Among other things, opposition may challenge the admissibility of survey evidence related to trademark infringement through a procedural hurdle imposed by Federal Rule of Evidence 702 as construed under Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993). This paper briefly reviews the history of this gatekeeping mechanism, synthesizes what the empirical and secondary literature shows about how often trademark surveys are challenged and excluded, and examines the methodological grounds on which courts have excluded or admitted survey evidence. After applying appropriate screening criteria, the analysis examines 88 court cases (2000 to 2024) involving Daubert motions related to trademark infringement surveys. Across 144 expert-level admissibility decisions, 56.3% of the challenged survey evidence was admitted in full, 11.1% was partially excluded, and 32.6% was excluded entirely. The most frequently raised methodological issues involved survey format (40.8%), question wording (36.6%), correspondence with marketplace conditions (30.5%), and universe definition (28.9%). No individual category of methodological challenge was significantly associated with overall admissibility outcomes, suggesting that exclusion is not readily attributable to the presence of any single type of methodological criticism.
Daubert and Rule 702
Survey evidence is central to trademark litigation. Whether it is ruled admissible as evidence turns on clearing any Daubert motion (see Daubert v. Merrell Dow Pharmaceuticals, Inc., 1993) that the opposing party may bring against a specific expert witness’s testimony. Once survey evidence clears this threshold, cross-examination, competing expert testimony, and jury instructions may influence how much weight it carries at trial. From a market research perspective, this matters because the questions federal judges ask when screening survey evidence substantially overlap with the questions a survey methodologist asks when evaluating the quality of a research instrument.
Ford (2005) presents an excellent review of the Daubert decision, and more specifically, a consideration of key cases that had an impact on shaping admissibility decisions. As noted before 1993, federal courts applied Frye v. United States (1923), admitting expert testimony only when the underlying technique had gained “general acceptance” in its field. Daubert replaced that standard with a reliability requirement under Federal Rule of Evidence 702. Under this, trial judges must independently assess whether an expert’s methodology is reliable, weighing factors such as testability, error rates, peer review, and field acceptance. For survey experts, this is the operative rule. It requires that the expert be qualified, that the testimony assist the factfinder, that it rest on sufficient facts or data, that more likely than not, it applies reliable principles and methods, and that the expert reliably applies those methods to the case’s facts.
Two later decisions matter especially for how this plays out with survey evidence. General Electric Co. v. Joiner (1997) held that a judge may exclude testimony not only for a flawed method but for an analytical gap between an otherwise sound method and the conclusion an expert draws from it. Further, Kumho Tire Co. v. Carmichael (1999) extended the same reliability screening to all expert testimony, not just laboratory science, putting survey methodologists on the same footing as bench scientists. Rule 702 was amended in 2000 to codify this, and again in December 2023 to make explicit that the proponent bears the burden of showing admissibility by a preponderance of the evidence and that an expert’s opinion must reflect the reliable application of a reliable method to the facts of the case. Weisgram v. Marley (2000) also provides guidance for an appellate court to overturn a verdict if material used at trial did not meet Daubert standards.
In practice, the bases on which survey evidence is most often challenged trace back to a handful of design choices: how the relevant universe of respondents is defined, whether a control group or stimulus isolates the effect at issue, how questions are worded and sequenced to avoid leading or demand effects, and how closely the format mirrors the real-world context of exposure. It is worth pausing on why trademark confusion surveys are disproportionately exposed to this kind of ground-by-ground scrutiny compared with expert testimony in other litigation contexts. Unlike some areas, where the expert must apply an accounting formula or evaluate a factual finding (e.g., a value on a biological or chemical test based), and the assessment boils down to whether the general methodology itself is reliable, and whether it was applied without obvious error. A trademark confusion survey, by contrast, has to be tailored in a manner to the specific context where the alleged trademark infringement is thought to be occurring. As such, there is no single accepted test (e.g., Squirt, Eveready, or otherwise) and the universe, stimuli, controls and general nature of the survey itself have to be evaluated within this specific context. Swann (2023) argues that Eveready and Squirt have survived as flexible templates precisely because they leave so many of these design choices to the expert’s judgment rather than reducing survey design to a single validated protocol. Each of those judgment calls is, in principle, a separate potential point of methodological failure and therefore a separate potential ground for a Daubert motion, which is part of why the “elements courts treat as critical” list below has so many independently litigated branches rather than collapsing into one or two dominant questions the way Daubert litigation over, say, a standard diagnostic biomarker test more often does.
Standards for Survey Research
With respect to surveys used in trademark infringement litigation, one of the foremost resources is the Reference Guide on Survey Research, published in the Federal Judicial Center and National Academies of Sciences’ Reference Manual on Scientific Evidence (Diamond, 2011). This work represents an excellent compilation and application of principles to the legal context, and the principles themselves are predicated on strong and grounded work in psychological survey procedures, as well as market research and consumer behavior standards. Diamond (2011) provides a systematic framework for evaluating the adequacy of survey evidence, organized around six principal elements of survey research: Purpose and Design of the Survey; Population Definition and Sampling; Survey Questions and Structure; Surveys Involving Interviewers; Data Entry and Grouping of Responses; and Disclosure and Reporting. These elements encompass the principal stages through which a survey is conceptualized, implemented, analyzed, and ultimately presented as evidence.
A series of specific operational characteristics is apparent from the questions and considerations within Diamond’s broader framework. These elements provide a concrete basis for evaluating how the general principles of survey quality are implemented in a particular study. They include:
- Expert Qualifications. Individuals responsible for designing, conducting, analyzing, or interpreting the research should possess expertise appropriate to those responsibilities.
- A properly defined universe. The population studied should correspond to the population relevant to the issue being litigated.
- Use of an appropriate sampling procedure. Respondents should be selected in a manner that permits appropriate conclusions about the relevant universe (i.e., the universe could be correctly specified, but the sample might not appropriately capture it). Sample size and screening are relevant here.
- Survey purpose and format. The survey should be designed to address the research question the resulting evidence is intended to answer. Concerns often arise about the appropriate survey format, often including whether to use Squirt or Eveready.
- Representative stimuli and conditions. The survey should present respondents with materials and circumstances reflective of the real-world where any alleged infringing behavior is occurring.
- Survey questions and structure. Questionnaire design should permit respondents to express their perceptions or beliefs without directing them toward a desired response (e.g., unbiased wording, clearly articulated questions, appropriate response options, sound survey structure and format).
- Proper control condition(s). Control conditions should allow the researcher to distinguish findings (e.g., confusion) attributable to alleged infringement from other elements, such as background noise or systematic responding.
- Coding and Analysis. Survey responses should be clearly coded, documented, analyzed and reported using appropriate evaluative methods (e.g., employing the correct statistics).
- Survey Administration. Procedures used to administer the survey should adhere to generally accepted standards (e.g., appropriate pre-testing, attention check questions, clear presentation of material on screen).
Examining Daubert Decisisons Involving Consumer Confusion Surveys from 2000 to 2024
No single comprehensive study answers what percent of reports containing surveys associated with consumer confusion are excluded under Daubert. Of cogent relevance is the work of Bird and Steckel (2012), which examined the prevalence and impact of surveys in trademark infringement cases from (2000 – 2006). Bird and Steckel’s data do not account for the intervening 15 years, and more importantly, do not specifically examine Daubert decisions. For a more recent, exclusion-specific benchmark, the closest work comes from outside the realm of survey evidence. Specifically, Sundararaman and Tyler’s (2023) looked at Daubert decisions involving intellectual property damages experts, which examined individual admissibility decisions from 2015 to 2020 and found an overall exclusion rate of 24 percent, with 46 percent of challenged experts facing at least partial exclusion and only 17 percent excluded in full.
The general method used by Bird and Steckel (2012) was employed here, employing court cases from Westlaw. The search terms used were “Daubert” and “Trademark Infringement” and “Survey” appearing jointly in the case record. Cases were further restricted to only those where an original survey was part of the report being challenged (e.g., not a product, damages or statistics expert), and were limited to those ranging from 2000 to 2024. Using the above criteria, there was a final number of 88 different cases, representing 144 Daubert decisions related to survey evidence. Of these, 54.2% were experts for the plaintiff, and 45.8% were experts for the defense. In terms of outcomes, 56.3% were admitted, 11.1% were partially excluded, and 32.6% were excluded in entirety.
Exhibit 1 provides a breakdown of the frequency of these cases by five-year increments. Daubert challenges to survey experts increased steadily across each five-year period, with the largest share occurring between 2020 and 2024 (31.9%). Nearly 60% of all observations occurred during the final 10 years examined.
Each judicial decision was reviewed at the individual-expert level to identify the specific methodological issues raised in connection with a Daubert or other admissibility challenge and to determine how the court treated each issue. This review was primarily conducted using ChatGPT, guided by categories and associated search terms contained in a coding guide (see Appendix), but was augmented by manual efforts for approximately 20% of the cases to check coding accuracy. Search terms were used as retrieval aids rather than as requirements for coding.
Coding was based on the court’s discussion of the individual expert rather than on issues elsewhere in the decision. This distinction was important in cases involving multiple experts because a methodological criticism directed at one expert was not assumed to apply to another expert in the same case. For each expert, the relevant portion of the judicial opinion was identified and reviewed, including the description of the expert’s work, the grounds offered in support of exclusion or limitation, the court’s analysis of those grounds, and the court’s disposition. Each methodological category was evaluated separately to determine whether it was raised.
Exhibit 1
Distribution of Survey Expert Observations by Five-Year Period, 2000–2024
| Five-Year Period | n | % |
| 2000–2004 | 8 | 5.6 |
| 2005–2009 | 21 | 14.6 |
| 2010–2014 | 29 | 20.1 |
| 2015–2019 | 40 | 27.8 |
| 2020–2024 | 46 | 31.9 |
| Total | 144 | 100.0 |
Note. N = 144 expert observations. Frequencies represent expert observations rather than unique cases and as such, cases involving more than one qualifying expert may contribute multiple observations.
Results by category are provided in Exhibit 2. The most frequently raised issues concerned survey format (40.8%) and question wording (36.6%), followed by correspondence with marketplace conditions (30.5%) and universe definition (28.9%). Expert qualifications and data-related issues were each raised in 25.4% of observations, followed by control design (23.9%) and sampling (21.1%). Issues involving survey administration were considerably less common (7.7%). Supplemental analysis using Chi-Square tests revealed that none of these differed as a function of the expert side (i.e., plaintiff or defense), or the overall outcome.
Exhibit 2
Survey Evaluation Categories, Definitions, and Frequency of Issues Raised
| Category | Definition | Percent |
| Expert Qualifications | Individuals responsible for designing, conducting, analyzing, or interpreting the research should possess expertise appropriate to those responsibilities. | 25.4 |
| Properly Defined Universe | The population studied should correspond to the population relevant to the issue being litigated. | 28.9 |
| Appropriate Sampling Procedure | Respondents should be selected in a manner that permits appropriate conclusions about the relevant universe. Sample size and screening are relevant considerations. | 21.1 |
| Survey Purpose & Format | The survey should be designed to address the research question the resulting evidence is intended to answer. Concerns often arise about the appropriate survey format, often including whether to use Squirt or Eveready. | 40.8 |
| Survey Questions & Structure | Questionnaire design should permit respondents to express their perceptions or beliefs without directing them toward a desired response, using unbiased wording, clear questions, appropriate response options, and sound survey structure. | 36.6 |
| Representative Stimuli & Conditions | The survey should present respondents with materials and circumstances reflective of the real-world setting in which the alleged infringing behavior occurs. | 30.5 |
| Proper Control Condition(s) | Control conditions should allow the researcher to distinguish findings attributable to the alleged infringement from background noise, guessing, or systematic responding. | 23.9 |
| Coding & Analysis | Survey responses should be clearly coded, documented, analyzed, and reported using appropriate evaluative and statistical methods. | 25.4 |
| Survey Administration | Procedures used to administer the survey should adhere to generally accepted standards, including appropriate pretesting, attention checks, and clear presentation of materials. | 7.7 |
Note. Percentages indicate the proportion of valid expert observations for which the corresponding issue was raised. A given expert might have had multiple items identified, hence the percentages do not sum to 100. No differences were found between the expert side (i.e., plaintiff or defense), or overall outcome (i.e., admitted, partially excluded or excluded).
Conclusion
Trademark confusion surveys occupy an unusual position in the application of Daubert and Rule 702. The basic principles underlying good survey research are well established, but their application in trademark litigation is necessarily context dependent. The appropriate universe, stimuli, controls, questions, and survey format depend on the particular theory of confusion and the marketplace circumstances the survey is intended to represent. As a result, the reliability of a trademark survey cannot ordinarily be established simply by identifying the use of a recognized methodology such as Squirt or Eveready. The more important question is whether the methodological choices made in a particular survey appropriately address the research question and the circumstances of the alleged infringement.
The empirical findings reinforce the importance of this distinction. Across 88 cases decided between 2000 and 2024, the analysis identified 144 expert-level admissibility decisions involving survey evidence. More than half of the challenged survey evidence was admitted in full (56.3%), while 11.1% was partially excluded and 32.6% was excluded entirely. Challenges were also distributed across a broad range of methodological issues rather than being concentrated on a single recurring defect. Survey format and question wording were the most frequently raised issues, followed by correspondence with marketplace conditions and universe definition. Concerns involving expert qualifications, data analysis, controls, sampling, and administration were also present to varying degrees.
Perhaps equally important is what the analysis did not reveal. The frequency with which particular methodological issues were raised did not differ significantly between plaintiff and defense experts, nor did the presence of individual categories of methodological criticism differ significantly across overall admissibility outcomes. This cautions against treating any single methodological issue as a simple predictor of exclusion. A criticism concerning the universe, control, wording, format, or another aspect of survey design does not by itself determine whether the resulting evidence will survive a Daubert challenge. The significance of a methodological limitation depends on its nature, magnitude, and relationship to the inference the survey is being offered to support.
Taken together, these findings suggest that Daubert review of trademark surveys is best understood as an evaluation of the fit between the research design, its execution, and the conclusions drawn from it. The practical implication for survey researchers is therefore broader than adherence to a checklist of accepted practices. Good litigation research requires defensible methodological choices throughout the research process and, importantly, the ability to explain why those choices are appropriate for the particular marketplace and theory of confusion being examined. Courts ultimately distinguish between limitations that affect the weight that should be given to survey evidence and deficiencies sufficiently consequential to undermine its admissibility. Designing surveys that survive that distinction requires not merely technically competent research, but research in which the connection between the question being asked, the marketplace being modeled, the methods employed, and the conclusions ultimately offered is clear and defensible.
The case records themselves comprised over 3,500 pages, hence, the use of AI to assist in the process was particularly valuable. Regarding the method used to code each expert and the associated coding guide, some elements may have been double-counted. For example, question wording may have affected participant screening and, in turn, the sample used. However, this limitation is not considered to be troublesome for the purposes of this report. If anything, this would likely increase the impact of a variable (i.e., a researcher using this paper as a guide to examine what is important would be erring on the side of caution). Second, while the numbers are meant to be descriptive, no legal or causal conclusions are being drawn. Hence, concern regarding variance surrounding any of the estimates is mitigated to some degree.