Open New Windows for PDF and other Non-Web Documents (Jakob Nielsen, 2005) 
The Top 10 Web Design Mistakes of 1999 (Jakob Nielsen, 1999) 
Top 10 Mistakes in Web Design (Jakob Neilsen, 2011) 
We gain nothing by "keeping the user on our site." The user doesn't care, to them it's all just content they want to look at. If they want to come back to our site they can use the back button, the most popular feature on a browser. If we open new windows we are: littering the user's screen; breaking their chain of navigation; and if they're on a mobile device, consuming extra RAM and causing other windows to disappear prematurely.
Since then a lot has changed, including the spread of multi-tabbed browsing (which one could imagine would have a huge impact on this issue) and drastic changes in the way web applications are designed. Is the "back" feature even still the most popular feature on a browser? While this claim would have been indisputable a few years ago, I seem to almost never use it these days unless I accidentally clicked on the wrong link. At least in my case, the evolutions I mentioned have changed the way I use my browser quite a lot. Less anecdotically, this does seem to be a more global trend since on Chrome for Android the back button has even been demoted to the "More" menu, while tab-switching has been given a more prominent spot (swipe from edge).
Not saying it has now become good to open links in new windows of course, but I'd be curious to see more recent research about the topic.
If anything, we can an extra issue and suggest it's even worse nowadays. Per parent, it gobbles RAM on mobile devices, and this leads other tabs to close prematurely. It also turns older devices into helicopters when the fan kicks in.
But if you really need one, here's a study:
Date: just now.
Author: yours truly.
Methodology: browse random news websites every now and then on an old laptop and on an old mobile device.
Sample: yours truly.
Results: far too much cursing to publish as a HN comment, and frequent rage against the punk designers who ship huge JS payloads.
Conclusion: it still sucks. :-)
"In God we trust, all others must bring data." Best practices should be grounded in data rather than opinions whenever possible, especially when they're based on claims about users' cognitive load like in the NN/g recommendations.
> Per parent, it gobbles RAM on mobile devices, and this leads other tabs to close prematurely. It also turns older devices into helicopters when the fan kicks in.
I don't find that to be true with mobile devices. I much prefer having my links on mobile opening in a new tab, because a) mobile internet connection can be unreliable (and often a much bigger problem than lack of RAM), so I may not still have reception when trying to reload the original page, b) slow page loads makes browsing not as enjoyable since transitions take time, so I'd rather have everything in separate tabs and c) being on the move means I often put my phone down to do something else, so I don't necessarily remember every tab's browsing history.
Besides, mobile browsers have gotten a lot better at managing RAM and most have been able to optimize background tabs for a long time, so I think the RAM issue is way overrated.
> Besides the offender's unwarranted sense of entitlement (by keeping you around against your will), the main issue is that it breaks the back button. 15 years later, it's still an unwarranted sense of entitlement, and it's still breaking the back button.
I can see several reasons except for "unwarranted sense of entitlement" for this practice. It's all about reasonable expectations on the users' end: if a user can reasonably expect the website will behave some way, then not doing it this way can be detrimental to him. For example, many people expect "help" links (like the one next to the HN comment box) to not make you leave the page and rely on the "back" function to restore your painfully typed comment once you realize it does. They may be wrong in expecting this, but this seems like a perfectly rational safeguard against accidental data loss, so going against this could very well "increase the user's cognitive load".
As for the breaking the back button, my previous point was that the back button is not as important anymore, in part because a) a lot of things break the back button because dynamic content has made it much easier to do so and b) multi-tabbed browsing is now widespread and users may not be as confused by the fact each window/tab has its own browsing history anymore.
There are plenty of contradicting effects (precisely the reason why you and I can legitimately feel different about this issue) and the point of doing an actual study is to be able to measure them so we can compare them and say which trade-offs are worth it and which aren't.
tl;dr: your snark is unwelcome
Indeed they should. Except that here, the case is bloody obvious. "Show me the data that says the grass is green!" "Show me that evolution exists!" The onus of finding the data is on whoever thinks the statements are unreasonable.
Look… Opening links in new tabs does break the back button; no extra evidence is needed here beyond stating the obvious. The hardware back button on Android phones do lead to unexpected behavior; no extra evidence needed either. Having a save button in a word processor is needlessly forcing 40-year old IO limits onto end-users; no extra evidence needed here either.
In each case, what you can argue with and request data for, is this: whether end-users successfully cope with it or not; not whether it's broken to begin with.
And sadly, we humans are wonderfully adaptable. Including to pathetically poor user interfaces decisions such as the need to click a Start button in order to shut down a Win95 computer.
> Besides, mobile browsers have gotten a lot better at managing RAM and most have been able to optimize background tabs for a long time, so I think the RAM issue is way overrated.
Not everyone has a modern device. A whole bunch of iOS 5.1 devices (3rd-gen iPhones, 1st-gen iPads) and old Android devices are still in the wild.
> For example, many people expect "help" links (like the one next to the HN comment box) to not make you leave the page and rely on the "back" function to restore your painfully typed comment once you realize it does.
And that would qualify as a session that shouldn't be interrupted — i.e. it's a perfectly valid use-case, including for the OP.
(I, for one, would nonetheless suggest that a floating div on the same screen might be superior in that case, because the current window/tab would not lose its focus — a hopefully obvious fact.)
> a lot of things break the back button because dynamic content has made it much easier to do so
Assuredly so. But two wrongs don't make a right, so posting that as a reason to break it further is dubious at best.
Speaking personally, I hate those sites because they break my back button. I actually use the damned thing a lot, come to think of it. Especially on my 1st-gen iPad, since RAM-starved device can't seem hold more than a tab or two in memory when browsing sites loaded with JS.
> the point of doing an actual study is to be able to measure them so we can compare them and say which trade-offs are worth it and which aren't
Indeed. But chances are you wouldn't actually be measuring these things due to the huge bias involved in the measurement itself.
By simply asking a user if he prefers A or B, even qualitatively, you may open up a whole new world that he was unaware of until you asked; if so, the odds are good that — of course! — he'll prefer it rather than his more familiar daily routine, since the latter is in shambles due to widespread bad practices.
But hold… Your profile states that you're a data scientist, and I see you attended IEP at that. So you necessarily know this, and the — obvious — fact that a great many studies out there are inconclusive due to dubious methodology.
My original argument was simple:
1) opening links in new windows is said to be bad because a) it breaks the back button, and b) the back button is the most important one
2) the back button does not seem to be as important as it is now because of multi-tabbed browsing. Furthermore, since what matters is not not the back button itself, but how users expect the back button to perform, the fact this functionality has become unreliable could change the user's expectations (it changed mine)
3) therefore, the original argument may not be relevant anymore
> Indeed. But chances are you wouldn't actually be measuring these things due to the huge bias involved in the measurement itself.
The burden of proof should be proportional to the size of the claims you make. You can only be as confident as your data: if you believe there are biases in the measurement, then you should stay measured, too.
By the way, you say:
> And that would qualify as a session that shouldn't be interrupted — i.e. it's a perfectly valid use-case, including for the OP.
Who defines what should and shouldn't be interrupted? In the case of an article, I could argue that when I click a link in the middle of an article (like in The Verge's case), it's generally just because I'm curious about the link, kind of like I would read a footnote in a book, but ultimately I don't want to lose my pace and that would interrupt my reading "session". It is definitely not an indisputable claim, but it sure is a justifiable one (unlike, say, auto-loading audio advertisements) so The Verge isn't evil for doing so. If they had an A/B test and it showed it increased retention and that this is worth irritating some users, that's their decision to make.
Anyway, I feel this discussion has grown way out of proportion to the original claim and that we're just arguing for the sake of being right at that point. I'll willingly concede that you're right to say that as of now, forcing new tabs may place an unnecessary burden on users using older hardware.