Spike fruiting efficiency (FE), defined as grains per unit of spike dry weight at anthesis (SDWa) is a promising trait for improving grain number (GN) in wheat (Triticum aestivum L.). It is often estimated at maturity as the grains per unit of chaff or FE at maturity (FEm). The fertile floret efficiency (FFE), defined as fertile florets per unit of SDWa, and grain set (GST), or the number of grains per floret, were studied to better understand FE determination for the first time. Two double haploid populations designed by crossing modern cultivars contrasting for FE ['Baguette 19' and 'Baguette Premium 11'(high FE) × 'BioINTA2002' (low FE)] were sown in five environments. The FE and FEm showed an unstable correlation (low or high) among genotypes within environments (caused by variable SDWa-chaff associations), resulting in a worse correlation between GN and FEm than between GN and FE. Therefore, the use of FEm as a surrogate for FE to improve GN may yield lower gains than those expected if FE were used. The narrow-sense heritability of FFE was high but the variability in fertile florets per spike among genotypes within environments was correlated with FFE only in the environments with high SDWa. Despite the close association between FE and FFE, the former was not totally set at anthesis, as GST greatly affected FE and GN. Selecting for higher FFE and GST, where genotype × environment effects determine heavy spikes at anthesis, is an alternative to breeding for improved GN.