Developing an algorithm requires expressing it in some (formal) language. The respective language is usually understood to be textual (conventional programming language) or partly graphical (design languages, and languages in programming environments for children). As writing and reading are capabilities not to be presumed from preschoolers, many educators claim that confronting such young kids with algorithmic concepts is beyond their abstraction capability. This paper reports on an experiment with kindergarten-groups requiring them to discover simple algorithms without resorting to reading and writing. It clearly showed that limited capabilities of abstractions are not a hurdle at all, if the problems are posed in a way corresponding to the limited experience base of the children, and if solutions are small enough to be kept in memory and allow expressing themselves in other forms than writing.