Abstract. The equilibration of polymer conformations tends to be sluggish in polymer layers adsorbed onto highly attractive substrates, so the structure of these layers must be understood in terms of the layer growth process rather than equilibrium theory. Initially adsorbed chains adopt a highly flattened configuration while the chains which arrive later must adapt their configurations to the increasingly limited space available for adsorption. Thus, the chains adsorbed in the late stage of deposition are more tenuously attached to the surface. This type of non-equilibrium growth process is studied for polymethylmethacrylate (PMMA) adsorbed on oxidized silicon where the segmental attraction is strong (4 k B T /segment) and for polystyrene (PS) adsorbed on oxidized silicon from a carbon tetrachloride solution where the segmental attraction is relatively weak (1 k B T /segment). Measurements were based on Fourier transform infrared spectroscopy in attenuated total reflection (FTIR-ATR). In both cases, the chains arriving first adsorbed more tightly, became flattened (as measured by the dichroic ratio), and occupied a disproportionately large fraction of the surface. This non-uniform structure persisted indefinitely for the strongly adsorbed PMMA chains, while the PS chains exhibited a gradual evolution, presumably reflecting an equilibration of the adsorbed layer occurring after long times. On the theoretical side, the initial heterogeneity of these adsorbed polymer layers is modelled using a random sequential adsorption (RSA) model where the size of the adsorbing species is allowed to adapt to the surface space available at the time of adsorption. The inhomogeneity in the size of adsorbing species (hemispheres) in this model is similar to the distribution of chain contacts in our measurements on adsorbed polymer layers. Owing to extensive variance around the mean, conformations having the mean number of chain contacts are least probable, which contrasts starkly with expectations based on equilibrium adsorption theory.