top of page
  • Writer's pictureDr. Cynthia Cohen

Proving Causality in Likelihood of Confusion Studies

– More than Eliminating Noise with Control Groups

Filter Market Share and Effects.

Social scientists trained in experimental design develop hypotheses about human behavior and look for causation or explanations by comparing subjects in test and control groups. In likelihood of confusion and dilution studies, control cells are necessary to confirm causality–to filter out market share and other effects.

Eliminating Noise.

“Noise” interferes with the signal that the likelihood of confusion percentage is correct. If the survey contains leading questions or there is a pattern of responses in multiple choice questions, a control group can measure that effect or noise. There is also a chance that a respondent is purely guessing or seeking a socially desirable answer. Survey-takers may name a company that is top of mind, even if uncertain whether that company is the source, or associated, or sponsors, that product/service. There could be an amount of systematic or random distortion in the group’s test responses. A good control cell assists in eliminating that pattern of bias, measurement error, or sampling error.

ACME vs. Rival Hypotheses.

When designing surveys, a control cell provides a standard of comparison to gauge the interpretation of the answers that the consumers give in the test cells. In this hypothetical, ACME alleges that the defendant's anvil is confusingly similar to its anvil and confusing consumers in the marketplace.

If 20% of the test group respondents in the likelihood of confusion survey mention ACME, the plaintiff company, when shown the defendant’s anvil, then one may suspect that there is likelihood of confusion. A competent expert parses out, “Was ACME top of mind because consumers believe ACME is the source or because of the survey process?” It is essential to measure and remove the “noise” or question-taking effects with a control group cell.

In the control cell, the subjects see another anvil that is not alleged to be infringing. If 10% of the respondents in the control group mention ACME as the source or associated with the anvil, then the net confusion rate is 20%-10%=10%. Control cells rule out rival hypotheses or competing explanations. Did mentioning ACME occur because the subjects felt the need to give an answer? Were they guessing? Or did they truly believe that ACME was the source or was associated with the anvil? 10% is a minimally accepted confusion rate by the courts. Gerald L. Ford, a pioneer consumer survey expert, addressed acceptable confusion rates in conjunction with control cells. A 55% confusion rate as seen in Union Carbide Corp. v. Eveready, Inc. is the high mark. Between 20 and 50% confusion rates generally are admitted.

Control Groups Subtract Noise from the Test Group
Calculating Causation in Likelihood of Confusion

Appropriate Control Cells.

When a test group indicates a likelihood of confusion, it is not enough that the study has a control cell. It is important that a study has an appropriate control cell. One can use multiple control cells in trade dress design, to rule out rival hypotheses. An appropriate control stimuli is close to the test control stimuli and is not an alleged infringing product. In trade dress comparisons, multiple control cells may be incorporated to extract design element percentages.

Zero Confusion.

Need for a Control Group? If a test cell reveals a zero or low likelihood of confusion percentage, a control cell’s importance in addressing causality is diminished since there is no causality to address. Courts may still favor a control group to show any questionnaire effects despite the lack of confusion revealed in the test group.

When there is zero or low confusion in the test group there is little difference to subtract.
When Control Groups Do Not Add or Subtract


36 views0 comments
bottom of page