Thomas Larimer, a salesman for IBM, was fired and brings suit against the company under both ERISA and the Americans with Disabilities Act. The district judge granted summary judgment for the defendant.

Larimer was hired in August of 2000, and in May of the following year his wife, who was also an employee of IBM, gave birth to twin daughters after only 29 weeks of pregnancy. At birth the two girls suffered from a variety of serious medical conditions owing to their prematurity, including respiratory distress, jaundice, apnea, and sepsis. One of the girls also had bleeding in the brain and the other had a lesion on her nose. They were hospitalized for almost two months at a total expense of almost $200,000, all of which IBM's employee health plan paid for. By the close of discovery in January 2003 the two children seemed to be healthy and normal, but there is some probability (how great a one is unknown) that they will develop serious physical or mental handicaps as they grow older.

Larimer was fired in August of 2001, shortly after the children came home from the hospital. His principal claim is that IBM violated the Americans with Disabilities Act, by firing him because his daughters are disabled. Are they? They seem fine at present, and so the question, left open in Goldman v. Standard Ins. Co., 341 F.3d 1023, 1026 and n. 2 (9th Cir.2003), and not elsewhere answered definitively, is whether a possible, or even probable, future disability can ever be a disability that triggers the protections of the Act. 42 U.S.C. § 12102(2); 29 C.F.R. § 1630.8; Den Hartog v. Wasatch Academy, 129 F.3d 1076, 1081-82 (10th Cir.1997); Tyndall v. National Education Centers, Inc., 31 F.3d 209, 214 (4th Cir.1994). The Supreme Court's decision in Sutton v. United Air Lines, Inc., 527 U.S. 471, 482-83, 119 S. Ct. 2139, 144 L. Ed. 2d 450 (1999), suggests (in dictum — the question before the Court was whether a person who has to wear glasses is disabled because without them he couldn't see) that the answer is "no" unless the individual is mistakenly regarded by his employer as having a disability; such a mistake is an alternative trigger of the Act's protections. 42 U.S.C. § 12102(2)(C); EEOC v. Rockwell Int'l Corp., 243 F.3d 1012, 1014-15 (7th Cir.2001).

Larimer must lose even if his daughters are disabled or regarded as disabled. He is suing not on their behalf but on his own, under a provision of the ADA that forbids discrimination against "a qualified individual because of the known disability of an individual with whom the qualified individual is known to have a relationship or association." 42 U.S.C. § 12112(b)(4). Notice first the oddity of requiring the plaintiff to show that he is a "qualified individual," since the only definition in the ADA of a "qualified individual" is the definition of "qualified individual with a disability" as "an individual with a disability who, with or without reasonable accommodation, can perform the essential functions of the employment position that such individual holds or desires." 42 U.S.C. § 12111(8). If this is the "qualified individual" to which the association provision (section 12112(b)(4)) refers, then Larimer cannot obtain any relief under that provision because he has no disability! The term "qualified individual" in that provision must simply mean qualified to do one's job, as assumed though nowhere discussed in the legislative history and the cases. H.R. Rep. 101-485, pt. 2, at 61-62 (1990), reprinted in 1990 U.S.C.C.A.N. 303, 343-44; 29 C.F.R. § 1630.8; Hilburn v. Murata Electronics North America, Inc., 181 F.3d 1220, 1230-31 (11th Cir.1999); Den Hartog v. Wasatch Academy, supra, 129 F.3d at 1083-85; Ennis v. National Ass'n of Business & Educational Radio, Inc., 53 F.3d 55, 59-60 (4th Cir.1995); Rocky v. Columbia Lawnwood Regional Medical Center, 54 F. Supp. 2d 1159, 1164-65 (S.D.Fla.1999).

Three types of situation are, we believe, within the intended scope of the rarely litigated (this is our first case) association section. We'll call them "expense," "disability by association," and "distraction." They can be illustrated as follows: an employee is fired (or suffers some other adverse personnel action) because (1) ("expense") his spouse has a disability that is costly to the employer because the spouse is covered by the company's health plan; (2a) ("disability by association") the employee's homosexual companion is infected with HIV and the employer fears that the employee may also have become infected, through sexual contact with the companion; (2b) (another example of disability by association) one of the employee's blood relatives has a disabling ailment that has a genetic component and the employee is likely to develop the disability as well (maybe the relative is an identical twin); (3) ("distraction") the employee is somewhat inattentive at work because his spouse or child has a disability that requires his attention, yet not so inattentive that to perform to his employer's satisfaction he would need an accommodation, perhaps by being allowed to work shorter hours. The qualification concerning the need for an accommodation (that is, special consideration) is critical because the right to an accommodation, being limited to disabled employees, does not extend to a nondisabled associate of a disabled person. 29 C.F.R. § 1630.8; Den Hartog v. Wasatch Academy, supra, 129 F.3d at 1083-85; Tyndall v. National Education Centers, Inc., supra, 31 F.3d at 214.

This case fits none of the categories. (2) can be ruled out peremptorily; the girls' premature birth and resulting medical afflictions are neither communicable to Larimer nor predictive of his becoming ill or disabled. Likewise (3): there is no evidence that Larimer was absent or distracted at work because of his wife's pregnancy or the birth and hospitalization of his daughters. As for (1), there is to begin with no evidence that health benefits are in the budget of the unit of IBM that employed and discharged Larimer. Cf. Rogers v. International Marine Terminals, Inc., 87 F.3d 755, 761 (5th Cir.1996). Benefits can be in a unit's budget in multiple ways. For example, IBM may charge every manager's budget with a fringe-benefit allocation for each employee in his unit that is equivalent to the premiums for health insurance allocable to the employee, or, alternatively, with the dollar amounts actually paid in benefits to the unit's employees or their dependents. In the latter case but not the former, the manager would care about the actual expense for health services to the relatives of an employee in his unit because that expense would be in his budget. But there is no evidence that expenses are accounted for in that fashion. If IBM has a profit-sharing plan or pays bonuses based in part on company-wide performance, all employees who participate in the plan or receive such a bonus — and presumably they would include Larimer's supervisors — have a financial stake in the company's performance and thus a stake, however attenuated, in the firing of an "expensive" employee. But Larimer has made no effort to pitch his case on such ground either.

Having no evidence, Larimer falls back on the ubiquitous McDonnell Douglas test for a prima facie case of employment discrimination. Den Hartog, the case with the most extensive discussion of the ADA's association provision, purports to use a version of the test that requires the plaintiff to show that "(1) the plaintiff was `qualified' for the job at the time of the adverse employment action; (2) the plaintiff was subjected to adverse employment action; (3) the plaintiff was known by his employer at the time to have a relative or associate with a disability; (4) the adverse employment action occurred under circumstances raising a reasonable inference that the disability of the relative or associate was a determining factor in the employer's decision." 129 F.3d at 1085; see also McGuinness v. University of New Mexico School of Medicine, 170 F.3d 974, 979-80 (10th Cir.1998). The test is sound, but it's not really a version of the McDonnell Douglas test because it requires the plaintiff to prove all the elements of what would be the prima facie case of discrimination against a relative or other associate of a disabled person even if there had never been a McDonnell Douglas case. This is apparent from the fourth element of the Den Hartog test, which requires the plaintiff to produce evidence that he was discriminated against because of the disability of a person with whom he has a relationship or other association. Compare this to the corresponding part of the McDonnell Douglas test, which in the case of employment discrimination based on race or some other invidious ground requires the plaintiff to prove only that he was replaced by someone of a different race — not that he was discriminated against because of his race. A plaintiff who can produce evidence of actual discrimination on the basis of disability has made out a prima facie case without regard to McDonnell Douglas.

A true parallel to McDonnell Douglas in the association setting would allow a prima facie case to be made out if the plaintiff, having shown that he was qualified in the sense of meeting his employer's expectations ( not a "qualified individual with a disability," as we explained earlier), went on to show that his employer knew he had a relationship or association with a disabled individual, that the employer fired him, and that he was replaced by someone who lacked such a relationship or association. This would not however be a very sensible test — which shows that there may be limits even to the cloning of McDonnell Douglas. When deciding whether to adapt McDonnell Douglas to a new legal setting, a court should ask not "how can we create a formula closest to that one?" but "what conditions imply a comparably high likelihood that the employer is violating the statute?" In the present setting a true McDonnell Douglas test, as distinct from the Den Hartog test, would generate the prima facie case of disability discrimination when only the most tenuous basis for an inference of discrimination was present — for example when the employer knew merely that the plaintiff had a second cousin who was sterile, or that he had shaken hands with a person who was HIV-positive. The latter example would be ruled out by cases that hold that casual associations with a disabled person are not protected by the ADA, Freilich v. Upper Chesapeake Health, Inc., 313 F.3d 205, 215-16 (4th Cir.2002); Oliveras-Sifre v. Puerto Rico Dept. of Health, 214 F.3d 23, 26 (1st Cir.2000); O'Connell v. Isocor Corp., 56 F. Supp. 2d 649, 653 (E.D.Va.1999), but probably not the former; for in 29 C.F.R. § 1630.8 we read that "it is unlawful for a covered entity to exclude or deny equal jobs or benefits to, or otherwise discriminate against, a qualified individual because of the known disability of an individual with whom the qualified individual is known to have a family, business, social or other relationship or association." Better to require, as Den Hartog does though not in precisely these words, that the plaintiff present evidence that his case falls in one of the three categories in which an employer has a motive to discriminate against a nondisabled employee who is merely associated with a disabled person.

Larimer's alternative claim is that his discharge violated ERISA. The usual ERISA claim is for benefits, but the expense of the girls' medical treatments was fully defrayed by IBM and what Larimer is arguing is that IBM fired him because of annoyance at having to pay so much, which may grow to be even more in the future should either or both of the girls develop serious physical or mental handicaps. In other words, the claim is that IBM retaliated against Larimer for exercising his rights under IBM's welfare benefits plan. 29 U.S.C. § 1140. He has no evidence of this, just as he has no evidence of disability discrimination, but in Stone v. City of Indianapolis Public Utilities Division, 281 F.3d 640, 644 (7th Cir.2002), we authorized the following "adaptation of McDonnell Douglas to the retaliation context": "the plaintiff [must] show that after filing the charge [in the present case, after applying for atypically large benefits] only he, and not any similarly situated employee who did not file a charge [in this case, did not apply for atypically large benefits], was subjected to an adverse employment action even though he was performing his job in a satisfactory manner." See also Rogers v. City of Chicago, 320 F.3d 748, 754-55 (7th Cir.2003). Stone was not an ERISA case, but the standard it sets forth is appropriate to such cases. It is true that one finds courts saying that the plaintiff in an ERISA retaliation case must show that the defendant had a "specific intent" to punish him for asserting rights under the plan. Bilow v. Much Shelist Freed Denenberg Ament & Rubenstein, P.C., 277 F.3d 882, 892 (7th Cir.2001); Lindemann v. Mobil Oil Corp., 141 F.3d 290, 295 (7th Cir.1998); Humphreys v. Bellaire Corp., 966 F.2d 1037, 1043 (6th Cir.1992). But all that this means is that since the forbidden retaliation is retaliation for claiming ERISA benefits, the plaintiff cannot prevail unless he shows that there was an ERISA plan and presents evidence from which the trier of fact can infer that the defendant's motivation in taking the adverse action of which the plaintiff is complaining was indeed to thwart his right to benefits. That is the ultimate question and is separate from what must be shown to establish merely the prima facie case.

Had Larimer identified a similarly situated employee of IBM who had not applied for substantial welfare benefits yet had been treated better than he, he would have made out a prima facie case of retaliation under Stone — provided, as in any McDonnell Douglas case, that he also showed that he was performing his job in a manner that satisfied his employer's legitimate expectations. Coco v. Elmwood Care, Inc., 128 F.3d 1177, 1178-79 (7th Cir.1997). He strikes out on both counts. Declining ostrich-like to mention, let alone try to distinguish, Stone, even though his opponent relies heavily on it, he makes no effort to identify a comparable employee of IBM who did not apply for atypically large welfare benefits and was treated better than Larimer was, though it would be easy to find such an individual if one existed. Nor does he show that he was performing up to his employer's expectations — in fact he was fired because he did not perform up to those expectations. He was a new hire and the uncontradicted evidence, described at length in the district court's opinion and unnecessary to repeat here, demonstrated that he failed to obtain an adequate understanding of the product that he had been hired to sell (Lotus software) and as a result was unable to convince prospective customers that the product was the answer to their needs. He was especially poor at convincing them to buy the various ancillary services that are an important part of the revenue of many software producers, including IBM. His discharge had nothing to do with the expense incurred by IBM with respect to his daughters.

AFFIRMED.