OPINION

EVAM. GUZMAN, Justice.

Prairie View A & M University (the “University”) appeals a judgment granted in favor of Eddie Ray Brooks for injuries Brooks sustained while repairing a pipe on the University’s campus. After conducting a legal sufficiency review of the evidence introduced at trial, we conclude there is no evidence that the University had actual knowledge of the dangerous condition that resulted in Brooks’ injury. For this reason, Brooks cannot establish a waiver of sovereign immunity under the Texas Tort Claims Act (“TTCA”), and the trial court lacked jurisdiction to enter judgment. We accordingly reverse the judgment of the trial court and render judgment dismissing the case for lack of jurisdiction.

I. Factual amd PROCEDURAL Background

The University hires contractors to maintain and service many of its mechanical systems. Appellant Bill Turner, d/b/a Turner Mechanical Services (“Turner”), was one of these contractors and at the time of trial had been servicing systems at the University sporadically for thirty years. In October 2000, Turner had a maintenance contract with the University to maintain and repair the University’s steam delivery system, among other systems. Appellee Eddie Ray Brooks was an independent contractor working for Turner. At the time of trial, Brooks had worked for Turner for approximately ten years, doing most of his work on the University’s campus.

Because of the length of time that Turner and Brooks had worked on the University’s campus, they knew some of the University’s employees well. Aaron Watson was the University’s Plant Superintendent and had worked for the University for twenty-eight years. Charles Muse was the University’s Chief Engineer and had worked for the University for twelve years. The testimony at trial established that Turner and Brooks had known and worked around Watson and Muse for a long time. In fact, at the time of the incident in this case, Watson’s cousin was actually working on Turner’s crew.

A. The University’s Health Center Valve Begins Leaking.

On October 25, 2000, Watson went to the University’s Health Center and heard steam venting. He investigated the sound and determined that the steam valve at the Health Center (“Health Center Valve”) was leaking and needed to be repaired. He contacted his supervisor, Muse, to inform him of the situation. Turner was contacted to perform the repair.

The Health Center Valve needed to be isolated prior to beginning the repair. Watson first tried to secure the Health Center .Valve by heading upstream on the campus steam system to find an upstream isolation valve that would deprive the Health Center of its supply of steam. The first upstream isolation valve that Watson *698 tried was at Evans Hall (“Evans Isolation Valve”). The Evans Isolation Valve was not working; although Watson tried to turn the valve, it would not close. Watson admitted that he had difficulties with the Evans Isolation Valve in the past, and he knew when he tried to turn it off that it probably would not work.

When Watson failed to secure the broken Health Center Valve by shutting off the Evans Isolation Valve, he went further upstream to find the next isolation valve that was in working order. The second isolation valve that Watson approached was the isolation valve for a low pressure line that provided steam and heat to the southwest campus (“Southwest Isolation Valve”). Under Watson’s supervision, a colleague used a pipe wrench to shut off the Southwest Isolation Valve. Watson testified that shutting off this isolation valve should have deprived steam to the low pressure line feeding steam to the southwest campus, which included the broken Health Center Valve. Watson further testified that he believed he had in fact shut down the steam that fed the broken Health Center Valve by turning off the Southwest Isolation Valve, and thereafter, he called his supervisor, Muse, to inform him that the steam to the southwest campus had been shut off.

Watson testified that, when he shut off the Southwest Isolation Valve, he heard or felt some turbulence in the pipe. Turbulence is the movement of steam through the pipes, and the turbulence or change in pressure that may be heard or felt when a valve is turned can predict whether a valve was effective in shutting off steam. 1

B. Turner’s Crew Arrives to Repair the Valve.

The next morning, Turner and his crew showed up to repair the leaking Health Center valve. Muse told Turner that the steam feeding the southwest campus had been shut off, and Turner informed his crew of this fact. In addition, Watson may have told Turner’s crew that the steam had been shut off. Turner’s crew accordingly began to repair the broken valve.

Repair work began, but Turner’s crew, which that morning included Watson’s cousin, was having difficulty reaching the bolts that needed to be removed to perform the repair. Because of Brooks’ relatively small size, Brooks climbed into the pit where the valve was located to assist with the repair. Brooks first checked the pipe leading to the broken valve to confirm that it was cool to the touch. A pipe being fed with steam would ordinarily be so hot that a person could not touch it, but could instead feel the heat inches away from the pipe. The fact that the pipe was cold assured Brooks that steam was not present in the pipe.

Brooks began his work by hammering two bolts on the pipe in order to remove the valve. When this effort was unsuccessful, he began to use his cutting torch. After working on the first bolt with his torch, he broke the first bolt loose with his hammer, and a small amount of water dripped out of it, again suggesting that there was no steam in the pipe. He then began cutting the second bolt. When he was finished, he extinguished his cutting torch. At this point, steam exploded from the pipe and burned Brooks. His coworkers removed him from the pit, and after assessing his injuries, took him to the local hospital.

C. An Open Bypass Line Caused the Accident

The testimony in the record indicates that Watson should have been able to shut *699 off the steam to the southwest campus by closing the Southwest Isolation Valve. The Southwest Isolation Valve should have closed off the low pressure line that fed steam to the southwest campus, including the Health Center. If steam was not being fed to the University’s southwest campus, the broken Health Center Valve should have been safe to repair.

But, the low pressure line feeding the southwest campus was not the only steam line on the University’s campus. Alumni Hall, which housed the University’s dining facilities, also received steam from a second line, a high pressure line, used to power the cooking kettles in the Alumni Hall kitchen. Although the Southwest Isolation Valve was located near this high pressure line, Watson testified that he did not believe that there was any reason to also turn off this second, high pressure line. By design, this second, high pressure line existed for the purpose of feeding the cooking kettles, and not to supply steam to the low pressure line feeding the southwest campus.

Apparently, however, a bypass line had been created for the purpose of providing backup steam to the cooking kettles (the “Bypass Line”) in the event that there was a problem with the high pressure line ordinarily used for this purpose. The Bypass Line tied the low pressure line and the high pressure line together, so that steam from the low pressure line could be used to power the cooking kettles if necessary. Because the Bypass Line tied into the low pressure line past the point of the Southwest Isolation Valve, it had the potential to permit steam to enter the low pressure line that should have been isolated by shutting the Southwest Isolation Valve.

This is in fact what occurred when steam severely burned Brooks during the repair of the Health Center Valve. After the incident, the University determined that the Bypass Line was open, permitting steam from the high pressure line to escape into the low pressure line feeding the southwest campus, and accordingly, to the broken Health Center Valve which Brooks had been repairing. The reason that this pipe had been cold to Brooks’ touch was likely that someone had turned off the broken Health Center Valve, preventing steam in the line from traveling in this direction. When Brooks effectively reopened this valve by knocking the bolts off of the pipe, steam from the high pressure line entered the low pressure line feeding the. southwest campus and escaped to burn Brooks.

D. What the University Knew About the Bypass Line.

Whether the University had actual knowledge of the Bypass Line was a subject of dispute at trial. Watson and.Muse both denied knowledge of the Bypass Line, testifying that they were not aware of its existence in the complex maze of piping underneath Alumni Hall. 2 According to them, the Bypass Line was not supposed to exist.

Although there was no testimony indicating that Muse had actual knowledge of the Bypass Line, Turner directly contradicted Watson’s testimony that Watson did not know the line existed. Turner had installed the Bypass Line about a year and a half earlier. Turner testified he believed Watson was aware of the existence of the Bypass Line because he recalled that Watson was present at the Bypass Line’s fa *700 brication and installation. Turner also testified that he would not have installed this line without Watson’s permission, therefore he believed Watson must have been aware of the Bypass Line.

There was no testimony at trial that anyone at the University had actual knowledge that the Bypass Line created the dangerous condition of permitting steam to enter the low pressure line that supplied the southwest campus. For example, there was no testimony that Watson or Muse or anyone else at the University knew. that the Bypass Line had been turned on, as opposed to remaining in an “off’ position for as-needed use. According to Turner, the Bypass Line could be turned on and off by a valve on the line itself. Turner testified that the Bypass Line valve should not have been turned on and that the Bypass Line itself should not have been open. At the time of the incident, the high pressure line feeding the cooking kettles was on and working, so there was no need to open the Bypass Line to obtain additional steam from the low pressure line. Turner stated that he did not know who had turned the Bypass Line on, but Turner acknowledged that Marriott, one of the University’s cooking contractors, opened and closed valves they needed for cooking regularly.

In addition, there was no testimony that Watson or Muse understood that the Bypass Line, which was not part of the piping system’s original design, connected to the low pressure line downstream of the Southwest Isolation Valve, rendering this valve ineffective as an isolating mechanism. Watson testified that he believed he had adequately shut off the steam to the southwest campus by turning off the Southwest Isolation Valve. No one contradicted Watson’s testimony, and there is no testimony indicating that Watson had actual knowledge of the presence of steam in the pipes leading to the valve that needed to be repaired.

E. Brooks’ Lawsuit Against the University.

Brooks sued the University to recover for his injuries. Brooks’ pleadings allege that the University was negligent by failing to shut off all the steam flowing through the piping system. Brooks also alleged a waiver of sovereign immunity because Brooks’ claim involved personal injury caused by the condition or use of property.

By its answer, the University asserted “full sovereign immunity from suit and liability,” and specifically denied that Brooks had asserted a claim under § 101.021 of the TTCA. The University also asserted a third-party claim against Turner for contribution, bringing him into the suit. 3 Thereafter, with respect to its assertion of sovereign immunity, the University filed special exceptions, arguing that Brooks had failed to clearly plead a claim that was actionable under the TTCA and requesting that the court either require Brooks to replead or dismiss the action.

Brooks filed an amended pleading in response, alleging alternative theories under the TTCA. He claimed that the University waived immunity because the use or misuse of the piping and shut-off valves constituted the negligent use or misuse of either tangible personal or real property. He also alleged that the University was responsible as an owner of defective or improperly designed real property, by failing to replace broken valves and by creat *701 ing a bypass line. At no point did Brooks allege in his pleadings that the University had actual knowledge of a dangerous condition existing on the University’s campus.

In response, the University filed a second set of special exceptions, again arguing that plaintiffs pleadings were inadequate, and stating:

Defendant would move that the allegations be stricken and that Plaintiff be required to plead with specificity the alleged dangerous condition of the premises in question.

(Defendant, Prairie View A & M University’s Special Exceptions to Plaintiffs Second Amended Petition (emphasis original)).

We cannot determine from the record on appeal whether the trial court considered the issue of sovereign immunity prior to trial. Apparently the trial court did not rule on the University’s special exceptions until after the first witness, Watson, had testified, at which time the court simply noted on the record that “[fit’s denied.” Prior to the close of the evidence, neither the trial court, nor the parties, focused on the issue of what the alleged dangerous condition was, or whether the University had actual knowledge of it.

At trial, Brooks sought to establish various lapses on the University’s part. Calling Watson adversely as his first witness, Brooks established among other things that Watson failed to follow University procedures precisely in attempting to secure the leaking valve; 4 that there was no updated master diagram detailing the University’s steam system, which valves were broken, and which valves needed to be shut off in order to effectively isolate part of the system; and that, if the first isolation valve Watson had approached, the Evans Isolation Valve, had been working, the accident would not have occurred.

At the close of the evidence, the University again urged the issue of sovereign immunity, arguing in a motion for a directed verdict that the evidence failed to establish a waiver of immunity under a premises theory. Among other things, the University challenged the existence of a dangerous condition and claimed that there was no evidence it had actual knowledge of the condition that caused the injury. The court took this motion under advisement, and postponed a ruling on the various issues raised until after the jury had returned a verdict.

The parties submitted the case to the jury solely under a premises theory. The charge expressly instructed the jury “that the only allegedly dangerous condition was steam in the pipes.” The charge further instructed the jury that the University was negligent with respect to the condition of the premises if, among other things, “Prairie View A & M University had actual knowledge of the danger.” 5

The jury returned a verdict for Brooks, finding the University and Turner were *702 negligent, but Brooks was not. The University and Turner both moved for judgment notwithstanding the verdict. The University again argued that it had no actual knowledge of the dangerous condition and that the evidence was legally and factually insufficient to support the jury’s verdict. The court denied these motions and entered a final judgment. The University and Turner appealed.

II. Analysis

The University raises five issues on appeal, asserting that (1) Chapter 95 of the Civil Practice and Remedies Code imposes additional limitations on the TTCA’s waiver of sovereign immunity, (2) Brooks failed to establish that the University had actual knowledge of the dangerous condition that caused his injuries, (3) the TTCA does not provide a cause of action based on the failure to use property, (4) the TTCA does not provide a cause of action based on a discretionary budgetary decision, and (5) the TTCA does not provide a cause of action based on an improperly designed piping system.

We sustain the University’s second issue and hold the evidence is legally insufficient to establish that the University had actual knowledge of the dangerous condition that caused Brooks’ injury. 6 Accordingly, there is no waiver of sovereign immunity under the TTCA, and we must render judgment dismissing the case for lack of jurisdiction. Because dismissal is appropriate, we do not address the University’s remaining issues on appeal. Similarly, because the only claim against Turner is the University’s third-party claim for contribution, which is derivative of Brooks’ claim against the University, it is unnecessary to address the issues Turner has raised on appeal.

A. A Court Must Address Subject Matter Jurisdiction.

The Texas Supreme Court has recently reiterated that courts must address questions of subject matter jurisdiction at the “earliest opportunity” to do so. Tex. Dep’t of Parks & Wildlife v. Miranda, 133 S.W.3d 217, 226 (Tex.2004). This is because such questions implicate a court’s authority to act. Id.; see also Harris County v. Sykes, 136 S.W.3d 635, 638 (Tex.2004). For the same reason, questions of jurisdiction “cannot be waived, and may be raised for the first time on appeal.” See Waco Indep. Sch. Dist. v. Gibson, 22 S.W.3d 849, 850 (Tex.2000) (holding that it was error for appellate court to decline to address jurisdiction on ground that issue had not been preserved for review “because subject matter jurisdiction is essential to the authority of a court to decide a case”).

Here, the University asserts its sovereign immunity, arguing that the TTCA’s limited waiver of immunity does not apply in this case. This allegation raises a question of subject matter jurisdiction that we must address as a preliminary matter on appeal. 7 See Miranda, 133 *703 S.W.3d at 224 (detailing the procedure for raising, and ruling on, a plea to the jurisdiction based on a waiver of sovereign immunity under the TTCA); Gibson, 22 S.W.3d at 850. Courts must remain “mindful that this [jurisdictional] determination must be made as soon as practicable,” utilizing their discretion to assess jurisdiction promptly, given the needs of a particular case. Miranda, 133 S.W.3d at 227.

The TTCA’s limited waiver of sovereign immunity, Tex. Civ. PRAC. & Rem. Code §§ 101.001-.109, provides that immunity from suit and immunity from liability are eo-extensive. Id. at § 101.025 (“Sovereign immunity to suit is waived and abolished to the extent of liability created by this chapter.”); see also Miranda, 133 S.W.3d at 224. This provision effectively creates a situation in which evaluating whether subject matter jurisdiction exists in a given case may require a court to examine the merits of a plaintiffs claims. Id. at 226. We are faced with such a circumstance here, and we must determine whether jurisdiction exists by examining the elements of Brooks’ claim under the TTCA.

Brooks’ claim against the University was tried to the jury. It is not clear from the record why jurisdictional questions were not addressed at an earlier stage of the proceedings, such as the pleading stage or by a motion for summary judgment, but perhaps the trial court was taking care not to invade the jury’s province, given the fact that the jurisdictional question was intertwined with the merits of the case. Although Miranda had not been issued when this case was tried, we note that Miranda thoroughly details the procedure that courts and litigants should follow in examining questions of subject matter jurisdiction under the TTCA. See 133 S.W.3d at 226-28.

B. A Deferential Standard of Review Applies to Jury Findings.

Whether a court has subject matter jurisdiction is a question of law. Tex. Natural Res. Conservation Comm’n v. IT-Davy, 74 S.W.3d 849, 855 (Tex.2002). Review of determinations of subject matter jurisdiction occurs de novo. Miranda, 133 S.W.3d at 226. However, Miranda recognizes that reviewing jurisdictional determinations may require appellate courts to examine the evidence supporting a claim. Id. at 228. In such a circumstance, when a trial court has examined the evidence and made a pre-trial ruling on subject matter jurisdiction, Miranda requires reviewing courts to employ the familiar summary judgment standard of review, holding that a reviewing court must “take as true all evidence favorable to the non-movant,” and “indulge every reasonable inference and resolve any doubts in the nonmovant’s favor.” Id. at 228.

This case, however, was tried to a jury, and accordingly reaches us in a procedural posture different from Miranda, and somewhat different from most other cases that raise challenges to subject matter jurisdiction under the TTCA. 8 As a conse *704 quence, Miranda does not clearly dictate the standard of review we must apply to jury findings that bear on the showing necessary to establish subject matter jurisdiction. We believe, however, that a review of Miranda illuminates the path we should take.

Miranda expressly holds that the existence of a fact issue on the jurisdictional inquiry precludes a court from granting a plea to the jurisdiction. The Court stated:

If the evidence creates a fact question regarding the jurisdictional issue, then the trial court cannot grant the plea to the jurisdiction, and the fact issue will be resolved by the fact finder. However, if the relevant evidence is undisputed or fails to raise a fact question on the jurisdictional issue, the trial court rules on the plea to the jurisdiction as a matter of law.

[[Image here]]

This standard allows the state in a timely manner to extricate itself from litigation if it is truly immune. However, by reserving for the fact finder the resolution of disputed jurisdictional facts that implicate the merits of the claim or defense, we preserve the parties’ right to present the merits of their case at trial.

133 S.W.3d at 227-28 (citations omitted). This reasoning makes clear that the court intends for jury findings to play a role in determinations of subject matter jurisdiction when disputed facts are relevant to the jurisdictional inquiry.

For this reason, we hold that the proper standard of review to apply to the jury’s findings in this case is the familiar legal and factual sufficiency standard for review of determinations by the finder of fact. Though not expressly ruling on a question of subject matter jurisdiction, our court has employed these standards to evaluate a jury verdict on the merits of a claim under the TTCA in the past. See, e.g., Harris County v. Gibbons, 150 S.W.3d 877, 881 (Tex.App.-Houston [14th Dist.] 2004, no pet.) (employing legal and factual sufficiency standards to review a verdict under the TTCA).

Relying on Texas Department of Criminal Justice v. Miller, 51 S.W.3d 583, 587-88 (Tex.2001), the University argues that this Court must not defer to the jury’s verdict, or to reasonable inferences that the jury may have drawn from the evidence produced at trial, because this case involves a question of jurisdiction. Miller does not stand for such a proposition. Rather, Miller involved an interlocutory appeal after summary judgment proceedings, not a jury verdict, and required only that the reviewing court examine “the evidence submitted by the parties.” Id. at 587. The University does not have any support for its position that a court must disregard jury findings, which in any event is contrary to the reasoning of Miranda.

We will accordingly employ the traditional legal sufficiency review of the jury’s verdict in this case under the legal sufficiency guidelines that the Texas Supreme Court has recently reiterated and further explained in City of Keller v. Wilson, 168 S.W.3d 802 (Tex.2005). 9 See also Harris County v. Vernagallo, 181 S.W.3d 17, No. 03-00619, 2005 WL 1771128 (Tex.App. *705 -Houston [14th Dist.] July 28, 2005, n.p.h.) (restating standard pursuant to Keller). As an appellate court, we consider the evidence in the light most favorable to the verdict and indulge every reasonable inference that supports it. Keller, 168 S.W.3d at 821-22. The evidence is legally sufficient if it would enable reasonable and fair-minded people to reach the verdict under review. Id. at 827-28. We credit favorable evidence if reasonable jurors could, and disregard contrary evidence unless reasonable jurors could not. See id. The trier of fact is the sole judge of the witnesses’ credibility and the weight to be given their testimony. See id. at 819-21. We cannot substitute our judgment for that of the jury, so long as the evidence falls within the zone of reasonable disagreement. See id. at 821-22. But, “if the evidence allows of only one inference, neither jurors nor the reviewing court may disregard it.” Id.; see also Vernagallo, 181 S.W.3d at 25, 2005 WL 1771128, at *5.

We may sustain a legal sufficiency, or no evidence, point if the record reveals one of the following: (1) the complete absence of a vital fact, (2) the court is barred by rules of law or of evidence from giving weight to the only evidence offered to prove a vital fact, (3) the evidence offered to prove a vital fact is no more than a scintilla, or (4) the evidence established conclusively the opposite of the vital fact. Keller, 168 S.W.3d at 810.

C. The TTCA Grants a Limited Waiver of Sovereign Immunity.

Before examining the evidence in this case, we begin by discussing the nature of sovereign immunity under the TTCA. The TTCA contains only a limited waiver of sovereign immunity. State enti-ties such as the University are ordinarily immune from suit and liability. 10 The TTCA creates an exception to the general rule of immunity “only in certain, narrowly defined circumstances.” Miller, 51 S.W.3d at 587. The TTCA “does not waive sovereign immunity for all negligence claims against governmental units.” Id. There will accordingly be cases in which a state entity was clearly negligent but will not be held liable. See, e.g., id. at 588 (failure to properly recognize and treat symptoms of meningitis was “allegation only of negligence,” which does not, without more, waive immunity under the TTCA).

The TTCA specifically waives immunity under the following circumstances. Section 101.021, entitled “Governmental Liability,” provides:

A governmental unit in the state is liable for:

(1) property damage, personal injury, and death proximately caused by the wrongful act or omission or the negligence of an employee acting within his scope of employment if:

(A) the property damage, personal injury, or death arises from the operation or use of a motor-driven vehicle or motor-driven equipment; and

(B) the employee would be personally liable to the claimant according to Texas law; and

(2) personal injury and death so caused by a condition or use of tangible personal or real property if the governmental unit would, were it a private person, be liable to the claimant according to Texas law.

Tex. Civ. Prac. & Rem.Code § 101.021. This waiver of immunity permits recovery against the State for injuries in three gen *706 eral circumstances, including (1) the use of motor driven vehicles, (2) the condition or use of personal property, and (3) premises defects, or the condition of real property. County of Cameron v. Brown, 80 S.W.3d 549, 554 (Tex.2002).

Recovery for premises liability based on the condition of real property is further restricted by section 101.022, which in relevant' part imposes upon the State the duty that a private landowner owes a licensee:

(a) If a claim arises from a premise defect, the governmental unit owes to the claimant only the duty that a private person owes to a licensee on private property, unless the claimant pays for the use of the premises.

Tex. Civ. PRAC. & Rem.Code § 101.022(a). This section provides the only means of establishing premises liability against the State. Miranda, 133 S.W.3d at 233 (stating that the TTCA’s “scheme of a limited waiver of immunity from suit does not allow plaintiffs to circumvent the heightened standards of a premises defect claim contained in section 101.022 by re-casting the same acts as a claim relating to the negligent condition or use of tangible property”); State v. Tennison, 509 S.W.2d 560, 561 (Tex.1974) (rejecting claim that TTCA “creates two entirely separate grounds of liability” involving premises).

A private landowner’s duty to a licensee “requires that a landowner not injure a licensee by willful, wanton or grossly negligent conduct, and that the owner use ordinary care either to warn a licensee of, or to make reasonably safe, a dangerous condition of which the owner is aware and the licensee is not.” State Dep’t of Highways & Public Transp. v. Payne, 838 S.W.2d 235, 237 (Tex.1992). Accordingly, the State may be hable for a dangerous or defective condition on its premises if it either (1) injured a person by willful, wanton or grossly negligent conduct, or (2) if State had actual knowledge of the dangerous condition, the injured person had no such knowledge, and the Stated failed to warn the person of the condition or make the condition safe.

Here, Brooks did not plead or prove that the University was grossly negligent or that it injured him by willful or wanton conduct. Instead, Brooks submitted this case to the jury only under the theory that a dangerous condition existed on the premises that the University failed to warn him about or make safe. To sustain a claim under this theory, Brooks was required to prove, among other things that the state had actual knowledge of a dangerous condition and that he had no actual knowledge of that condition. See State Dep’t of Highways & Public Transp. v. Kitchen, 867 S.W.2d 784, 786 (Tex.1993) (“[Liability for a premise defect requires a finding that the State actually knew of the dangerous condition.”).

D. There Is No Evidence of Actual Knowledge of the Dangerous Condition.

As the Texas Supreme Court noted over thirty years ago in Tennison, “[a]ctual knowledge rather than constructive knowledge of the dangerous condition is required” to sustain a premises claim against the State. 509 S.W.2d at 562 (“The clear intent of the legislature was to limit the State’s immunity in tort claims arising from premise defects by imposing the same duty upon the State as that owed by private persons to a licensee on private property.”). This requirement has been reaffirmed through the years and continues to be a limitation on the State’s liability under the TTCA. See, e.g., See City of San Antonio v. Rodriguez, 931 S.W.2d 535, 536 (Tex.1996) (city’s knowledge of a leaky roof did not establish its awareness *707 of a wet floor, which was the dangerous condition that caused injury); Kitchen, 867 S.W.2d at 786 (“The icy bridge in this case was a premise defect, and since the jury failed to find that the State was aware of the defect before the accident, the State is not hable to plaintiffs.”); see also Creek v. Tex. State Dep’t of Highways & Public Transp., 826 S.W.2d 797, 802 (Tex.App.Houston [14th Dist.] 1992, writ denied) (requiring proof of actual knowledge of damaged sign, not knowledge of improper sign installation).

Proof of actual knowledge requires a finding that the State knew of the dangerous condition that caused the injury, not just proof that the State was aware of a related condition that may create a danger at some time in the future. In Rodriguez, for example, the Texas Supreme Court reversed a jury verdict premised upon a city’s knowledge that a roof leaked, stating the “leaky roof was not itself a dangerous condition; it could only cause a dangerous condition.” 981 S.W.2d at 536. The court therefore remanded for a determination of whether there was actual knowledge of the dangerous condition that had caused plaintiff’s injury, which was water on the floor. Id. at 536-37 (“On retrial, the jury should be instructed that the allegedly dangerous condition was the water [plaintiff] claims was on the floor.”); see also Dyall v. Simpson, 152 S.W.3d 688, 710 (Tex.App.-Houston [14th Dist.] 2004, pet. filed) (holding by en banc court that employer’s awareness that liquid was being pumped through a pipe did not establish awareness of leak in pipe); Payne, 838 S.W.2d at 238 (holding that the fact a claimant knew of a culvert’s existence did not establish as a matter of law his actual knowledge regarding the dangerous location of the culvert).

The requirement that there be proof of actual knowledge of the dangerous condition that caused the injury is disposi-tive in this case. Here, the dangerous condition that caused Brooks’ injury was the entry of steam into the section of pipe that was being repaired. See, e.g., Rodriguez, 931 S.W.2d at 537 (holding as a matter of law that the dangerous condition was the condition that actually caused the injury, and not the preliminary “condition [that] could only cause a dangerous condition”); see also Dallas County Mental Health & Mental Retardation v. Bossley, 968 S.W.2d 339, 343 (Tex.1998) (“Property does not cause injury if it does no more than furnish the condition that makes the injury possible.”). Contrary to Brooks’ arguments on appeal, it was not the fact that the Evans Isolation Valve did not work, or the presence of steam in the pipes generally, 11 that was the dangerous condition. Rather, it was the fact that the Bypass Line was open and permitted steam to enter the section of the pipe that was being repaired. As in Rodriguez, which required a showing of actual knowledge that water was on the floor, here, the necessary showing was that the University had actual knowledge that steam would enter the section of pipe that was being repaired.

There was no evidence at trial that Watson, or anyone else at the University, actually knew steam would enter the section of pipe being repaired. Watson testified that *708 he believed he had shut off steam to the southwest campus, and no one contradicted this testimony. In fact, as part of his case, Brooks established that Watson and Muse told Turner and his employees that steam to the southwest campus had been shut off. In addition, because Watson had turned the steam off the previous day, any steam remaining in the pipe should have bled off or cooled by the time the repair was to occur the following morning. Finally, it was undisputed that the pipe that was being repaired was cool to the touch, and actually dripped water as the first bolt was removed, suggesting to everyone that it was not filled with steam. There is no evidence in the record showing or suggesting that anyone actually knew steam would enter the pipes near the broken Health Center Valve, but advised repair anyway.

Compounding the absence of direct evidence of actual knowledge was Brooks’ counsel’s argument to the jury. 12 Despite the clear requirements of the TTCA, during closing, counsel repeatedly argued that the showing of “actual knowledge” could be satisfied by demonstrating that the University should have been aware of a particular condition. Counsel argued that “actual knowledge is something a reasonable person could or should have known under the circumstances,” and that “actual knowledge is knowledge that the person should have, could have, would have known with any little effort.” Counsel also argued as follows:

... dangerous steam got into a pipe where it should not have been. And if they would have, should have, could have, would have known. They should have known. Don’t get hung up on, “Oh, I didn’t know steam was in it.” They should have known steam was in it. Every bit of evidence offered in this case, every bit of credible evidence tells you they should have known that line was live. They didn’t know it. No doubt about it. Aaron Watson is not a monster. He’s not credible, but he’s not a monster. Neither is Charles Muse. But they should have known. Actual knowledge. They should have known steam was in that line.

Finally, counsel urged the jury not to “get hung up on a technicality” like actual knowledge.

Notwithstanding his position at trial, Brooks nevertheless insists on appeal that the University had actual knowledge of the presence of steam in the pipe being repaired and claims that this knowledge was established by inference. We disagree. Considering all of the evidence in the record, as we must when state of mind is at issue, Keller, 168 S.W.3d at 817-18 (mandating inclusive standard for “consciousness evidence”), we hold that none of the facts relied upon by Brooks creates an inference that the University had actual knowledge that steam would enter the section of the pipe that was being repaired. As the Texas Supreme Court stated in Keller, it is error to disregard evidence that is relevant to a showing of knowledge if a reasonable jury could not have disregarded such evidence. See id. at 829 (“The critical question in this case was the City’s state of mind — [plaintiffs] had to prove that the City knew (not should have *709 known) that flooding was substantially certain. A reviewing court cannot evaluate what the City knew by disregarding most of what it was told.”).

1. Knowledge of the Bypass Line

Brooks first points to the fact that Watson knew of the existence of the Bypass Line to supply the inference that Watson had actual knowledge that steam would enter the section of pipe being repaired. Although Watson claimed that he was not aware of this particular Bypass Line, Turner testified that he believed Watson was aware of the Bypass Line based on his presence when it was fabricated and installed. Because it is the “province of the jury to resolve conflicts in the evidence,” Keller, id. at 820, on review, we assume that Watson knew of the Bypass Line’s existence.

Watson’s knowledge of the existence of the Bypass Line, however, does not establish actual knowledge that, via the Bypass Line, steam would enter the pipes under repair. There is uncontradicted testimony in the record supporting the assertion that, because the cooking kettles were working, the valve on the Bypass Line should have been turned off, rendering the Bypass Line ineffective. Turner testified that the Bypass Line was created as a backup steam supply for the cooking kettles, and that the Bypass Line’s valve should not have been open at the time of the accident, because the cooking kettles were otherwise being supplied with steam from the high pressure line. Turner also testified that the University’s third party cooking contractors, such as Marriott, had access to this valve and sometimes turned valves on and off for cooking purposes.

This uncontradicted evidence undermines Brooks’ claim that Watson, by knowing of the Bypass Line, necessarily knew that steam continued to supply the low pressure line, and accordingly, the broken Health Center Valve. In order to support such a claim, Brooks would have to establish that Watson knew the valve on the Bypass Line was open. But, there is absolutely no evidence of this. Rather, assuming Watson was aware of the Bypass Line, he may have thought it was closed, as Turner testified that it should have been. For the jury to conclude that Watson also knew the Bypass Line was open would have required speculation. Given the uncontradicted evidence that the Bypass Line should have been closed, no reasonable jury could have inferred that Watson’s knowledge of the existence of the Bypass Line, without more, meant that Watson wás also aware that the Bypass Line was open and permitted steam to enter the low pressure line where the broken valve would be repaired. 13

*710 2. The Existence of Turbulence

Brooks also claims that there was no turbulence when Watson shut down the Southwest Isolation Valve and that this fact should have alerted Watson that steam continued to flow to the southwest campus. As an initial matter, we question whether there was testimony at trial actually showing that Watson heard no turbulence when shutting down the Southwest Isolation Valve. Watson testified that he heard turbulence, and in response to an attempt to impeach him with his earlier deposition testimony, the University’s counsel read into evidence part of Watson’s deposition, in which he stated that he “heard some of [the turbulence] cease.” We will assume for the purposes of review, however, that this evidence could be viewed as indicating that Watson did not hear turbulence when shutting the valve.

Even if Watson heard no turbulence when shutting the Southwest Isolation Valve, an absence of turbulence when shutting this valve does not establish actual knowledge of the dangerous condition, i.e., that steam would enter the section of pipe being repaired. The configuration of the University’s piping system was undisputed, and demonstrates that the Bypass Line tied into the low pressure line below the Southwest Isolation Valve. Because of this configuration, whether the Southwest Isolation Valve was open or closed, the Bypass Line, if open, could continue to supply steam to the low pressure line (and hence, the Health Center Valve). 14 Therefore, it is not material whether Watson heard turbulence when shutting the Southwest Isolation Valve, since this fact, without more, has no bearing on whether the Bypass Line was open and permitted steam to enter the low pressure line where the repair would be conducted. Given these facts, whether Watson heard turbulence or not upon shutting the Southwest Isolation Valve does not permit a reasonable jury to infer that Watson, or anyone at the University, had actual knowledge of the dangerous condition created by the open Bypass Line.

3. Problems with Watson’s Credibility

These conclusions are not altered by any questions that the jury may have had concerning Watson’s credibility. We note that in conducting our legal sufficiency review, we have assumed that the jury disbelieved Watson, both with respect to whether he knew of the Bypass Line and whether he heard turbulence when shutting the Southwest Isolation Valve. However, even if questions about credibility permit a jury to disregard a witness’ testimony on a given point, they do not permit a fact finder to disregard uncontra-dicted facts that a reasonable juror could not ignore. Keller, 168 S.W.3d at 820. “Jurors cannot ignore undisputed testimony that is clear, positive, direct, otherwise credible, free from contradictions and inconsistencies, and could have been readily controverted.” Id. In this case, even assuming that the jury disbelieved Watson and found that he was not a credible witness, this disbelief does not overcome other, undisputed facts that were introduced into evidence. In particular, questions about Watson’s credibility do not undercut Turner’s testimony that a valve on the Bypass Line, which should have been closed, was actually open. Neither do such questions confer significance on the aspect of turbulence, when it was undisputed that the steam was being supplied from a source downstream of the isolation valve.

*711 Put simply, “[c]rediting all favorable evidence that reasonable jurors could believe and disregarding all contrary evidence except that which they could not ignore,” we hold that there is no evidence that the University had actual knowledge of the presence of steam in the pipes to be repaired. See Keller, id. at 830. On the evidence introduced at trial, a reasonable jury could not have found that the University had actual knowledge of the dangerous condition that steam would enter a section of pipe that would be repaired.

We have carefully reviewed the evidence supporting the element of actual knowledge in this case, because it is clear that the Legislature insisted upon a showing of actual knowledge, rather than constructive knowledge, as a factor limiting the State’s liability for premises defects under the TTCA. See Tex. Civ. PRAC. & Rem.Code § 101.022(a) & (b); Payne, 838 S.W.2d at 237 (“premise” defects cases require actual knowledge, though “special” defect cases do not). In light of this legislative intent, expressed in the clear language of the statute, it would be error for us to uphold a jury verdict finding negligence based on anything less than a showing of actual knowledge. As the Texas Supreme Court noted in Bossley, “the waiver of immunity in the Tort Claims Act is not, and was not intended to be, complete. Arguments for application of the Act that would essentially result in its waiver becoming absolute must therefore be rejected as contrary to the Act’s fundamental purpose.” 968 S.W.2d at 342.

III. Conclusion

For the foregoing reasons, we find the evidence legally insufficient to establish that the University had actual knowledge of the dangerous condition of steam entering the section of pipe that would be repaired. Absent proof of actual knowledge, the University was entitled to sovereign immunity under the Texas Tort Claims Act. We sustain the University’s second issue, reverse the trial court’s judgment, and render judgment dismissing this case for lack of subject matter jurisdiction.

. Brooks attempted to impeach Watson’s testimony on this point by relying on earlier deposition testimony and maintains on appeal that Watson did not hear any turbulence.

. The Bypass Line is in the Alumni Hall basement with several other pipes that are used for purposes including electricity and hot and cold water. Watson testified that this Alumni Hall piping system has between 12 and 20 different pipes in the basement, with approximately 30 to 40 different valves controlling them.

. Brooks did not file a claim against Turner, so the only claim against Turner is the University's.

. Although the University did not have written procedures in place prior to the accident in this case, it drafted a “Procedure for Steam Shutdown/Start-Up” after the accident that reflects procedures that the University has historically followed. Considering this written procedure, Watson testified that he should have contacted Muse for permission to shut off an isolation valve but that he simply decided to take the required action himself. In addition, Watson admitted that procedures ordinarily would have required him to "bleed” the system after shut off, and then confirm that the system was no longer pressurized. He stated that he did not follow this procedure, though, because the Health Center Valve was already leaking and should have bled down completely by the next morning when the repair would likely be made.

. Neither party alleges any charge error on appeal.

. At trial, the case was submitted to the jury solely on a premises defect theory. Accordingly, this is the only claim that we evaluate on appeal. See, e.g., County of Cameron v. Brown, 80 S.W.3d 549, 554 n. 2 (Tex.2002) (reviewing only premises defect claim where lower court and parties focused only on that claim, despite broader allegations in the pleadings).

. Although it is not necessary to our determination that we must evaluate subject matter jurisdiction on appeal, see Gibson, 22 S.W.3d at 850, we note that there is no danger of unfair surprise arising from a jurisdictional ruling at this stage of the case. A plaintiff unquestionably bears the burden of pleading and proving facts demonstrating that immunity has been waived. Dallas Area Rapid Transit v. Whitley, 104 S.W.3d 540, 542 (Tex. *703 2003). Here, the fact that the University was asserting immunity from suit, and challenging the trial court's jurisdiction to determine the case, was apparent from its original answer. In addition, the University re-asserted its right to immunity pre-trial in two sets of special exceptions, at the close of the evidence in a motion for a directed verdict, and again, after the verdict, in a motion for judgment notwithstanding the verdict.

. Many denials of sovereign immunity reach appellate courts prior to any trial on the merits, via interlocutoiy appeal. See Tex Civ. Prac. & Rem.Code § 51.014 (2003); Sykes, 136 S.W.3d at 638 ("If the trial court denies the governmental entity's claim of no jurisdiction, *704 whether it has been asserted as a plea to the jurisdiction, a motion for summary judgment, or otherwise, the Legislature has provided that an interlocutory appeal may be brought.").

. In this case, we review only for legal sufficiency. It is not clear that the University properly preserved any factual sufficiency complaints that it may have had. Brooks urges that the University waived these complaints, and the University has not disputed this argument.

. It is undisputed that the University, which is part of the A & M University system, is a state entity.

. Without citing any authority, Brooks argues that this court cannot determine the dangerous condition because the University did not object to a jury charge that instructed that "the only allegedly dangerous condition was steam in the pipes.” However, this instruction is not inconsistent with our determination that the dangerous condition was the entry of steam into the section of pipe that was being repaired, caused by the Bypass Line being open.

. Counsel made these arguments without objection from opposing counsel or correction by the trial court. However, we note this problematic argument because, as a court reviewing a question of subject matter jurisdiction, we are obliged to assure ourselves of jurisdiction irrespective of any concerns about waiver. See Gibson, 22 S.W.3d at 851. We believe that statements made during closing argument place the jury’s decision into context, but we note that they do not affect the standard of review we apply to examine the evidence in this case.

. Brooks relies on Rodriguez and Keetch v. Kroger, 845 S.W.2d 262 (Tex.1992), for the proposition that knowledge of factors creating the dangerous condition supports an inference of knowledge regarding the dangerous condition itself. This is not, however, what the Court held in either case. Rather, in Rodriguez the Court held that, on remand, the plaintiff should be permitted to try to establish that the City had actual knowledge of water on the floor. 931 S.W.2d at 537 (“Depending on the position of the leaks above the floor and the amount of rain, the jury might have inferred that the person in charge knew that there would be water on the floor.”). Keetch, in contrast to this case, permitted a finding of liability based on either actual or constructive knowledge. 845 S.W.2d at 265 ("The fact that the owner or occupier of a premises created a condition that posed an unreasonable risk of harm may support an inference of knowledge. However the jury still must find that the owner or occupier knew or should have known of the condition.”) As discussed above, unlike Rodriguez or Keetch, this case involves uncontradicted evidence that undercuts the claim that the University had actual knowledge of the dangerous condition.

. We also note that, aside from Brooks’ arguments about turbulence, there is no evidence in the record that the Southwest Isolation Valve failed to close.