Sandhi is the general name for a number of effects in which sound changes at the boundaries between morphemes (the minimal units of meaning in language, like the in-, describe, and -able mentioned earlier), depending on the sound or grammatical function of neighboring morphemes. The word sandhi is from the Sanskrit संधि, "joining".
There are two forms of sandhi. Internal sandhi is what happens when two morphemes join together and the sounds in those morphemes affect each other in some way. An example of internal sandhi would be the change of /n/ to /m/ in the word <sympathy> which comes from the Greek roots <syn> and <pathy>.
External sandhi is a change that happens at a word boundary rather than inside a single word. All languages will do this to some extent, but it is often not reflected in that language's writing: <ten books> is usually pronounced as /tEm bUks/ in fast speech and features the same /n/ to /m/ change as with <sympathy>.
The word ‘sandhi’ is a term originally used to refer to a set of Sanskrit phonological processes, such as how certain combinations of vowels happen to change when they meet at morpheme boundaries: <hita> + <upadesha> → <hitopadesha>, ‘friendly instruction’.
So it's all very well to say that sounds change when they meet, but we still need to understand how they change. What kinds of transformations can occur? What patterns are there?
Perhaps the most frequent form of sandhi that languages have is called assimilation. Assimilation happens when two different sounds come together and one of the sounds changes to become more like the other. We've already seen an example of this in the word <sympathy> where the /n/ in <syn> changed its place of articulation from dorsal to labial to match the labial consonant /p/ at the beginning of <pathy>.
Similarly the word <assimilation> is itself an example of assimilation. It comes from the Latin roots <ad> and <simulo>. The /d/ changes its manner of articulation to become a fricative and it also changes its voicing to become unvoiced. These are both properties that /s/ has, and it so happens that an unvoiced dorsal fricative is /s/ meaning that the assimilation is so complete that the /d/ is completely swallowed up by the /s/ and the only remnant left is a doubled <s> in the written form of the word.
Another common form of sandhi is lenition, which means ‘softening’. Lenition usually occurs to consonants when they get surrounded by vowels. If you can remember back to the sonority hierarchy, you might remember that vowels are more sonorous than consonants and you can think of lenition as consonants assimilating to the higher sonority of the vowel that they are now butting up against.
Lenition can be broken down into several more specific processes where certain specific kinds of consonants undergo certain kinds of changes:
- voicing, which involves a consonant moving from unvoiced to voiced, as with /k/ → /g/
- affrication, which involves a stop becoming an affricate, as in /t/ → /ts/
- spirantization, which involves a stop or affricate becoming a fricative, as with /ts/ → /s/, or /b/ → /f/
- debuccalization, which involves moving the place of articulation, usually to the back of the mouth, such as /s/ → /h/
No matter the specific details, all of these involve taking a consonant from lower down on the sonority hierarchy and moving it a little further up.
The opposite of lenition is fortition. While less common, it does occur; as in the Italian /kasa/ ‘house, home’ → [ka:sa] but /a kasa/ ‘at home’ → [ak:a:sa].
In tonal language the tones will often interact via a process called tone sandhi, where one of the tones will become another under certain situations, often under the influence of other tones.
Here a few examples from modern Chinese (Mandarin, Pinyin writing system) :
When any number of tone 3 syllables directly proceed a tone three syllable the first syllables will change to a tone 2.
- xiang3 mai3 hao3 ma3 -> Sandhi -> xiang2 mai2 hao2 ma3
Another language exhibiting complex tone sandhi is Amoy.
Quoted from Wikipedia : Tone sandhi
"Amoy has five tones, which are reduced to two in syllables which end in a stop consonant. (These are numbered 4 and 8 in the diagram above.) Within a phonological word, all syllables but the last one change tone. Among unstopped syllables (that is, those which do not end in a stop), tone 1 becomes 7, tone 7 becomes 3, tone 3 becomes 2, and tone 2 becomes 1. Tone 5 becomes 7 or 3, depending on dialect. Stopped syllables ending in /p/, /t/, or /k/ take the opposite tone (phonetically, a high tone becomes low, and a low tone becomes high), whereas syllables ending in a glottal stop (written h in the diagram above) drop their final consonant to become tones 2 or 3."