The Journal of Things We Like (Lots)
Select Page
Zhihao Zhang, Maxwell Good, Vera Kulikov, Femke van Horen, Mark Bartholomew, Andrew S. Kayser & Ming Hsu, From Scanner to Court: A Neuroscientifically Informed "Reasonable Person" Test of Trademark Infringement, 9 Sci. Advances 1 (2023).

In trademark litigation, consumer surveys can determine a number of important doctrinal questions, including the most important one: whether consumers are likely to be confused into thinking that the defendant’s product was made or licensed by the plaintiff. Recently, scholars have questioned the validity and reliability of standard trademark surveys, suggesting that they are easy to manipulate and biased in favor of one party or the other. Wouldn’t it be great, then, if there was a reliable way to determine whether a survey was biased or not? Using neuroscientific imaging, an interdisciplinary group of researchers (including law professor, Mark Bartholomew) has proposed just such a possibility in a new paper, From Scanner to Court: A Neuroscientifically Informed “Reasonable Person” Test of Trademark Infringement.

Trademark surveys can suffer from a number of flaws. They may be explicitly biased in favor of one party or another, for example, by describing the defendant as a “copycat” or the plaintiff as a “trademark bully.” They may exhibit more subtle biases in how they frame questions about similarity and confusion. And, finally, survey participants always know the nature of the survey they are taking, so participants may exhibit “demand effects,” providing what they anticipate are the surveyor’s desired answers rather than their true responses.

The researchers began by demonstrating the manipulability of survey instruments. With two different plaintiff products, they manipulated the survey language to create both “pro-plaintiff” and “pro-defendant” surveys, as well as a putatively “neutral” survey. For each plaintiff product, they included a list of other products that varied in their degree of apparent similarity. For example, Reese’s Peanut Butter Cups served as the plaintiff product in one group, and the other products included Toffee Crisp (an actual defendant in litigation brought by Reese’s) as well as less similar products like Snickers, Justin’s, and Ghirardelli. When they tested a group of participants recruited from Amazon Mechanical Turk, the biases had the expected effect. Participants thought the defendant’s product (Toffee Crisp) was more similar to the plaintiff’s (Reese’s) in the “pro-plaintiff” survey than in the “pro-defendant” survey, and the “neutral” survey produced intermediate results.

To generate a more objective measure of product similarity that does not rely on participant reports, the researchers exploited an intriguing feature of human perception and cognition known as “repetition suppression.” The idea is simple: when we are presented with a stimulus that is very similar to one that we have just seen, our perceptual response to it diminishes. This is effectively a visual heuristic. Having seen something once, our brains devote less cognitive capacity to seeing it the second time. The empirical strategy, then, compares the degree of neural diminution across various stimuli to objectively measure stimuli similarity. The more similar two stimuli are, the more participants’ neural responses to the second stimulus will be diminished.

The study entailed functional magnetic resonance imaging (fMRI) scans of 26 participants who viewed the various products described above. fMRI scanning measures relative changes in brain blood levels as a proxy for neural activity. By focusing on regions of the brain known to process visual stimuli, the researchers could measure the degree of diminished neural activity associated with the Reese’s-Toffee Crisp pair compared to the Reese’s-Snickers pair.

When the researchers compared the neural similarity measures detected by fMRI to the self-reported similarity measures from the prior surveys, they found a strong correlation between the fMRI data and the “neutral” survey but no significant correlations between the fMRI data and either the “pro-plaintiff” or “pro-defendant” surveys. This suggests that the “neutral” survey is, in fact, a good proxy for participants’ actual experiences of visual similarity.

To be clear, the researchers’ methodological contribution isn’t to suggest that all trademark cases should require incredibly expensive neuroscientific studies. Rather, by using techniques like this one, scholars can develop a set of “best practices” or “gold standards” for trademark survey research. The goal is to use neuroscience to validate much cheaper behavioral surveys.

Of course, this is just the beginning. There is much that the current study doesn’t tell us. It can tell us that participants think the overall visual impression of certain trade dress is more or less similar to other trade dress. But it cannot tell us, for example, whether the participants were paying attention only to the protectible aspects of the trade dress or not. Nor do we know if some degree of similarity is consistent with consumers being confused as to source. The Mercedes logo and a “peace” sign look very similar, but people may not be confused by them. But this study seems a step in the right direction, and I’m excited to see where this kind of research will go next.

Download PDF
Cite as: Christopher J. Buccafusco, Can Neuroscience Fix Trademark Surveys?, JOTWELL (June 7, 2023) (reviewing Zhihao Zhang, Maxwell Good, Vera Kulikov, Femke van Horen, Mark Bartholomew, Andrew S. Kayser & Ming Hsu, From Scanner to Court: A Neuroscientifically Informed "Reasonable Person" Test of Trademark Infringement, 9 Sci. Advances 1 (2023)), https://ip.jotwell.com/can-neuroscience-fix-trademark-surveys/.