By Emilie Reas
A recent article from PLOS ONE, The Influence of Partner-Specific Memory Associations on Picture Naming: A Failure to Replicate Horton (2007), by Sarah Brown-Schmidt and Sid Horton, reports an unsuccessful replication of their prior research.
This paper has received considerable attention (including 1,576 views and 123 tweets since its October 3 publication), as the authors have been commended for their willingness to openly share their failed replication.
In an earlier study, coauthor Horton reported that the presence an individual who was associated with a previously learned object increased the speed at which the object was named. In other words, the partner’s presence served as an associative memory cue to enhance lexical processing. In a follow-up paper published this month in PLOS ONE, the researchers aimed to replicate these findings as a foundation upon which to further explore the mechanisms by which associative cuing facilitates naming. But to their surprise, a series of experiments modeled after the originals failed to replicate their prior results – the presence of the partner from the learning phase did not influence the speed of object naming.
First author Brown-Schmidt proposes several possible reasons for these inconsistencies:
“The fact that our direct replication was run at 99% power suggests that if the effect were real, it is substantially smaller than originally estimated. Other possibilities include that the original finding represents type-1 error, or that the original finding is of limited generalizability.”
Considering our current scientific environment, in which the most novel, positive findings are lauded, many researchers might hesitate to report a failed self-replication for fear of interfering with their research trajectory or compromising their reputation. However, Brown-Schmidt and Horton (2014) has served as an exemplar of transparency in scientific reporting, and the authors’ open sharing of their null findings has been received with overwhelming positivity from the scientific community.
From Dr. Horton’s perspective, “I decided that I would rather be part of the publication process than simply reacting to. For my part, the reaction has generally been one of empathy mixed with pats on the back. I think the public reaction is generally positive, especially in light of the growing appreciation of the potential importance of replications to certain areas of science.” Brown-Schmidt adds, “I had no hesitations about publishing a failed replication; in this case science is best served by having these results in the literature.”
Beyond eliciting public praise, their surprising results have also put them one step closer to better understanding the influence of associative memory on language fluency. Brown-Schmidt explains, “The partners had no real relation to the task or to the words and pictures used. This is quite unlike our daily experience, in which the people with whom we interact are usually directly related to what we are talking about or where we are at the time. An unanswered question is whether a different result might have obtained if partners as individuals were more integral to the task. Moving forward, understanding the ways in which memory guides language use is a key goal for future research.”
Brown-Schmidt S, Horton WS (2014) The Influence of Partner-Specific Memory Associations on Picture Naming: A Failure to Replicate Horton (2007). PLoS ONE 9(10): e109035. doi:10.1371/journal.pone.0109035
What are your thoughts on reporting null findings and replications? Would you publish a failed self-replication? Share your comments on this study below.