Various constructs named associative or content-addressable memories have existed for about thirty years, and they are extensively used in computer technology, especially as fast buffer memories. The role of associative memories as the basic storage medium of large relational data bases, or for the storage and handling of text has been much less significant than initially expected. Two reasons for this are (1) the still relatively high price of the associative memory elements, (2) theoretical problems in the description of associations. The cost problem could be solved soon by extra-large-scale integration of the electronic circuits or possibly by new optical methods. On the other hand, natural experiences, for example, the sensory data from imaging devices, are seldom describable by discrete relations. A genuine associative memory ought to take into account and be able to represent the statistics of the occurrence of various items. This task falls into the category of pattern recognition problems. A natural solution would be a massively parallel, collectively operating computer resembling the biological brain. These aspects and the basic technologies of associative memories are surveyed in this presentation.