In NVVK Info issue July 2015 my review of Hollnagel's latest book "Safety I And II" was published. Read my rather critical take on the book (not on the subject!) here.


Or check the English translation below, of course:


Safety-I And Safety-II: The Past And Future Of Safety Management, by Erik Hollnagel

It’s a strange situation to be totally in favour of an idea (in this case Safety-II/creating safety/learning from success) while being extremely critical and rather unconvinced about a book on the subject - at least after the first few chapters, and only slightly more positive after finishing…

I am a major fan of Hollnagel’s work, and I love especially his “Barriers And Accident Prevention” and “The ETTO-Principle”, but I was struggling seriously with his latest book “Safety-I And Safety-II: The Past And Future Of Safety Management”. So I’m now sitting with mixed feeling about what (at least potentially) could be the most important book in Safety of this year.

What is Safety-II?

Let’s first see what it’s all about. Summing up quickly, Safety-I is the traditional way of safety management, taking accidents/incidents as the point of departure and trying to prevent bad things from happening. Safety-II is taking another point of view, focusing on things that go right and ensuring that as much possible goes right. To quote Andy Evans: “Safety-I is minimising the bad and Safety-II is maximising the good”. Important is that Safety-II sees variation in human performance as an essential factor for success because reaction to variability is necessary, while Safety-I rather tends to restrict variability and focuses on error, or what is perceived as error.

Ideas like this are not necessarily new, and indeed we find a synthesis of ideas and practices of things that have been around for a while. The reader may recall things discussed in the Resilience Movement (of which Hollnagel is an important thinker, of course), Dekker and Perrow are other names that pop up. And of course Deming (interestingly missing from the references in this book) discussed variability quite extensively. And Hollnagel himself has written quite a bit. In a way it’s a logical continuation from ETTO. After some (white) papers and articles now there’s a book on Safety-I and II.

Selling in

In reflection I must conclude that the various papers do a better job of ‘selling in’ Safety-II than this book. This may be a personal thing because I didn’t need any particular convincing that it would be wise to have something else in addition to the traditional (Safety-I) approach. A large part of the book is devoted to telling about the shortcomings of Safety-I. This negative approach to Safety-I doesn’t convince me and thereby actually harms the arguments in favour of Safety-II. Especially because I don’t find all arguments very convincing, like stating that: “Risk management is on the whole reactive rather than proactive because it is a response to risks that have been found through a risk analysis or in the aftermath of an accident”. Come again?

Another problem I have with some of the passages in the book is that some practices, by some people ‘within Safety-I’ are generalized as being practices of Risk Management even though I can tell from personal experience that this generalization is not correct. At times there is a total lack of nuance. It's not black, it's not white, it's shades of grey. Yes, Safety-I does have shortcomings and yes some applications are below par but that doesn’t mean we should throw out the baby with the bathwater.

So, as far as I’m concerned the first half of the book could be shortened down seriously, e.g. without going through pyramids (of which Hollnagel does a better job than Manuele, but that wasn’t really a challenge, of course) and stuff like that. Why so much space and ink is devoted to this? We would have to ask Hollnagel. Probably he feels that he has to build up the argument. And isn’t it customary to present something ‘new’ by telling how bad the stuff that came before actually was? But again, I find the negative tone of the first half unnecessary and it actually undermines the real message, namely that Safety-I and II are not in conflict, but that both are needed and that they are complementary. Regrettably one has to get until halfway the book before this is revealed and then it feels a bit out of balance with what has come before…

Another point of criticism is that I think that Hollnagel restricts himself a bit to the academic audience by throwing around terms like phenomenology, aetiology and ontology quite liberally. Yes, it fits his story, but sacrifices accessibility. And that’s a shame because it may give Safety-II an appearance of being elitist, which it really shouldn’t regarding the “pay attention to what’s really happening out there” message that runs like a main theme through the book.

Positive elements

Enough of that, let’s have a look at the positive elements. Even despite of my criticism the first five chapters do contain some gems, most of all the discussion of work-as-imagined versus work-as-done, which by the way is one of the recurring themes throughout the book.

Chapter 6 discusses "The Need To Change". Those who already have realized that there could be more than the traditional ways of doing safety management might start reading here. It’s not hard to draw some parallels to Perrow’s work and there is a really interesting discussion of how we should regard (modern) systems wider than before. Not really new either, but a good discussion.

Chapter 7 proved to be a treat because it is all written in a positive tone which more enjoyable to read. The chapter explains how Safety-II in principle works (in a nutshell: create with people) and can serve as a good introduction to the matter. Chapter 8 is also worthwhile. Here it’s stressed that we need both Safety-I and II (which the book should have started with). We also get some guidance of how we can do things with people by having our focus on the Work-As-Done rather than on Work-As-Imagined. Pages 160 to 162 are of special interest. Here Hollnagel builds a rather strong case in favour of the ‘safety pyramid’ based on General Learning Theory. A shame the chapter ends a bit weakly with some not too solid discussion of cost/benefit of Safety-I and II. It would have been much, much better with a real example instead of just the general idea.


There are not too many real practical examples in the book, but then, I think the aim of the book is more about creating consciousness and laying out the general concept rather than being a ‘do book’. That leaves the practical implementation for now in our hands which is a challenge we may or may not chose to adopt as early adopters (which is a term open to debate depending whether or not you find Safety-II a new concept). Good luck, I’ll be working on it myself.

For those whose interest is tickled, start by downloading a few of the papers from Internet (freely available) since they give a more accessible and sometimes better introduction. Especially the Eurocontrol White Paper does have some clear advantages over the book: 1) much, much more compact and easier to digest, 2) better wording, more positive and stressing the co-existence 3) pictures in colour and last but not least 4) practical Safety-II examples from Eurocontrol (very brief, but still). Decide then for yourself if you want or need the book to flesh out things further.

The book is available in various formats from I’ve read the paperback: ISBN 978 1 4724 2308 5

The Eurocontrol White Paper is found at:

More papers on the subject are found here.