PIE *(h2)(s)kwal-o-s
Gloss: ‘catfish’ (item 78 in Bjørn 2017)

Lat. squalus ‘large fish’; Germ. *hwali– ‘whale’; OPrus. kalis ‘wels catfish’; Gr. (Hes.) ασπαλος ‘fish’; YAv. kara ‘name of a fish’, MPers. kar māhīk ‘mythical fish’.

The inclusion of the Greek form demands an extension of the root with initial *h2s not otherwise warranted, although possible. De Vaan’s suggestion that Latin is an internal derivation to homonymous squālus ‘unkempt, dirty’ (2008: 584) is semantically unconvincing, and the Italic s- form gains plausibility in light of Greek. The semantic development of a riverine ‘fish’ to a maritime ‘whale’ in Germanic, Old Prussian ‘catfish’ or Italic ‘large sea fish’ is not surprising in light of the native distribution of the catfish, a large river fish, historically absent in Western Europe ( The circumstantial evidence in favor of the Graeco-Aryan sub-set is presented by Rodriguez (1989). Kroonen reconstructs the word internally in PIE as a derivation from the verbal root *kwelh1– ‘to turn’ (item 80) → *kwol– ‘the turner’ (2013: 262), but this fails to explain the consistent a-vocalism, which presents an additional problem for the age of the stock. In defense of an old PIE item, and given the distribution of the wels catfish on the Pontic steppes, loss may be expected in emigrant branches such as Anatolian, Celtic, Armenian, etc., although this is very ad hoc and fails to account for the lack of Slavic evidence. There are no reflexes in Anatolian and Tocharian, but the intricacies regarding this particular sphere are salient, especially in the former branch (cf. Mallory & Adams 1997: 204f.).

Distribution of the wels catfish

External comparanda:
Uralic: *k(w)ala


There are several phonemic features that are relevant to the comparison with the Uralic material. Most salient is the unquestionable IE labiovelar which does not exist in traditional reconstruction of Proto-Uralic, and there is no vocalism that could round the velar from assimilation. Although strictly tentative, Samoyedic Selkup does alternate two different reflexes of Uralic initial *k-, viz. q– and k-, that are seemingly unconditioned by the phonetic environment (Collinder 1960:50), and further seems to correspond to IE labiovelar comparanda, cf. *gwelh1– ‘die’ (item 26) with q- for PIE *gw– opposing k- for *-, e.g. *k̂uon (item 72) and pronominal kut ‘who’ for PIE *ku– (traditionally reconstructed as PIE *kw-, cf. Bjørn 2016: 9ff.); Attractive, of course, in light of the Indo-Uralic hypothesis, the ramifications to Uralic linguistics of this analysis are palpable as it demands an expansion of the traditionally reconstructed phoneme inventory of Proto-Uralic, and it is important to stress the caution with which this proposal is put forward, including this author’s very limited familiarity with the Samoyedic languages in general, and Selkup in particular. Confounding factors include some variation within the interrogative pronoun [see further in the comments below]. Rédei suggests further connections among the steppe languages, cf. Proto-Tungusic *kcolo ‘fish’ (UEW: 119), that similarly could provide a comparable phonetic environment. Within such a paradigm, ancient loan relations may work in either direction. A substrate origin has been suggested by Schrijver (2001: 423), but it has to be very old to account for widespread Uralic occurrences, incl. Samoyedic, as well as a diverging PIE dialect continuum. A very early loan Uralic > PIE seems to be the most probable solution, although chance resemblance in some of the more southern IE branches may strengthen the case for a later borrowing with more limited regional distribution in Northeastern Europe.

Published by


PhD stud. at the Max Planck Institute for the Science of Human History (Jena) MA in Indo-European Studies (2017) from the University of Copenhagen. Graduated with the thesis "Foreign elements in the Proto-Indo-European Vocabulary" that forms the basis of my blog. Particular fields of interest are the formation and dissolution of PIE with reference to Uralic and Caucasian languages.

2 thoughts on “Catfish”

  1. Your suggestion that PIE *Kʷ ~ Selkup /k/ seems to be refuted right away: ‘fish’ in Selkup is ⁽*⁾quələ with /q/, not /k/ as your tentative *kʷala would predict.

    /k/ versus /q/ depends in a majority of cases on the following vowel: /k/ appears before original front vowels and *u, /q/ before original non-close back vowels (including Proto-Samoyedic *ə). This distribution has often been disturbed by later vowel changes though — thus in /quː-/ ‘to die’, from PSmy *kåə-.

    On the other hand, PU *a is reflected as a labial vowel either conditionally or unconditionally in a majority of branches, often merging with *o. It could be quite feasibly reconstructed as labial *å to begin with, as has been sometimes suggested (see e.g. Janhunen 1982, On the structure of Proto-Uralic; Finnisch-Ugrische Forschungen 44).

    1. Dear Juho,
      This is the exact reason I decided to upload the content to a blog format. Thank you so much for your input, incl. references (I think I have a print of Janhunen somewhere).
      I haven’t looked at this far-fetched hypothesis for quite a while, and I have never been completely convinced myself either (as I hope stands clear from my presentation). With regards to PIE labiovelars (LV) and their tentative represenation in Selkup, this word, like most Selkup comparanda, has q- as it should according to the idea (or am I confused?). What’s curious is that there are very few Uralic comparanda to PIE words with palatal or plain velars (*k’uon ‘dog’ is a possibility, but A LOT of pre-proto phonology is in play, so it’s obviously iffy territory, but I consider it our job to formulate the best explanation to observed phenomena). Could they have palatalized (except in front of *u, hence Uralic dog *k- and the pronominal difference between animate kut and inanimate q-; I have submitted an article discussing PIE vs PU pronouns where I argue that the PIE interrogative complex is based on *ku- rather than a true LV *kw, based on an old paper) while the LV were retained as plain velars except in Selkup where the distinction was retained, just like Sam. *t and *c’ (looking at Helimski’s Selkup ch. in Routledge Uralic, p. 550)? I passed it by Salminen who conceded that the q – k distinction remains problematic, but I’m obviously very keen to see an explanation. I do my utmost to get the full picture, but it is very likely that I’ve missed crucial pieces of the puzzle; do you know where else the topic is treated? It would be nice to put it to sleep 🙂
      So, thanks again for the inputs.
      Best regards,

Leave a Reply

Your email address will not be published. Required fields are marked *