Why is "salient" pronounced with a "long a" sound?

The word salient is pronounced with a "long a" sound; Wiktionary gives the US pronunciation as /ˈseɪ.ljənt/, /ˈseɪ.li.ənt/. Is there any reason why the vowel letter here receives its "long" pronunciation (IPA /eɪ/) rather than its "short" pronunciation (IPA /æ/)?

The vowel does not seem to have been long in the Latin ancestor of the word, saliens. We use a "short a" in the word valiant /ˈvæljənt/, which has the same spelling pattern.

Solution 1:

Lengthening rule for {a, e, o, u} before CiV or CeV

The long vowel in salient is caused by a lengthening rule that originally applied in Middle English to stressed vowels followed by a single consonant (not including x) and two unstressed vowels. In most cases, the first unstressed vowel was i, as it is here, but it could also be e, as ocean. In other words, the general resulting pattern is V̄CiV or V̄CeV, where V̄ represents a long vowel (ā, ē, ō or ū), C represents any single consonant letter apart from x, and V represents any vowel letter. Notably, this lengthening rule did not apply when the stressed vowel was i: that was subject to an opposite, shortening rule in this environment (described in my answer to the following question: Why do we pronounce a long second vowel in “decide”, but a short second vowel in “decision”?)

In Modern English, the unstressed i or e in these words has often been reduced to a palatal glide /j/ or absorbed altogether by the preceding consonant. However, the stressed vowels remained lengthened. This lengthening rule may also apply by analogy to words with this spelling pattern that were taken from Latin into Modern English.

I found a quite illuminating discussion of this rule, and its historical application, in Otto Jespersen's Modern English Grammar on Historical Principles: Part 1, Sounds and Spelling (1954). Valiant happens to be an exception.

The rule only applies to some types of words

First, a note about the scope of this rule. It only applies to words of Classical origin (ultimately from Latin or Greek, possibly brought to English via French), or in some cases to words where VCiV or VCeV occurs due to the addition of a Classical suffix (such as -ial or -ium) to a non-Classical base (for example, nobelium may be pronounced in accordance with the rule, as nobēlium, or with a short “e”, as nobĕlium).

It does not apply to words where VCiV or VCeV occurs within a native English word, or due to the addition of a native English suffix such as -er or -est.

Example words where the rule does not apply:

• burial < Old English byrgels
• cŏpier < cŏpy + -er: the same would apply to any other verb ending in a short vowel, a single consonant and "y" followed by the suffix -er (a hypothetical wrd “studier,” for example)
• busier, busiest < busy + -er, -est: the same would apply to any other adjective ending in a short vowel, a single consonant and "y" folwd by the suffix -er or -est

It also doesn’t necessarily apply to un-nativized loanwords that entered English later than the Middle English period, such as Soviet. However, in some cases it’s hard to identify loanwords, so I’ll just list them along with the other violations of the rule.

List of words that violate the VCiV lengthening rule

The first two categories I’ll discuss contain very few words, but they are among the most commonly used words that violate this rule.

with "short e" before -ci-/-ti-:

• discrĕtion, prĕcious, spĕcial. Jespersen notes these, but doesn’t explain why they developed differently from other words such as matērial, impērial, spēcious. It may just be coincidence, but all of these occur before the sound /ʃ/.

ending in -ătional:

• rătional, nătional. These could be considered to exhibit trisyllabic laxing. In fact, Jespersen says the long vowels in other adjectives ending in -VCional, such as occāsional, congregātional, devōtional, are due to analogy with the vowels in the corresponding nouns, rather than regular phonetic development. Despite this history, in Modern English the productive pattern is the one with the long vowel, so these two words can be considered irregular.

The next two categories are larger, and contain some words that are common and others that are quite rare. They consist of words spelled with VniV or VliV. Jespersen says a number of words in this category don’t follow the lengthening rule because in Old French, they had a palatal consonant sound /ɲ/ or /ʎ/ rather than a Ci sequence. A number of other words spelled like this seem to be pronounced with short vowels for unclear reasons, possibly due to analogy.

before -ni-:

• compănion, onion (="ŭnion"), pŏniard, Spăniard, spăniel. According to Jespersen, from French words with /ɲ/.

• bŭnion. Jespersen doesn’t mention this word. Its etymology is obscure.

• grŭnion, cănion. Both or these are probably from Spanish words with /ɲ/ (spelled ñ). The dictionaries I have checked say the usual etymology of the fish “grunion” is from Spanish gruñón, and the rare/archaic word "canion" (also spelled “cannion”) is from Spanish cañon.

• Proper nouns: Dăniel, Nathăniel, Hăniel, Ăniel. These are names from the Bible ("Haniel" is an angel's name). I don’t know how or at what time they received their current English pronunciations.

The remaining words in this category are all unimportant, obsolete, or have alternate regular pronunciations; I’m only listing them for completeness.

1. Two words that end in -ănia, possibly:

   • Oceănia (regular pronunciation also exists). Although this pronunciation is fairly common, there are also pronunciations with “broad a” or “long a”.
   • Brittănia. Variant of Britannia. Likely to be considered a misspelling in modern English, although variation between -t- and -tt- and between -n- and -nn- in this word dates back to Latin. I don’t know if a pronunciation with long a was ever used historically.

2. Three words from French that in modern French have /nj/ rather than /ɲ/

   • mănioc. The modern French word manioc actually has /nj/ rather than /ɲ/ (insofar as these sounds are distinguished). However, the OED says the spelling “magnioc” was used (among others) in French in the 17th century, which may indicate past confusion between these sounds.
   • fănion. Rare. From French fanion, diminutive of fanon.
   • lăniard. Alternate spelling of lanyard. From French lanière.

3. frănion. Obsolete and rare. Of uncertain origin. Not used in modern times, so I’m not sure how anyone would know how it used to be pronounced, but Collins says /ˈfrænjən/.

before -li-:

• văliant, battălion. According to Jespersen, from French words with /ʎ/. Regularly-pronounced words with similar spelling: sālient, Pygmālion.

• Itălian. Jespersen doesn’t mention this word, but perhaps the same explanation applies here. The OED lists a variety of historical spellings, some of which have -ll-, -ill- or -ly-, which to me seem to suggest /ʎ/. There are regularly-pronounced words with similar spelling: Austrālian, Episcopālian, Vitālian.

• retăliation, retăliate. French /ʎ/ doesn’t seem like it would be relevant here, since these words were taken from Latin, not French. John Walker (1824) lists the modern, irregular pronunciation, so it doesn’t seem to be a recent thing. According to the OED, the noun retaliation is older. It may be relevant that in this word, the vowel only has secondary stress; it seems possible to me that this might inhibit vowel lengthening. There don’t seem to be any other words with similiar spelling to compare these to.

• tălion. (MW, Dictionary.com) Rare, but it seems it did come to English through French. Possibly this explains the pronunciation. Of course, the modern pronunciation could actually be affected by the pronunciation of the related words “retaliate/retaliation”; it’s not clear to me that the historical pronunciation of this word managed to be passed down through oral transmission.

Proper nouns:

• Vălium. It’s a brand name, so it’s not too suprising it violates a pronunciation rule. Obviously not from Old French, so /ʎ/ is irrelevant. Possibly influenced by the pronunciation of valiant. There is a regularly-pronounced word with similar spelling: dentālium.
• Ĕliot. A name. I don’t know how it got its current pronunciation. It seems it may come from Scotland originally; it doesn’t seem to be Latinate. There is a variant spelling "Elliot" that corresponds better to the pronunciation.

The remaining words in this category are all unimportant, obsolete, or have alternate regular pronunciations; I’m only listing them for completeness.

• triskĕlion. This is a synonym of triskele, from Greek. Both are very rare. It seems unclear why -ion was suffixed to the word. Regularly-pronounced words with similar spelling: aphelion, perihelion.
• Evangeliar? OED says /iːvanˈdʒɛlɪɑː/
• Evangelion? I don’t know the usual pronunciation.
• glossolălia. Only the OED lists this; all other dictionaries I’ve checked give the regular pronunciation glossolālia. I don’t know how the OED determined this pronunciation. It might actually be a mistake.

Miscellaneous other words:

• chăriot, clărion. (Only relevant for speakers without the Mary-marry merger.) Jespersen says the pronunciations of these words were influenced by now-obsolete synonyms from French, charet and clarine. There are various other words from Romance languages that are exceptions like this, such as căviar (which the OED says had a stress shift, with a variety of historical forms), cămeo (from Italian and French, with a variety of forms), lăriat (derived from Spanish “la reata") and pătio (derived from Spanish patio).

• glădiator. From the etymology, I would expect this word to rhyme with radiator; it’s unclear to me why it does not. (Actually, apparently, for some people they do rhyme, because "radiator" is pronounced with a short vowel!) The same irregularity is present in one pronunciation of the related word glădiolus (the other pronunciation has a stressed long i and reduces the vowel of the first syllable; compare the stress pattern of areola).

• găseous. The word gas is of relatively recent origin. This may explain why the a is not lengthened in this adjective.

• one common pronunciation of rătion. There is also a regular pronunciation rātion (with /eɪ̯/) that seems about as common. The pronunciation with "short a" seems like it might be due to influence from the pronunciation of the related word rational.

• one common pronunciation of hăgiography. There is also a regular pronunciation hāgiography, although it doesn’t seem to be as common. Another irregularity about the pronunciation of this word is that the “g” is frequently pronounced as “hard g” /g/ rather than “soft g” /dʒ/. It seems that the pronunciation with "soft g” only occurs for speakers who use a “long a” in this word.

• the usual pronunciation of gĕriatric (with /ɛ/, /e/ or /eə̯/). According to MW and dictionary.com, a regular pronunciation gēriatric (with /ɪə̯/, /i/ or /ɪ/) also exists, but it does not seem to be common at all.

• one common pronunciation of the prefix stĕreo-. There is also a regular pronunciation stēreo-, although it doesn't seem to be as common.

• one pronunciation of hystĕria. There is also a regular pronunciation hystēria that seems about as common. The pronunciation with "short e" seems like it might be due to influence from the pronunciation of the related word hysterical.

• one common pronunciation of sacrilegious (="sacrilĭgious"). There is also a regular pronunciation sacrilēgious (with /iː/), although it doesn't seem to be as common. The pronunciation with short stressed /ɪ/ uses a vowel quality generally associated with the letter "i" rather than with the letter "e", and seems to have arisen by analogy with the word religious (although the words are generally not thought to be related etymologically).

• one uncommon pronunciation of epidĕmiology; probably due to influence from the related adjective and noun epidĕmic. The regular pronunciation epidēmiology seems to be more common.

• Another general class of words where “irregular” pronunciations are common is adjectives ending in -ian, or nouns ending in -ium, derived from proper nouns that end in a single vowel followed by a single consonant. Often these show some variation between a pronunciation with a short vowel (which is presumably felt to be closer to the pronunciation of the independent name) and a pronunciation with a long vowel (presumably by analogy to other words in English with this spelling pattern). Examples: the elements nobelium and meitnerium, and the adjective Wagnerian, have pronunciations with ĕ as well as ē. The adjective Brobdingnagian seems to only be pronounced with “short a.” A relevant question: Adding an L when appending an -ium suffix to a word? (Metallium vs. Metalium)

Additional Sources:

• English Grammar for Dutch Schools, By C. van Teil, revised by P.C. van Heynsbergen, 1934 (provided list of some words with the spelling _CiV and exceptional pronunciations)
• Oxford English Dictionary (pronunciations and etymologies of some words)
• OneLook Dictionary Search (used wildcard search to find some words with this spelling pattern)
• Phonetic Word Search (used wildcard search to find some words with irregular pronunciations)