While learning (a little) Cantonese, I was annoyed by the fact that every initial [n] was converted to [l], so that the word "you", written

néih hóu

in guidebooks is universally pronounced

léih hóu

People would, if pressed, say that the pronunciation is "néih hóu", and that they were using "lazy pronunciation" (see Cantonese phonology). The sound-shift was common to words with initial [n]. It did not appear in the superficial list of cross-linguistically regularly observed sound changes that I saw on Wikipedia, which were b>p>f, d>t>th and g>c>ch. These sound changes are just voicing/unvoicing toggles and turning into a fricative, both of which are simple transformations. Initial n to l is just strange (and, possibly coincidentally, it relates the semitic and Indo-European words for "no").

I was wondering if [n] -> [l] is a sound change which is common cross-linguistically, whether it is directional (is it observed going the other way), if it is confined to the initial position, and more generally which transformations are known with what direction and probability.

  • 3
    Is there a comprehensive cross-linguistic list of regular sound changes? Anyone has any references? – kamil-s Mar 15 '12 at 9:35
  • 1
    Could you provide a link to that list of regular sound changes you mentioned? – Otavio Macedo Mar 15 '12 at 12:06
  • 3
    A "regular" sound change is one without exceptions. Asking whether initial-/n/-to-/l/ is "regular" would mean asking whether all the words that used to start with /n/ now start with /l/ in Cantonese. Is that what you want to know? Or are you (for instance) wondering whether it's a cross-linguistically common sound change? – Leah Velleman Mar 15 '12 at 16:54
  • 1
    It occurs syllable-finally in Thai, as in PhumipoN AduNyadet. – Mechanical snail May 23 '12 at 23:40
  • 3
    I can't say much about other languages, but it's really common among the Chinese languages. Nanjing dialect, Sichuan dialect, HK Cantonese (but not all varieties of Cantonese), etc. This usually bleeds over to their Modern Standard Mandarin pronunciation. – Stumpy Joe Pete Jun 22 '12 at 9:59

I'd actually been preparing the latter part of your question myself: what diachronic sound changes are common cross-linguistically, and are there databases documenting their occurrence? (I was prompted by remarks about identifying common changes from personal experience working with the comparative method). I haven't gotten around to posting the question, but I did stumble upon a database which is relevant:

The DiaDM Project: A web-based platform for Diachronic Data and Models, with one of the three sub-projects covering historical sound change: UniDia.

Although determining to the elaboration of models that are directly applicable to (real) case studies, there are no compendia that explicitly survey the types of sound changes, their directionality, their relative prevalence, their immutable tendencies and the dimensions structuring their specific behaviours. All this is admittedly of common knowledge, but for someone external to the field - mathematicians, biologists or physicists that could inspire new quantitative approaches to recurrent issues in historical linguistics - it is not. Moreover, accessing this knowledge can prove difficult and excessively time consuming as it is scattered in a multilingual literature and transcribed under various and diverging symbolisms.

UniDia targets this gap by associating the compilation of published sound change hypotheses from a large geographically and genetically diverse sample to distribution and tendency analysis methods. The symbolism is unified and all transcriptions are harmonized to the IPA – transcription conventions are presented on the left-hand menu bar. No prior typology is imposed on the data but rather, it is the aim of UniDia to explore the overall consistency of the sound change hypothesized across language areas and families and to distinguish the specific from the ‘universal’.

This of course doesn't (yet) cover all language families (and the interface isn't ideal), but it looks interesting.

P.S. meta - does this more general question merit a separate post or should it be answered here?

  • The database looks like it will completely answer this question, once it goes online. I am stunned by the amount of data! On the order of 5000 change events! This is enough for good statistics. thanks for the answer, +1. If it were online already I would accept. – Ron Maimon Mar 16 '12 at 2:56
  • I figured I can't get a better answer, since this database seems to be in the works. Thanks again. – Ron Maimon Mar 23 '12 at 2:41
  • 2
    Four years later, the web site is still "under construction". Looks like a daed project, alas. – jk - Reinstate Monica Aug 2 '16 at 11:10

A sound change can only be "regular" with respect to a particular language. There are sound changes in languages which seem completely off the wall to me, but are "regular" in the sense that they are consistent with the diachronic evidence (i.e. the forms in a daughter language can be derived regularly from rules applied to a mother language). As such there is no list of "regular sound changes", where a sound change is somehow impossible if it is not in the list.

As for the body of your question, [n] and [l] are said to be in free variation in Cantonese, which means that they are used interchangeably by native speakers, who consider both pronunciations acceptable. There are also some socio-linguistic factors involved — the [l] pronunciation is associated with female speech.

[Incidentally, I personally say nei5 for 你 'you', but loeng4 instead of the "correct" noeng4 for the word 娘 'young girl'. This particular pair of pronunciations is very much in flux at the moment, and Cantonese may eventually shift largely to one pronunciation over time. Cantonese used to distinguish retroflex [ʂ] from apical [s] (e.g. 傷 shoeng1 'wound' vs 相 soeng1 'mutual'), but that distinction too has gradually been erased over the last 50-100 years.]

An example of a quirky sound change, and I really do enjoy these, is from Taishanese, a Chinese topolect which, like Cantonese, is also spoken in Guangdong:

t -> ∅ / #_

b -> ∅ / #_

  • 3
    Another quirky sound change, but one well-attested many times in many language families: [kʷ] -> [p] in Welsh and other so-called P-Celtic languages. This is why the Welsh word for "who" is pwy, cognate to Latin quis. – jogloran Mar 15 '12 at 10:17
  • 3
    My favorite one is illustrative of the NW NA Sprachbund. Three geographically separated languages in the NW culture area (Lushootseed on E Puget area, Quileute on the Olympic coast, and Dididaht on Vancouver Is) simultaneously, and recently, changed all the nasal resonants into voiced stops. m became b, n became d, just like a cold in the head. – jlawler Mar 15 '12 at 16:39
  • 3
    To top it off, the three are of three different language families. Lushootseed is Salishan, Dididaht is Wakashan, and Quileute is the sole remaining Chimakuan language. And the change can be shown to have come after contact with Europeans was initiated, at least in Lushootseed. – jlawler Mar 15 '12 at 16:40
  • @jogloran: that change has happened independently at least three times in the development of Indo-European languages: in (pre-classical) Greek before back vowels (kwod -> pou); in P-celtic as you say; and in Romanian (quattuor -> patru, aqua -> apa). – Colin Fine Mar 15 '12 at 17:17
  • 2
    @RonMaimon: X -> Y / #_ means X becomes Y in initial position. – jogloran Mar 15 '12 at 22:23

[n] and [l] are similar acoustically and aerodynamically, in that both are made with periodic vibration and no buildup of pressure, but have more complex formant structures than vowels due to the presence of antiformants. They are also similar articulatory, both being made with the front of the tongue raised towards the alveolar ridge. The only difference in tongue shape is that the tongue body is narrowed laterally in the production of [l] to allow air to flow over the edges of the tongue. Given their phonetic similarity, it is not surprising if they change into one another. Interesting work on phonetic similarity can be found in Jeff Mielke's phonetic similarity database. A more technical treatment will be found in Stevens' (1998) Acoustic Phonetics.

Unfortunately I cannot find a good database of attested sound changes in the languages of the world. Maybe the closest thing that can be found is a database of phonological alternations, which are themselves "sound changes in progress." The sound change is however quite frequent cross-linguistically in my estimation. What is unusual about the Cantonese case is that the change has resulted in massive homophony, which typically doesn't happen to such an extent.

The most interesting part of the question concerns the transition probabilities of the change and its reverse. I have no idea about this but the question could be answered with a rich enough database.

  • Meilke's phonetic similarity database places "f","v", and "ch"(German ch, chi) about equally distant from "p","b", and "k" and these closer to each other than to any of the fricatives, when I know from coinciding sounds in Hebrew letters and reconstructed PIE that the transition p>f(unvoiced) b>v(voiced) k>ch(unvoiced) is common. There is also the voiced analog of k>ch, g>r, (which theoretically could occur in languages where r is the voiced version of ch), but postulating "g>f" is probably nonsense (it should require 3 consecutive transitions g>k>p>f, or g>b>v>f, with a different permutation) – Ron Maimon Mar 15 '12 at 20:00
  • I (amateurishly) believe that the African click is close to the nonoccuring consonent cluster "tl", and is close (in different directions) to "kl", "ts" (like the Hebrew tzadik), "tsh" (English ch), "dzh", "zh", and so on, by various steps. This transition table would allow a quantitative distance between consonents, which would place "m" and "tl" (click) as maximally distant, "b" and "p" as neighbors, and "n" and "l" (surprisingly) as neighbors. I have nowhere near enough data to make such a table, I was hoping someone else does. Thanks for the answer, by the way, +1. – Ron Maimon Mar 15 '12 at 20:06
  • 3
    "The" African click? There are dozens of them, alas! – Mark Beadles Mar 17 '12 at 20:22
  • 1
    @jknappen the organizer seems to have changed to a post at NC State English department. Check whether he's relocated the databases to the new university's server. english.chass.ncsu.edu/faculty_staff/jimielke – user483 Aug 2 '16 at 16:22
  • 1
    @jknappen it won't hurt to contact the author if you have some research that could be improved by the database. – user483 Aug 3 '16 at 20:45

From Chapter 2 of Historical Linguistics: An Introduction by Lyle Campbell:

(8) In Old French livel (from which English borrowed level), the sequence of two l’s dissimilated, giving nivel, which became Modern French niveau ‘level’ through subsequent sound changes which affected the final l.

So there's a [n] <- [l] case!

  • Another case going in this direction: Ojibwe has inini, from Proto-Algonquian *elenyiwa (man). – OmarL May 10 '19 at 8:39
  • Compare Spanish nivel, Catalan nivell, Galician nivel, Portugese nível, Albanian nivel, Romansch nivel/livel, but not Italian livello or Venetian łiveło. It's interesting how the [l] -> [n] case seems to be the norm, rather than the exception in many Romance languages. – Toby Mak Aug 4 '19 at 3:45
  • But this is dissimilation, so not really comparable to the OP's example. – TKR May 10 '20 at 20:22

I believe that l>n and n>l are rather easily explained through nasalization. Nasalize an 'l' and you get 'n.' Many language families have dialects that vary in degree of nasalization, and this creates an alternation between the two sounds in various dialects. This clearly explains the French case cited above as well as the Algonquian case -- Algonquian dialects vary widely in degree of nasalization and this corresponds to the prevalence of both l-dialects and n-dialects among Algonquian languages. The same alternation is present in the Tungusic and related Amuric languages of eastern Siberia, which are related to Algonquian through common ancestry. It is even possible that Cantonese borrowed this distinction from ancient Amuric.

  • Welcome to Linguistics! This post would benefit from adding further details. It suggests an interesting and convincing concept, but with no prooflinks, it may attract downvotes and criticism. Please edit it to add further references to credible sources. – bytebuster Nov 2 '19 at 21:22

There is a good reason why such a database (of sound changes) doesn't exist. As Crowley and Bowern 2010 argue, "given a sufficient period of time, any sound can change into any other sound by a series of changes such as those discussed in this chapter [chapter 2, Types of sound change; emphasis mine - AB]" (p. 48). How sounds are changed is more important than what sounds are changed.

  • 1
    Of course any sound can change into any other, but by a series of transformations, which are locally sensible, for example click>ts(guttaral)>tsh>dzh>zh>sh>s>z, or b>p>k>f>v. It is extraordinarily unlikely that s changes to g by one step, so much so that I would call it impossible. I want to know which sounds are close, and which sounds are far. – Ron Maimon Mar 15 '12 at 20:38
  • I'm not sure what you mean by "which sounds are close/far". However, you might want to have a look at the IPA charts langsci.ucl.ac.uk/ipa/ipachart.html – Alex B. Mar 15 '12 at 20:45
  • 1
    I think the database could easily exist. Any historical linguistics textbook will give a nice-sized list of examples of sound changes, but what we are interested in here is one which can help to compare relative frequencies of different sound changes. We are mostly interested in sound changes involving one or just a couple of steps. – user483 Mar 15 '12 at 21:04
  • 1
    Compare direction. @RonMaimon wondered about directionality in /n/-->/l/; rhotacization, for example, usually only goes one way. – user483 Mar 15 '12 at 22:23
  • 3
    @jlovegren, as a matter of fact, there is assibilation of r ("reverse rhotacism"), e.g. Polish and Icelandic in Sole 1992:260, French in Lodge 2004:140, Liddicoat 1994 etc). The most famous example is French chaise<=chaire. – Alex B. Mar 16 '12 at 1:02

[l] is related to [n] the same way as [b] is related to [m]. And [b]->[m] changes are common (for instance, PIE -bhios ending becoming -mos in Slavic and Germanic).

The PIE has l/n and r/n alternations in some stems, for instance, in the word for Sun (sea̯u̯ol).

  • [d], not [l], is to [n] as [b] is to [m] (plosive : nasal). Also it's far from uncontroversial that those two PIE case endings are related. – TKR May 10 '20 at 20:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.