1. Māl ‘speech’ < *maþla. According to the orthodox view þ disappeared without change before l. According to Sievers þ was first shifted to χ > h, as in WGmc., with loss of h before l; cf. *maþl > *maχl > *mahl > ON māl: OHG OS mahal. It is certain that in ON þ did disappear without the intermediate shift to χ > h after a long radical vowel, thus in *nāþlu > nǭl ‘needle’ = Goth. nēþla, where in WGmc. this loss did not occur (OE nǣdl, OFris. nēdle, OS nādla, OHG nādala). If ON þ after a short radical vowel underwent the shift in question, we should expect the shift to have taken place under the same conditions in WGmc. According to Sievers WGmc. þl remained wherever the l continued unvoiced, but was shifted to χl > hl wherever the l became voiced: thus nom. sg. *maþl (with voiceless l) but gen. sg. *mahles (with voiced l due to the following vowel); the type of OHG mahal, beside madal = OE mæÐel, he explains as due to the influence of the verbal form mahalen < *mahali̯en < *mahali̯an. That these conditions do not hold for ON is apparent from the example of *maþliÐō > *mǣlþa > mælta, for if the l before the vowel i in *maþliÐō had become voiced we should have expected a form *mǣlÐa > *mǣlda. This discrepancy between North and West Gmc. led Sievers to assume that ON þ before l in accented syllables was everywhere shifted to χ > h, i.e. without the restrictions characteristic of WGmc. But since this hypothesis has no factual support and is inconsistent with his theory regarding WGmc. conditions, there is no necessity for accepting it, especially as it in no way invalidates the orthodox view.