Getting Proto-Germanic stress: phonetics and historic accentology

In this guest talk given at the University of Stockholm in October 2008, I challenge the mainstream interpretations of Verner's law and Holzmann's law and argue that free stress existed after the breakup of the Germanic protolanguage. I adopt Dybo's (1961a, 1962b, 2002, 2008) diachronic analysis according to which Germanic words with Holzmann's law correspond to Baltic-Slavic oxytona and combine it with the revision of Verner's law (1877) in Zimmerling (2007), where I argue in detail that the voicing of the intervocalic consonants (PIE stops || PG fricatives) directly reflects the remnants of the free stress in Early Germanic languages but not the effects of its loss as was wrongly assumed by Verner himself and his followers. The problem is that correct diachronic correspondenses between PIE and PG consonants do not reconstruct any reliable-looking phonetic process that could be instrumental in Modern and Ancient languages. However, if one considers the external accentological parallels to Germanic verbs with Holzmann's law, the analysis is most straightforward. The variation of voiceless vs voiced fricatives (Verner's law) and standard vs reinforced glides j ~ JJ > gjj (djj) on the same segmental basis is explained uniformly by the placement of the word accent. The main position of the accentological mutation was the onset of the non-initial stressed syllable in words from the Baltic-Slavic mobile paradigm: voiceless fricatives got voiced, glides -j-, -w- got lengthened. Long-vocalic roots with post-root stress could undergo vowel shortening. A plausible phonetic interpretation of this distribution is that the words displaying the Verner or Holzmann effects had early timing, i. e. anticipating tonal movement starting before the stressed vowel. Where neither phonologically relevant pairs of phonemes nor glides were present, early timing did not trigger any changes in the segmental inventory. Lexical divergencies between different Old Germanic languages regarding Verner's law (Wood 1895) show that different Germanic languages generalized either the variants with early timing (>> Verner's effects) or with standard timing (no Verner's effects). Finally, obsolete cases of Verner's law in the forms of one of the same word, cf. Gothic sais-lep ~ saiz-lep can reflect not the retraction of stress to the initial syllable but coexisting variants with normal and early timing.2008年10月于斯德哥尔摩大学（University of Stockholm）所作的这场客座演讲中，我对维尔纳定律（Verner's law）与霍尔茨曼定律（Holzmann's law）的主流解读提出了质疑，并主张日耳曼原始语解体后仍存在自由重音。我采纳了戴博（Dybo）1961a、1962b、2002、2008年的历时分析框架，该框架指出符合霍尔茨曼定律的日耳曼词汇对应波罗的-斯拉夫语（Baltic-Slavic）的oxytona，并结合齐默林（Zimmerling）2007年对维尔纳1877年提出的定律的修订版本展开论证。在该修订中，我详细阐释了：词间辅音（印欧语（PIE）塞音 || 原始日耳曼语（PG）擦音）的浊化，实则直接反映了早期日耳曼语言中自由重音的残余，而非维尔纳本人及其追随者所误判的重音消失后的影响。当前的问题在于，印欧语与原始日耳曼语辅音间的准确历时对应关系，无法重构出任何可在古今语言中发挥作用的、看似可靠的语音过程。但倘若将符合霍尔茨曼定律的日耳曼动词与外部重音学（accentology）的平行现象相结合进行分析，整个推导过程将极为清晰直观。基于相同的音段基础，清擦音与浊擦音的交替（即维尔纳定律现象），以及标准滑音j与强化滑音j~jj>gjj（djj）的差异，均可通过词重音的位置得到统一解释。重音音系变异的核心位置，出现在源自波罗的-斯拉夫语可移动范式的词汇中，非首音节重读音节的起首位置：清擦音发生浊化，滑音-j-、-w-发生延长。词根后带有重音的长元音词根可能会发生元音缩读。对于这一分布特征，一种合理的语音学解读是：呈现维尔纳定律或霍尔茨曼定律效应的词汇，早期具备“早时序”（early timing）特征，即重读音节元音之前便已出现预期的音调移动。而当既无与音系相关的音位对立，又无滑音存在时，早时序则不会触发音段系统的任何变化。 不同古日耳曼语言在维尔纳定律上的词汇分歧（Wood 1895）表明，不同日耳曼语言分别泛化了早时序变体（即维尔纳定律效应）或标准时序变体（即无维尔纳定律效应）。最后，同一词汇的不同形式中出现的维尔纳定律过时用法，例如哥特语（Gothic）sais-lep与saiz-lep的对比例证，其成因或许并非重音后移至首音节，而是共存了正常时序与早时序的两种变体。