The funding councils have been consulting on proposals to permit academics to submit one fewer than the standard four outputs to the 2014 REF if they have been absent from the academy for more than 14 months in the 2008-13 census period for a variety of reasons, including maternity, ill health or career breaks.
However, the UCU's submission to the consultation, which was due to close on 5 October, says the 14-month threshold is too high in the case of maternity.
Ann Blair, chair of the UCU women members' standing committee and vice-president of the union's branch at the University of Leeds, said the impact of maternity on research productivity began during pregnancy and continued after maternity leave due to the difficulties of parenthood and the potential need to find new funding.
"Even people who don't get very angry about these things generally are absolutely astounded (that the funding councils) could be so naive," she said.
"The idea that you would effectively have to have two pregnancies before even one output was remitted seems bizarre."
Ms Blair, a senior lecturer in law at Leeds, also dismissed the funding councils' suggestion that universities could make a case to the REF Equality and Diversity Advisory Panel for a reduced number of submissions from a female academic whose maternity had been "accompanied by additional constraints".
She feared that women might be reluctant to disclose all of their personal difficulties to research managers. She was also concerned that some departments and institutions would not use their discretion "positively".
She described an alternative proposal, mentioned in the consultation document, to permit one fewer output for each period of pregnancy during the census period as a "pragmatic, sensible and workable approach".
She said the aim of the REF was not to create "a workforce under immense stress and pressure to work too fast and publish things that aren't ready - but this is the danger if you don't allow a realistic period for maternity".