Someone had described Pinker's latest book to me when it came out, and I foolishly chose to base my opinion off of this person's interpretation of Pinker's analysis and conclusion at first. Luckily, I happen to be working the day Pinker was giving a talk about his book, so I went down to hear him. I was glad to be wrong, and subsequently read his book.
There were, however, a few things (both in his presentation and his book) which were not exactly accurate and/or were somewhat distorted. Some were, I think, due to the simplification necessary to write a book on so broad an issue and make it non-technical. Others were more likely a result of Pinker's specialty (and the fact that he discussed much which has little or nothing to do with this specialty).
For example, he deals a lot with frequencies of violence, classified into various categories.There are currently a several international organizations which attempt to provide comprehensive reports on violence in various regions, such as EuroStat, WHO, HEUNI, etc. In the 2011 World Development Report: Conflict, Security, and Development, which alone is about half the size of Pinker's book and is only concerned with understanding modern violence, crime, etc., we can't even get to the Intro without the following: "One of the greatest challenges in researching lessons on violence prevention and recovery is the lack of available quantitative and qualitative data, due to challenges of security and access, along with low statistical capacity. Even in the World Banks comprehensive data sets, countries most affected by violence often register empty data columns."
HEUNI's chapter "Homicide" in it's Insternational Statistics on Crime and Justice (2010), like Pinker, deals with the issue of categorizing violence. Unlike Pinker, however, the concern is only with the most recent data. And homicide in particular "is perhaps the most widely collected and reported crime in law enforcement and criminal justice statistics". So we'd expect to see the least issues here. Yet instead we find "the challenges of cross-national comparability are considerable" (p. 7). Even for the EU (see EuroStat's Crime and Criminal Justice report), not only is classification a problem (different legal systems), but so too is missing data.
In other words, the leading organizations concerned documenting, classifying, and analyzing current trends, tendencies, and frequency of violence in the world continuously note the major roadblocks.
Pinker, on the other hand, is not primarily concerned with actually analyzing data (e.g., ensuring that the comparisons he makes are accurate), but taking datasets and interpreting them on a global scale throughout human history.
The most glaring example of an almost complete failure to properly understand violence trends is his treatment of genocide. Pinker discusses the "genocides" in the Homeric epics, the Bible, as well as other ancient literature. The problem here is that the destruction of Troy, or slaughter of the Hittites, and ancient warfare in general isn't genocide. At least not in the way the term is used to day. Perhaps the most glaringly obvious way to demonstrate this is etymologically: the word is a combination of the Latin for killing/murder with the Greek genos. For the Homeric epics in particular, but also for ancient Greek in general, the word meant "tribe" or "family" more than "race." Ancient warfare (like most warfare in human history) was characterized by going into some other city, or the region occupied by some other tried, and slaughtering them. Usually this included rape, enslavement, etc,. but sometimes just destruction and wholesale slaughter. Yet the only way we can classify this as genocide is by thinking that "race" means living in a particular city.
The anthropological models of genocide do tend to make comparisons with the past, but only in terms of certain common characteristics which can help inform underlying psychological mechanisms at play. Hinton's study "The Dark Side of Modernity: Toward an Anthropology of Genocide" (the introductory essay in the edited volume Annihilating Difference: The Anthropology of Genocide; vol. 3 of California Series in Public Anthropology) makes this quite clear: "Prior to the twentieth century, the concept of genocide did not exist." (p. 3). It's true that althought "the concept of genocide is a twentieth-century invention, the types of destructive behaviors it references go far back in history" but with "the advent of modernity, however, genocidal violence began to be motivated by a new constellation of factors" (p. 7).
Not that all historians or anthropologists agree. There are some who would define, as Pinker does, older massacres of cities and so forth as "genocides". But as Gellately & Kiernan note in a similar essay to the once cited above (in that it too is the introduction to an edited volume on genocide), even scholars who argue genocide is as old as war, this tendency to "underline continuities in the human condition as explaining the recurrence of mass murder" (p. 9) misses extremely important factors. The authors (who are the volume's editors) write "In this book, Omer Bartov, Marie Fleming, and Eric Weitz focus on the specific modernity of genocide. In their essays here they insist that there is something very new about many (if not all) of the twentieth-century mass murders, such as those inflicted on the Armenians or the Jews. Many of us would agree with the point made by Isabel Hull in her essay in this volume. On the basis of what happened to the Herero tribe in German South West Africa before the First World War, she argues that the vastness and totality of recent genocides or final solutions aimed at what she terms problem populations is such that they can be pursued only by an institution like the modern state." (p. 9).
Pinker cites Chalk and Jonassohn's book both in support of the idea that genocide is ancient, and to explain why people are under the (false) impression it is new. Using their work as back-up, he writes "historians have never found genocide particularly interesting." It's hard to find interesting a concept that didn't exist until recently. And one need only look at the samples of definitions of genocide from just the 80s onward in Jones' textbook (Genocide: A Comprehensive Introduction 2nd ed.), which require about 3 pages even though some are only one line long, to not that "never found interesting" can only be true if we ignore the past 30 years.
To compare the warfare in Homer to the Holocaust, the "ethnic cleansing" in Yugoslavia, the slaughter in Rowanda, and similar hallmarks of the 20th century in particular (it has not stopped and there are comparable stains in the history of humanity from e.g., colonization) is at best flawed, and at worse to trivialize these horrors.
Let's just say that Pinker did some great work trying to present a balanced view of history, but the secondary nature historiography played in his work, his lack of familiarity with the study of history and historiography, and the incredible complexities inherent in his project all resulted in some rather unfortunate distortions. Hopefully, his conclusions aren't too affected by them.