A sceptical view is advanced for the future of optical methods for signal processing, as a counterbalance to the implicit optimism usual in research papers concerning techniques with fascinating and elegant properties. Although optics is an extremely active field, and not yet fully exploited in providing general techniques for signal and image processing, it already appears to be conceding ground to the far less mature engineering discipline of electronic circuit integration. The glittering prizes of optics, in the shape of large bandwidths, high-order parallelism and freedom from limitations such as parasitic capacitances, carrier mobilities and crosstalk, are distanced from us by its mundane constraints of inadequate input-output techniques, large physical size, parametric inflexibility and dynamic-range limitations. Significant exceptions occur in specific cases such as where the signal itself is optical, or where a commonly required and complicated system function like an integrating spectrum analyser can be realised in a relatively simple opto-acoustic form.