbon exit cones with carbon-phenolic cones).

In 1986, certain insurers brought a subrogation action in which they sought recovery on negligence and strict liability theories against McDonnell Douglas, Thiokol, and HITCO (the Thiokol supplier that manufactured its carbon-carbon exit cones). Appalachian Insurance Co. v. McDonnell Douglas Corporation, 214 Cal.App.3d 1 , 262 Cal.Rptr. 716 (1989). The state trial court granted summary judgment in favor of the defendants. Id. at 718. The California Court of Appeals affirmed. Id.

Following the state court action, McDonnell Douglas brought this diversity action in federal district court. The district court conducted an eleven-day bench trial. At the conclusion of the trial, the court issued detailed findings of fact and conclusions of law. It entered judgment against McDonnell Douglas. McDonnell Douglas filed a timely notice of appeal.

Standard of Review

We review de novo the district court’s application of the principles of contract interpretation to the facts. O’Neill v. United States, 50 F.3d 677, 682 (9th Cir.), cert. denied., — U.S. --, 116 S.Ct. 672 , 133 L.Ed.2d 521 (1995). Whether the seller’s representations formed part of the basis of the parties’ bargain is a question of fact. See Royal Business Machines Inc. v. Lorraine Corp., 633 F.2d 34 , 43 (7th Cir.1980) (whether seller affirmed a fact amounting to an express warranty is a question of fact). We review the district court’s factual determinations for clear error. Erdman v. Cochise County, Ariz., 926 F.2d 877, 879 (9th Cir. 1991).

Discussion

Under California law, any affirmation of fact or promise relating to the subject matter of a contract for the sale of goods, which is made part of the basis of the parties’ bargain, creates an express warranty. Cal. Com.Code § 2313(1)(a). California courts use a three-step approach to express warranty issues. Keith v. Buchanan, 173 Cal. App.3d 13 , 220 Cal.Rptr. 392, 395 (1985). First, the court determines whether the seller’s statement amounts to “an affirmation of fact or promise” relating to the goods sold. Id. Second, the court determines if the affirmation or promise was “part of the basis of the bargain.” Id. Finally, if the seller made a promise relating to the goods and that promise was part of the basis of the bargain, the court must determine if the seller breached the warranty. Id.

Here, McDonnell Douglas argues that Thiokol made two separate promises, both of which constitute express warranties. McDonnell Douglas maintains that Thiokol promised to deliver goods that (1) were “free from defects in labor, material, and manufacture,” and (2) complied with “any drawings or specifications incorporated [into] or referenced” by the contract. We analyze each promise separately.

I. Defects in Labor, Material and Manufacture

We turn first to Thiokol’s promise to deliver goods free from defects in labor, material, and manufacture. There is no question that this promise relates to the goods and was made a part of the basis of the parties’ bargain. It therefore constitutes an express warranty, Cal. Com.Code § 2313, and we move directly to the final step of the analysis. We must determine whether the WESTAR VI and Palapa B-2 satellite mo tors suffered from defects in labor, material, or manufacture.

The district court concluded that the motor failures were caused by severe and extensive interply density variations. It further concluded such variations did not amount to a breach of Thiokol’s warranty because, “[w]here the buyer qualifies and approves product acceptance specifications, a product which satisfies all [such] specifications is not ‘defective.’” The district court thus held that the interply density variations could not be considered a defect “because [they] did not [violate] the product acceptance specifications qualified and approved by [McDonnell Douglas].”

A further inquiry is necessary, however, because Thiokol did not limit its warranty to technical compliance with the specifications set forth in the contract. Thiokol made a separate promise — that the goods would be free of defects in labor, material, and manufacture. We must then make the further inquiry of whether, under the terms of the contract, the interply density variations in the motors delivered constitute a defect in labor, material or manufacture.

McDonnell Douglas offers alternative definitions of the term “defect.” First, it contends that a defective product is one that “differs from the manufacturer’s intended results or from other ostensibly identical units of the same product line.” Barker v. Lull Engineering Co., Inc., 20 Cal.3d 413, 429 , 143 Cal.Rptr. 225 , 573 P.2d 443 (1978). This argument lacks merit. While this definition of “defect” is firmly rooted in the law of products liability, McDonnell Douglas presented no evidence that the parties intended to incorporate this definition, and all of the social concerns regarding distribution of risk that it connotes, into their contract. In fact, all of the evidence is to the contrary. See discussion infra at 12076-79.

McDonnell Douglas advances a second definition of the phrase “defect in labor, material, and manufacture.” Relying on S.M. Wilson & Co. v. Smith Int’l, Inc., 587 F.2d 1363, 1372 (9th Cir.1978), McDonnell Douglas maintains that a defect in labor, material, or manufacture is a “defect in quality.” McDonnell Douglas argues that Thiokol breached its warranty by delivering motors that contained “severe and extensive interply density variations,” which constituted a defect in quality.

McDonnell Douglas’ definition of the critical language finds superficial support in our case law. However, a close reading of Lombard Corp. v. Quality Aluminum Products Co., 261 F.2d 336 (6th Cir.1958), the ease relied on in S.M. Wilson, demonstrates that a defect in labor, material, or manufacture is a defect or flaw in the quality of the labor, material, or manufacture of the product. See Lombard 261 F.2d at 338 (holding that there was no defect in material where “[t]he steel [used] in the tie rods was of excellent quality”). This definition of defect is compelled by the plain meaning of the terms selected by the parties because it gives meaning to the limiting language “in labor, material, or manufacture.” McDonnell Douglas’ definition, to the contrary, renders the limiting language superfluous.

We therefore reject McDonnell Douglas’ definition of a “defect.” We hold that, to prove a breach of Thiokol’s warranty against defects in labor, material, or manufacture, McDonnell Douglas had to demonstrate that the WESTAR and Palapa satellite motors suffered from a flaw in the quality of the material used in their construction. See S.M. Wilson, 587 F.2d at 1372 . McDonnell Douglas failed to do this.

At trial McDonnell Douglas attempted to prove that the rayon cloth provided by Union Carbide, which was used in the manufacture of Star 48 motors, was “dirty” and of substandard quality. The district court expressly found “[n]o evidence ... that the cloth supplied by Union Carbide was defective.” This conclusion is amply supported by the record and is in accord with the findings of the failure investigation committee. Additionally, there was conflicting evidence regarding the effect of “dirty” cloth. A NASA expert testified that, although NASA pursued the “dirty cloth” theory for a “number of months,” it eventually concluded that the condition of the cloth at the beginning of the manufacturing process had no significant ef- feet on the strength of the carbon-carbon exit cone.

The record is otherwise devoid of any evidence that the failures were caused by a defect in labor, material, or manufacture of the carbon-carbon exit cones. There is, rather, substantial evidence to the contrary. The failure investigation committee concluded that “[c]hanges and variations in materials or [the manufacturing] process were not a determinative cause of the failures.” It also found that “[equipment, work procedures, personnel and production and support systems were not proven to be contributory to the failures.” Although the committee found that there was a shift in the quality of the exit cones, it concluded that “[a] single cause of density variations has not been isolated.”

In light of the foregoing, we conclude that McDonnell Douglas failed to demonstrate that a flaw in labor, material, or manufacture caused the interply density variations which resulted in the failures. Accordingly, the district court properly entered judgment against McDonnell Douglas on this issue.

II. Failure to Comply with Drawings and Specifications

McDonnell Douglas also contends that Thiokol warranted that all Star 48 motors delivered under the contract would comply “with any drawings or specifications incorporated [into] or referenced” by the 7047 contract. The Specification Control Drawing was attached to the 7047 contract. Section 3.0 of the statement of work, which was attached to the Specification Control Drawing, stated that “[a]ll rocket motor components shall be suitable for the purpose for which they are intended.” Additionally, section 3.5.4 of the statement of work stated that “[t]he nozzle [which includes the exit cone] shall be capable of withstanding the thermal mechanical loads during motor bum without any detrimental structural failure.” McDonnell Douglas argues that these provisions were incorporated into and referenced by the 7047 contract, and that Thiokol breached those provisions.

The essential inquiry is whether these statements in a drawing attached to the contract constitute additional express warranties that were understood and bargained for by the parties. These provisions would constitute performance warranties with serious financial risks involved and it would be very unusual to find the parties intended to tuck them away in drawings attached to the contract. The determinative question is whether the parties intended the warranties to be confined to those expressed in the body of the contract (products free from defects in labor, material, and manufacture) or expanded to performance warranties by drawings attached to the contract.

The district court determined that “it was the true understanding of the parties that a performance warranty ... was not part of the agreement or a basis of [McDonnell Douglas’] bargain.” The facts and circumstances surrounding the making of the contract, the testimony of McDonnell Douglas agents, understandings prevalent throughout the aerospace industry, and McDonnell Douglas’ own conduct all support the conclusion that McDonnell Douglas did not bargain for a performance warranty. Accordingly, we conclude that the statements in the Specification Control Drawing were not intended to be additional warranties forming a part of the bargain of the parties.

The state-of-the-art of carbon-carbon technology, which is one of the facts and circumstances surrounding the formation of the contract, indicates that a performance warranty was not technically feasible. McDonnell Douglas was aware, at the time it prepared the Specification Control Drawing, of the state-of-the-art of carbon-carbon exit cones. Additionally, McDonnell Douglas knew that the NASA/Air Force investigation of the test failures concluded that a meaningful margin of safety could not be maintained. Finally, McDonnell Douglas knew that the carbon-vapor densifieation process did not yield uniform density throughout the exit cone. McDonnell Douglas could not have bargained for a performance warranty because it knew that, given the state-of-the-art of carbon-carbon technology, such a promise was impossible to fulfill. See Royal Business Machines, 633 F.2d at 44 (“An affirmation of fact which the buyer from [its] experience knows to be untrue cannot form a part of the basis of the bargain.”).

Moreover, juxtaposing the costs of a satellite with the costs of the Star 48 motor, indicates that a performance warranty was economically unfeasible. In an internal memorandum commenting on a NASA inquiry, a McDonnell Douglas employee stated that McDonnell Douglas “decided early on in the PAM program that the financial risks associated with performance warranties were too high considering the price of the satellite compared to the price of the PAM.” McDonnell Douglas accordingly wrote its sales contracts “in terms of what the product is, not what it does” because “[t]o prepare' a contract in terms of what the hardware does would indeed be a performance warranty.” McDonnell Douglas did not include the cost of a performance warranty in the price of its product. Because McDonnell Douglas knew that its profit margin was greater than Thiokol’s and it knew what it paid Thiokol for the Star 48 motors, McDonnell Douglas knew that it was not being charged for a performance warranty.

The testimony of John Willacker, the McDonnell Douglas employee who drafted the specification control drawing, also supports the district court’s conclusion that the parties did not bargain for a performance warranty. He testified at trial that in 1977 it was McDonnell Douglas’ practice to set forth only technical requirements in specification control drawings. Willacker was asked if “it was a standard industry practice that a specification, like a specification control drawing, was not to be used to contain warranty provisions?” Willacker answered, “That’s correct.” More importantly, Willacker also testified that he intended only to set forth technical requirements in the document. Willaeker’s testimony further supports the district court’s conclusion that the parties did not intend to create a performance warranty by incorporating the “drawings and specifications” of the Specification Control Drawing.

Finally, McDonnell Douglas’ post-failure conduct is inconsistent with the understanding of the contract that it now advances. On February 20, 1984, D.H. Hauver, a negotiator/administrator for McDonnell Douglas, wrote to Thiokol, “Should this investigation determine that the PAM-D (STAR 48) rocket motors failed to achieve specified performance levels due to defects in their manufacture, we will expect you to comply with the warranty provisions of our subcontract.” (emphasis added). Had McDonnell Douglas understood Thiokol to have provided a performance warranty, it would not have included the limiting language “due to defects in their manufacture” in the letter. This subsequent practical construction of the warranty provisions further supports the district court’s conclusion that the affirmations in sections 3.0 and 3.5.4 were not a basis of the parties’ bargain.

In sum, McDonnell Douglas did not bargain for a warranty of performance, knew that it was not paying for such a warranty, and acted as if it understood the contract not to include such a warranty. To create a performance warranty under the circumstances of this case, the contract would have to be much more specific. Accordingly, the district court correctly concluded that statements in the Specification Control Drawings were not intended by the parties to be additional performance warranties.

Conclusion

Thiokol did not breach either of its express warranties. The judgment of the district courtis

AFFIRMED.

. Neither McDonnell Douglas nor Thiokol make res judicata arguments grounded on the state court proceedings.