Yield increases by use of phosphorus (P) fertiliser has been a defining aspect of last century's agriculture, but in many countries P inputs are now being regulated to avoid ecological damage and improve agricultural sustainability. We tested the hypothesis that P forms and functions in agricultural soils have changed over a period of 50–70 years of agricultural change using topsoils from 35 agricultural sites in North East Scotland compared at ‘original’ and ‘resampled’ timepoints. The original dried and archived samples were collected between 1951 and 1981, and resampling took place in the autumn of 2017 with all samples reanalysed at the same time and by same methods. A range of soil extractants, from labile (water soluble) P, carbon (C) and nitrogen (N) forms, to surface exchangeable P (acid ammonium oxalate extraction), to complexed P (NaOH-EDTA extraction), were compared with agronomic soil test P (modified Morgan; MM-P), soil pH, and, on a subset of five time pairs (n = 10), specific P compounds using 31P NMR spectroscopy. Whilst MM-P had declined significantly in time, all soils were originally and remained at high soil P status. In contrast, no change in labile P pools, nor in the more refractory P pools that replenish labile P, were observed. In the subset analysed by NMR, orthophosphate P consistently increased in concentration. This suggests that soil P shows a slow (multidecadal) response to the documented reductions in fertiliser P for the study region. Significantly greater soil pH (by 0.4 units) over time and reduced water-soluble C and N forms suggests that labile P in contemporary soils exists in a biogeochemical environment of enhanced solubility and in excess to other available macronutrients that control its microbial cycling. This study represents a long and important period of changing drivers acting on soil P change with potential to improve understanding future soil P trajectories. The value of archived historical soils samples is shown, if care in interpretation of storage implications is made. Highlights: Phosphorus fractions in agricultural soil were compared at two timepoints 50–70 years apart. Whilst soil test P declined with time, other labile and refractory P pools did not change. The readjustment of P pools is slow even to substantial change in fertiliser inputs. The investigation of long-term land use to inform future policy is enabled by soil archives.