Grouping pupils according to ability only exacerbates the divide between high-performing and underachieving pupils - with the divide often falling along class and economic lines.
Dr Ed Baines of the University of London's Institute of Education analysed existing research into setting and streaming in order to determine the effectiveness of the different types of ability grouping used in schools.
In the run-up to most UK general elections, he says, politicians talk about grouping pupils by ability, supposedly to improve standards in schools.
"Ability grouping is regularly presented as the silver bullet that will enable schools and teachers to get back to basics," Baines says.
Setting, which involves allocating children to classes on the basis of attainment in one specific subject, is the most common method of ability-grouping in British schools.
A recent survey found that 94 per cent of pupils between 11 and 16 were placed into sets for at least one subject. Two-thirds of pupils were put into sets for all core curriculum subjects. And another study showed that more than a third of primary pupils were placed into sets for maths, literacy or both.
An increased divide
A longitudinal study, involving 6,000 pupils across 45 comprehensives, revealed that setting had little effect on pupils' scores in key stage 3 tests or in core subjects at GCSE. But low-ability pupils who had been grouped by ability were found to perform slightly worse in key stage 3 maths tests.
"Those in the high-ability sets made greater progress, while those in the low-ability sets dropped further and further behind," Baines says.
Streaming involves allocating pupils to ability-based classes for most or all curriculum subjects, according to overall attainment.
In the UK, this is primarily used at whole-school level, for example when selecting pupils for grammar school. But recent research has also revealed that more than 16 per cent of primary pupils are streamed. Exact figures at secondary level are unknown, although it is likely to be more than 11 per cent of pupils.
Several studies cited by Baines show that streaming has no effect on academic performance in most cases. Pupils in high-ability classes will benefit when taught a curriculum specifically matched to their ability. However, even when given a tailored curriculum, low-ability pupils perform worse when taught in separate streams than when taught in mixed-ability classes.
The main form of differentiation used in primary schools is ability grouping within mixed-ability classes: around 60 per cent of primary pupils will sit with classmates of a similar ability. This approach is more flexible than other forms of grouping: pupils can interact with one another across the groups, for example, and can more easily move between them.
Baines concludes: "Low-ability students appear to benefit more from mixed-ability grouping, while high-ability students benefit equally from both approaches.
"Mixed-ability groups promote the use of elaboration, explanation and collaborative discussion between peers - all essential ingredients for developing high-level understanding and high-level thinking skills."
He points out that ability groups often - if unintentionally - divide pupils according to social class, ethnicity and sex. This can make it harder for pupils to conquer social divisions, especially when grouping is introduced from primary school onwards.
"The system of ability grouping can create an educational rut that is difficult for pupils to get out of," he says.
Baines, E. "Grouping pupils by ability in schools", Bad Education: debunking myths in education (Open University Press, 2012).
Ed Baines, senior lecturer in psychology of education, Institute of Education, University of London.
Pupils in lower-ability groups are often allocated the least-experienced teachers, who expect little of them, new research suggests.
Ability grouping also encourages pupils to make friends only with others of the same ability, thus affecting their ambitions and attitudes to school.
Research reveals the "low-level, conceptually weak and fragmented nature of teaching" in low-ability groups, Dr Ed Baines of the Institute of Education, University of London says. "Work is more structured, slow and repetitive, and thus is tedious for those who may need to be inspired by education."
While high-ability groups worked on interactive, challenging activities, their low-ability peers were rarely expected to use analytical skills or to demonstrate independent thinking.
Often, Baines suggests, schools allocate the most knowledgeable and experienced teachers to high-ability groups. "But lower-ability groups will be better served by teachers ... who can provide rich learning experiences," he says.
And teachers' lowered expectations of bottom-set pupils can affect those pupils' ambitions. Research has shown that children placed in sets above their assessed level of achievement tend to make more progress than their peers of equal ability who are placed in the correct sets.
Ability grouping can affect the friends children make. "Adolescents often look to their immediate peers to help them think about how they should see themselves, how they should behave, how they should think about learning and school," Baines says.
Mixed-ability groups allow pupils to mix with classmates from a range of social, economic and ethnic backgrounds. Studies show this can create a more positive attitude to school among less-able pupils.
By contrast, low-ability classes can lead to "a counter-productive, anti-learning peer culture, that can exacerbate poor behaviour and eventual alienation from school," Baines says.