Two putative tetrapod humeri of Devonian age, ANSP 21350 from the late Famennian of Pennsylvania and GSM 104536 from the late Frasnian of Scat Craig, Scotland, are reinterpreted in the light of recent discoveries. The morphology of ANSP 21350 can be more fully homologized with those of elpistostegids and early tetrapods than previously recognized. Unique features include distally displaced dorsal muscle attachments and a ventrally rotated distal face of the bone. This suggests that a weight-bearing ventrally directed forearm was created, not by means of a flexed elbow as in other tetrapods, but by distorting the humerus. The olecranon process on the ulna was probably poorly developed or absent. Primitive characters that are absent in other tetrapods add support to the contention that ANSP 21350 is the least crownward of known tetrapod humeri. Contrary to previous claims, Acanthostega has a characteristic tetrapod ulnar morphology with an olecranon process; it does not resemble an elpistostegid ulna and is not uniquely primitive for tetrapods. This suggests that the flexed tetrapod elbow with ulnar extensor muscles attached to the olecranon evolved simultaneously with the large rectangular entepicondyle typical for early tetrapods, probably as part of a single functional complex. GSM 104536 is definitely not a primitive tetrapod humerus, nor a sarcopterygian branchial bone, but cannot be positively identified at present.