.

Measuring Confusion For Intellectual Property Disputes

Marc Green


Click For A pdf "Human Factors In Intellectual Property Law Disputes", presented to the Toronto Intellectual Property Group, April 2007.

The criterion for trademark infringement is possibility of confusion. One of the best ways to prove this possibility is to perform a test which demonstrates actual confusion. In some cases, courts may even reject an infringement claim simply on the grounds that the plaintiff failed to show any evidence of actual confusion. In other cases, failure of either side to produce empirical evidence to support/refute confusion has been looked upon with suspicion. The court may infer that failure to produce evidence indicates that the litigant had, in fact, performed a study but that the results were not favorable.

Devising a scientifically-credible confusion test, however, is more difficult than might first appear. In this article, I briefly describe and critique the four major methods which are typically used.

I should first clarify some terminology. The person running the test is the "researcher," and the person responding takes the role of "consumer." (In order to avoid influencing the consumer, accidentally or on purpose, it is critical that the researcher not know the reason for the test.) In most cases, the dispute is between a well-established company with the "senior brand" and a new company or product, the "junior brand." The most typical contention is that consumers are confusing the junior brand's mark with the senior one. There are issues in intellectual property law, such as determining whether a term is generic, that are outside the scope of this discussion.

Free-Association

The first method resembles the technique of free association often used in psychology and psychiatry. The consumer is shown the junior mark and asked to say what comes immediately to mind. After response, the researcher then may ask a subsequent question that inquires into the reason for the response. (Courts virtually make this a requirement.) The percentage of the consumers giving the senior brand as the association is the measure of confusion.

This method has several drawbacks:

1. The most obvious problem is that association does not imply confusion. If I say "black" and you free-associate "white," this does not mean that you confuse black and white. Words and concepts may be associated for many reasons, not just similarity or confusion. However, a properly phrased follow-up question may help determine whether the association was based on similarity/confusion.

2. The value of the follow-up question, however, is uncertain. Free-association is largely an automatic response made without conscious thought. If the researcher asks for an explanation, the consumer is likely to provide some rationale which is manufactured post hoc to explain the answer.

3. There is a strong possibility of "demand characteristics." This is a term from psychology research, where it has been found that the method of testing inadvertently communicates the response answer to the consumer.

4. The method cannot find confusion in the case of visual identity. That is, if the marks are so similar that the consumer cannot distinguish them. The consumer believes him/herself to be viewing the senior mark and therefore does not give its name as the association.

Directed-Association

The second method resembles the first, except that consumers are asked for a specific type of association, rather than a free-association. Most commonly, the researcher shows the consumer the junior mark and then asks for the name of the company which puts out, authorizes, etc. the product. The percentage of consumers naming the senior brand is the measure of confusion.

This method solves one problem of the free-association method because the basis for the response is presumably known - it is based on a specific type of association. Moreover, visual identity is no longer a problem if the consumer is familiar with the senior mark; misidentifying the junior mark as being to the senior company clearly indicates confusion.

However, this method too has drawbacks. If the consumer is not familiar with the senior mark, then confusion rates will be low. This highlights the importance of using the appropriate population. To test confusion among brands of denture cream, don't use young consumers because they are likely unfamiliar with the companies which produce it.

There is also the problem that in asking the question, the researcher may induce the consumer to consider differences which would otherwise have gone unnoticed. For example, when buying familiar items, a consumer will probably reach for the product based on simple cues such as color or shelf location and will not look at it directly, read the label completely or pay much attention. By showing a junior mark and asking for the name of the company that puts it out, the researcher invites the consumer to examine the product closely and to read the writing. The consumer is then likely to notice specific identifying features that would otherwise have been missed. The result is a gross underestimation of confusion in the natural store setting.

Another problem lies in the issue of "base rates," the likelihood that the consumer would name the senior mark, even in the absence of the junior mark, or any other information. Let me clarify with an example. When I lecture about measuring confusion, I often start by holding up an unmarked floppy disk and asking the class, "What company made this?" Invariably, the overwhelming response is "Microsoft," usually at a rate over 80%. They say this because the a priori probability of any given piece of software being made by Microsoft is very high. Hence, Microsoft is a good guess. It did not occur because the 80+% of the class confused the generic disk in my hand with a disk from Microsoft. They would have made the same response if I had merely asked them to imagine that I was holding a floppy disk in my hand. In absence of specific knowledge, and feeling compelled to guess, they simply gave the most plausible answer. In order to correctly measure confusion, there must be a "control procedure" to assess the effects of base rate responding.

Same-Different

In this technique, the researcher simultaneously shows the consumer both the junior and the senior mark and asks whether they are from the same or different sources. At face value, this method may seem the most direct method for determining confusion, but it has numerous pitfalls.

There is the problem, described above, that the consumer is likely to scrutinize the marks more closely than s/he would in a normal situation. However, the effect is strengthened by the simultaneous viewing of both marks. People are far more acute at comparing two images which are viewed simultaneously than when one is present and the other must be recalled. Memory is "low resolution" and people are relatively poor at recalling shapes, shades of color, similar sounding names, etc. As a result, differences are likely to be even more obvious and measured confusion correspondingly lower.

On the other hand, the consumer might notice a similarity which would not have been noticed under normal circumstances. The degree of confusion would then be overestimated. I've seen people make associations, which never would have occurred until the question brought them to their attention.

Lastly, if the consumer says that the products come from the same source, it is possible that s/he is thinking of the junior brand, not the senior brand (or possibly a third party.) This is great for the senior mark because degree of confusion is overestimated, unless the researcher takes steps to have the consumer identify the source by name.

Behavioral Experiment

The final technique is not a survey method, but more resembles an actual psychology experiment. It tests the consumer in a real or mock purchasing task. By hiding the reason for the test until after the purchase, it avoids biasing the consumer or drawing attention to similarities or differences.

Here's how it might work: suppose the junior mark is "Fred's Beer" and the senior mark is "Fern's Beer." A researcher unobtrusively stands near a cash register in the beer store and waits for someone to buy "Fred's Beer." The researcher then follows the consumer outside and asks for the brand of beer that was just bought. (It's important not to let the consumer look in the shopping bag to check.) If the consumer says "Fern's Beer," then there was confusion; s/he bought Fred's Beer thinking that it was Fern's beer. Although it does not explicitly deal with mark confusion, it is powerful evidence because it occurred in a natural setting with no attempt to direct the consumer. It also demonstrates an economics loss for the senior brand and provides a measure of its magnitude.

The downside to the experiment method is the greater expensive and longer time to perform than a survey. Moreover, although the example may seem simple, experimental design is a complex task which requires much experience and knowledge. It is easy to induce inadvertently (or purposely!) bias the data or to create experiments which have alternative interpretations. One common problem, as with base rates, is that many experimenters fail to include the proper controls procedures necessary to make a valid statistical analysis. The beer example described above, for example, is by itself insufficient because there is no control. In practice, the study would have to measure other variables such as the number of times that Fern's beer is correctly purchased.