joined by Justice MEDINA and Justice JOHNSON, dissenting.
The Fowlers and the Eppses signed a contract agreeing that if either sued the other, the “prevailing party” would be entitled to recover reasonable attorney fees. The Fowlers sued the Eppses, but after the Eppses had incurred $22,950 in attorney fees defending the suit, the Fowlers suddenly nonsuited. Did the Eppses prevail?
Because the parties were undisputedly free to agree on what would happen in this situation, the answer depends entirely on what they meant when they signed the contract. “In construing a written contract, the primary concern of the court is to ascertain the true intentions of the parties as expressed in the instrument.”1 But the Court is not primarily concerned with, or even especially interested in, ascertaining the Fowlers’ and the Eppses’ intentions from the text of their contract. The Court’s primary concern is whether recovery of attorney fees from a plaintiff who nonsuits is good policy, and it presumes the Fowlers and the Eppses must have shared its view of the subject.
The Court begins, as it should, with the presumption that the Fowlers and the Eppses intended to give the word “prevailing” its ordinary meaning but then turns to federal case law for that meaning. The Court finds one case: the Fifth Circuit’s 2001 decision in Dean v. Riser.2 There, the issue was whether a defendant sued in a civil rights action was a prevailing party within the meaning of 42 U.S.C. § 1988(b) when the plaintiffs claim was nonsuited *873with prejudice.3 The determinative consideration was “the general policies and competing interests that prompted Congress to enact [Title VII of the Civil Rights Act of 1964] authorizing district courts to award attorney’s fees to prevailing parties in civil rights litigation.”4 The court finally concluded:
The policy considerations surrounding the law of attorney’s fees for prevailing civil rights litigants demand a flexible rule. It should empower trial courts to balance the concerns for encouraging vigorous enforcement of civil rights against discouraging frivolous litigation within the specific and unique context of each individual case. Accordingly, we hold that a defendant is not a prevailing party within the meaning of § 1988 when a civil rights plaintiff voluntarily dismisses his claim, unless the defendant can demonstrate that the plaintiff withdrew to avoid a disfavorable judgment on the merits.5
For two reasons, the federal cases the Court cites do not give guidance. First, the cases cited all deal with legislative policy reflected in public statutes, not with private parties’ intentions in ordering their personal affairs by contract. The Court notes this problem, observing that “it might be improper to look to cases” construing statutes based on legislative policy choices for guidance in determining what private parties intended in a contract,6 but then does it anyway. Riser could not be clearer in explaining that the availability of attorney fees there depended on policy considerations in the Civil Rights Act. There are no legislative policy choices involved in deciding what “prevailing party” means in a private agreement, even a standard form agreement like the one in this case. Second, nothing suggests that private parties like the Fowlers and the Eppses would have federal case law in mind in reaching an agreement that attorney fees should go to a prevailing party.
The place to look for the ordinary meaning of words is not federal case law but a dictionary.7 According to Webster’s Third New International Dictionary, to prevail means “to gain victory by virtue of strength or superiority: win mastery: TRIUMPH”.8 Now surely it is beyond argument that, policy considerations aside, when a plaintiff decides to abandon his lawsuit, the defendant, thereby relieved of the further worry and expense of defending himself, thinks he won. Common experience teaches that the challenger who forfeits, loses, and his opponent wins. Imagine the conversation between the Eppses and their lawyer: “Good news! The Fowlers dropped their suit.” “Wow! So we won!” “No, you didn’t win. The Fowlers just gave up.” “But we said all along the case had no merit, and now they’ve effectively conceded it. We didn’t win?” “Well, you have to understand that a federal case construing the Civil Rights Act has held that....”
*874The Court’s problem with the Eppses’ common-sense, dictionary understanding of “prevailing” is that it “appears to promise simplicity of application.”9 The Court’s notion of what it means to prevail has the virtue of complexity. The defendant prevails if the plaintiff takes a nonsuit with prejudice (because further suit would be barred by res judicata, an affirmative defense10), but not if the nonsuit is without prejudice, even if further action would be barred as if the nonsuit were with prejudice (by limitations, for example, another affirmative defense11), except when the nonsuit is taken to avoid an unfavorable judgment. It is impossible to think that parties like the Fowlers and the Eppses would ever have all this in mind when agreeing that a prevailing party should recover attorney fees.
The internal inconsistencies in this new test cannot be reconciled. One is between nonsuits of weak claims that should not be discouraged and nonsuits of claims to avoid unfavorable judgments that should be discouraged. The Court reasons that awarding attorney fees against a plaintiff who nonsuits a “weak claim[ ]” that “should be abandoned” would “ ‘penalize the plaintiff for doing precisely what should be done’ ”.12 “At the same time,” the Court concludes that attorney fees should be awarded against a plaintiff who “nonsuit[s] in order to avoid an unfavorable judgment.” 13 What is the difference between a weak claim that should be abandoned, which can be nonsuited with impunity, and a claim that is likely to result in an unfavorable judgment, which cannot? Says the Court: “the determination has been made largely based upon inferences drawn from the course of events in the lawsuit.”14 In other words: there is none. Then why differentiate between them? The pieces of this puzzled ruling have no unifying principle but are supported instead by a somewhat cynical pragmatism: the defendant who is nonsuited without prejudice cannot recover attorney fees unless he is willing to continue litigating to prove that the plaintiff would have lost anyway. And so a contractual provision unquestionably intended to discourage unnecessary litigation is construed either to foment it or do nothing at all.
Another inconsistency is in the Court’s differentiation between post-nonsuit litigation that is barred by res judicata and post-nonsuit litigation that is barred by limitations. The Court concludes that attorney fees may be awarded against the nonsuiting plaintiff in the former instance but not the latter. Why? Because, the Court explains, “the mere possibility that limitations would bar future suits does not effect a change in the parties’ relationship that confers prevailing party status on a defendant.”15 But the bar of res judicata is also a “mere possibility” in the sense that both it and limitations are affirmative defenses that are waived if not raised. If either is successfully raised, the effect is the same: suit is barred. Yet the Court goes out of its way to treat them differently. As hard as it is to understand why the Court would differentiate between the two defenses, it is impossible to think the Fowlers and the Eppses did.
*875The Court doubts that the Fowlers and the Eppses intended for a defendant to be awarded attorney fees after a nonsuit without prejudice because the result could be two different prevailing parties if the plaintiff refiles the same suit and wins. But the difficulty the Court sees in this example is not avoided by sparing the nonsuiting plaintiff from an attorney fee award. In the Court’s view, if the plaintiff nonsuits without prejudice a claim that is then barred by limitations, the defendant has not prevailed. But if the plaintiff refiles the same action, the defense is raised, and the defendant wins, has he prevailed? Of course. The Court does not avoid the difficulty it raises.
In reality, the difficulty the Court sees should not have been any problem at all for the Fowlers and the Eppses, had they thought about it before signing their contract. It is neither illogical nor unreasonable for parties to agree that a plaintiff who abandons litigation should make everyone whole, even if he tries again and wins. The law may afford a mulligan, but the parties can decide it should not be free. And if a prevailing attorney is to recover attorney fees, it makes perfect sense to award them to a defendant both when he is nonsuited, and again later when a second suit is dismissed based on res judicata or limitations.
Finally, the Court worries that to enforce a fee-shifting provision like the one in this case will result in satellite litigation over attorney fees. Usually, determining the amount of a party’s reasonable attorney fees does not require much litigation. But the important point, here and throughout, is that any cause for concern belongs to the parties in reaching agreement, not to the Court in setting policy.
In the end, the Court forces parties who desire a broader fee-shifting agreement than it thinks is good policy to use clearer words than “prevailing party”. “Just party” would only encourage more judicial subjectivism. I don’t think “escaping party” would do it, because the nonsuiting plaintiff may be escaping the defendant’s becoming a prevailing party. “Fortunate party” might work, though it’s very general. Viewed from another angle, the provision might award fees to the “oppressed party”, though it, too, is very general and also injects a moral tone. But if “prevailing” is not clear enough, probably no one word is. To be safe, parties will have to spell out their intentions in more detail. An agreement to shift attorney fees will require more attorney fees to draft. But it will be worth it.
“A court must be careful not to substitute its own view of what should have been intended for what was intended.”16 In accordance with the parties’ agreement, I would award the Eppses reasonable attorney fees. Accordingly, I respectfully dissent.
. J.M. Davidson, Inc. v. Webster, 128 S.W.3d 223, 229 (Tex.2003).
. 240 F.3d 505 (5th Cir.2001).
. Id. at 506.
. Id. at 507.
. Id. at 511.
. Ante at n. 3.
. The Court accuses me of hypocrisy, or at least a faulty memory, pointing out that I joined the majority in Intercont'l Grp. P’ship v. KB Home Lone Star L.P., 295 S.W.3d 650 (Tex.2009), which considered case law in determining when a plaintiff is a prevailing party. But we held in KB Home that a plaintiff who sues for damages and recovers nothing does not prevail. The case law the Court cited only supported the dictionary meaning of "prevailing”.
. Webster’s Third New International Dictionary (1961).
. Ante at 870.
. Tony Gullo Motors I, L.P. v. Chapa, 212 S.W.3d 299, 314 (Tex.2006).
. Id.
. Ante at 869.
. Ante at 864.
. Ante at 870.
. Ante at n. 8.
. Lane Bank Equip. Co. v. Smith S. Equip., Inc., 10 S.W.3d 308, 321 (Tex.2000) (Hecht, J., concurring) (emphasis in original).