Saturday, March 1, 2014

Forensic Anthropology and Race

Looks like I'm temporarily going back to the topic of human races again. This is probably going to cause me to categorize my posts from now on, but that will come later, if at all.

This is stemming from my post on "Lewontin's Fallacy" and the topic of race as a biologically valid category for humans (or, rather, why it is not). Within the last week, I continued to receive a small number of comments, many of them branching from an ad nauseam argument regarding forensic anthropologists' ability to identify the race of an individual by either examining certain genetic clusters that correlate with self-identified race, or by examining bone fragments or other anthropometric features. I've disabled comments now, because I had already stated that I wouldn't further entertain the discussion, but I feel that a further explanation is needed to kill this line of reasoning.

To begin, the reason a forensic anthropologist can identify "race" with a high level of confidence can be narrowed down to two things: (1) a preconception of what "race" is, and (2) an informative prior.

The first should seem fairly self-explanatory, but it needs context. In order to meaningfully identify "race," a forensic anthropologist needs to be aware of the circumstances they're working under. Definitions of different races vary from society to society, thus if a forensic anthropologist were to say "this is the body of a black male" in America, it would be different than if he had said the same thing in South Africa.

This is fairly well discussed in the article "Understanding Race and Human Variation: Why Forensic Anthropologists are Good at Identifying Race" by Ousley, Jantz and Freid in 2009. The authors do concede that there are rather unambiguous morphological differences between American blacks and American whites; however, they explain why this can be taken out of context:

Forensics: Useful for practical reasons.
"Part of the reason for the disagreement between forensic and biological anthropologists has been in their different approaches and goals. Forensic anthropologists answer practical questions of age, sex, and race to construct the biological profile and narrow down possible identifications. In examining American blacks and whites, forensic anthropologists would naturally think in terms of two biological races because of the concordance between social and morphological race. Identifying social race, available in missing persons reports, would be the stopping point. Biological anthropologists would explore within-group variation further. These findings illustrate the essential difference between a forensic analysis and a biological analysis: forensic analysis produces practical information useful for forensic identification, while a biological analysis provides insight about relationships among arbitrarily defined populations, which may be defined by social races, breeding populations, language, nationality, time periods, and other criteria." (pp. 73)

These comments serve to echo and confirm the findings of Sauer in 1992 in a paper entitled "Forensic Anthropology and the Concept of Race: If Races Don't Exist, Why are Forensic Anthropologists so Good at Identifying Them?" Humorously, I would like to note that the only mentioning you'll find of this study on Metapedia (or any study featured in the special edition of the American Journal of Physical Anthropology Race Reconciled) is that the name of this study shows how forensic identification has "bewildered race deniers." How selectively vague and uninformative, right? But enough of mocking "race realists" (lest I break my rule).

So, this is what it means to have a preconception of what race is. It means that a forensic anthropologist can both disagree with the biological concept of race, but also work very well within the confines of his/her work to identify someone's social race, and remain completely consistent. But, there remains a question to answer: if the biological concept of race is inaccurate, how can a forensic anthropologist find such high concordance rates between self-identified race and morphological race? Surely that means the traditional racial categories we use are useful, right?

They are meaningful, for identification purposes, and that only works within a given society. This goes into what I alluded to earlier: an informative prior. This one is a bit less obvious to work with, so I'll defer to a study also featured in Race Reconciled which explains this function of forensic anthropology: a 2009 study by Konigsberg et al. entitled "Estimation and Evidence in Forensic Anthropology: Sex and Race." Not only was this study an easy read, but it comprehensively explains what so many people don't understand.

This study discusses the identification of one "Mr. Johnson" from Iowa. Using craniometric features and matching them to the samples in FORDISC 2.0 and Howells' sample using linear and quadratic discriminant function analysis, forensic anthropologists were able to identify the gender and race of Mr. Johnson rather unambiguously. The posterior probability of the body being male ranged from 0.996 to 0.999 -- noting this, I would like to make it clear that it is not easier to identify race than it is to identify gender, unlike what some might suggest. This part was rather uncomplicated by comparison, however, when we get into the identification of race, my talk of an informative prior becomes relevant.

By matching to the sample data, the typicality of Mr. Johnson relative to the Easter Islander sample was 0.8415, indicating a high probability that Mr. Johnson could be correctly identified, if he were alive, as an Easter Islander. This is, however, using a uniform prior, i.e. no informative information about the demographics of the area Mr. Johnson was found in. Given that Mr. Johnson was found in eastern Iowa, the probability of him being an Easter Islander significantly decreases -- that's just a matter of knowing the demographics of Iowa. To continue, what does this do to the typicality of Mr. Johnson's craniometric features?

By using the 2000 US census for Iowa as an informative prior, the highest posterior probability for Mr. Johnson is for "American white" at 0.6976, whereas the probability of him being an Easter Islander is now at 0.0449. This is an incredibly significant change considering the posterior probability we had using the uninformative prior. How do these probabilities change depending on the prior we use?

Mr. Johnson's race somewhat 'transforms' from state to state.
By using the 2000 US census data for Hawaii, the posterior probability for Mr. Johnson being an Easter Islander is now 0.9068, while the probability for being an American white is 0.0188. Using data for Gary, Indiana, the probability of Mr. Johnson being American white is 0.2728, but the posterior probability of him being American black is the highest at 0.5342. To summarize, the authors state in the discussion:

"From an evidentiary standpoint, sex of the individual is only of much use if the sex ratio is very biased for the population at large and the identification is for an individual of the lesser represented sex. Regarding the use of race in estimation and evidentiary problems, this, like many problems in forensic anthropology is very context specific. We have shown that prior information can be very important both in estimation and evidentiary problems. But in estimation problems, we are ultimately faced with posterior probabilities which we must decide how how to implement." (pp. 84)

Let me stress the significance of those words. Using race in forensic anthropology is largely reliant on probabilities, and even then, it requires context.

This is why I get somewhat tired and irritated when, while debating about race, a sentence starts with "a forensic anthropologist can-" or even mentions forensic anthropology at all. Within the field of anthropology, there exists "social race" and "biological race." Forensic anthropologists more often than not are using morphological features to identify social race, that is, the race the individual would've been identified as had they been alive. This is why informative priors can change the probability estimates -- because an individual's social race is quite fluid; therefore, it does not justify race as a biological concept.

Thank you for reading.



Follow me on social media!

Twitter: https://twitter.com/AlexisDelanoir
Google+: https://plus.google.com/+AlexisDelanoir0/
YouTube: https://www.youtube.com/AlexisDelanoir



References:

ResearchBlogging.org Konigsberg, L., Algee-Hewitt, B., & Steadman, D. (2009). Estimation and evidence in forensic anthropology: Sex and race. American Journal of Physical Anthropology, 139 (1), 77-90 DOI: 10.1002/ajpa.20934

Ousley, S., Jantz, R., & Freid, D. (2009). Understanding race and human variation: Why forensic anthropologists are good at identifying race. American Journal of Physical Anthropology, 139 (1), 68-76 DOI: 10.1002/ajpa.21006  

Sauer, N. (1992). Forensic anthropology and the concept of race: If races don't exist, why are forensic anthropologists so good at identifying them? Social Science & Medicine, 34 (2), 107-111 DOI: 10.1016/0277-9536(92)90086-6

13 comments:

  1. "Using race in forensic anthropology is largely reliant on probabilities, and even then, it requires context."

    I think you mean that context provides the basis for the probabilities.

    Rare populations (Polynesians, Australian aborigines, Khoi, etc.) are even more rarely found outside of their historic homelands. So when presented with indirect forensic evidence that a member of one of those groups might have died in an area not known for those populations, you have to immediately discount its likelihood.

    That's not the same thing as saying forensics doesn't deal with real biological differences when assessing race.

    "Forensic anthropologists more often than not are using morphological features to identify social race, that is, the race the individual would've been identified as had they been alive."

    But that individual is dead and unidentified, and so his "social" race is a moot point.

    A forensic anthropologist ought to look only at the evidence and the context, and he ought to interpret the latter with care. Mr Johnson in Iowa, for example, is still looking very Polynesian, even after running through the priors. Why else would he demonstrate such a high probability (4.5%) for being an "Easter Islander" in a state whose 2000 census shows fewer than 7 NHPI for every 10,000 Iowans?

    So a good forensic anthropologist shouldn't look at those numbers and say, "Gee, this individual is native Hawaiian in Hawaii, but white in Iowa, and therefore there is no biological basis for race."

    He ought to look at them and say, "You know, there's a good possibility this guy is white because races are fuzzy sets and the underlying scientific data is insufficient to provide more clarity in this context, but I'd better alert law enforcement to be on the lookout for any reports of missing Pacific Islanders at the local colleges."

    ReplyDelete
    Replies
    1. "I think you mean that context provides the basis for the probabilities."

      I mean that in order to give a probability that is accurate and serves its purpose, it requires a context.

      "Rare populations (Polynesians, Australian aborigines, Khoi, etc.) are even more rarely found outside of their historic homelands. So when presented with indirect forensic evidence that a member of one of those groups might have died in an area not known for those populations, you have to immediately discount its likelihood."

      I didn't disagree with this.

      "That's not the same thing as saying forensics doesn't deal with real biological differences when assessing race."

      I also didn't disagree with this; I said that there exist morphological differences between people, but how these people are classified depends on the social construction of race that the forensic anthropologist is operating under.

      "But that individual is dead and unidentified, and so his "social" race is a moot point."

      A forensic anthropologist's job is to identify a body/skeleton. That hardly makes social race a moot point. If you were referring to physical anthropologists, your argument would have been more accurate.

      "A forensic anthropologist ought to look only at the evidence and the context, and he ought to interpret the latter with care. Mr Johnson in Iowa, for example, is still looking very Polynesian, even after running through the priors. Why else would he demonstrate such a high probability (4.5%) for being an "Easter Islander" in a state whose 2000 census shows fewer than 7 NHPI for every 10,000 Iowans?"

      Because Mr. Johnson's typicality represented Easter Islander at a higher rate than an even rarer group -- it's a posterior probability, not a uniform probability. Table 9 on page 82 shows precisely what I mean: in Iowa, Mr. Johnson's remains were more likened to an Easter Islander than, by comparison, a Mokapu or an Ainu. I can't even begin to understand why you would consider a posterior probability of 4.5% to be high.

      "So a good forensic anthropologist shouldn't look at those numbers and say, "Gee, this individual is native Hawaiian in Hawaii, but white in Iowa, and therefore there is no biological basis for race.' He ought to look at them and say, 'You know, there's a good possibility this guy is white because races are fuzzy sets and the underlying scientific data is insufficient to provide more clarity in this context, but I'd better alert law enforcement to be on the lookout for any reports of missing Pacific Islanders at the local colleges.'"

      That isn't anywhere near what I was suggesting. Nowhere did I suggest that a forensic anthropologist's need of an informative prior is proof that there is no biological basis for race -- you're arguing an obverse position. What I suggested is that a forensic anthropologist's ability to identify "race" is not proof that there IS a biological basis for race.

      What a forensic anthropologist ought to do is test the priors and get a posterior probability and then, in the example of Indiana, say "this body most likely belongs to an American black male, approximately X' Y'' and Z years old; therefore, he seems to match the description of the person who went missing 2 weeks ago in Gary."

      Delete
  2. "A forensic anthropologist's job is to identify a body/skeleton. That hardly makes social race a moot point. If you were referring to physical anthropologists, your argument would have been more accurate."

    So long as the forensic anthropologist is dealing only with a body and not with other types of evidence found at the scene, I don't see how he can make a determination about "social" race.

    I met a black guy recently who claimed to be Jewish. No, he didn't claim to be a black Jew. He claimed to be a Jew and not black. That was how he identified socially. He didn't want to be identified as black.

    And in American society, we generally respect those individual determinations. Tiger Woods can be "Cablisian." Halle Berry can say her baby daughter is black, even though the father is white and Ms Berry herself almost certainly has more than a little European blood in her.

    But those social determinations shouldn't matter to the forensic anthropologist. He's looking at physical evidence that has nothing (or little) to do with race as it's used socially.

    Does he have to look at geographical context? Yes, of course. But social race? No, he doesn't care.

    "Because Mr. Johnson's typicality represented Easter Islander at a higher rate than an even rarer group -- it's a posterior probability, not a uniform probability. Table 9 on page 82 shows precisely what I mean: in Iowa, Mr. Johnson's remains were more likened to an Easter Islander than, by comparison, a Mokapu or an Ainu. I can't even begin to understand why you would consider a posterior probability of 4.5% to be high."

    It's high because Mr Johnson almost certainly doesn't look like the typical Iowa white boy.

    In a state with a white population over 93 percent in 2000, he manages only 70 percent probability of being white? Something's off. Even if he is verifiably white, he may not look it. Or perhaps he's hapa.

    And remember those priors are susceptible to change at a moment's notice. The science underlying the morphological determinations is not subject to such easy change - although one would hope the science improves over time, as most sciences do.

    For example, what if our fearless forensic anthropologist is told that two visiting Chilean students at a local university went missing recently, and that the background for those students is that they were originally from Hanga Roa. You don't want him to revisit his priors and work out new probabilities?

    That's the problem with relying too much on Bayesian probabilities as opposed to the underlying science. Your priors can change at a moment's notice. Hopefully, that's not true of the anthropologist's forensics work.

    ReplyDelete
    Replies
    1. He does have to look at social race, because that's synonymous with self-identified race. Anecdotal evidence of someone who would, under normal circumstances, be classified as "Black" in America but declared himself Jewish is an exception to the rule. Self-identified race has a very high correspondence with morphological race, and that is why forensic anthropology works the way it does. I've discussed this already.

      Just as well, using examples of people with European admixture doesn't counter anything I've said. Under normal circumstances, a discriminant function analysis would be used, but in such cases of overlapping categories (admixture or overlapping races), finite mixture analysis would be the alternative (pp. 78).

      "It's high because Mr Johnson almost certainly doesn't look like the typical Iowa white boy."

      It's high because it's a posterior probability and races are, even as you admitted, fuzzy categories.

      "And remember those priors are susceptible to change at a moment's notice. The science underlying the morphological determinations is not subject to such easy change - although one would hope the science improves over time, as most sciences do."

      Again, you're confusing the work of a forensic anthropologist with that of a physical anthropologist. I will repeat this ad nauseam until it sinks in: a forensic anthropologist will use morphological features to identify a corpse or a skeleton or what have you based on what that person would have been identified as (most likely) had they been alive. They look not to make any determinations of that person's true biological ancestry (although that plays a role, and nobody denies this), but for practical identification purposes.

      "For example, what if our fearless forensic anthropologist is told that two visiting Chilean students at a local university went missing recently, and that the background for those students is that they were originally from Hanga Roa. You don't want him to revisit his priors and work out new probabilities?"

      Again, exceptions do not disprove the rule. In the example of Mr. Johnson, it would not be "an Easter Islander visiting from Hawaii came to Iowa recently and has been found dead -- you must now swap your priors." It would, instead, be: "we've found a body and we want to see if it matches the identifications found in any missing person's reports from Iowa."

      You're trying to force an example of using uniform probabilities into this situation, but practically and realistically that's not what we're looking at.

      Delete
  3. You missed Edwards's main point, which is that genetic differences between human populations mostly reflect correlations between small differences in numerous allele frequencies. This is the reason why you can take as few as 100 random genetic polymorphisms and reliably classify people into races or other groups. Classification is not dependent on the existence of large allele frequency differences. Correlations between small frequency differences suffice just fine.

    Why is this important? It's because most trait variation in humans is influenced by a large number of small-effect genetic loci. Thus the probability of there not being genetic differences between races (however you choose to define them) in any given complex trait is low. Each race occupies a different location in multidimensional genetic space, reflecting their different evolutionary histories.

    Taxonomy is not fundamentally based on shared phenotypes, such as skin color or skull shape or whatever in the case of humans. It's based on phylogenetics, i.e., actual shared ancestry. Visible markers of such ancestry may be used to infer taxonomic relations, but the actual basis of classification is shared ancestry. Therefore it's a great mistake to think that because the distributions of phenotypic traits cut across traditional racial lines, any classification based on any trait is valid. It's like saying that whales could be grouped with the fish rather than land mammals because they look and behave the same.

    Taxonomy is about the propinquity of ancestry, not about genetic or phenotypic differences, although all these of course usually go together to some extent. As it happens, the racial categories of traditional anthropology show a nice match to DNA-based phylogeny, which is not surprising considering that racial classification has always been based on geographic ancestry.

    ReplyDelete
    Replies
    1. I couldn't tell if this was a comment addressing this post or my other post on race. If it's the latter, I'd advise you and other commenters to not shoehorn further arguments, as I've concluded them due to a lack of any productive discussion. If your comment is addressing this post, I'd advise making a comment that addresses the arguments made. If it's a combination of both, I'd further advise you to read the two points above.

      This post doesn't say that race can't be distinguished, but that forensic anthropology is a bad argument to support it if you believe it can be. Nevertheless, I'll reply.

      I didn't miss Edwards's point -- I cited a more recent argument of the same line of thinking (Rosenberg et al., 2002), which shows that genetic clusters correlate with geography. The problem with the conclusions that people draw from these studies (such as the conclusion you seem to draw), is that they're misinterpreting the results. I'll echo the words of Michael Muehlenbeing in "Human Evolutionary Biology," pages 270-271:

      "Perhaps the most celebrated confusion of geographic difference for race followed the publication of Genestic Structure of Human Populations (Rosenberg et al., 2002). The authors studied genetic variation in 1052 people from 52 populations and then asked a computer program called Structure to group the samples. When they asked it to produce two groups, Structure gave them EurAfrica and East Asia-Oceiania-America. When asked for three groups, Structure gave Europe, Africa, and East Asia-Oceania-America. When asked for four, it gave Europe, Africa, East Asia-Oceania, and America. When asked for five, it gave roughly the continents. And when asked for six, it gave the continents and the Kalash people of Pakistan. When asked for more (up to twenty groups), it gave more (Bolnick, 2008).

      This was more or less what population geneticists had been doing with the human gene pool since the pioneering work of Cavalli-Sforza and Edwards (1965). On the face of it, once again, this would seem to have little relevance for race. The user specifies the number of groups, and geographic proximity is the strongest predictor of similarity, so asking the computer to break the human species into five gropus might reasonably be expected to yield groups roughly corresponding to the continents."

      Which begs the question: how many human races are there?

      This is why, in my opinion, using "race" as the baseline for studying human variation is problematic and relatively uninformative. As I suggested to one individual in the past, if we accept that such groups are "races," then we should accept that the Amish are their own race or that there are dozens of races in India due to endogamy.

      While we can find differences between groups genetically, the way we partition these groups is arbitrary and should remain context-specific. We shouldn't treat human variation as being divided among easily distinguished groups, because this is far from correct from a practical viewpoint. To understand why I feel this way, one should read a study by Long et al. in 2009 entitled "Human DNA Sequences: More Variation and Less Race." It highlights how Sub-Saharan Africa has the greatest genetic diversity out of anywhere in the world, and that "a classification that takes into account evolutionary relationships and the nested pattern of diversity would require that Sub-Saharan Africans are not a race because the most exclusive group that includes all Sub-Saharan African populations also includes every non-Sub-Saharan African population" (pp. 32). Ultimately, they agree with Lewontin's conclusions about race, but from an evolutionary perspective.

      I'd recommend you read the special 2009 issue of the American Journal of Physical Anthropology: "Race reconciled?: How biological anthropologists view human variation." It's very informative.

      Delete
    2. "Classification is not dependent on the existence of large allele frequency differences."

      So if there's any genetic structure within or between populations they can be classified separately? That's just awful, both scientifically and pragmatically!

      "Taxonomy is not fundamentally based on shared phenotypes, such as skin color or skull shape or whatever in the case of humans. It's based on phylogenetics, i.e., actual shared ancestry."

      Taxonomy is based on observable evolutionary differences between groups. Phylogeny plays a part in this. So do cladistics and systematics. The point is this: it's complex, and your methods alone aren't sufficient in displaying the biological validity of race according to modern systems of classification.

      "As it happens, the racial categories of traditional anthropology show a nice match to DNA-based phylogeny, which is not surprising considering that racial classification has always been based on geographic ancestry."

      Which is incorrect because then we're assuming that human races have arisen as a result of isolation by distance, that these races have been isolated throughout history. This isn't true.

      Delete
  4. Well you can get black and white twins. Mixed people can be white, black or anything in between. Thats the ultimate combination possible no? Twins?

    https://www.google.co.za/search?q=black+and+white+twins&source=lnms&tbm=isch&sa=X&ei=3uM2U-jjOq6M7Ab_14DABQ&sqi=2&ved=0CAYQ_AUoAQ&biw=1366&bih=667

    I think its just shallowness from the race crowd that causes all of this.

    ReplyDelete
    Replies
    1. I'm not exactly sure what you're getting at, but I can agree on the shallowness part. A lot of issues that come from racial classification are due to superficial values in what percentage of admixture a person can have in them before being called a "mutt."

      Delete
    2. Sorry I forgot to mention that my comment was not really too related to this specific article. It was meant for the Lewontin's Fallacy one but comments have been disabled there(not complaining).

      I wanted to point out that even twins can be considered different races.

      Delete
    3. Oh, in that case I'm going to have to play the Devil's Advocate and say that exceptions don't disprove the rule. At the same time, I'd ask you to not try forcing in more discussion on my Lewontin's Fallacy and Race post. I closed the comments for a reason and thus I can't treat you differently from my detractors.

      Delete
    4. "Forensic anthropologists more often than not are using morphological features to identify social race, that is, the race the individual would've been identified as had they been alive."

      I read this against the backdrop of linguistic differences in a multilingual country that's otherwise ethnically uniform; all the population appear to have more or less similar physical features but due to the linguistic differences and therefore the cultural differences, they might sport different types of clothes, hang around different regions of the country and so on. The perceived differences would make mr Johnson racially distinct were he alive but once deceased there'd be no way to tell who he had been.
      East African countries such as Sudan, Ethiopia, Eritrea and even the western parts of the Arabian peninsula would have such examples of perceived ethnicities.
      Just an observation from having worked in genomics sequencing of samples out of there.

      Delete
    5. Excellent point, bio! In that scenario, a forensic anthropologist's job would probably be pretty difficult; however, I wonder if these perceived ethnic groups based on superficial/temporary characteristics alone has resulted in some kind of isolation, which could result in modest morphological differences like a difference in forehead shape or nose width. I'm thinking of even fewer differences than between a Hutu and Tutsi.

      I guess for argument's sake the scope of this post should be limited to "Western" usage of forensic anthropology, since our methods of racial classification tend to excuse themselves as looking at (at least what we think are) more static characteristics.

      Delete

WARNING: Please read the Comment Guidelines page before posting!

Sometimes comments won't go through properly, so if you write a lot and are concerned about losing your work, please save your comment in a separate text document before posting. Keep it saved until you're sure your comment has been received/published.