Policy-makers have spent years worrying about the differences between the best and worst performing schools, while neglecting the far bigger problem of pupils in the same schools doing better in some subjects than in others.
School leaders, too, have often shied away from tackling what can be a sensitive issue - or thought that there wasn't much they could do about it. But early results from research into what's known as "within-school variation" suggest that there are solutions.
Of the 22 schools taking part in the first phase of the National College for School Leadership project, more than half - seven secondary schools and five primaries - have managed to reduce variations between subjects or departments.
With little previous research in this area to draw on, schools are using the project to find their own ways of identifying and reducing variation.
So it is striking that those now reporting success have come up with similar strategies. These include giving subject heads or co-ordinators a key role in dealing with the problem, making sure that data on pupil performance is interpreted in the same way throughout the school and promoting standardised approaches to teaching.
"A lot of the work has been about team building and ironing out the inconsistencies and variabilities in the way teachers approach their work,"
says Ray Tarleton, principal of South Dartmoor community college, Devon and co-ordinator of the NCSL's Leadership Network, which is running the project.
"That isn't to say we are not allowing teachers to carry on being creative, but it does mean that there has to be an absolutely rock-bottom series of approaches and standards and ways of working, and then you can be creative after that."
Mr Tarleton's own school started narrowing the gap between GCSE results in core and option subjects after discovering that not all staff understood the data system they were using. This meant that their predictions were often inaccurate and they were targeting the wrong students for extra support.
The school's response was to devote three whole training days to making sure middle leaders understood how the data system worked and had a shared vocabulary for discussing students' performance. As a result, they are now making better predictions, and the average GCSE pass rate for core subjects has gone up from 61.2 per cent in 2003, to 70.8 per cent in 2005 - reducing the gap with the options from 22 points to 11.
International comparisons show that once factors such as social background have been taken out of the equation, performance variations within British schools are wider than in most other developed nations.
"The shattering thing is that they get bigger and bigger as you go through the key stages," says David Reynolds, professor of education at Exeter university, who has helped to lead the NCSL research. He points to a Department for Education and Skills study of its own value-added data for 2003, which found that, at key stage 2, variations within schools are five times greater than between schools. At key stage 3, they are 11 times greater and at key stage 4, they are 14 times greater.
The traditional autonomy of individual teachers and departments in UK schools, which still hasn't disappeared, offers one possible explanation for the stark contrasts in pupils' attainment. The under-developed role of middle managers could also be a factor, according to Professor Reynolds.
"If you haven't got people in the middle of the school driving this, what you get is good people doing good things in a department and less good people doing less good things in another department and then you get all the variation under the sun."
Findings from the NCSL research confirm that middle leaders can make a real difference to how well students do in different subjects. Thomas Sumpter school in Scunthorpe, for example, managed to reduce some huge variations between subjects and improve performance overall after putting heads of department through a development programme that gave them the skills they needed to coach and mentor their staff.
The 11-16 comprehensive has also given departmental heads more time to work with teachers on devising individual learning plans for pupils, who from September will have personal tutorials with their form teachers every other week.
In addition, LEA consultants have worked with middle leaders on analysing departmental data, and the school's assessment tracking system is now being tweaked to provide clearer information on both effort and attainment.
"It is a way of flagging up students who aren't performing as well as they should be and aren't putting in the effort required," explains headteacher Angela Briggs.
But isn't there a danger that highlighting variations within a school will demotivate teachers in under-performing departments and make matters worse?
Ms Briggs believes it depends on the way schools "sell" this kind of work. "We've always sold it in terms of support," she says. "We have a departmental review every term when we pull together all the pieces of information about that department and then we say what we are going to do to support this area or person, and we put in strategies to support them."
The findings from the first phase of the NCSL's project will be discussed today (Friday, March 3) at the annual Leadership Network conference at the Hilton Metropole in Birmingham. Another 25 schools taking part in the second phase of the project will report this autumn. For further information, see www.ncsl.org.ukleadershipnetwork.