Why are "sugar" and "sure" pronounced with an SH?

Question

As far as I know, those are the only two. They should be pronounced Soogher and Soor, shouldn't they? I looked them up on Dictionary.com, and their etymologies reveal no trace of an SH, except where the listing for sugar had:

Middle English sugre, sucre (noun) < Middle French sucre < Medieval Latin succārum < Italian zucchero < Arabic sukkar; obscurely akin to Persian shakar, Greek sákcharon

I see an obscure kinship to shakar, but the word morphed so many times since then that the SH disappeared pretty much completely. And sure is even worse, with no sign of SH:

Middle English sur ( e ) < Middle French sur, Old French seur < Latin sēcūrus

Why are these the only two like this?

Answer 1

This interesting page explains that sugar used to be pronounced originally with a common su sound, but (emphases mine):

(...) sometime in the Middle English period the initial letters su shifted to the pronunciation they now have.

If you relax the mouth and tongue somewhat when you are saying the older form, your pronunciation shifts to the modern one, as you’ll realise if you try out the two sounds in turn; the modern version is actually rather easier for slack-jawed English speakers to say. (...) The same change happened with other words, such as sure, and also to words in which the sound occurred in the middle, such as pressure and nation. By the time this shift in pronunciation was taking place, the spelling of the words had already become fixed (...)

Answer 2

The sound of French /u/ (a close front rounded vowel) is denoted [y]. Sugar is from French sucre [sykʀ(ə)], and sure is from French sur (e) [syːʀ].

Middle English kept that vowel originally—as [sykrə] and [syːr(ə)], respectively—but over time it was transformed into [ju] (like modern you). What was previously [sy-] thus became [sju-] (RP sue) and eventually [ʃʊ-] (as in modern sugar).

This happened in other contexts: -tion is pronounced [-sjɔ̃] in French, with a nasal vowel. This became [-sjɔn] (syon) or [-sjən] (syun), and finally [-ʃən] (shun).

Answer 3

As you say, there doesn't seem to be any way the pronunciation of "s" as "sh" in sugar (and sure) could be related to the “sh” in Persian shakar. Rather, it seems to have developed (apparently irregularly) within English.

Palatalization of /sj/ to /ʃ/, which could also occur in other words historically

The use of the “sh” sound (in IPA, /ʃ/) in sugar and sure developed after they had entered English, via palatalization of the sound /s/ before the palatal glide /j/ (like “y” in “you”). As Jon Purdy explains, these words developed a palatal glide due to the change from the French vowel /y/ (spelled "u") to English /juː/.

The sound /ʃ/ occurs for the same reason in many other words related to sure, such as surety and assure, ensure etc. An analogous change affected the voiced counterpart of /s/, /z/, causing /zj/ to develop in many words to /ʒ/ (the sound at the start of the French name Jacques).

I found a good Quora post by Uri Granta that mentions a third word starting with the letter “s” that is pronounced by modern English speakers with the sound “sh” somewhat often: sumac. However, modern dictionaries also list a pronunciation with /s/ for this word, so it is not as categorical as the other two words. (In fact, the book The Pronunciation of Standard English in America, by George Philip Krapp (1919), calls [ˈʃuːmæk] for sumach [sic] a dialectal pronunciation (p 124), and says it is not standard.)

Granta explains:

Sugar, sure and sumac are the only three in Modern English, but historically there were others. In the sixteenth century a phonetic change of sy- to sh- was attested (in the shape of sh- misspellings) not just in the words sugar and sure, but also in words like suit (variously spelled shute, shutte, shuite and shuett), suet (spelled showitt, shewet and shuet) and sue (spelled shue). By the nineteenth century, the sy- pronunciations won out for all but sugar, sure and sumac (and was later replaced by plain old s-), though the sh- pronunciation also survived in the middle of some words like issue, tissue, assure, ensure, insure, pressure, etc.

Another source I found that describes some of the history of this pronunciation is English Pronunciation in the Eighteenth Century: Thomas Spence's ‘Grand Repository of the English Language’, by Joan C. Beal (1999). According to Beal,

much of the seventeenth-century evidence for the palatalization of preceding alveolar consonants [...] is from ‘vulgar’ or dialectal sources: Cooper, for instance, in his semi-phonetic notation, lists shugar with the [j] assimilated, as a ‘barbarous’ pronunciation. By the end of the eighteenth century, this palatalization, with its subsequent assimilation of the /j/, is accepted even by Walker in sugar and with palatalization but not assimilation in sure, but elsewhere it is heavily stigmatized, at least in stressed syllables. (144)

Some dictionary authors, such as Sheridan, did list pronunciations with word-initial /ʃ/ for some other words, but these pronunciations were criticized by other orthoëpists:

We have already seen that the author of A Caution to Gentlemen Who Use Sheridan’s Dictionary attacked Sheridan for his palatalizations and, in our discussion of the words in Appendix 6a, that Sheridan had more instances of palatalization than Spence or Burn before the unstressed ⟨ure⟩ ending. It is certainly the case that Sheridan admitted this palatalization of alveolars before earlier /juː/ to an extreme degree: he has palatalization of /s/ to /ʃ/ with consequent loss of following /j/ in all words with the prefix super-, whereas Spence, Walker and Burn all have /sjuː/ here, and in all similar words except sugar, sure and surety, in which /ʃ/ occurs in PDE. (150)

Interestingly, Beal considers it possible that this palatalization was a feature of Irish English (which would support the relevance of redbmk's answer):

The author of A Caution and, indeed, Jespersen (1909-49: 344, 347) attribute Sheridan’s pronunciation of words such as tune and suicide to interference from Irish English. In many cases, the accusation of ‘Irishism’ is a knee-jerk reaction from critics who simply did not agree with Sheridan’s recommendations, but, in this case, there may be some substance to it. Jespersen (1909-49: 347) points out that ‘B Shaw writes Choosda and schoopid as Irish for Tuesday and stupid (John Bull’s Isl. 12, 38)’. On the other hand, initial /tʃ/ and /ʃ/ for /tj/ and /sj/ are attested in seventeenth century English (as opposed to Irish) sources: Dobson (1957: 706-7) shows suit: Shute: shoot as homophones in Hodges; and tulip: julip as well as dew: due: jew in Brown (none of our sources shows initial /dj/ becoming /dʒ/). These instances are, however, few and far between and in homophone lists, which, as Dobson explains, often give more colloquial pronunciations. If Sheridan’s palatalizations in supreme, tune etc. are not Irishisms, perhaps they represent a more colloquial kind of London English. (150-151)

However, she thinks it is more likely that these variants arose in London:

I am informed (Karen Corrigan, personal communication) that this palatalization does occur in Irish, but only in the Ulster dialects, so that it is very unlikely that the Dublin-born Sheridan would have been influenced by it. What is more likely is that the palatalization was a colloquialism common in London, which may have been widely used in less careful speech, then as now. (150-151, note 33)

I'm not sure precisely what she means by "then as now." I have heard that some, perhaps even many, present-day British English speakers merge /tj/ into /tʃ/ (i.e. their "tune" sounds like "choon") and /dj/ into /dʒ/, but I haven't heard of any that have /ʃ/ where standard British English has /sj/ or /s/.

It's unclear why these words in particular show this change, and others don't in present-day English

I don't know of any explanation for why sugar and sure (and for some people sumac) ended up being pronounced with /ʃ/, but other words starting with ⟨su⟩ (such as suit and super) ended up being pronounced with /s/ by modern speakers.

Example words showing the variable outcomes of historical /sjuː/ and /zjuː/ in present-day English

Before a stressed syllable:

• /s(j)/ or /z(j)/: suit, assume, (resume, presume etc.) supreme, suture, exude, exuberant, esurient. I found a number of other words like these, many of them words starting with super-.

• /ʃ/ or /ʒ/: sugar, sure (surety, assure, ensure etc.), sumac (pronunciations with /ʃ/ and /s/ both exist), luxurious, luxuriant (these last 2 might have been influenced by luxury, which is stressed earlier), usurious (MW lists pronunciation with /z/ also; influence from usury seems plausible), caesura (MW lists pronunciation with /z/ also). I don't know of any other words like these.

Before an unstressed syllable:

• /ʃ/ or /ʒ/: pressure, fissure, seizure, treasure, enclosure, exposure, usury, censure, tonsure, licensure, azure, sensual, sensuous, sexual, flexuous. I found a number of other words like these, many of them nouns with the suffix -ure.

• /s(j)/: insulate, insular, peninsula, insulin, consulate, consular, capsule, encapsulate, Ursula. There are a few more words I've found like these.

• variation between /s(j)/ and /ʃ/ and/or /z(j)/ and /ʒ/: exudate, exudation, lapis lazuli, mensurable, mensural, commensurate.

Answer 4

The first part of 'sugar' used to be pronounced more like ''sue''. The vowel in that syllable is a high, back vowel and sometimes consonants change to be more like the vowels that are near them. It's called assimilation.

Answer 5

I've noticed many Scottish and Irish Gaelic words to be spelled with an s, followed by a vowel, and pronounced like sh. Think about the way Sean Connery speaks (not to mention how the Se in his name is pronounced).

Read the section on nomenclature here and click on some of the links about Anglic and Scots (Yes, I realize I'm quoting Wikipedia, but there is a source attached to this quote). Note how the s is pronounced sh:

Prior to the 15th century, the Anglic speech of the Lowlands was known as Inglis ("English"), with Gaelic being called Scottis ("Scottish"). From the late 15th century, however, it became increasingly common to refer to Scottish Gaelic as Erse ("Irish") and the Lowland vernacular as Scottis.

There is also some good information about Scottish Gaelic orthography and phonology that you might find useful if you can read IPA.