According to Bartholomae (1885:206–7), when an Indo-European aspirated stop was followed directly by one or more unaspirated obstruents, the aspiration was transferred from the beginning of the cluster to the end, as in the conventionally cited example IE PGk. *pat-skhō > Gk. páskhō ‘I suffer’ beside Gk. páthos ‘suffering, disease’. Bartholomae's formulation stipulates further that the entire cluster became voiced when its first component was voiced, e.g. IE *lubh-tós > * lub-dhós > Skt. lub-dhás ‘covetous’ beside lubh-yāmi ‘I yearn’. As textbook descriptions of the process regularly indicate, the Indo-European languages as a group honor Bartholomae's Law considerably more in the breach than in the observance. Thus, on purely phonologic grounds, IE *légh- + -trom might be expected to become *lég-dhrom, but the corresponding historic Greek form léktron shows the reflex of IE /kt/, not /gdh/. In such instances, which are so numerous as to constitute the rule rather than the exception, the absence of change is generally ascribed to analogy. Thus, IE *légh-trom, under the influence of related forms like *légh-os, would not become *lég-dhrom, but would merely undergo devoicing and de-aspiration of /gh-t/ to /kt/, hence Gk. léktron ‘couch’ beside lékhos ‘bed’.