User:ThaesOfereode/Szemerényi's law

Szemerényi's law is an early Proto-Indo-European sound law that causes vowels to undergo compensatory lengthening after the loss of a word-final fricative in some contexts. In short, when a vowel is followed by a sonorant and then a fricative word-finally, the final fricative is dropped and the vowel is lengthened. The law serves as an explanation for several otherwise unexplainable phenomena in Proto-Indo-European nominals, namely the behavior of vowels in a word's final syllable. It is named after Hungarian-British linguist Oswald Szemerényi, who first described it in 1956.

History
Oswald Szemerényi first described the law in 1956 in a paper published in the Journal of Comparative Linguistic Research in the Field of Indo-European Languages (Zeitschrift für vergleichende Sprachforschung auf dem Gebiete der Indogermanischen Sprachen) entitled "Latin rēs and the Indo-European long-diphthong stem nouns". In it, he disputes the reconstruction of several terms, both following his proposed law and not. He proposes that some of the long vowels used to reconcile Latin and Sanskrit cognates are not the result of accent or morphology, but are the result of phonological processes alone. He explains:

"We must conclude then that not even the monosyllabic nasal (or liquid) stems are capable of justifying the assumption of vṛddhi combined with ' in the nominative, and that is for very good reason. [...] The difference between the nominative of ' stems on the one hand, and ' stems on the other, is not due to accent or morphological factors, but is purely phonetic: the vṛddhi is, as it were, compensation for the absence of '."

Szemerényi argued that there was a disparity between certain Ancient Greek and Sanskrit nouns that could not be accounted for with contemporary reconstructions. For example, some monosyllabic words in Proto-Indo-European had exhibited both a  marked nominative form, called the sigmatic nominative, and vṛddhi, morphophonological process which affected vowels in certain environments; it was well-established that Vedic Sanskrit क्षास् 'ground, earth' and Greek χθών  'soil, land' both derived from a common Proto-Indo-European root word, but Szemerényi objected to the constructions put forth, namely those of Karl Brugmann. Szemerényi dismissed F. B. J. Kuiper's claim that χθών 'soil, land' and χιών  'snow' were "new formations that have supplanted the old nominatives in -ως []", arguing that disagreement about the original term was that linguists were attempting to pattern the terms on the idea that a stem would end in a "long-diphthong" – that is, one where the initial vowel is long and the glide vowel is short – and the nominative ', while insisting the final ' in the accusative.

Szemerényi argued instead that if this were true, then Greek terms such as χθών and χιών would correspond to Indo-Iranian cognates that are unattested, such as Sanskrit *kṣā and *h(i)yā, respectively. These unattested forms must have existed at one point, but were ultimately changed to align with the other declensions based on analogical similarity, leading to nominative forms such as क्षाः.

For example, he argued that Vedic Sanskrit क्षास् 'ground, earth' was indeed cognate with Greek χθών  'soil, land', but rather than deriving from a ' source – as Karl Brugmann had postulated in 1897 – an earlier ' was a better fit, given full evidence. This explained the long vowel in the nominative of Greek words like χθών and χιών  'snow', where a sigmatic nominative would produce a short vowel (i.e., -ων ).

Szemerényi defended the law several times throughout the 1950s, 1960s, 1970s, 1980s, and 1990s.

Overview
The law is triggered when ' or any of the three laryngeal sounds represented by ', ', and ' is lost word-finally after a sonorant – which includes liquids such as ' or nasals such as ' – itself preceded by a vowel which is thereby lengthened. In other words, in a cluster where "V" represents any vowel, "R" represents any sonorant, and "H" represents either  or one of the three laryngeals, the law is only demonstrated in the sequence "VRH". In this context, the fricative  or laryngeal is lost and the vowel is lengthened to compensate. One example includes the following, explaining the long ' in ' 'mother':



Most linguists, including Szemerényi himself, agree that it is likely that the sounds represented by "H" were not dropped altogether, but rather assimilated to the preceding sonorant, leading to a geminate, the shortening of which is what actually led to the vowel lengthening to compensate. Some linguists prefer this explanation because the gemination followed by the compensatory lengthening is typologically common, though Robert S. P. Beekes – admitting the typological frequency of the sound change – calls this intermediate gemination unnecessary.

This law helps to demonstrate why Proto-Indo-European nominal inflections depart from their expected values. For one, the final ' is often nominative case marker, whereas ' is associated with the accusative. Many words in the language lack the sigmatic nominative, but retain an accusative-marking final . In these cases, however, the accusative form's vowel is short and Szemerényi's law thereby explains this discrepancy by explaining that the nominative form's vowel is long by a regular phonological process. For example, the word ' 'man', shown here in the nominative case, has the long vowel '. However, in its accusative form, the reconstructed form is ', with a short vowel and the accusative-marking final '. In Proto-Indo-European, a noun like ' 'father' is not marked with the sigmatic nominative, but the law explains irregularities like this by arguing that at a previous stage, sometimes termed "Pre-Proto-Indo-European", the term for 'father' would have been '. After Szemerényi's law took effect, the final ' was lost and the ' became lengthened.

The law largely affects the nominative singular forms of many masculine and feminine nouns, as well as the nominative and accusative forms of neuter collective nouns. Often, this law is followed by the loss of the sonorant as well, particularly ' when the preceding vowel is an unaccented ', though other linguists have argued that the assimilation may convert sounds like ' to ' before  before Szemerényi's law is applied.

Objections
Szemerényi's law is accepted by most linguists, but the law does not enjoy universal acceptance and what agreement exists often differs in the details.

In 1975, Frederik Kortlandt argued that, at least in some cases, it is an assumption that Szemerényi's Proto-Indo-European examples must have had a final  to mark the nominative case. Similarly, in 1990, building on work from 1988, Robert S. P. Beekes echoed Kortlandt's criticism, arguing that "there is no necessity to assume an original nominal -s", further dismissing the law entirely as a "hypothesis" which "has always seemed to me most improbable". Beekes also points to the largely domineering role of , which usually devoices sounds around it – rather than being assimilated to voiced [z] in voiced environments – except in narrow circumstances, making this kind of assimilation less probable. Instead, he supports Kortlandt's proposition, which built on earlier work by Jacob Wackernagel, which states that vowels are lengthened before a word-final resonant. According to Beekes, this position is more defensible because it is more phonetically likely and explains the facts more simply without any further assumptions, namely that there is cause to assume a nominative final .

Some disputants point to the lengthening of some monosyllabic terms which preserve the final ' and have a phoneme other than a sonorant in the "R" position, including ' 'voice'.

Response
In 2014, Ryan Sandell and Andrew Miles Byrd defended the law in a presentation at a conference on Indo-European linguistics. In it, they point out several issues with what they call the "Wackernagel–Kortlandt hypothesis", which is the combined points of Wackernagel's 1896 paper, Kortlandt's 1975 paper, and Beekes' 1990 paper. They take issue with Kortlandt's supposition that vṛddhi affects the vowel of all monosyllabic words, which fails to account for monosyllabic words with short vowels like  'night'.

Sandell and Byrd admit, however, that Szemerényi's formulation still needs revision and thereby reanalyze the law with what they called a "broad" or "final" form of Szemerényi's law, presented as follows:

$$\text{*VC(C)F →}$$ $$\begin{cases} \mathrm{\bar{V}C} \text{ }/ \text{ }\_\# \\ \text{VCC} \text{ }/ \text{ }\_\# \\ \text{VC} \text{ }/ \text{ }\_\$\text{C}_0 \end{cases}$$

The Pre-Proto-Indo-European on the left preceding the arrow demonstrate the syllable structure involved: vowel (V), consonant (C), optional consonant (C in parentheses), and fricative (F), meaning that the environment must either be VCF or VCCF. The right half demonstrates the result of the law (near the bracket) in three different environments (following the forward slashes). In short, this updated form of the law states that the fricative is deleted word-finally in VCF or VCCF contexts and the vowel is lengthened to compensate for the loss if the loss occurs after a single consonant.

Relative chronology
The law is thought to have affected early Proto-Indo-European for a few reasons. First, the lengthened vowel spread by analogy in forms that would be unaffected by the law, such as in  'foot'. Next, word-final sonorants other than ' were dropped in nominative singular forms, seemingly following the law even where it was not justified, such as in ' 'companion', which was originally an i-stem.