The cell volume permissible for a specified degree of loss of efficiency can be computed from response volume considerations. For an open tubular column, the permissible illuminated length can be computed for the unretained peak. Similar estimations can be made for the maximum permissible cell volume with a packed column for a given column efficiency and flow rate. The packed column case does require an assumption on the degree to which the cell behaves as a mixer. An altogether different question is if the data from a long cell, with its considerable advantage in S/N, can be used and the associated dispersion mathematically removed. Experimental data for a variable path length (0-60 mm) HPLC detection cell indicate that an exponential model fits the observed dispersion. Once fit parameters are determined, the same can be applied to a peak, not part of the original training set, obtained in a longer path cell and the effects of dispersion mathematically reversed without major loss of S/N. Results with shorter path cell dispersion characteristics are then obtained with much higher S/N. A comparison is made with Fourier transform-inverse Fourier transform based deconvolution that can be used to achieve the same ends.