Frank Special, as the personal representative of his wife’s estate, appeals a final judgment in favor of the defendants below, Dr. Ivo Baux, his related corporations, and West Boca Medical Center, Inc. Special raises three claims. We affirm on all three, but write to discuss Special’s contention that the trial court erred in limiting the cross-examination of one of the defendants’ expert witnesses.
In considering that issue, we take up this case en banc to reconsider other decisions of this court describing the harmless error test in civil cases. We hold that our cases using an outcome determinative, “but-for” test for harmless error are contrary to the Florida Supreme Court’s interpretation of the harmless error statute. We recede from those cases and adopt the following standard for harmless error in civil cases: To avoid a new trial, the beneficiary of the error in the trial court must show on appeal that it is more likely than not that the error did not influence the trier of fact and thereby contribute to the verdict. Applying this test, we find that harmless error occurred in the trial court and affirm the judgment.
Facts
Susan Special became pregnant at age 38. Five weeks before her due date, she underwent a cesarean delivery. She was wheeled into the operating room at the Center’s labor and delivery suite. Dr. Baux, the anesthesiologist, administered spinal anesthesia. A moment after the placenta was removed, Susan became unresponsive, her blood pressure fell precipitately, and she went into cardiopulmonary arrest. Dr. Baux and hospital staff attempted to revive her. She was temporarily resuscitated and transferred to the Intensive Care Unit, where another ear-diopulmonary arrest occurred. Susan died five hours after the delivery.
Susan’s estate sued the defendants for negligence. The claim was that Dr. Baux and the hospital were negligent in administering anesthesia, in monitoring her system and controlling her fluids during surgery, and in responding to her cardiopulmonary arrests. The defendants denied the allegations; they alleged instead that Susan’s death was caused by amniotic fluid embolus (AFE), an allergic reaction from a mother’s blood mixing with amniotic fluid, sometimes causing heart-lung collapse.
At trial, the plaintiffs expert testified that Susan died because of the departures from the requisite standard of care. The AFE diagnosis figured prominently. Most notably, the plaintiff called Dr. Barbara Wolf, the chief medical examiner of Palm Beach County at the time of Susan’s death. Dr. Wolf conducted the autopsy on Susan and concluded that there was no evidence of AFE in her body. She explained that in a majority of cases where someone dies from AFE, the autopsy provides evidence of AFE, and that was not the case with Susan.
Special also presented the testimony of Dr. Mark Adelman, a pulmonary specialist, who was called in when Susan went into distress. He diagnosed AFE at the time based upon her clinical signs. Special asked him about the number of patients diagnosed with AFE at West Boca. He testified that he saw all such patients. He estimated that he saw about one or two cases per year at the center. During his testimony, Special was able to elicit national statistics showing incidence of AFE diagnosis at West Boca was about 15 times the rate elsewhere. Dr. Adelman, however, contended in his answers that he was only estimating the number of cases he *758saw and had no medical records to back up his recollection.
The defendants called Dr. Gary Dildy as their expert. Dr. Dildy opined that Susan died of AFE. He based this on his analysis of the medical records and tests. He explained that AFE is a diagnosis of exclusion. In other words, a doctor will look at all the circumstances and test results to determine likely causes for the patient’s condition. Where no other circumstances account for the patient’s distress during or after a delivery, a diagnosis of AFE can result.
On cross-examination, the plaintiff elicited from Dr. Dildy that the probability of AFE is approximately 1 in 20,000 births, but can range between 1 in 8,000 and 1 in 80,000. The plaintiff then tried to begin a line of cross-examination of Dr. Dildy about the reliability of the Adelman diagnosis that AFE had actually occurred in Susan, in light of the unusually high incidence of it at the hospital. The defendants’ objection on relevancy grounds was sustained.
Special responded that this line of questioning was sought to impeach Dr. Adelman’s testimony. The trial court sustained the objection, noting that the plaintiff could inquire about the statistical occurrence of AFE and make argument about disproportionate diagnoses in closing, but could not question Dr. Dildy using the substance of Dr. Adelman’s testimony and its reliability to explore the trustworthiness of the AFE diagnosis. The court concluded that doing so would amount to improper collateral impeachment. We understand the trial court’s characterization of the proposed impeachment as “collateral” as being merely another way of saying that the line of questioning was irrelevant.1
The plaintiff proffered Dr. Dildy’s testimony on this issue. The expert stated that, assuming Dr. Adelman’s recollection of the incidence of AFE at the hospital was accurate, he would be concerned that AFE was being over-diagnosed at the Center. Yet even when confronted with statistics documenting this' possibility, Dr. Dildy persisted in his opinion that Susan presented a case of AFE. He testified, “But this case here, we’re talking about, it doesn’t matter what all these other cases are, this case is the case, and this case is an amniotic fluid embolism.”
In closing argument, the plaintiff vigorously argued that the hospital either had an epidemic of AFE or was over-diagnosing it:
[Dr. Adelman] said, I see one to two a year at West Boca Medical Center. I didn’t put the words in his mouth. He said, I see one to two a year at West Boca Medical Center.
*759[I]f you take his numbers, and you believe they have this many amniotic fluid emboluses at West Boca Medical Center every year, it is somewhere between 15 and 80 times the national average they’re diagnosing amniotic fluid embolus at West Boca Medical Center, between 15 and 80 times the national average.
So, it was either an epidemic, which there isn’t, at West Boca Medical Center, or they’re overdiagnosing amniotic fluid embolus. They’re calling things that aren’t amniotic fluid embolus, like he did in this case, ... because they’re not bothering to look at autopsies, they’re not bothering to look at other records, they’re not bothering to investigate why...:
It’s not the epidemic, it’s that he’s overstating the diagnosis, and that’s wrong, ladies and gentlemen, that is flat out wrong to do, and that’s what they did in this case.
The jury found no negligence by the defendants and the trial court rendered a final judgment in their favor.
The Evidentiary Ruling
Again, the principal dispute at trial was the cause of Susan’s death. In response to the plaintiffs claims of negligence, the defendants contended that regardless of their handling of the emergency from cardiopulmonary arrest, it was AFE that caused Susan’s death. The presence of AFE was thus the essential issue at trial. The trial court abused its discretion in failing to allow the cross-examination.
Three sections of the evidence code provide the framework for evaluating questions of relevance. The general rule is that “[a]ll relevant evidence is admissible, except as provided by law.” § 90.402, Fla. Stat. (2009). “Relevant evidence is [defined as] evidence tending to prove or disprove a material fact.” § 90.401, Fla. Stat. (2009). Section 90.403, Florida Statutes (2009), establishes a limitation on the introduction of relevant evidence: “Relevant evidence is inadmissible if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of issues, misleading the jury, or needless presentation of cumulative evidence.”
When, on cross-examination, a piece of evidence is offered to attack the credibility of a witness on a material issue, such evidence is “relevant” under section 90.401 because credibility is central to the truth seeking function of a trial. Under subsection 90.608(5), Florida Statutes (2009), any party “may attack the credibility of a witness by ... proof by other witnesses that material facts are not as testified to by the witness being impeached.”
The object of the proposed cross-examination of the defense expert was to elicit answers leading to proof of the cause of death, the crux of the lawsuit. Dr. Adelman and Dr. Dildy both testified that the cause of death was AFE. Counsel sought to impeach Dr. Adelman’s diagnosis with evidence showing that the incidence of diagnosed AFE at West Boca, all done by Dr. Adelman, was grossly in excess of national statistics, thus impeaching Dr. Adelman. Where the diagnosis is one of exclusion,2 the frequency with which one comes to that conclusion is a “material fact” bearing upon the credibility of the diagnosis. The cross-examination was also relevant to Dr. Dild/s direct examination where he testified to the incidence of AFE in births and its rarity. The trial judge *760abused his discretion in refusing to allow the cross-examination.3
The central question to this appeal is whether the exclusion of the cross-examination amounted to harmless error. To consider that issue, it is necessary to review the development of the harmless error standard in Florida.
Harmless Error Prior to State v. DiGuilio
We first review the history of the harmless error rule contained in section 59.041, Florida Statutes (2009) — the circumstances leading to its enactment and how the interpretation of it has evolved since 1911.4 The Florida cases describe a general trend away from a “correct result” test, utilized in the earliest common-law decisions and in earlier interpretations of the harmless error statute, and toward an “effect on the fact finder” test, as embodied in the Supreme Court’s landmark decision in State v. DiGuilio, 491 So.2d 1129 (Fla.1986).
According to the “orthodox” English rule, an error in admitting or rejecting evidence was not a sufficient ground for a new trial unless it appeared, looking at all the evidence, that the truth had thereby not been reached. 1 Wigmore, Evidence § 21 (3d ed.1940); see also Doe v. Tyler, (1830) 130 Eng. Rep. 1397, 1399 (C.P.) (orthodox rule). In contrast, under the more stringent “Exchequer” rule, which took hold in English and in many American courts after the 1830s, an error at trial created per se a right to reversal. See Crease v. Barrett, (1835) 149 Eng. Rep. 1353, 1360(Ex.).5 The earliest Florida cases followed the orthodox rule,6 though *761by the turn of the century some cases applied the more rigid Exchequer rule in narrow circumstances.7
The Exchequer rule and its influence on American courts were widely criticized for making reversal too easy. See, e.g., 1 Wig-more, Evidence § 21. A reform movement in the United States gained steam in the early twentieth century,8 spurred by an influential address by Roscoe Pound, in which he opined that “the worst feature of American procedure is the lavish granting of new trials.” Roscoe Pound, The Causes of Popular Dissatisfaction with the Administration of Justice, 29 A.B.A. Rep. 395, 413 (1906). The American Bar Association studied the problem and suggested statutory reforms, which were adopted at the state and federal levels. See 33 A.B.A. Rep. 542 (1908). Florida’s harmless error statute, originally enacted in 1911, see Ch. 6223, Laws of Fla. (1911), was almost identical to the A.B.A.’s proposed statute,9 and has remained unchanged since:
No judgment shall be set aside or reversed, or new trial granted by any court of the state in any cause, civil or criminal, on the ground of misdirection of the jury or the improper admission or rejection of evidence or for error as to any matter of pleading or procedure, unless in the opinion of the court to which application is made, after an examination of the entire case it shall appear that the error complained of has resulted in a miscarriage of justice. This section shall be liberally construed.
§ 59.041, Fla. Stat. (2010) (formerly § 54.23, Fla. Stat.).
Two aspects in the wording of the statute are significant. First, the statute applies in both civil and criminal cases. Second, the trigger for reversible error is the occurrence of a “miscarriage of justice”; how the courts have defined this term has determined the scope of the statute’s application since its enactment.
The 1911 harmless error statute differs in one important respect from the A.B.A. model set forth in footnote 9. The Florida statute adds the last sentence: “This section shall be liberally construed.” While a “strict construction” of a statute would consider “only the literal words of [the] writing,” a liberal construction is “[a]n interpretation that applies a writing *762in light of the situation presented and that tends to effectuate the spirit and purpose of the writing.” Black’s Law Dictionary 332 (8th ed. 2004). The purpose of the harmless error statute is to enhance finality by limiting the granting of new trials. However, by insisting on a liberal construction, the statute allows for discretion and flexibility in its interpretation; the term “miscarriage of justice” should not be construed so narrowly that reversal is a rarity.10
In the years following the passage of the harmless error statute in 1911, the Florida Supreme Court used two tests to define a “miscarriage of justice” giving rise to a reversible error: a “but-for,” “correct result,” test that is oriented on the outcome, and the more forgiving “effect on the fact finder” test that is oriented on the process.
A “coi’rect result” approach asks whether, despite the error, the trial court reached the correct result. It assumes that when the result was correct, there cannot have been a “miscarriage of justice.” But see Traynor, supra note 5, at 18-22 (criticizing this approach). The question is, would the result have been the same without the error? Or, but for the error, would the result have been different? An “effect on the fact finder” approach, on the other hand, asks whether the error influenced the trier of fact and contributed to the judgment, not just whether it changed the result. Looking at the record as a whole, did the error mislead the trier of fact? See id. at 22-23 (discussing benefits of this approach). The former approach effectively narrowed the class of cases that could be reversed; the latter broadened it.
The most commonly used test, the “but-for” formulation, focused on whether the result of the trial would have been different but for the error. This outcome oriented approach considered whether the “wrong” result was reached as a result of the error. A typical criminal case, Henderson v. State, 94 Fla. 318, 113 So. 689 (1927), illustrates the Supreme Court’s early interpretation of the harmless error statute:
The language of the statute ... makes it clear that it was the purpose of the Legislature that verdicts and judgments of trial courts should not be overturned and set aside by this court on account of mere errors committed in the court below unless it is made to appear to this court, after inspection of the entire record, that the errors complained were prejudicial and injurious in them nature and tendency and resulted in a miscarriage of justice. This statute was, no doubt, based upon the idea that if the result of a trial, the verdict and judgment, was just and right, even though there were technical errors committed by the trial court, no good purpose could be subserved by the labor, expense, and delay of trying the case over again. And to make [this] intention effective, the statute was so framed as to require it to be made to appear to the reviewing court that the error complained of caused, or at least contributed to causing or reasonably tended to cause, the *763result, and that the result was wrong — a miscarriage of justice.
Id. at 697-98 (emphasis added). This equation of a “miscarriage of justice” with a “wrongful result” characterizes much of the Supreme Court’s early harmless error jurisprudence, and harkens back to Florida’s earlier application of the orthodox English rule.11
The same outcome oriented analysis also prevailed in some early civil cases. In E.O. Roper, Inc. v. Wilson & Toomer Fertilizer Co., 116 Fla. 796, 156 So. 883 (1934), the Supreme Court held that even if the trial court committed technical errors, under the harmless error statute, its judgment would not be set aside where
the record as a whole shows that the judgment rendered accords with justice in the premises, and that a reversal of the cause for the correction of such technical errors as may have occurred must inevitably lead to the rendition of a new judgment identical with that now appealed from ....
Id. at 884 (emphasis added). This interpretation of the harmless error statute focused on the legislative purpose of conserving judicial resources; because the statute was designed to reduce the waste caused by needless retrials of cases reversed for technical error, it was therefore applied to prevent reversal whenever errors would not have altered the outcome.12
Other early civil and criminal cases focus less on the correctness of the outcome and more on whether the decision-making process was compromised; these cases apply an “effect on the fact finder” test for harmless error. For example, Eggers v. Phillips Hardware Co., 88 So.2d 507 (Fla.1956), involving an action for injuries to a pedestrian caused by a truck driver, held it was error to admit into evidence the testimony of the investigating officers that, following the investigation, they had not arrested the driver for breaking any of the city’s traffic ordinances. The Court reversed the trial court’s judgment for the defendant and remanded for a new trial on the grounds that the erroneously admitted evidence might have influenced the jury’s verdict:
There was a direct conflict in the evidence at the trial on this vital point [whether the defendant ran a red light] and it may well be that the fact of the non-arrest of defendant might have balanced the issue in favor of the defendant. We think that the ends of justice would best be served by submitting this issue to another jury, so that it can be decided without the defendant’s having the benefit of the inadmissible evidence in question.
Id. at 508 (emphasis added).
Further, an early criminal case anticipates the DiGuilio test of a “reasonable possibility” of an effect on the verdict. See *764infra pp. 11-13. In Pearce v. State, 93 Fla. 504, 112 So. 83 (1927), the defendant challenged his conviction for murder on the grounds that improperly excluded evidence of a pair of bloody brass knuckles found at the crime scene, together with evidence of the defendant’s head wounds, would have supported his claim of self-defense. The Supreme Court agreed and used an analysis that emphasized the effect that the excluded evidence might have had on the jury:
It is impossible to say with any degree of certainty that, if the evidence of the finding of a ‘piece of a pair of knucks’ [sic] near the scene of the difficulty had been admitted for consideration by the jury ..., the jury would not have accounted for the wounds on the head of the defendant upon the theory that the deceased had attacked him with metallic knuckles.
Id. at 86.
And, fifteen years before DiGuilio, the Supreme Court applied an “effect on the fact finder” harmless error test in a civil case, but without explicitly characterizing its approach. In Stecher v. Pomeroy, 253 So.2d 421 (Fla.1971), a personal injury action, the trial court erroneously admitted evidence about the extent of the defendant’s insurance coverage; the Supreme Court held that the error was harmless “in light of the fact that the verdict was $19,000 despite policy limits of $100,000/ $300,000; where there was a disc involvement with serious and prolonged disability, traction and hospitalization; and where the injuries were permanent.” Id. at 422. The Supreme Court emphasized that the error was harmless, and not a basis for reversal, when considered in the context of the whole trial, because the record showed that it did not contribute to the judgment.
This recognition of harmless error in these particular circumstances is not to be regarded as approval by this Court of the mention of policy limits to a jury. This should not be done. Nor is it approval of the trial court’s refusal to grant the requested instruction to disregard, which should have been given. It is simply held to be harmless error here where an examination of the entire record reflects a tone which indicates in no wise any adverse effect upon the jury’s verdict.
Id. (footnote omitted). In essence, the Supreme Court set the defense oriented verdict against the abundant evidence favorable to the plaintiff and concluded that the erroneous admission of the defendant’s insurance coverage had little effect on the jury’s verdict.
From Eggers and Stecher, we distill two general propositions about harmless error analysis in civil cases: First, to determine whether an error is harmful, the appellate court must examine the entire record. Second, the central issue is whether the error had an adverse effect upon the jury’s verdict; in other words, whether the error contributed to the judgment.13 Such was the state of the law before DiGuilio.14
*765 State v. DiGuilio
State v. DiGuilio, 491 So.2d 1129 (Fla.1986), is the touchstone for harmless error analysis in Florida. In it, the Supreme Court firmly establishes an “effect on the fact finder” harmless error test for criminal cases.
In DiGuilio, testimony from a police officer about his arrest of an alleged cocaine trafficker was interpreted as a comment on the defendant’s silence. Id. at 1180-31. The Fifth District ordered a new trial, applying a rule of per se reversal for comments on a defendant’s silence. Id. at 1134. The Supreme Court rejected a rule of per se reversal, and instead adopted the harmless error test announced by the U.S. Supreme Court in Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967). The Court explained the test:
The harmless error test, as set forth in Chapman and progeny, places the burden on the state, as the beneficiary of the error, to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict or, alternatively stated, that there is no reasonable possibility that the error contributed to the conviction.
DiGuilio, 491 So.2d at 1135 (citation omitted). This “effect on the fact finder” test focuses on the likelihood that an error at trial influenced the trier of fact and contributed to the judgment. If it is reasonably possible that the error contributed to the verdict, then the verdict must be set aside, even when, in the reviewing judge’s opinion, the verdict would have been the same without the error. The error and its probable effects must be evaluated in light of the other evidence:
Application of the test requires an examination of the entire record by the appellate court including a close examination of the permissible evidence on which the jury could have legitimately relied, and in addition an even closer examination of the impermissible evidence which might have possibly influenced the jury verdict.
Id. Following Chief Justice Traynor,15 the Supreme Court emphasized that applying the harmless error test is not simply a matter of reviewing the evidence left untainted by error to determine whether it is sufficient to support the judgment. Id. at 1136. Instead, the appellate court places the error in the context of the other evidence to estimate the effect of the error on the trier of fact. The purpose of the analysis, in other words, is not to retry the case without the error, but to reconstruct the original trial to determine what role, if any, the error played in the judgment. As the Court said:
The test is not a sufficiency-of-the-evidence, a correct result, a not clearly wrong, a substantial evidence, a more probable than not, a clear and convincing, or even an overwhelming evidence test. Harmless error is not a device for the appellate court to substitute itself for the trier-of-fact by simply weighing the evidence. The focus is on the effect of the error on the trier-of-fact. The question is whether there is a reasonable possibility that the error affected the verdict.
Id. at 1139. Thus, even abundant evidence in support of a verdict will not prevent reversal when the appellate court cannot say, after reviewing the whole record, that there is no “reasonable possibility that the *766error affected the verdict.” Id. The “burden to show that the error was harmless must remain on the state. If the appellate court cannot say beyond a reasonable doubt that the error did not affect the verdict, then the error is by definition harmful.” Id.
The DiGuilio test for harmless error, which draws heavily on Chief Justice Tray-nor’s insights, contrasts sharply with the “correct result” test applied by the Supreme Court in the decades following the enactment of the harmless error statute in 1911. Under the “correct result” test, a judgment generally could not be reversed unless the appellate court concluded that the outcome of the trial would have been different, but for the error. Under the DiGuilio test, a judgment should be reversed, and a new trial granted, whether or not the outcome of that trial is likely to be different whenever the appellate court believes there is a reasonable possibility that the error influenced the trier of fact and contributed to the verdict.
The differences between a “correct result” test and an “effect on the fact finder” test are subtle but important. An “effect on the fact finder” test asks the appellate judge to look closely at the error and estimate its effect on the trier of fact. A “correct result” test asks the judge to look at everything but the error and determine whether the verdict in a trial without it would have been different. In short, one test focuses on process; the other on the end result. Moreover, a “correct result,” or “but-for,” test asks the judge to exclude the wrongly admitted evidence (or include the wrongly rejected evidence) and weigh the evidence anew — precisely what DiGui-lio forbids. See DiGuilio, 491 So.2d at 1136.
Supreme Court Civil Cases After DiGuilio
While the Florida Supreme Court has not explicitly adopted a standard for harmless error in civil cases after DiGuilio, three cases employed an “effect on the fact finder” test akin to the one that the court applied in DiGuilio.
Gormley v. GTE Products Corp., 587 So.2d 455 (Fla.1991), established two things about harmless error analysis. First, the court expressly placed the burden on the beneficiary of an error in the trial court to demonstrate on appeal that the error was harmless. Second, the court used an effect on the verdict analysis to determine whether harmless error had occurred. In Gormley, the Supreme Court ordered a new trial after finding that the introduction of collateral source evidence may have influenced the jury’s verdict for a defendant. The court explained why the burden to prove the harmlessness of the error was on the defendant-appellee, which injected the improper evidence into the trial.
Equity and logic demand that the burden of proving such an error harmless must be placed on the party who improperly introduced the evidence. Putting the burden of proof on the party against whom the evidence is used ... would simply encourage the introduction of improper evidence.
Id. at 459. The Court held that the defendant-appellee had failed to meet its burden to establish that the erroneous introduction of the collateral source evidence was harmless — because the issue of liability was close, the Supreme Court “[could not] say that the jury’s verdict on liability was not improperly influenced by the evidence of the [plaintiffs’] insurance claim.” Id.16
*767A second case applying DiGuilio’s “effect on the fact finder” analytical framework is Sheffield v. Superior Insurance Co., 800 So.2d 197 (Fla.2000). There, the trial court denied the plaintiffs motion to exclude collateral source evidence, and the plaintiff, after stipulating that she would have a standing objection to the introduction of the evidence, introduced her own rebuttal collateral source evidence. Id. at 199. Although the jury found for the plaintiff, they found no permanent injury and awarded only her past medical expenses and $6,554.61 for future medical expenses. Id. The first district affirmed, holding that Sheffield invited the error by introducing her own collateral source evidence. Id.
The Supreme Court reversed. It held that (1) allowing any collateral source evidence was error because of “the inherently damaging effect of the jury hearing collateral source evidence on the issues of liability and on issues of damages:” and (2) that Sheffield did not waive her objection to that evidence by introducing her own collateral source evidence following the trial court’s denial of her motion in limine. Id. at 203 (citing, inter alia, Gormley). The court explained the reversal with language that evaluated the effect of the improper evidence on the jury:
[Gjiven the inherently prejudicial effect of such evidence, which is the very reason the collateral evidence rule was first established, we cannot conclude that in this case the introduction of collateral source evidence was harmless. The jury certainly could have concluded that because Sheffield had group insurance available to cover future medical expenses, there would be no need to award substantial damages for the future.
Id. (emphasis added). The italicized language demonstrates the Supreme Court’s conclusion that the error was not harmless, because the appellee had failed to demonstrate that it was more likely than not that the error did not contribute to the verdict.
A third post-DiGuilio civil case is Linn v. Fossum, 946 So.2d 1032 (Fla.2006). In that medical malpractice case, the court did not explicitly apply a harmless error test, but held that a trial court’s error in allowing an expert witness to testify that she had consulted with colleagues before forming her opinion “was not harmless because the competing expert opinions on the proper standard of care were the focal point of this medical malpractice trial.” Id. at 1041. This reasoning is consistent with an “effect on the fact finder” test because it recognizes that in a “battle of the experts” the trier of fact would likely be influenced by the credibility of an expert witness which had been enhanced by the hearsay confirmation of other doctors.
In summary, in civil cases after DiGuilio, the Supreme Court has utilized an “effect on the fact finder” test for harmless error in civil cases, even though it has not explicitly declared so.17 The court has expressly declared that on appeal the burden of proving the harmlessness of an error is on the beneficiary of the error in the trial court, who improperly introduced the offending evidence.
*768 District Court of Appeal Harmless Error Cases
Without specific guidance from the Supreme Court, the district courts of appeal have drifted in different directions in applying a section 59.041 harmless error test to civil cases.18 There are three principal lines of cases applying tests for harmless error in the district courts. The most stringent test, occurring primarily in this district, derives from language contained in the earlier Supreme Court cases, and asks whether the result would have been different, but for the error.19 Another strain of decisions, from the first and third districts, lowers the bar for harmful error, and asks whether the result may have been different had the error not occurred.20 Finally, a third line of cases, mostly from the second district, asks whether it is reasonably probable that the appellant would have obtained a more favorable verdict without the error.21 The last two tests are arguably similar to each other, but the test most frequently applied by this court is clearly more stringent.
Under this court’s stringent “but-for” formulation, it is difficult for an appellant to establish harmful error, that a “miscarriage of justice” occurred within the meaning of section 59.041. The line of cases applying this “but-for” test began with Anthony v. Douglas, 201 So.2d 917 (Fla. 4th DCA 1967). Though it has often been cited by this court, Anthony rests on shaky footing. Anthony cites two cases in support of its test for harmless error, i.e., “whether, but for the error complained of, a different result would have been reached by the jury.” Id. at 919.
The first, Cornelius v. State, 49 So.2d 332, 335 (Fla.1950), is a criminal case that *769predates DiGuilio. Following other criminal cases from the same period, see supra, Cornelius states the test for harmful error as “whether the result would have been different had the improper evidence been excluded.” Id. The persuasiveness of Cornelius has been undercut by the different direction the Supreme Court took in DiGuilio.
The second case cited as authoritative in Anthony, Banco Nacional de Cuba v. Steckel, 134 So.2d 23, 25 (Fla. 3d DCA 1961), does not articulate any test for harmless error, holding only that, “[w]hile the defendant contends the trial court erred in striking his defensive motions, this could constitute no more than harmless error where summary final judgment was properly entered.” In fact, the holding in Banco Nacional does not appear to support any one test for harmless error, so it is unclear why we cited it as authoritative in Anthony. This stringent “but for” test, which characterizes almost every error as harmless, encourages evidentiary gambles on questionable evidence in the trial court, placing a premium on winning at all costs, because only the most egregious evidentiary errors will result in reversal.
Like the outcome oriented approach in this district, the second line of cases, from the first and third districts, focuses on the impact of the improperly admitted evidence on the outcome of the trial. These cases appear to have sprung from a footnote in Marks v. Delcastillo, 386 So.2d 1259, 1267 n. 15 (Fla. 3d DCA 1980), which stated, without citation;
We tentatively suggest the following as a shorthand-rule of thumb approach to this and related questions as applied to civil cases: fundamental error occurs when the result would have been different; reversible error, when the result might have been different; harmless error, when it would not have been different.
In Katos v. Cushing, 601 So.2d 612, 613 (Fla. 3d DCA 1992), this “tentative suggestion” morphed into persuasive authority for the proposition that “[t]he test for harmless error is whether, but for the error, a different result may have been reached.” Katos in turn has often been cited as stating the proper test for harmless error in civil cases.22 This test eases the difficulty of the strict “but-for” test by requiring some lesser degree of probability that the result in the case would have been different.23
The third line of cases, starting with Damico v. Lundberg, 379 So.2d 964 (Fla. 2d DCA 1979) (on rehearing), uses somewhat different language to put a finer point on the test of the probability of a different result. In Damico, the second district held that an “error is reversible only when, considering all the facts peculiar to the particular case under scrutiny, it is reasonably probable that a result more favorable to the appellant would have been reached if the error had not been committed.” Id. at 965 (citing Stecher, 253 So.2d at 422).24 This test differs from the DiGu*770ilio test for harmless error in two ways. First, it requires a “reasonable probability,” rather than a mere “reasonable possibility.” Second, it focuses on the probability of a different outcome on retrial rather than the probability that the error contributed to the outcome in the actual trial.
We believe that the district courts of appeal have primarily used a variation of outcome-oriented analysis in approaching the harmless error conundrum instead of employing the process-oriented “effect on the fact finder” approach that the Supreme Court adopted in DiGuilio and reaffirmed in Goodwin v. State, 751 So.2d 537 (Fla.1999) and Ventura v. State, 29 So.3d 1086 (Fla.2010).
At least one of our civil cases appears however to apply an “effect on the fact finder” test similar to the one applied in DiGuilio. Mattek v. White, 695 So.2d 942 (Fla. 4th DCA 1997) was a personal injury action arising from an auto accident. The trial court allowed a physicist, who was an accident reconstruction and biomechanics specialist, to offer his opinion that the collision could not have caused permanent injury to the plaintiff. Id. at 943. We held it was error to admit the physicist’s testimony about permanent injury because the physicist was not a qualified medical expert. Id.
Holding that the error was harmful, we said: “We cannot find the error in admitting this testimony to be harmless because there was ample evidence in this case that plaintiff did have a permanent injury, and the admission of [the physicist’s] opinions regarding permanency could well have been what persuaded the jury to find no permanency.” Id. at 944 (emphasis added). Here, as in DiGuilio, “[t]he focus [was] on the effect of the error on the trier-of-fact.” 491 So.2d at 1139.25 But Mattek, with its “effect on the fact finder” test, stands as an island in a sea of cases applying outcome-oriented, “but-for” anal-yses.
Harmless Error in Civil Cases
In formulating a harmless error test in civil cases, it is important to recognize that DiGuilio derived its formulation from the elevated burden of proof in criminal cases:
The harmless error test ... places the burden on the state, as the beneficiary of the error, to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict or, alternatively stated, that there is no reasonable possibility that the error contributed to the conviction.
DiGuilio, 491 So.2d at 1135 (emphasis added) (citation omitted). This elevated test acknowledges (1) the higher burden of proof in criminal cases, which amplifies the potential effect of an evidentiary error on the trier of fact, and (2) the special concern for the legitimacy of criminal convictions expressed in the constitutional and statutory protections accorded to criminal defendants. A harmless error test for civil cases should acknowledge the particular attributes of those cases.
*771As in a criminal case, the approach to harmless error analysis in a civil case should begin with an examination of the entire record by the appellate court,26 including a close examination of both the permissible evidence upon which the jury could have relied and the impermissible evidence which may have influenced the verdict. The focus of the analysis is to evaluate the effect of the error on the trier of fact to determine whether or not the error contributed to the judgment. We agree with Chief Justice Traynor that a “reasonableness” standard is inappropriate for a harmless error analytical framework because it does not specify a degree of probability:
The nebulous test of reasonableness is unlikely to foster uniformity either in the application of standards, should there be any, or in the pragmatic exercise of discretion. Discretion is at least under better control within tests that focus on the degree of probability as more probable than not, highly probable, or almost certain.
Traynor, supra note 5, at 34-35.
Just as the Supreme Court used the burden of proof in a criminal case to describe the harmless error standard in DiGuilio, so should the burden of proof in civil cases inform the harmless error standard here: harmless error occurs in a civil case when it is more likely than not that the error did not contribute to the judgment. To avoid a new trial, the beneficiary of the error in the trial court must show on appeal that it is more likely than not that the error did not influence the trier of fact and thereby contribute to the verdict.
This test for harmless error is consistent with the way the Supreme Court approached the issue in DiGuilio, Gormley, Sheffield, and Linn. Because section 59.041 applies to both criminal and civil cases, the same type of “effect on the fact finder” harmless error analysis should be used in both types of cases, with the adjustment in civil cases that takes the lower burden of proof into consideration. The “more likely than not” burden is not insurmountable for an appellee contending that a trial error was harmless; it is consistent with the “liberal construction” of the statute mandated by the legislature.
The lower burden also effectuates the statutory goal of enhancing finality in a way that recognizes the different stakes involved in criminal and civil cases. Criminal cases involve a deprivation of liberty, not merely financial loss, so the procedural and substantive law emphasizes the goal that the end result in a criminal case be just and right. Social policy places a greater premium on finality in civil cases than in criminal cases, a finality that should come sooner rather than later. Put differently, society is willing to tolerate more mistakes in civil cases than it will in criminal ones. This policy preference for a quick finality in civil cases supports our decision to require the appellee to demonstrate not that there was a high probability that the error did not affect the verdict, or that there was a reasonable probability that it did not, but that, more likely than not, the error had no such harmful effect.
We therefore recede from the line of cases in footnote 19, which apply a strict, outcome-determinative “but-for” test for harmless error. We also certify the following question to the Supreme Court as being of great public importance:
IN A CIVIL APPEAL, SHALL ERROR BE HELD HARMLESS WHERE IT IS MORE LIKELY THAN NOT *772THAT THE ERROR DID NOT CONTRIBUTE TO THE JUDGMENT?
Applying the Harmless Error test in This Case
The question here was whether the trial court’s refusal to allow the proposed cross-examination of Dr. Dildy was harmless error. The ultimate purpose of the proposed cross-examination was to call into question the hospital’s AFE diagnosis by suggesting that the hospital diagnosed that condition about 15 times more than the rate elsewhere. This issue was presented to the jury through the testimony of Dr. Adelman and in part from Dr. Dildy. This evidence allowed the plaintiffs attorney in closing argument to hammer on the significance of the statistical abnormality. During the proffer of Dr. Dildy, he said that if the incidence of AFE at the hospital were accurate, he would be concerned that AFE was being over-diagnosed. Yet, even when confronted with the statistics documenting this possibility, he persisted in his opinion that Susan presented a special case of AFE. He testified, “But this case here, we’re talking about, it doesn’t matter what all these other cases are, this is this case, and this case is an amniotic fluid embolism.”
Considering all of the testimony, the jury had the full ability to take the statistical anomaly into consideration; the omitted testimony added little to the plaintiffs case. Having reviewed the entire record, we conclude that it is more likely than not that the restriction on the cross-examination of Dr. Dildy did not contribute to the verdict. The error was harmless.
Accordingly, we affirm the judgment entered below. We withdraw the panel opinion previously issued in this case and substitute this opinion in its place.
MAY, C.J., WARNER, POLEN, STEVENSON, TAYLOR, and CIKLIN, JJ., concur. DAMOORGIAN, J., concurs specially with opinion, in which MAY, C.J., concurs. CONNER, J., concurs in majority opinion only in result and specially with opinion, in which LEVINE, J., concurs. HAZOURI and GERBER, JJ., recused.. In the field of evidence, another use of the term ''collateral'' concerns the ability to offer extrinsic evidence to contradict a witness’s answer to a question posed on cross examination. As the first district observed in Faucher v. R.C.F. Developers, 569 So.2d 794 (Fla. 1st DCA 1990), overruled on other grounds by Ullman v. City of Tampa Parks Dep’t, 625 So.2d 868 (Fla. 1st DCA 1993):
The law is well settled that it is improper to litigate purely collateral matters solely for the purpose of impeaching a party or witness. Once a question is put to the party or witness on a purely collateral matter for the purposes of impeachment, the proponent of the question is bound by the witness’s answer; it is inappropriate to then try the truth or falsity of the answer on the collateral matter by adducing independent proof through other witnesses.
Id. at 804. Here, the plaintiff was attempting to ask Dr. Dildy a type of fact that could bear on his opinion under section 90.704, Florida Statutes (2009). The plaintiff's cross examination did not violate the rule stated in Faucher.
. Dr. Dildy also referred to this as a "wastebasket” diagnosis.
. We also reject the trial court’s explanation that the evidence was unfairly prejudicial under section 90.403. This provision is not a general grant of authority to trial judges to bar evidence adversely impacting a party’s position at trial; rather the concept of "unfair prejudice” pertains to "evidence which is directed to an improper purpose, such as evidence that inflames the jury or appeals improperly to the jury’s emotions.” Charles W. Ehrhardt, Florida Evidence § 403.1 (2006 ed.); see also Westley v. State, 416 So.2d 18, 19 (Fla. 1st DCA 1982) (same). Unfair prejudice within the meaning of section 90.403 does not arise from relevant inquiries directed at experts offering contrary opinions relevant to a material issue at trial.
. In addition to section 59.041, section 90.104, Florida Statutes (2009) provides that a court may reverse a judgment or grant a new trial on the basis of admitted or excluded evidence "when a substantial right of the party is adversely affected” and the point is properly preserved in the trial court. The primary contribution of the statute to the law is its requirement of preservation. Section 90.104 adds little to harmless error analysis; if admitted or excluded evidence does not adversely affect "a substantial right of a party,” its admission cannot be a "miscarriage of justice” under section 59.041
Nonetheless, some cases involving evidentiary errors apply a harmless error test based on "injury to substantial rights.” See, e.g., Tormey v. Trout, 748 So.2d 303 (Fla. 4th DCA 1999); Wall v. Alvarez, 742 So.2d 440 (Fla. 4th DCA 1999); Centex-Rooney Constr. Co. v. Martin Cnty., 706 So.2d 20 (Fla. 4th DCA 1997). See also Prince v. Aucilla River Naval Stores Co., 103 Fla. 605, 137 So. 886, 887 (1931) ("A judgment should not be reversed or new trial granted in any case for error in rulings upon the admission or rejection of evidence unless it shall appear to the court from a consideration of the entire case that such errors injuriously affect the substantial rights of the complaining party.”) (citations omitted)).
. But see Roger J. Traynor, The Riddle of Harmless Error 4-8 (1970) (arguing that later cases applying a rule of per se reversal misinterpreted the Exchequer’s decision in Barrett ).
. O’Steen v. State, 92 Fla. 1062, 111 So. 725, 730 (1926) ("This jurisdiction appears to have followed what is known as the 'orthodox English rule,’ rather than the rule announced by the Court of Exchequer in 1830 ...”). See also McKay v. Lane, 5 Fla. 268, 276 (1853) ("This court has uniformly proceeded upon the practice not to reverse a judgment, how*761ever erroneously an isolated point may have been ruled by the Judge below, when it is clearly apparent that the party complaining has been in no degree injured by the improper ruling.”); Hooker v. Johnson, 10 Fla. 198, 203 (1860) (same); Randall v. Parramore, 1 Fla. 409, 486, 1847 WL 1060 (1847) (same).
.See, e.g., Mayer v. Wilkins, 37 Fla. 244, 19 So. 632, 637 (1896) (holding with regard to erroneous jury charge that "injury is presumed” and reversal appropriate where Court could not say "that the misdirection of the court did not influence the result of the verdict”); Walker v. Parry, 51 Fla. 344, 40 So. 69, 71 (1906) (reversing for an erroneous jury charge, citing Mayer). See generally Wadsworth v. State, 201 So.2d 836, 841-43 (Fla. 4th DCA 1967) (Willson, Assoc. J., dissenting) (summarizing early history of harmless error in Florida), rev’d, 210 So.2d 4 (Fla.1968).
. See generally Roger A. Fairfax, Jr., A Fair Trial, Not a Perfect One: The Early Twentieth-Century Campaign for the Harmless Error . Rule, 93 Marq. L. Rev. 433 (2009) (chronicling the movement to curb excessive reversals by reforming harmless error rules).
. That proposed model provided:
No judgment shall be set aside, or new trial granted, by any court of the United States, in any case, civil or criminal, on the ground of misdirection of the jury or the improper admission or rejection of evidence, or for error as to any matter of pleading or procedure, unless, in the opinion of the court to which application is made, after an examination of the entire cause, it shall affirmatively appear that the error complained of has resulted in a miscarriage of justice.
33 A.B.A. Rep. 542, 550 (1908).
. There are two plausible explanations for the legislative softening of the language from the model statute. One explanation is that the problem of excessive reversals does not seem to have been as serious in Florida where the infamous Exchequer rule never fully took hold, compared to other jurisdictions. Legislators may have worried that a radical cure would be worse than the mild disease. Another explanation is that in a state like Florida, with a strong. tradition of electing state judges, legislators expected that the ballot box would be a more effective check on abuses of judicial discretion than statutory rules, and therefore saw little purpose in tying judges’ hands.
. See also Johnson v. State, 61 So.2d 179, 179 (Fla.1952) C'[A]ny error in allowing such statements to remain in the confession was harmless when considered in context with the entire record, and we cannot find that it could have had any effect whatsoever in the ultimate outcome of the case.”); Cornelius v. State, 49 So.2d 332, 335 (Fla.1950) (“In determining whether the error ... was harmful or prejudicial, we must decide upon examination of all the evidence whether the result would have been different had the improper evidence been excluded.”); Banks v. State, 116 Fla. 834, 156 So. 905, 906 (1934) ("[UJnder the facts shown by the record, the jury should not have returned any other verdict than that which was returned.”).
. Other civil cases applying the outcome oriented analysis are Rance v. Hutchinson, 131 Fla. 460, 179 So. 777, 780 (1938); Herman v. Peacock, 103 Fla. 438, 137 So. 704 (1931); Routh v. Richards, 103 Fla. 757, 138 So. 72 (1931).
. See also Josey v. Futch, 254 So.2d 786, 787 (Fla.1971) (following Stecher) ("[T]he essential consideration is evidence of influence on the jury....”).
. In 1985, the year before DiGuilio, Justice Overton noted that the Florida Supreme Court had "never expressly set forth a harmless error test for the appellate courts of this state to apply in civil cases.” Fla. Patient's Comp. Fund, Inc. v. Von Stetina, 474 So.2d 783, 793-94 (Fla.1985) (Overton, J., concurring in part and dissenting in part). But Justice Overton added that, in general, "[t]he application of the harmless error statute requires an appellate court to consider the entire record and determine whether the verdict was affected by the error.” Id. at 793.
. See Traynor, supra note 5, at 18-22 (arguing against a "correct result” test for harmless error); see also People v. Ross, 67 Cal.2d 64, 60 Cal.Rptr. 254, 429 P.2d 606, 620-21 (1967) (Traynor, C.J., dissenting) (same), rev'd, 391 U.S. 470, 88 S.Ct. 1850, 20 L.Ed.2d 750 (1968) (citing, inter alia, Chapman).
. See also Flores v. Allstate Ins. Co., 819 So.2d 740, 751 (Fla.2002) ("[I]f there has been error in the admission of evidence, the burden is on the beneficiary of the error to *767establish that the error was harmless.” (citing Sheffield v. Superior Ins. Co., 800 So.2d 197, 203 (Fla.2001)).
. See also Tallahassee Mem'l Reg’l Med. Cir. v. Meeks, 560 So.2d 778, 782 (Fla.1990) ("Considering the totality of the evidence, we conclude that the introduction of this one privileged statement did not prejudicially affect the jury's determination of negligence and that no reversible error occurred in its admission.”).
. Recently, now-Chief Justice Canady acknowledged the split in the lower courts over the test for harmless error:
The requisite prejudice to support overturning the judgment based on the jury’s verdict can be established neither under a harmless error standard requiring a showing of a reasonable probability of a result more favorable to the appellant if the error had not occurred, see Damico v. Lundberg, 379 So.2d 964, 965 (Fla. 2d DCA 1979), nor under a standard requiring a showing that the appellant might have obtained a more favorable result but for the error, see National Union Fire Ins. Co. of Pittsburgh v. Blackmon, 754 So.2d 840, 843 (Fla. 1st DCA 2000).
Saleeby v. Rocky Elson Constr., Inc., 3 So.3d 1078, 1089 (Fla.2009) (Canady, J., dissenting) (emphasis in original).
. See Hayes v. State, 55 So.3d 699 (Fla. 4th DCA 2011) (civil commitment); Petit-Dos v. Sch. Bd. of Broward Cnty., 2 So.3d 1022 (Fla. 4th DCA 2009), rev. denied, 19 So.3d 311 (Fla.2009); Marshall v. State, 915 So.2d 264 (Fla. 4th DCA 2005) (civil commitment); Kammer v. Hurley, 765 So.2d 975 (Fla. 4th DCA 2000); Pascale v. Fed. Exp. Corp., 656 So.2d 1351 (Fla. 4th DCA 1995); Nationwide Mut. Fire Ins. Co. v. Vosburgh, 480 So.2d 140 (Fla. 4th DCA 1985); Aristek Cmtys., Inc. v. Fuller, 453 So.2d 547 (Fla. 4th DCA 1984); Anthony v. Douglas, 201 So.2d 917 (Fla. 4th DCA 1967). See also Dessanti v. Contreras, 695 So.2d 845, 849 (Fla. 4th DCA 1997) (Hauser, Assoc. J., concurring in part and dissenting in part) (citing Pascale, Aristek, and Anthony ).
. See Witham v. Sheehan Pipeline Constr. Co., 45 So.3d 105 (Fla. 1st DCA 2010); Hogan v. Gable, 30 So.3d 573 (Fla. 1st DCA 2010); Healthcare Staffing Solutions, Inc. v. Wilkinson ex rel. Wilkinson, 5 So.3d 726 (Fla. 1st DCA 2009); Gold v. W. Flagler Assocs., Ltd., 997 So.2d 1129 (Fla. 3d DCA 2008); Jones v. Goodyear Tire & Rubber Co., 871 So.2d 899 (Fla. 3d DCA 2003); Nat’l Union Fire Ins. Co. of Pittsburgh v. Blackmon, 754 So.2d 840 (Fla. 1st DCA 2000); Katos v. Cushing, 601 So.2d 612 (Fla. 3d DCA 1992).
. See In re Commitment of DeBolt, 19 So.3d 335 (Fla. 2d DCA 2009); Esaw v. Esaw, 965 So.2d 1261 (Fla. 2d DCA 2007); Fla. Inst. for Neurological Rehab., Inc. v. Marshall, 943 So.2d 976 (Fla. 2d DCA 2006); Damico v. Lundberg, 379 So.2d 964 (Fla. 2d DCA 1979) (citing Stecher) (on rehearing).
. See, e.g., Hogan, 30 So.3d at 575; Gold, 997 So.2d at 1130-31 (also citing Marks); Blackmon, 754 So.2d at 843. See also Gencor Indus. Inc. v. Fireman’s Fund Ins. Co., 988 So.2d 1206, 1209 (Fla. 5th DCA 2008); USAA Cas. Ins. Co. v. McDermott, 929 So.2d 1114, 1117 (Fla. 2d DCA 2006).
. It is unclear exactly what degree of probability the test requires. But we can safely assume that "may” implies a lesser degree of probability than "would,” which implies near-certainty.
. This interpretation of the harmless error statute accords with the longstanding interpretation of a similar constitutional provision in another jurisdiction. Cf. Cal. Const. art. VI, § 13 (formerly art. VI, § 4 1/2); People v. Watson, 46 Cal.2d 818, 299 P.2d 243, 254 *770(1956) ("[A] 'miscarriage of justice’ should be declared only when the court, ‘after an examination of the entire cause, including the evidence,’ is of the 'opinion' that it is reasonably probable that a result more favorable to the appealing party would have been reached in the absence of the error.” (citation omitted)).
. We also looked at the effect of the error on the trier of fact in another recent civil case. See Gen. Motors Corp. v. McGee, 837 So.2d 1010, 1036 (Fla. 4th DCA 2002) (“We agree with GM that two errors occurred during the trial, but we find those errors to be harmless in the context of this case ... The jury was not swept away by the emotions of the attorneys. The jury’s verdict separated the issues of liability and damages from that of punitive damages.”).
. See also Medina v. Peralta, 724 So.2d 1188, 1189-90 (Fla.1999) ("When examining an evi-dentiary ruling under section 59.041, we are required to look at the entire record.”).