The sampling system ministers want to introduce to replace national tests could narrow the curriculum and trigger problems of its own, new research has warned.
Ed Balls, the Schools Secretary, said he wanted "national-level sampling" at key stage 3 to monitor the performance of the education system as a whole when he announced the end of compulsory tests for 14-year-olds in October.
Those now campaigning for the end of national tests for 11-year-olds are enthusiastic backers of sample tests.
They argue that sampling would not involve the high stakes inherent in Sats that encourage teaching to the test and a narrowing of the curriculum.
And because sampling would not involve collecting comparative results from all schools, it could not be used to judge them or compile league tables, campaigners say.
But a report from the influential National Foundation for Educational Research warns that the "low-stakes" system can bring problems of its own.
The researchers looked at existing sampling systems, including the National Assessment of Educational Achievement used in the United States, and the Scottish Survey of Achievement.
On the US system, they found: "The low-stakes nature has been linked to some issues with lower-than-desired response rates and concerns that low motivation may affect the reliability of the results.
"Even with this system there is still the view that it has led to a narrowing of the curriculum."
The researchers also expressed concerns about the impact of the Scottish low-stakes system on performance.
The report examined the Assessment of Performance Unit national monitoring system used in England in the 1970s and 1980s before national testing was introduced, and international comparative studies, Timss and Pisa.
The potential problems identified include increased teacher workload; controversy over analysis of findings; difficulties in keeping the measures constant, relevant and reliable over long periods; misinterpretation of results by media and public; low participation rates among some pupil groups; high costs; and goalposts being moved once monitoring has begun.
The NFER says ministers must be clear about the exact purpose of a KS3 sampling system before they start to develop it.
It argues that sampling should be used to monitor absolute changes in standards over time and to investigate areas of strength and weakness in the curriculum.
On the size of the sample, the researchers say the need for precision must be balanced against the requirement not to overburden schools. They recommend a sample structure of one class per school.
Their report calls for a low-stakes system, with no schools identified in results even though that could lower motivation in the sample tests. They advise that participation should be made compulsory.
The researchers suggest sample tests at KS3 should cover maths, English and possibly ICT.
By testing different groups of pupils on different areas of knowledge, national standards across the curriculum could be monitored.
The report has been submitted to expert panel advising the Government on sampling and other assessment issues.