DNA and Negative Evidence

In today's blog post at http://www.legalgenealogist.com/2016/12/21/the-dog-that-did-bark/, Judy Russell corrected her response to a question posed at the end of her BCG webinar on negative evidence last night. As she writes, "The question was whether DNA results could be negative evidence, and, in my answer, the example I used of a case that could be negative evidence… isn’t," and she goes on to explain why.

I'm wondering under what circumstances DNA evidence might in fact be negative evidence. Would a non-paternity event where there evidence of "a break in the link between an hereditary surname and the Y-chromosome resulting in a son using a different surname from that of his biological father" (from http://isogg.wiki/Non-paternity_event) be an example of negative evidence? Or is this simply direct evidence?

Any other examples?


Submitted byEEon Thu, 12/22/2016 - 08:29

What a great question, Lesley!  The "problem" here is that complex uses of DNA evidence to solve problems of historical identity and kinship are so new that we don't have a convenient body of literature. To my knowledge, no one has pulled together a bibliography of peer-reviewed case studies, analyzing  the specific techniques used or types of evidence engaged, so that a question like that can be answered on the fly. I'm headed out for a bit of business elsewhere but will be mulling this while I'm gone--not only your for-if but also sifting through past cases of my own to see what might be the best examples. Meanwhile, I'll post an alert on our FB page and see what other answer we might get.

Submitted byyhoitinkon Thu, 12/22/2016 - 12:45

Interesting question. Take this example:

To prove a theory, you want to know if two people are third cousins. Both take an autosomal DNA test. Third cousins have a 90-98% chance that they would share detectable amounts of DNA. The test comes back, and they have no shared segments. 

I think most people will agree that this DNA result provides evidence that the theory is likely to be incorrect. You would still need to use other evidence to come to a conclusion, but by itself it suggests that the answer to the research question whether these two people are third cousins is "probably not.

But what type of evidence is this?

  • Is it direct, because it directly addresses the question? Direct evidence does not need to be specific. If the question is: "when is A born?" then his age at a given time is direct evidence even though it does not specify the birth date. I don't think the absence of any shared segments qualifies as direct evidence, because you have to use knowledge about statistics to arrive at the answer "probably not." 
  • Is it indirect, because you need to combine it with other evidence (knowledge of statistics) to answer the question and need to have other evidence to arrive at a conclusion? I think this could be the case, but since you use the absence of segments rather than the presence of segments, I don't think it is indirect evidence.
  • Is it negative evidence, because you find a possible answer based on the absence of information? I think that is the answer here, since the absence of matching segments combined with knowledge of cousin statistics provides the answer "probably not."

What do you think?

Submitted byEEon Thu, 12/22/2016 - 22:37

Lesley, about noon-time, I started out to answer your question, thinking I could do it in fairly simple terms. But then I kept thinking of things I needed to add. Before long, it morphed into a QuickLesson. https://www.evidenceexplained.com/content/quicklesson-24-evaluating-dna-evidence. 

I appreciate that serious thought needed to make sense of how DNA results fit into evidence analysis. The literature in this area is still very thin. The link to your new QuickLesson doesn't appear to be working, Elizabeth, but I look forward to reading the lesson when it does.

Thanks for your response, too, Yvette. I never studied statistics and have a way to go in understanding the science behind DNA evidence. GEDmatch offers a number of helpful tools that aren't found on FTDNA, etc.--for example, the ability to lower the cM threshholds (segment and total) in order to expand one's list of matches. In some cases, doing that reveals DNA evidence that apparently could be used to make a case for kinship or identity. But how reliable is that data if I lower the segment threshold from 10 cM to 9, 8, or even 7 when looking for matches to both of two kits? What if I lower the threshold for One-to-Many matches from 7 cM to 5 cM? I clearly need to learn more. Just ordered Genetic Genealogy in Practice.

I'm also wondering what difference it makes to a proof argument that relies on both DNA and historical evidence to have evidence from two DNA matches as opposed to one. Would it be a stronger case? Would it be stronger if the two matches lie of collateral lines instead of the same line to the MRCA?

Clearly, lots of questions, including, "When Is Enough Enough?"



Submitted byEEon Fri, 12/23/2016 - 17:01

For certain, Lesley, the literature on the subject is thin! 

Re the link, I'm puzzled. The link is https://www.evidenceexplained.com/content/quicklesson-24-evaluating-dna-evidence — if you want to try clicking it here. It's the same URL I posted this morning on the EE's FB page, which has been used by a thousand or so folks today. I can't explain why it didn't work.

As for your question about working with DNA test results and third-party tools, that's a field of weeds EE would be better off not wading into! There are quite a few forums that dive deeply into genetic genealogy and could give you more robust guidance.