The relationship between earlywood width (EW) and latewood width (LW) is investigated using 197 tree-ring collections representing several tree species from across the North American continent. Chronologies of LW have limited paleoclimate value when they have low variance or very high correlation with EW from the same site. The correlation of LW and EW can be removed by taking the residuals from linear regression to provide a chronology of discrete latewood growth free from the carryover effects of prior EW (the so-called adjusted latewood chronology, LWa). The correlation between EW and LW, along with LW_a variance, varies dramatically across North America. The lowest correlations between EW and LW chronologies can be found in Pseudotsuga menziesii in the summer monsoon region of northwestern Mexico. Low correlations between EW and LW chronologies are also noted for Pinus echinata and Quercus stellata in the south-central United States. Q. stellata also displays the highest LW_a variance among any species in the dataset. For three conifer species, correlations between EW and LW appear to increase with the biological age of the tree. An age-related decline in LW_a variance was also detected for Douglas-fir, bald cypress and ponderosa pine older than 200 years. These results imply that heavy sampling to produce “age-stratified” chronologies based on trees ≤ 200 years in age throughout the record may produce the best quality LW chronologies with the highest variance and most discrete growth signal independent from EW.