We employ unconditional quantile-decomposition methods to analyze the gender wage gap (GWG) in the urban region of twelve Latin American countries. Using data from harmonized household surveys we decompose the GWG into an explained component (dif- ferences in human capital) and an unexplained component (different rates of return to human capital). We find evidence of sticky floors (larger GWG at the tenth percentile than at the median) and glass ceilings (larger GWG at the ninetieth percentile than at the me- dian). The former are more frequent and their magnitude is generally larger. Working women are more educated than working men all along the wage distribution, which par- tially hides the ‘unexplained’ pay difference. Finally, we find that poorer countries and countries with more income inequality have higher GWG at the tenth percentile of the wage distribution. Richer countries and countries with lower inequality present larger GWG at the ninetieth percent of the wage distribution.