Chemical analyses of late-type stars are usually carried out following the classical recipe: LTE line formation and homogeneous, plane-parallel, flux-constant, and LTE model atmospheres. We review different results in the literature that have suggested significant inconsistencies in the spectroscopic analyses, pointing out the difficulties in deriving independent estimates of the stellar fundamental parameters and hence, detecting systematic errors. The trigonometric parallaxes measured by the Hipparcos mission provide accurate appraisals of the stellar surface gravity for nearby stars, which are used here to check the gravities obtained from the photospheric iron ionization balance. We find an approximate agreement for stars in the metallicity range -1.0{<~}[Fe/H]{<~}0, but the comparison shows that the differences between the spectroscopic and trigonometric gravities decrease toward lower metallicities for more metal-deficient dwarfs (-2.5{<~}[Fe/H]{<~}-1.0), which casts a shadow upon the abundance analyses for extreme metal-poor stars that make use of the ionization equilibrium to constrain the gravity.