There are a few ways of interpreting that term. Isaiah Berlin has an article called "The Concept of Scientific History" that takes a very broad view of what that could mean:
But I think in this case he probably has in mind people like the Annales school (mentioned in another comment here) - not so much Fernand Braudel, who wasn't particularly scientific although he was dabbled in economic theories like Kondratiev waves. But more the followers of Braudel who became very focused on crunching numbers relating to land usage patterns, trade, and other socioeconomic data from early modern Europe, to the point that they ended up a bit detached from reality, in much the way that some economists seem to be working in an ideal world rather than a real one. There's also Lawrence Stone who as far back as the 1960s was using mainframes and punch cards to figure out trends relating to the decline of the English aristocracy and the rise of the middle class.
I figured the HN crowd might be interested in what Stone had to say about this, so I looked up an old article of his from the 70s on his methods:
"'the 'new' historians have also borrowed from the social scientrists a series of new techniques...: quantification, conscious theoretical models, explicit definition of terms, and a willingness to deal in abstract ideal types as well as in particular realities. The one new tool they have borrowed is the computer... about 1960 historians suddenly gained free access to this immensely powerful but very obtuse machine, which can produce enormous quantities of data at fabulous speed but only if they are presented to it in limited, often rather artificial, categories, and if the questions are extremely clearly, precisely, and logically framed. Fifteen years of varied experience with the machine has led to a greater appreciation among historians both of its potential uses and its real defects... The computer is a machine in the elementary use of which most professional research historians should henceforth be trained -- a six-week course is ample for the purpose -- but it is one which should only be employed as the choice of last resort. Wherever possible, quantitative historians are well advised to work with smaller samples and to use a hand calculator."
http://berlin.wolf.ox.ac.uk/published_works/cc/scihist.pdf
But I think in this case he probably has in mind people like the Annales school (mentioned in another comment here) - not so much Fernand Braudel, who wasn't particularly scientific although he was dabbled in economic theories like Kondratiev waves. But more the followers of Braudel who became very focused on crunching numbers relating to land usage patterns, trade, and other socioeconomic data from early modern Europe, to the point that they ended up a bit detached from reality, in much the way that some economists seem to be working in an ideal world rather than a real one. There's also Lawrence Stone who as far back as the 1960s was using mainframes and punch cards to figure out trends relating to the decline of the English aristocracy and the rise of the middle class.
I figured the HN crowd might be interested in what Stone had to say about this, so I looked up an old article of his from the 70s on his methods:
"'the 'new' historians have also borrowed from the social scientrists a series of new techniques...: quantification, conscious theoretical models, explicit definition of terms, and a willingness to deal in abstract ideal types as well as in particular realities. The one new tool they have borrowed is the computer... about 1960 historians suddenly gained free access to this immensely powerful but very obtuse machine, which can produce enormous quantities of data at fabulous speed but only if they are presented to it in limited, often rather artificial, categories, and if the questions are extremely clearly, precisely, and logically framed. Fifteen years of varied experience with the machine has led to a greater appreciation among historians both of its potential uses and its real defects... The computer is a machine in the elementary use of which most professional research historians should henceforth be trained -- a six-week course is ample for the purpose -- but it is one which should only be employed as the choice of last resort. Wherever possible, quantitative historians are well advised to work with smaller samples and to use a hand calculator."