Angela Horton and Kevin Houser v. the Kansas City Southern Railway Company

Supreme Court of Texas
══════════
No. 21-0769
══════════

Angela Horton and Kevin Houser,
Petitioners,
v.
The Kansas City Southern Railway Company,
Respondent

═══════════════════════════════════════
On Petition for Review from the
Court of Appeals for the Fifth District of Texas
═══════════════════════════════════════

JUSTICE YOUNG, joined by Justice Blacklock, dissenting.

Justice Abbott clearly described respect for the jury-trial right as
the principle underlying the decision he announced for the Court in
Crown Life Insurance Co. v. Casteel, 22 S.W.3d 378 (Tex. 2000). “It is
fundamental to our system of justice that parties have the right to be
judged by a jury properly instructed in the law.” Id. at 388. When “it is
impossible for us to conclude that the jury’s answer was not based on one
of the improperly submitted theories,” id. at 389, therefore, a new trial is
required because only a new trial can ensure that the jury, rather than
judges substituting themselves for the jury, resolves the parties’ dispute.
On rehearing in this otherwise routine case, the Court retreats from this
principle. Apparently in a misguided sense that things will be easier or
more efficient in cases with an erroneous broad-form submission, the
Court exchanges clear rules for a case-by-case gut-check after appellate
justices review “the entire record.”
I am all for reconsidering our decisions. A humble Court should
acknowledge error—if it has erred. But the error comes today, not last
term. We were then unanimous in agreeing that this case required a
new trial under the principles of Casteel. The reason was that, over
objection, the trial court submitted the case to the jury in a way that
prevents anyone else from knowing which theory of negligence the
verdict turned on—specifically, whether it depended on one we found
(and still unanimously find) legally insufficient. Our original holding
was a straightforward application of Texas Rule of Appellate Procedure
61.1(b) and the Casteel line of cases. Put simply, because it was denied
the opportunity to show the appellate court the ground on which the jury
rested, Kansas City Southern by definition cannot properly present its
case on appeal. The Court’s original Part IV (which I attach as an
appendix so that it is not lost to the mists of time) followed settled law,
which respects the jury and the jury-trial right. Only a new trial can
ensure that a final judgment is compelled by a jury’s finding rather than
imposed despite a jury’s finding. The Court today, however, reverses
course with its new Part IV. We can only hope—we cannot know—that
the jury did not rely on the legally insufficient missing-yield-sign theory.
In the new Part IV, the Court’s principal error is to revise Casteel
into something it never was. Casteel always required applying Rule
61.1(b)—under which error is harmful, not just presumptively harmful—

2
when objected-to charge error makes it impossible for the appellate
courts to know what the jury actually decided. To dilute Casteel, the
Court latches onto the casual use of the English word “presumption” in
a few of Casteel’s progeny and now proclaims that Casteel no longer
stands for. The Court compounds this misstep by saying that an error
is not even “presumptively” harmful if the comingling error includes a
theory where the evidence was legally insufficient (i.e., as opposed to a
theory that for some other legal reason was invalid).
This distinction is immaterial and contrary to our cases. It makes
no sense, either, because what threatens the jury-trial right is a judgment
that relies on something a jury did not find. Whatever label one gives
that “something” is of no importance. What matters is not the reason
for the invalidity but the risk that the jury relied on a theory (any theory)
that is legally invalid. The cause of the legal invalidity is immaterial
because the result is obviously the same: a jury cannot rely on it.
Comingling any legally invalid theory within a broad-form question
means that we cannot know if the jury did or did not rely on it.
As applied here, Horton exhorted the jury to rely on the missing-
yield-sign theory. Horton still, as a ground for rehearing, argues that
the missing-yield-sign theory is enough to reinstate the verdict. Horton’s
playing up the missing-yield-sign theory is rather inconvenient for the
Court, which rules for Horton by doing the exact opposite—it busily
downplays the missing-yield-sign theory and assures everyone that the
jury was “unlikely” to have been motivated by the hours of testimony and
the video that pushed so hard on that theory. The whole endeavor is
contrary to the oft-intoned statement that judges must never, ever, ever

3
put themselves in the place of the jury. We should instead keep Casteel’s
promise of letting the jury speak for itself.
Both the cause and the likely effect of today’s decision is not taking
the jury-trial right all that seriously. Today’s decision would be
impossible if there were a greater commitment to ensuring that the
judgments of our courts rely exclusively on what a jury actually finds.
The consequence of today’s decision will diminish confidence that Texas
civil judgments are the mandate of a jury and not the guess of a court.
Compelling one citizen to transfer money to another should only happen
when the court knows that the ground on which the jury ruled was lawful.
The Court thinks it is making things better. I know that my
colleagues are motivated by that desire. There is no getting past today’s
error of law, but I genuinely hope that, in practice, it all works out better
than I fear. In particular, I do not want to diminish the Court’s
seriousness of purpose with respect to its emphatic reminders to the lower
courts that they should not erroneously submit charges like this one. But,
the Court continues, if it happens anyway, the harm analysis—with or
without a “presumption” of harm—is no mere show. If the lower courts
do not respect that warning, I believe that this Court will respond. In the
meantime, however, by stepping back from Casteel’s promise, the Court
injects more uncertainty into the law and trial practice and disincentivizes
the kind of clarity that litigants deserve and that our law demands.
Except as to Parts IV and V, therefore, I continue to join the Court’s
opinion. Because Part IV’s errors lead the Court to an improper outcome,
I must respectfully dissent from the Court’s judgment.

4
I
Casteel explained that “when a trial court submits a single broad-
form liability question incorporating multiple theories of liability, the
error is harmful and a new trial is required when the appellate court
cannot determine whether the jury based its verdict on an improperly
submitted invalid theory.” Casteel, 22 S.W.3d at 388 (emphasis added).
We applied this principle in subsequent cases, making clear that the key
determinant to whether Casteel applies is the appellate court’s ability to
determine whether the jury based its verdict on something that is a
legally impermissible basis for a verdict. The Court now speaks of that
decision’s progeny as creating just a “rebuttable presumption” of harm—
essentially just like any mine-run error in a trial. And as to whether the
presumption is “rebutted”—or, if the presumption was not triggered but
the submission was nonetheless erroneous—the Court focuses on the
likelihood that an erroneous broad-form submission has caused a
particular kind of harm (specifically, how likely it was that the verdict
was actually based on an impermissible ground).
Casteel has been reduced to asking whether this “presumption” of
harm “applies” or not. But I acknowledge that the Court understands
this to have weighty consequences. Once the presumption is determined
to apply, a new trial is mandatory unless the party who prevailed at trial
can point to the record to show that the error does not prevent the
appellant from adequately presenting the appeal. Ante at 47. The Court
does not diminish how tall a task that is. The prevailing party at trial, in
other words, must show how the record as a whole does not merely make
it arguable that the jury did not reach its decision on an impermissible

5
ground, or even more likely than not that it relied on only proper
grounds—that party must show, based on the record, that the court can
be “reasonably certain” that the jury did not rely on improper grounds.
Ante at 42, 47. The parties can push back with dueling references to the
record.
Appellate courts, moreover, must take their duty seriously when
determining whether the presumption has been rebutted. See id.
Moreover, even when the error does not lead to “presumptive harm”—as
the Court concludes that the error here does not—a similar analysis is
required. The appellant does not start off with the same sort of presumed
harm, but because the reviewing court must still be “reasonably certain”
that the error does not prevent the appellant from presenting its case
under Rule 61.1(b), see id. at 56, the parties’ diligent efforts to show
whether the record can make the court confident that the jury only
relied on proper grounds remains essential to the appellate courts’ work.
So although I think that these standards are not what the law
requires—Casteel has always required more—I do not suggest that the
Court has left this area of the law utterly unprotected, either. My
dispute with the Court concerns why we are watering down the law and
why the new standards are neither grounded in our precedent nor an
improvement in practice, as I discuss below.
A
Despite its desire to make things better, and an effort to give the
presumption real weight, the Court’s interpretation of Casteel and its
progeny is erroneous. The Court at least acknowledges (although it
relegates the history to a footnote) that Casteel did not think of itself as

6
creating a “presumption.” See ante at 44 n.29. Rather, it recognized the
occurrence of a particular type of harm—a petitioner’s inability to
present its case to the appellate courts.
In Casteel, the court of appeals had concluded that submitting
multiple theories in broad form was harmless—that it could not be
shown that the jury likely relied on the unlawful theories.1 But this
Court recognized that the erroneous broad-form submission is itself
what made it impossible to determine whether the jury rendered an
improper judgment—that is, one predicated on a legally invalid theory.
The best any court could do, as here, was to assume that the jury based
its decision only on valid theories (subject, I suppose, to the ordinary
“harmful error” analysis). Such an approach would be easy and efficient,
but wrong. We therefore said that “[t]o hold this error harmless would
allow a defendant to be held liable without a judicial determination that
a factfinder actually found that the defendant should be held liable on
proper, legal grounds.” Casteel, 22 S.W.3d at 388.
Casteel thus relied on Rule 61.1 to “hold that when a trial court
submits a single broad-form liability question incorporating multiple
theories of liability, the error is harmful and a new trial is required when
the appellate court cannot determine whether the jury based its verdict

1 Texas Rule of Appellate Procedure 81(b)(1) had already been repealed

and replaced by Rule 61.1, but it governed the case and, at that time, read as
follows: “No judgment shall be reversed on appeal and a new trial ordered
[because of error] . . . unless the appellate court shall be of the opinion that the
error complained of amounted to such a denial of the rights of the appellant as
was reasonably calculated to cause and probably did cause rendition of an
improper judgment . . . .” Our current rules, of course, make even more
expressly clear that an error that prevents “properly presenting the case to the
appellate courts” is a basis for reversal. Tex. R. App. P. 61.1(b).

7
on an improperly submitted invalid theory.” Id. (emphasis added) (citing
Tex. R. App. P. 61.1). That is, this kind of error is not presumed harmful.
It just is harmful. Casteel was a straightforward application of Rule 61.1.
This view of Casteel animated our conclusion in Harris County v.
Smith, 96 S.W.3d 230 (Tex. 2002). There we held that the trial court
had “erred in overruling Harris County’s timely and specific objection to
the charge, which mixed valid and invalid elements of damages in a
single broad-form submission, and that such error was harmful because
it prevented the appellate court from determining ‘whether the jury based
its verdict on an improperly submitted invalid’ element of damage.” Id.
at 234 (emphasis added) (quoting Casteel, 22 S.W.3d at 388). We noted
that the Rules provide for “two conditions upon which a judgment may
be reversed for an error of law.” Id. (emphasis added) (citing Tex. R.
App. P. 61.1). Those are (1) errors which “probably caused the rendition
of an improper judgment” or (2) errors which “probably prevented the
petitioner from properly presenting the case to the appellate courts.”
Tex. R. App. P. 61.1.
The Harris County dissent—much of which has been converted
into law today, giving me high hopes for the future of this dissent—
incorrectly focused on the first condition. It suggested that the Court
must have presumed harm to reverse. 96 S.W.3d at 235 (citing 96 S.W.3d
at 237 (O’Neill, J., dissenting) (“[T]he Court presumes reversible harm.
To do so the Court must assume, contrary to longstanding precedent, that
the jury failed to follow the trial court’s instructions.” (citation omitted)).
But the dissent was mistaken. The Court did not need to presume
harm—it found harm. The trial court’s charge error caused actual harm

8
because it prevented Harris County from properly presenting its case to
the appellate courts. Whether there was additional harm—that is, the
distinct harm that would have befallen the defendant if the jury in fact
had based its verdict on something it lawfully could not do—was thus
beside the point. And that is why Rule 61.1 today has two parts—Rule
61.1(a) focuses on improper judgments, while Rule 61.1(b) focuses on the
inability of the appellate courts to review those judgments. If the Court
today has not quite conflated them into a single vague rule, it has all
but done so.
Today’s decision is our first to turn Casteel’s straightforward
application of the rules into a special, limited-application presumption.
“We first suggested that Casteel creates a presumption in Cortez ex rel.
Est. of Puentes v. HCCI-San Antonio, Inc., 159 S.W.3d 87, 91 (Tex.
2005).” Ante at 44 n.29 (emphasis added). The Court cites each of our
“Casteel cases” that has used the English word “presumption.” But the
use in those cases suggests little more than loose language. For
example, Cortez was simply an application of Casteel’s conclusion that
when the Court cannot determine the basis for the jury’s verdict, the
charge error is harmful:
No one except the jurors themselves knows exactly what
transpires in the jury room; we know only the verdict. We
held in Crown Life Insurance Company v. Casteel that
when a jury bases its finding on an instruction that
“commingles invalid theories of liability with valid
theories,” we do not hold the error harmless because the
most that a reviewing court can say is that the verdict
might have been reached on a valid theory. 22 S.W.3d 378,
388 (Tex. 2000). Here, we do not know why veniremember
7 was objectionable. But as in Casteel, we cannot know for

9
certain that his inclusion did not affect the verdict, so we
presume harm.

Cortez, 159 S.W.3d at 91 (emphasis added). Adequate presentation of the
appeal was impossible because we could not determine the impact of the
trial court’s error. So the error was harmful—not presumptively so. No
amount of reviewing the entire record could cure it, and nobody suggested
such a thing. Indeed, the fact that the above-quoted statement ended our
analysis shows that we were not presuming but concluding that there
was harm.
As with Casteel, the foregoing analysis was a straightforward
application of now-Rule 61.1(b). Other cases to which the Court points
in support of the new approach refer to a “presumption,” but likewise
without any suggestion that this was consequential in any particular
way—the label did no work at all.
B
The best that can be said for the “presumption” is what Casteel
itself said: “when questions are submitted in a manner that allows the
appellate court to determine that the jury’s verdict was actually based
on a valid liability theory, the [broad-form submission] error may be
harmless.” 22 S.W.3d at 389 (emphasis added). The examples that
followed this statement show how the “questions” could dispel
harmfulness: “answers to other questions” might reveal that the jury did
not credit an impermissible theory, for example, or there might be
“independent grounds” for a verdict, so the broad-form error was
obviously not harmful. Id. (emphasis added; citations and quotation
marks omitted). Similarly, an instruction to jurors not to rely on an
improper ground could prevent any confusion about whether they did,

10
as this Court explained in Benge v. Williams, 548 S.W.3d 466, 476 (Tex.
2018), and as the Court today properly confirms, see ante at 43 n.28, 50.
So obviously a broad-form “error” was not necessarily a kill shot if
something else in the verdict eliminates what otherwise would be a
Rule 61.1(b) problem.
Not in a single case until today, however, have we ever suggested
that judicial speculation would play the role that other questions or
instructions in the verdict form could play—the whole point of
Rule 61.1(b) is to avoid such speculation. Thus, in some cases that use
the word “presumption,” like Thota v. Young, 366 S.W.3d 678 (Tex. 2012),
we rejected the “presumption” because we could determine the basis for
the jury’s verdict. The charge in Thota did exactly what Kansas City
Southern requested here—it included a second answer blank. Id. at
691–92. We therefore concluded that “Casteel’s presumed harm analysis
does not apply because the separate answer blanks allow us to determine
whether the jury found Dr. Thota negligent.” Id. at 691 (emphasis added).
From the Court’s mistaken historiography, it declares that when
the Casteel “presumption” applies, it is rebuttable. Notably, not even one
of the cases that it has mined for the word “presumption” uses the word
“rebut” or “rebuttable.” Today, the Court cautions, rebutting the
presumption of harm requires the court to be “reasonably certain that the
jury was not significantly influenced by issues erroneously submitted to
it.” Ante at 47 (quoting Romero v. KPH Consolidation, Inc., 166 S.W.3d
212, 227–28 (2005)). At least in this sense, Casteel as reimagined is not
nothing. If the presumption applies, then the burden on the party who
won at trial is serious: that party must displace that presumption by

11
establishing how the record as a whole proves that the jury ruled only
on permissible grounds so that the reviewing court can reach that
conclusion with “reasonable certainty.” This exercise does not offer the
clarity that Casteel and its progeny demanded, but in practice, I assume
that it will be next to impossible to dislodge the presumption of harm in
circumstances where the Court today says that the presumption applies.
C
It would be better, though, to just stick with Casteel rather than to
rebuild it into something much less. So in addition to describing why the
Court is wrong to latch on to the occasional word “presumption”—because
we have never treated it as one—I describe how I think that word came
into our Casteel lexicon. As I do so, I will also describe why it is a mistake
to say that some broad-form errors implicate the new “presumption” while
other broad-form errors do not (those in which a jury may have heard
hours of what it thought was “evidence” but that we conclude is legally
insufficient). Any theory that would lead to rendition of judgment in a
single-issue case is a theory that, if comingled with a valid theory, ought
to “presume” harm (if that must be our new theory).
Based on the Court’s telling, this “rebuttable presumption”
innovation stems, at least in part, from a misreading of Romero, 166
S.W.3d 212. See ante at 47. In Romero, the Court rightly concluded that
“[e]ven if the jury could still have made the same apportionment of fault,
the error in the [jury] question is nevertheless reversible because it
effectively prevents Columbia from complaining on appeal that they
would not have done so.” Romero, 166 S.W.3d at 226. The Court noted
that the error is reversible “unless the appellate court is ‘reasonably

12
certain that the jury was not significantly influenced by issues
erroneously submitted to it.’ ” Id. at 227–28 (emphasis added) (quoting
Braun v. Flynt, 731 F.2d 1205, 1206 (5th Cir. 1984)). That strikes me
as correct. But it does not set up the rebuttable presumption that the
court establishes today. Rather, if the Court can tell from its review of
the record whether the error had any impact on the jury’s decision, then
Rule 61.1(a) applies, and Rule 61.1(b) does not. But if there is an error,
as there is here in submitting the missing-yield-sign theory to the jury,
and the record and the charge are unclear about the impact, then Rule
61.1(b) applies and the error is reversible.
The Court, despite its aims, does not “clarify” the law today.
Having turned Casteel into a rebuttable presumption, the Court refuses
to extend that presumption to charge errors that are based on legal
insufficiency of the evidence. See ante at 54. But the resulting harm is
exactly the same—the inability to know whether the jury grounded its
verdict on something that the law forbids. The Court draws this artificial
line to harmonize the reasoning of an inapposite and hastily explained
per curiam opinion. See id. at 52–53 (discussing Ford Motor Co. v.
Castillo, 444 S.W.3d 616 (Tex. 2014)). In truth, the Court jettisons what
came before Castillo and after Castillo. Now those cases are
aberrations—and all to elevate what the Court mistakenly regards as
Castillo’s approach.
Castillo devoted one whole paragraph to the question of whether
Casteel applied—and held that it did not apply because there were no
improperly commingled theories. 444 S.W.3d at 621. The case was about
the sufficiency of evidence supporting an allegation of egregious juror

13
misconduct that threatened the integrity of judicial proceedings, not
about jury instructions on alternative theories of liability. See id. at 618.
Unsurprisingly, the briefing in Castillo spent little time on Casteel. The
discussion was confined to a few pages of the response brief. The Court
rejected the Casteel challenge because there was no submission error in
the first place. In this case, there is no evidence of the missing-yield-sign
theory, but in Castillo, there was evidence for everything necessary for
the verdict. See id. at 621. Nothing more was meant or needed. Castillo
hardly purported to upend all the many statements that the Court today
acknowledges—in pre-Castillo cases like Romero and in post-Castillo
cases like Benge—that expressly treat legal insufficiency as materially
indistinguishable, for Casteel purposes, from any other legal defect.
Ante at 50–51 & n.34. Treating Castillo as the One True Holding is both
ahistorical and wrong. Castillo did not say it was overruling existing
cases, it did not need to overrule existing cases, and it could not
prospectively overrule future cases.
Casteel was a bit player in Castillo—yet the Court today would
treat Castillo as a key player in our Casteel jurisprudence. And based on
it, the Court concludes, without any support for the proposition, that for
legal insufficiency, “harm is far less likely because nothing prevents the
jury from reaching a valid and proper finding based on the evidence it
heard.” Id. at 53.
The theory behind the Court’s newfound approach is that a jury
can simply ignore a theory that is insufficiently supported. Yes, it could
do so. And maybe it is even “more likely” to avoid error in this context
than when the claim is legally foreclosed for other reasons. Id. But for

14
purposes of Rule 61.1(b), the distinction is one without a difference. It
is perhaps “more likely” that, compared to me, the most recent Olympic
gold-medal long-jump champion could leap from one side of the Grand
Canyon to the other. So? We both end up at the bottom, with him a few
feet closer to the other side. The other side of the canyon is where the
substantive review of a verdict begins—when we know what the verdict
was based upon. When the jury hears a theory of liability that should
not have been submitted—whether because the theory is a legal claim
that is invalid or because it is a theory that has no legally sufficient
evidence—the appellate court cannot leap to the other side of the canyon.
That point is what Rule 61.1(b) memorializes. Whether one of them gets
us a tad closer to the other side is immaterial.
After all, it is hardly breaking news that juries sometimes render a
verdict that is erroneous solely because of the lack of legally sufficient
evidence. Human beings are fallible. Lawyers are persuasive. Something
that looks like evidence may not really be. The jury in this case surely
tried its very best to render a proper verdict. But it may have failed. Or
maybe it succeeded. We simply cannot know, and that is because Kansas
City Southern asked for a separate line to ensure that, if the jury was
motivated by the missing-yield-sign theory, we would know it. Horton did
not agree to that, and the trial court agreed with Horton. The Court now
agrees that this ruling was error. Ante at 39. Kansas City Southern was
right all along about the yield sign. Congratulations—and as vindication
for being right, now pay up, because the Court assumes that the jury
relied on the humped-crossing theory. We would know that, however,
only if the requested extra question had not been erroneously denied.

15
I again emphasize that our lack of knowledge has nothing to do
with what kind of legal error infected the broad-form submission. A
theory bereft of legally sufficient evidence is just as lawless a ground on
which to base a verdict as a totally made-up cause of action.
To illustrate the point, Horton prevails on rehearing today
because the Court now believes that the jury could not have been
hoodwinked by the substantial effort to push the missing-yield-sign
theory to the jury. Who, the Court wonders, could be persuaded by
something that, to the eighteen eyes of this Court, constitutes no legally
sufficient evidence? Surely not the jury, the Court opines—which means
that the jury instead must have believed the humped-crossing theory.
The Court makes this assumption even though the missing yield sign
was far more than just a stray comment by a lawyer or witness.
According to the parties, the simple missing-yield-sign theory was fully
10% of the evidence at trial. That includes Horton’s expert witness who
testified that the presence of yield signs would more likely than not have
made a difference. As the Court acknowledges, the missing yield sign
was emphasized in Horton’s opening statement. Ante at 56. The Court
denigrates the opening statement about the missing yield sign (the
lawyer said that the humped-crossing issue was worse!) and the hours
of testimony (more testimony was about other things!). Id. at 56–57. It
does not mention that Horton played a video to emphasize the absence
of the yield sign. The Court thinks it significant that the trial court did
not expressly tell the jury to consider the yield sign, id. at 57, but that
has nothing to do with whether the jury accepted Horton’s repeated
invitations to rely on it. The Court claims to be “reasonably certain”

16
that the jury was unmoved by the missing-yield-sign theory, id. at 56,
which it characterizes as an insignificant sideshow.
Guess who does not think it was a big pile of nothing? Horton—
who seeks rehearing in this Court on the ground that . . . the missing
yield sign does provide sufficient evidence. And the court of appeals—
three trained lawyers, and our colleagues in the judiciary—who sent the
case back for a new trial because they thought that . . . the missing yield
sign provided sufficient evidence. Despite all that, today’s opinion treats
finding out if an actual juror was persuaded by the missing-yield-sign
theory rather than the humped-crossing theory as pointless. Even if it
is plausible that things happened the way the Court now thinks, it is a
problematic basis for a judgment for several reasons.
First, whether we think the jury was unlikely to have been
motivated by the missing yield sign is simply not the relevant question
for Casteel and Rule 61.1(b). Instead, that is a question for Rule 61.1(a)—
whether the error “probably caused the rendition of an improper
judgment.” As I have discussed above, the only question for us is whether
we can determine whether harm occurred. If we cannot, then the charge
error is harmful by definition under Rule 61.1(b). That is itself the
relevant “harm.” The solution, as the rule says, is through reversal—
not through a review of the record to guess about whether Rule 61.1(a)
harm existed.
Second, the Court’s opinion implicitly adopts the U.S. Supreme
Court’s approach for federal criminal cases expressed in Griffin v.
United States, 502 U.S. 46 (1991). That case held that, in the criminal
context, evidentiary insufficiency regarding an alternative legal theory

17
does not constitute a legal error in the same way as other errors that
encompass the presumption of harm. The precedent of this Court is to
explicitly reject that approach in Texas civil trials, as explained in detail,
for example, in both Harris County2 and Romero.3 And we rejected the
analogy for good reason, too. Criminal cases involve much greater
protections against an erroneous verdict. Jurors must be certain of guilt
beyond a reasonable doubt; they cannot vote to convict based on a mere
preponderance of the evidence. And a judgment of acquittal is final—no
matter what, there will be no new trial.
The language in Romero that is now identified as a basis to
reimagine our jurisprudence in a more federal way comes from several
cases in the federal courts of appeals. Those cases also do not purport
to follow the Texas Rules of Appellate Procedure, and I do not think we
should feel any obligation to follow them. The cases include Braun, 731

2 In Harris County, we responded this way:

Instead of Casteel, the dissent urges that we follow the United
States Supreme Court’s decision in Griffin v. United States, 502
U.S. 46 (1991), and hold the charge error in this case to be
harmless. . . . Griffin did not make any new criminal law, nor
did it purport to extend its view of constitutional requirements
to civil procedure. The dissent here acknowledges as much, but
nevertheless suggests that Griffin’s logic should apply equally in
state civil procedural questions as in federal constitutional law.
But the United States Supreme Court itself has acknowledged
that a different reversible error analysis applies in civil cases.
96 S.W.3d at 234 (citations omitted).
3 In Romero, we said: “The argument was made in Harris County that

even if it is reversible error to include legally invalid claims with legally valid
ones in a single jury question, the same rule should not apply when all the
claims are valid but some lack support in the evidence. . . . We specifically
rejected this argument . . . .” 166 S.W.3d at 227 (emphasis added).

18
F.2d 1205, which Romero cited once in a footnote, and E.I. du Pont de
Nemours v. Berkley & Co., 620 F.2d 1247 (8th Cir. 1980). Those cases
pointed to Collum v. Butler, 421 F.2d 1257 (7th Cir. 1970), which was
not cited in Romero, but quite clearly illustrates the Court’s problematic
line of reasoning.
In Collum, the Seventh Circuit concluded that “[t]he dominant
issue, as the record shows, involved the injuries suffered by the plaintiff
as a result of the alleged beating inflicted upon him by the defendants.”
421 F.2d at 1260. Thus, it “would not serve the interest of justice” to
allow “other issues which occupied positions of such relative insignificance
in the trial to be treated now as so important as to make their submission
to the jury prejudicial.” Id. (emphasis added). That is, like the Court
today, the Seventh Circuit swept away the error based on its assessment
of the relative significance of the issues at the trial it did not attend. But
this is based, at least in part, on the Federal Rules of Civil Procedure.
See Fed. R. Civ. P. 61 (“Unless justice requires otherwise, no error in
admitting or excluding evidence—or any other error by the court or a
party—is ground for granting a new trial, for setting aside a verdict, or
for vacating, modifying, or otherwise disturbing a judgment or order.”
(emphasis added)). Cf., e.g., WickFire, L.L.C. v. Woodruff, 989 F.3d 343,
359 (5th Cir. 2021) (“We ‘trust the jury to have sorted the factually
supported from the [allegedly] unsupported.’ ” (alteration in original)
(quoting Nester v. Textron, Inc., 888 F.3d 151, 160 (5th Cir. 2018))).
But again, we have expressly rejected that view. Romero could not
have implicitly meant what the Court now says it meant, because it
explicitly said the opposite. 166 S.W.3d at 227. The Court never

19
acknowledges that it is effectively adopting the Supreme Court’s
approach in Griffin. But what the Court does is just what the Seventh
Circuit did in Collum—write off evidence as so “relative[ly]
insignifican[t]” that it could not possibly have infected the verdict. I
have no problem with such an approach if a defendant did not ask for a
separate question or an instruction. Such an omission would itself
confirm that the supposed evidence was relatively insignificant, and it
is too late to complain about it later in hindsight. But when an error is
timely identified, as here, precisely because of how substantially the
legally insufficient theory was presented to the jury, it is wrong to treat
the error so casually. Texas may have different rules on harm; perhaps
Texas cares more about the integrity of jury verdicts and about other
principles, too.4
Texas law is different, after all, in other ways. “No evidence” is a
legal issue, not a factual one. For that reason, this Court reviews legal
sufficiency but not factual sufficiency. See Tex. Const. art. V, § 6 (the
courts of appeals’ decisions regarding “all questions of fact brought before
them on appeal or error” are “conclusive”); Tex. Gov’t Code § 22.225(a) (“A
judgment of a court of appeals is conclusive on the facts of the case in all
civil cases.”). By contrast, the U.S. Supreme Court can review factual
sufficiency for clear error, so its approach does not raise any serious

4 The right to a jury trial in Texas is unique among states: “Virtually all

state constitutions contain some guarantee of a right to jury trial,
but . . . [n]one contain two separate and unrelated provisions comparable to
those of the Texas Constitution.” George D. Braden et al., The Constitution of
the State of Texas: An Annotated and Comparative Analysis 421 (1977); see Tex.
Const. art. I, § 15; id. art. V, § 10. The Texas Constitution provides civil
litigants a jury right broader than that conferred by its federal counterpart,
permitting jury trials for all causes—not only those at law, but also in equity.

20
concerns. If the line between legal- and factual-sufficiency review in
Texas courts is rendered indistinct, the result could be unpredictable.
Indeed, Horton’s motion for rehearing argues that the Court’s analysis
of the missing-yield-sign theory was an improper reweighing of the facts.
Her skeptical gaze is perhaps more appropriately cast on the Court’s new
opinion.
* * *
Applying Casteel here does not and has never undermined Texas
Rule of Civil Procedure 277’s preference for broad-form submission. As
Casteel emphasized, “Rule 277 is not absolute; rather, it mandates
broad-form submission ‘whenever feasible.’ ” 22 S.W.3d at 390. The rule
uses this language because broad-form submission sometimes is not
feasible. Adding a single question, as in Romero or in this case, would
have eliminated the problem. Or, as the Court notes, adding a single
instruction here—to not consider the yield sign—would have solved the
problem too. Such a modest price to ensure that the courts know what
a jury did. The extra question or extra instruction would hardly be a
departure from broad-form practice, much less constitute a significant
burden on anyone. But refusing to add a question or to instruct the jury
to disregard the challenged theory or to remand for a new trial when
that theory turns out to have been invalid all along is highly prejudicial
without any benefit for the rule of law.
II
The Court’s error today is an error regarding how we read
precedent and how we apply our technical rules of procedure. That is bad
enough. But the error is worse: today’s new approach also dishonors the

21
role of Texas juries—the very thing that Casteel emphasized as so
important. Today’s decision drags appellate justices into an independent
fact-finding role—assessing how the jury would have apportioned fault
based solely on the humped-crossing theory.
Today’s approach is anything but deferential to a jury. The courts’
coercive power to transfer property from one party to another (or to refuse
to do so) is justified only if both sides know that the jury actually found
the facts that warrant that result. “Upholding” a verdict by tethering a
result to a theory that the jury itself may well have rejected does not
respect the jury. A patient under the misimpression that a cancer screen
was clear would naturally refuse an offer of chemotherapy. A doctor
would hardly show respect to then say, “Well, some patients don’t want it
even if they know they have cancer, so no point in asking her—she said
she didn’t want it, and she’s the boss.”
It might be different if we could be sure that a patient emphatically
rejected chemotherapy under any circumstance. And it would surely be
different if we could be sure that the jury did not rely on any unlawful
ground. I agree—and so did Casteel—that there would then be no harm
and thus no need for a new trial if we knew, rather than guessed. See 22
S.W.3d at 389. It would disrespect a jury to pretend that we do not know
what we do know—just as it disrespects a jury to pretend that we know
what we cannot possibly. But there is a massive analytical gap between
concluding that (1) no legally sufficient evidence supports the missing-
yield-sign theory and (2) the jury did not believe and rely on it. A verdict
unsupported by evidence cannot lawfully be the basis for a judgment—
but that does not empirically mean that a verdict cannot issue without

22
legally sufficient evidence. Much experience proves the contrary, which
the Court can hardly deny. So the Court identifies various reasons to
deem it less likely that the jury relied on the missing yield sign but not
the humped crossing. Fair enough—but I need not establish that the jury
did prefer the missing-yield-sign theory to agree with Horton that there
was more than enough thrown at the jury to think that it may well have
made that choice.
Professional judges disagree about the legal sufficiency of this
evidence—the court of appeals concluded that the evidence was not just
legally but even factually sufficient. Despite this, the Court seems to
think the jury must not have relied on the theory. In doing so, it discards
our strong tradition of not presuming the grounds on which a jury ruled.
Honoring the jury’s role requires a new trial. Casteel would have sent it
back—and so would all our cases until today. We should not have
changed course. The new “presumption” should apply to this case—but
whether it applies or not, the error was harmful and the Court should
remand for a new trial.
* * *
Although there are many unanswered questions about today’s
new turn, the Court’s promise is important: that despite the change in
the law, the results should not be materially different. I take the Court
at its word that the record-based harmful-error review will be rigorous
and thorough, and that courts will not give the form while denying the
substance of that review. Albeit with less certainty than under Casteel,
error in a broad-form submission should still generally be recognized as
harmful—whether the old or the new version of Casteel is followed, and

23
whether the “presumption” applies or not. Today’s mistaken decision is
a bad first step on the new journey. But if the courts of appeals wave off
submission error, and if the trial courts become more willing to commit
such error, parties should alert this Court. If the Court can change the
law so quickly on rehearing in today’s case, based on its assumption that
this new way will be better, it can presumably change again if that
assumption is betrayed.
In short, my willingness to hope for the best notwithstanding, I
would stick with our well-reasoned decision from last June and remand
for a new trial consistent with our holding in Casteel and subsequent
cases. I must therefore respectfully dissent.

Evan A. Young
Justice

OPINION FILED: June 28, 2024

24
APPENDIX: THE ORIGINAL PART IV
IV.
Harmful Error
Having concluded that the ICCT Act does not preempt Horton’s
negligence claim based on the humped crossing, but no evidence
supports the negligence claim based on the missing yield sign, we must
finally determine whether the trial court’s submission of both negligence
theories through a single broad-form negligence question constituted
harmful error. KC Southern objected to the broad-form question on the
ground that it would permit the jury to find negligence on an invalid
liability theory, and it offered a proposed charge that included two
blanks for the jury to separately determine the parties’ negligence
liability for the humped crossing and the missing yield sign.
The trial court overruled the objection and refused the proposed
question, believing it improperly granulated a single negligence cause of
action. The court of appeals—after agreeing with KC Southern that
federal law preempts the humped-crossing claim and agreeing with
Horton that the evidence supports the yield-sign claim—concluded that
the trial court erred by submitting both theories in one question and
that the error was harmful under Crown Life Insurance Co. v. Casteel,
22 S.W.3d 378 (Tex. 2000), because the court could not determine
whether the jury found negligence on a valid or invalid theory. See 666
S.W.3d at 19.
We have also concluded that only one of Horton’s two allegations
can support the jury’s negligence finding, albeit the opposite one. Thus,
we too must determine whether submission of the broad-form question

25
over KC Southern’s objection resulted in harmful error. KC Southern
argues it did because the question “commingle[d] valid and invalid
liability theories . . . and a proper reason for the verdict cannot be
ascertained from the record.” Horton argues it did not because the
question properly submitted Horton’s single claim for negligence,
leaving it to the jury to determine what acts would support that claim.
We agree with KC Southern.
We held in Casteel that “when a trial court submits a single
broad-form liability question incorporating multiple theories of liability,
the error is harmful, and a new trial is required when the appellate court
cannot determine whether the jury based its verdict on an improperly
submitted invalid theory.” 22 S.W.3d at 388. Horton argues Casteel does
not apply here, however, because that case involved multiple, distinct
“theories of liability,” some of which were valid and others of which were
not. But here, Horton contends, the court submitted just one liability
theory—negligence. According to Horton, Casteel does not apply because
“[f]ailing to maintain tracks and failing to post a yield sign are different
negligent acts, not separate theories of liability.” We disagree.
It is true that Casteel involved a single broad-form liability
question with instructions addressing “thirteen independent grounds
for liability,” four of which we concluded were invalid. Id. at 387. But in
holding that the error was harmful because appellate courts could not
determine whether the jury based its verdict on an invalid theory, we
relied on and reaffirmed our prior decision in Lancaster v. Fitch, 246
S.W. 1015 (Tex. 1923), in which “the trial court submitted a single
general negligence issue with instructions regarding three distinct

26
theories of negligence liability.” Casteel, 22 S.W.3d at 389. Lancaster
applied the same rule in a case in which the plaintiff “pleaded three
separate acts of negligence as the proximate cause of his injury” and the
trial court submitted a single negligence question. 246 S.W. at 1015–16.
We held that the submission of one invalid negligence theory along with
two valid theories, where it was impossible to tell which theory the jury
relied on, was harmful error. Id. at 1015–17.
We have since applied Casteel’s harmful-error rule in cases
involving a variety of circumstances that created the same problem for
the appellate courts, including the broad-form submission of multiple
elements of damages, Harris County v. Smith, 96 S.W.3d 230, 231 (Tex.
2002), the inclusion of two theories within a single
apportionment-of-responsibility question, Romero v. KPH Consol., Inc.,
166 S.W.3d 212, 215 (Tex. 2005), and a trial court’s refusal to submit
necessary instructions with a broad-form question, Columbia Rio
Grande Healthcare, L.P. v. Hawley, 284 S.W.3d 851, 865 (Tex. 2009).
Most recently, and most importantly, we specifically rejected Horton’s
argument in Benge v. Williams, 548 S.W.3d 466 (Tex. 2018).
At the trial in Benge, the plaintiff “argued and offered
evidence that her physician was negligent both in using an
inexperienced resident to assist with performing her surgery and in not
disclosing the resident’s level of involvement.” Id. at 467–68. But the
plaintiff only claimed a right to recover based on the physician’s
negligent use of the inexperienced resident and did “not claim a right to
recover for the nondisclosure.” Id. at 468. The trial court submitted a
broad-form negligence question, after refusing the physician’s request

27
for an instruction that the jury should disregard the plaintiff’s
arguments and evidence regarding the nondisclosure, and the jury
found in the plaintiff’s favor. Id. at 470. We held that the court erred in
refusing to submit the requested instruction, and we found that error
harmful under Casteel because the jury could have found negligence
based on the nondisclosure even though the plaintiff “does not assert
that claim.” Id. at 474. We did so even though the question the trial court
submitted, “unlike the one in Casteel, did not include multiple theories,
some valid and some invalid. It inquired about a single theory:
negligence.” Id. at 475.
To the extent Horton contends that the Casteel rule applies only
when a broad-form question permits a liability finding based on a theory
or ground that is legally “invalid” as opposed to, as here, a ground
lacking sufficient evidence, our precedent has also rejected that
argument. We have applied the rule not only when the question permits
a finding based on a legally “invalid” theory, but when it permits an
erroneous finding based on a ground the evidence does not support,
Harris County, 96 S.W.3d at 231; Romero, 166 S.W.3d at 227–28, a
ground that is “jurisdictionally barred,” Tex. Comm’n on Hum. Rts. v.
Morrison, 381 S.W.3d 533, 535 (Tex. 2012), and, in Benge, a claim the
plaintiff simply “does not assert,” 548 S.W.3d at 474. As we explained in
Hawley, “[s]ubmission of an invalid theory” simply “involves ‘[a] trial
court’s error in instructing a jury to consider erroneous matters.’” 284
S.W.3d at 865 (quoting Harris County, 96 S.W.3d at 233).
Horton relies on our opinion in Dillard v. Texas Electric
Cooperative, where we stated that, under broad-form submission rules,

28
“jurors need not agree on every detail of what occurred so long as they
agree on the legally relevant result. Thus, jurors may agree that a
defendant failed to follow approved safety practices without deciding
each reason that the defendant may have failed to do so.” 157 S.W.3d
429, 434 (Tex. 2005) (citing Burk Royalty Co. v. Walls, 616 S.W.2d 911,
924 (Tex. 1981)). Indeed, we went on to say in Dillard that jurors “could
have unanimously found [the defendant] negligent, even if half believed
the negligent act was overloading his truck and half believed it was
failing to warn oncoming traffic—acts that preceded two different
collisions.” Id. We read Horton’s reliance on Dillard to argue that the
trial court’s broad-form submission of his two negligence allegations was
not error at all, much less harmful under Casteel.
But Dillard involved a completely different issue: “whether the
trial court abused its discretion in refusing to submit one of two different
instructions on the defendants’ inferential rebuttal defenses.” Id. at 430.
We held that the court sufficiently instructed the jury on those defenses
and thus committed no error, and we made the statements on which
Horton relies to explain why additional instructions would have been
duplicative and unnecessary. Id. Our discussion addressed only the
defendants’ defenses, not the plaintiff’s claims, and the statements on
which Horton relies presumed that each of the acts supporting a
negligence finding were themselves valid and supported by the evidence.
Here, by contrast, we have concluded that the evidence does not support
one of the acts on which Horton relied for a negligence finding. Jurors
finding negligence may not all have to agree on the same valid and

29
supported grounds to find negligence, but they cannot rely on invalid or
unsupported grounds.
Finally, Horton argues that application of the Casteel rule in this
case would undermine our strong preference for broad-form
submissions, as set forth in Texas Rule of Civil Procedure 277. We think
this argument goes too far. Although Rule 277 is “intended to simplify
jury charges for the benefit of the jury, the parties, and the trial court,”
it “was certainly never intended to permit, and therefore encourage,
more error in a jury charge.” Romero, 166 S.W.3d at 230. As we
explained when addressing this argument in Romero, Rule 277 requires
that issues be submitted to a jury in broad form “whenever feasible.” Id.
(quoting TEX. R. CIV. P. 277). We adhere to that rule today, but “Rule
277 is not absolute,” and “[s]ubmitting alternative liability standards
when the governing law is unsettled might very well be a situation
where broad-form submission is not feasible.” Casteel, 22 S.W.3d at 390
(quoting Westgate, Ltd. v. State, 843 S.W.2d 448, 455 n.6 (Tex. 1992)).
Our holding does not overhaul the general preference for
broad-form submission. Rather, it emphasizes that, despite our rules’
preference for broad-form jury questions, “broad-form submission
cannot be used to broaden the harmless error rule to deny a party the
correct charge to which it would otherwise be entitled.” Romero, 166
S.W.3d at 230. Where, as here, true doubt exists as to the validity of one
underlying theory and the trial court must resolve a close call20 as to

20 We note that this case does not present an issue of whether KC
Southern preserved its objection to the trial court’s error. KC Southern
objected that there was a Casteel-type defect in the form of the negligence

30
whether sufficient evidence supports a separate act of negligence,
submitting either separate questions or separate blanks within the
same question may be helpful. Separate jury questions are not the only
means to avoid a Casteel problem. In some cases, rephrasing the
question or giving an instruction not to consider theories that are
unpled, invalid, or lacking in evidentiary support will be sufficient.21
And that alternative is preferable to separate questions when it is
feasible. Again, we emphasize that this holding does not undermine the
general preference for broad-form submission, but rather provides
additional guidance as to how courts should approach instances where
broad-form submission is not feasible.
Because the trial court submitted Horton’s negligence claim as a
broad-form question subsuming both his humped-crossing theory and
his missing-yield-sign theory, we are unable to discern whether the jury
found KC Southern negligent based on the yield-sign theory, which the
evidence does not support. Because the question “allows a finding of

question because the humped-crossing theory was preempted, and the
missing-yield-sign theory was not supported by the evidence. Although this
objection was sufficient to make the court aware of its complaint, KC Southern
also tendered an alternative charge that separated the theories.
21 See, e.g., Benge, 548 S.W.3d at 474–76 (holding it was error to deny a

jury instruction not to consider an unpled negligence theory regarding failure
to disclose a resident’s level of involvement in surgery); Morrison, 381 S.W.3d
at 535–36 (holding that it was error to deny a request to rephrase an
employment discrimination liability question to specify discriminatory
termination rather than a term that encompassed actions that had not been
administratively exhausted); Hawley, 284 S.W.3d at 863–65 (holding it was
error to deny a limiting instruction that the jury should not consider actions of
a doctor who was not the hospital’s agent in determining hospital liability).

31
liability based on evidence that cannot support recovery,” Casteel’s
“presumption-of-harm rule must be applied.” Benge, 548 S.W.3d at 475.

32