We constructed diagnostic diagrams using emission line ratios and equivalent widths observed in several independent samples of H ii  galaxies .
Significant trends are seen , both in the line ratio diagrams , and in diagrams relating line ratios to the equivalent width of H \beta .
The diagrams are compared to predictions from photoionization models for evolving starbursts .
This study extends the work of Stasińska & Leitherer ( 1996 ) by including objects with no direct determination of the metallicities , and by using updated synthesis models with more recent stellar tracks and atmospheres .

We find that H ii  galaxies from objective-prism surveys are not satisfactorily reproduced by simple models of instantaneous starbursts surrounded by constant density , ionization bounded H ii  regions .
The observed relations between emission line ratios and H \beta  equivalent width can be understood if older stellar populations generally contribute to the observed optical continuum in H ii  galaxies .
In addition , different dust obscuration for stars and gas and leakage of Lyman continuum photons from the observed H ii  regions can be important .
As a result , H ii  galaxies selected from objective-prism surveys are not likely to contain significant numbers of objects in which the most recent starburst is older than about 5 Myr .
This explains the success of the strong line method to derive oxygen abundances , at least in metal poor H ii  galaxies .

The observed increase of [ O i ] /H \beta  with decreasing H \beta  equivalent width can result from the dynamical effects of winds and supernovae .
This interpretation provides at the same time a natural explanation of the small range of ionization parameters in giant H ii  regions .
The classical diagnostic diagram [ O iii ] /H \beta  vs [ O ii ] /H \beta  can not be fully understood in terms of pure photoionization models .
The largest observed [ O ii ] /H \beta  ratios require additional heating .

The [ N ii ] / [ O ii ] ratio is shown to increase as the H \beta  equivalent width decreases .
A possible explanation is an N/O increase due to gradual enrichment by winds from Wolf-Rayet stars on a time scale of \sim 5 Myr .
Alternatively , the relation between N/O and O/H could be steeper than N/O \propto O/H ^ { 0.5 } , with a previous stellar generation more important at higher metallicities .