Transcending the retreat from rationalityOne of the aims of this issue is to provide something for planning practitioners. Unlike their academic cousins, planning practitioners have seldom had the luxury of 'retreating from rationality'. They have always had a job to do, rationally; in this job they were once comprehensively assisted by researchers into computer-aided planning techniques. Yet the intensity of academia's retreat from rationality and the popularity of its rejection of 'rational comprehensiveness' have tended to inhibit such assistance for at least two decades.Recently, however, there have been signs of a regrouping. New forms of 'organic' computing have emerged and their application to planning problems has been widespread. They therefore have the potential to restore academia's technical assistance back to something like the levels enjoyed by practitioners during the 1950s and 1960s. Accordingly, although it hardly proclaims that a second quantitative revolution is almost at hand (Openshaw, 1993), this special issue attempts to present, clearly, some cutting-edge techniques which have great promise.The techniques of interest can be loosely grouped under the umbrella term of 'connectionist' computing techniques. Each focuses on connections within data. They include simulated neural networks, which 'learn' to connect inputs with their associated outputs, and genetic algorithms, which connect parts of some problem solutions to parts of other problem solutions in order to 'breed' improved solutions. They also include cellular automata, which connect the simple behaviours of individual elements in order to generate complex models of the overall system.To appreciate fully the true potential of these techniques, it is necessary to understand how their planning applications evolved. And all applications were able to evolve only because of at least four decades of progress within three separate disciplines-computer-aided planning, artificial intelligence, and connectionist computing itself. Each of these fields passed through similar phases of euphoria in the 1950s and 1960s, disillusion during the 1970s and 1980s, and revival throughout the 1990s. But there were differences in the depths of unpopularity to which their techniques sank during the dark days, there were differences in their levels of 'born again' enthusiasm during the 1990s, and so today there are differences in the nature and potential of their respective contributions.Accordingly, in this editorial essay I will trace the evolution of each field. The conclusion is that connectionist-type computer-aided planning is far less ambitious, and more modest, than the modelling which occurred during the heady days of the 1960s.But I will also argue that the chances of a genuinely productive partnership between humanist and boffin are far greater this time around. Moreover, I will suggest there are signs that computer-based comprehensive contributions to planning, as distinct from the segmented and restricted contributions of today, may one day becom...