We present high-quality ground-based spectroscopic observations of 54 supergiant H II regions in 50 low-metallicity blue compact galaxies with oxygen abundances 12 + log O/H between 7.1 and 8.3 .
We use the data to determine abundances for the elements N , O , Ne , S , Ar and Fe .
We also analyze Hubble Space Telescope ( HST ) Faint Object Spectrograph archival spectra of 10 supergiant H II regions to derive C and Si abundances in a subsample of 7 BCGs .
The main result of the present study is that none of the heavy element-to-oxygen abundance ratios studied here ( C/O , N/O , Ne/O , Si/O , S/O , Ar/O , Fe/O ) depend on oxygen abundance for BCGs with 12 + log O/H \leq 7.6 ( Z \leq Z _ { \odot } /20 ) .
This constancy implies that all these heavy elements have a primary origin and are produced by the same massive ( M \geq 10 M _ { \odot } ) stars responsible for O production .
The dispersion of the C/O and N/O ratios in these galaxies is found to be remarkably small , being only \pm 0.03 dex and \pm 0.02 dex respectively .
This very small dispersion is strong evidence against any time-delayed production of C and primary N in the lowest-metallicity BCGs ( secondary N production is negligible at these low metallicities ) .
The absence of a time-delayed production of C and N is consistent with the scenario that galaxies with 12 + log O/H \leq 7.6 are undergoing now their first burst of star formation , and that they are therefore young , with ages not exceeding 40 Myr .
If very low metallicities BCGs are indeed young , this would argue against the commonly held belief that C and N are produced by intermediate-mass ( 3 M _ { \odot } \leq M \leq 9 M _ { \odot } ) stars at very low metallicities , as these stars would not have yet completed their evolution in these lowest metallicity galaxies .
In higher metallicity BCGs ( 7.6 < 12 + log O/H < 8.2 ) , the Ne/O , Si/O , S/O , Ar/O and Fe/O abundance ratios retain the same constant value they had at lower metallicities .
By contrast , there is an increase of the C/O and N/O ratios along with their dispersions at a given O .
We interpret this increase as due to the additional contribution of C and primary N production in intermediate-mass stars , on top of that by high-mass stars .
The above results lead to the following timeline for galaxy evolution : a ) all objects with 12 + log O/H \leq 7.6 began to form stars less than 40 Myr ago ; b ) after 40 Myr , all galaxies have evolved so that 12 + log O/H > 7.6 ; c ) by the time intermediate-mass stars have evolved and released their nucleosynthetic products ( 100–500 Myr ) , all galaxies have become enriched to 7.6 < 12 + log O/H < 8.2 .
The delayed release of primary N at these metallicities greatly increase the scatter in the N/O abundance ratio ; d ) later , when galaxies get enriched to 12 + log O/H > 8.2 , secondary N production becomes important .

BCGs show the same O/Fe overabundance with respect to the Sun ( \sim 0.4 dex ) as galactic halo stars , suggesting the same chemical enrichment history .
We compare heavy elements yields derived from the observed abundance ratios with theoretical yields for massive stars and find general good agreement .
However , the theoretical models are unable to reproduce the observed N/O and Fe/O abundance ratios .
Further theoretical developments are necessary , in particular to solve the problem of primary nitrogen production in low-metallicity massive stars .

We discuss the apparent discrepancy between the N/O abundance ratios measured in BCGs and those in high-redshift damped Ly \alpha galaxies , which are up to one order of magnitude smaller .
We argue that this large discrepancy may arise from the unknown physical conditions of the gas responsible for the metallic absorption lines in high-redshift damped Ly \alpha systems .
While it is widely assumed that the absorbing gas is neutral , we propose that it could be ionized .
In this case , ionization correction factors can boost the N/O ratios in damped Ly \alpha galaxies into the range of those measured in BCGs .