Conceptual Metaphor in German Computer Terminology: Mechanisms of Terminologisation and the Problem of Transposed Motivation

The article offers a cognitive-onomasiological analysis of metaphorically motivated computer terms in present-day German. Most existing studies classify such terms by source domain, but this classification does not explain why some metaphorical nominations become established terms while others remain confined to journalism. In German the task is complicated by the fact that many metaphorical terms arose through semantic borrowing or loan translation from English, so their motivation was formed outside German. The material comprises 124 terminological units selected from Duden online and DWDS according to the criterion of lexicographic attestation. The study applies cognitive-onomasiological analysis as developed by O. O. Selivanova, reconstruction of metaphorical mappings, analysis of dictionary definitions and usage labels, and collocational data from DWDS-Wortprofil. The extensive mapping of the road domain in Datenautobahn did not secure terminological status for the word, since it denoted a general image of network infrastructure and lacked a delimited technical referent. Stable terms are those in which the transferred feature corresponds to a function or a technically defined characteristic of the object (Ordner, Papierkorb, Bandbreite, Datenstrom). For these the notion of an operationally verifiable metaphor is proposed. Scenario-based nominations (Trojaner, Wurm, Virus) are treated as products of conceptual blending. The metaphorics of artificial intelligence (halluzinieren, trainieren, füttern) and the term Virus show that transferring inferences without a counterpart in the referent does not prevent terminologisation but creates misconceptions among non-expert addressees. The distinction between operationally verifiable and inferentially redundant metaphors is therefore proposed as a criterion for evaluating terms. A hypothesis is advanced that metaphorical names of interface elements tend to function as semantic loans or calques, whereas names of infrastructural objects more often retain the English form.