concurring.
The panel reversed the judgment in favor of the EEOC in this sexual harassment suit because the district court had excluded evidence of the employer’s obligations under its collective bargaining agreement to the employee who had harassed the women on whose behalf the EEOC sued. The employer wanted to introduce this evidence both to show that it had acted reasonably in dealing with the harassment and to show that, at the least, it had not acted so unreasonably that it should have to pay punitive damages. The evidence was clearly relevant to the issue of punitive damages, less clearly to the issue of liability — so much less clearly, indeed, that I think the panel erred in ordering a new trial on liability. But we do not take cases en banc merely because of disagreement with a panel’s decision, or rather a piece of a decision, least of all a piece of a decision concerning the relevance of a particular item of evidence to a particular issue in a particular case. We take cases en banc to answer questions of general importance likely to recur, or to resolve intracircuit conflicts, or to address issues of transcendent public significance — perhaps even to curb a “runaway” panel — but not just to review a panel opinion for error, even in cases that particularly agitate judges, such as cases in which sexual harassment is charged. In attempted or at least apparent conformity to this policy, we asked the parties (when we granted rehearing en banc) to brief the following general questions: “whether, as a matter of law, evidence regarding arbitration and a company’s collective bargaining agreement is inadmissible in a Title VII suit to show: 1) that an employer’s response to sexual harassment was reasonable for the purpose of determining employer liability; and 2) that the employer did not act with the state of mind necessary for the imposition of punitive damages.”
Although these questions have the form appropriate for en banc determination, such determination is premature. The questions have not arisen in any other reported case, and for all I know may never arise; and they need not be answered even in this case, because the error in excluding the evidence relating to the collective bargaining agreement with respect to the defendant’s liability for sexual harassment, if there was an error (clearly the exclusion of that evidence with respect to punitive damages was erroneous, as the panel determined), was harmless in view of the overwhelming evidence of the employer’s inadequate response to a serial harasser. The grant of rehearing en banc was gratuitous as well as premature. And it was premature not only in the sense that there was no need to answer the questions yet, but also because, as a consequence of there being no other cases, we lack an adequate basis in experience for answering general questions about relevance; we lack sufficient particulars to be able to generalize intelligently.
The majority opinion answers the first and more far-reaching of the two questions that we put to the parties — “whether, as a matter of law, evidence regarding arbitration and a company’s collective bargaining agreement is inadmissible in a Title VII suit to show ... that an employer’s response to sexual harassment was reasonable for the purpose of determining employer liability” — with a flat “yes,” stating that “this evidence is not relevant to liabili*530ty and therefore is inadmissible under Fed.R.Evid. 402 with respect to that subject.” Yet nowhere does the opinion attempt to justify so categorical, so unqualified, an answer. All it tries to show is that the particular dispute-resolution procedures used by the employer in this case pursuant to the employer’s collective bargaining agreement are not relevant to this employer’s liability in the particular, indeed peculiar, circumstances that this case presents. The result of this approach is a yawning gap in logic, since it obviously does not follow that other dispute-resolution procedures used by other employers pursuant to other collective bargaining agreements in other factual settings can never be relevant to an employer’s liability for sexual harassment.
It might well be thought foolhardy to lay down a blanket rule, not justified by considerations of privilege, that an entire class of evidence is “irrelevant,” and therefore inadmissible, in a broad class of cases having, potentially at least, diverse facts, without some sense of what those facts might be. A narrower rule might be unexceptionable. Clearly it is no defense against a charge of failing to prevent an employee from harassing other employees that the employer had tied his hands by his contract with the harasser. If a collective bargaining agreement stated that an alleged harasser could not be terminated unless his guilt was proved beyond a reasonable doubt, and in conformity with that agreement the employer refused to take action against a harasser though his guilt was clear, albeit not clear enough to have convicted him in a criminal proceeding, the agreement would not be a defense in a suit under Title VII, or even relevant to liability, since it would add nothing to a defense based on other evidence, such as that a prompt and thorough investigation had faded to substantiate the complaint of harassment. (Evidence that an employer had schemed with the union to make it impossible to fire an alleged harasser would be relevant — as evidence for the plaintiff!) But the issue, so far as justifying the blanket rule announced at the outset of the majority opinion is concerned, is not whether a collective bargaining agreement is a defense to liability in such a suit, or whether it is always relevant to liability, but whether it is never relevant to liability; and I do not see how that question can be answered sensibly against relevance without a greater exercise of the imagination than attempted in the opinion.
Let me propose a hypothetical case, not altogether remote from the present one though readily distinguishable from it, to which the existence and terms of a collective bargaining agreement would be relevant. A worker complains to a supervisor that another worker is harassing her. The supervisor responds by immediately transferring the alleged harasser to another part of the workplace. But because of the design of the workplace the transfer does not keep the two workers apart all the time and the worker who complained insists to the supervisor that the harasser be fired. The supervisor replies that the company’s collective bargaining agreement protects workers from being fired other than for cause and entitles any worker sought to be terminated for cause to notice and a hearing before a joint labor-management committee with appeal to an arbitrator, the entire procedure from complaint to final arbitration to be completed within 21 days. The complaint is filed and the arbitrator rules that the accusation of harassment is false and indeed malicious and therefore that there is no basis for terminating the harasser or even for separating him from the complainant.
Suppose the complainant then brought suit against the company under Title VII, charging that it had acted unreasonably in *531failing to fire her alleged harasser, as she had urged. She offers to prove that the arbitrator’s decision was wrong. She succeeds in proving it. In presenting its own case, however, the employer tenders the collective bargaining agreement, with its provision on grievance and arbitration, and further offers to prove that this provision is standard in the industry. Would that evidence be irrelevant to whether the company had responded reasonably to her complaint? The company could of course just describe the steps it had taken without mentioning the collective bargaining agreement, but then the jury might wonder why the alleged harasser had any procedural rights. The majority opinion contains no reasons that would justify the exclusion of the evidence in my hypothetical case, and yet it falls under the blanket ban announced at the outset of that opinion.
I have nothing against bright-line rules, even though the inevitable consequence of such rules is to bring about substantively unjust results in certain cases. The gain in certainty will often exceed the loss that results when a rule is inflexibly applied to circumstances unforeseen when the rule was adopted and unprovided for in it. But nowhere in the majority opinion is there an effort to justify the adoption at the very outset of the development of the case law dealing with the interplay between collective bargaining and sexual harassment of a rule broader than necessary to decide this case, a rule that sweeps into the dustbin the hypothetical case that I have described as well as others as yet unimagined. A flat rule is stated at the outset of the opinion; reasons justifying a narrow rule are then presented; no effort to connect the two logically distinct parts of the opinion is made.
We are in danger of forgetting that not all complaints of sexual harassment are well founded, that not all alleged harassers are guilty as charged, and that companies do not typically enter into collective bargaining agreements in order to curtail their employees’ legal rights; nor do unions seek to bargain away those rights. It apparently is easy to lose sight of these elementary points in a case such as the present one, but it is precisely the lurid facts of this case that threaten to distract us from the question whether we should lay down a blanket rule that the grievance and arbitration provisions in a collective bargaining agreement are never relevant to an employer’s response to a complaint of harassment. There is no support in previous cases, let alone in Title VII or in the majority opinion, for such a rule, and it makes very little sense that I can see.