Why there are three different sounds for -ed?
Preamble
Syllables have two main parts, an ᴏɴsᴇᴛ and a ʀʜʏᴍᴇ. The onset consists of any initial consonants and the rhyme is the rest of the syllable. In the word cat /kæt/, the onset is the consonant /k/ and the rhyme is the section /æt/.
The rhyme itself also has its own internal structure. The main, first part of the rhyme is usually a vowel and is more sonorant (musically louder) than the rest of the syllable. It's called the ɴᴜᴄʟᴇᴜs. The last part of the rhyme, if there is one, is less sonorant and usually consists of consonants. It's called the ᴄᴏᴅᴀ. In the word /kæt/, the nucleus is /æ/ and the coda /t/.
The Original Poster's question
The Original Poster asks why there are three different past tense endings for regular verbs in English. It's a good question. Why not just the one?
Well, there are good reasons for supposing that the past tense ending in English is basically (or underlyingly) the phoneme /d/, which gets tacked on to the end of the base form of the verb. So in the most basic case, we have a word like /kli:n/, clean, and we stick the /d/ ending on to get /kli:nd/, cleaned.
However, things get more complicated when we want to stick this ending onto words that end in a voiceless consonant. If we tacked a /d/ onto the end of such words, we would end up with words like /kɪkd/ for the past tense of /kɪk/, kick. In other words, the word would have to have a voiced /d/ like we find at the beginning of the word dogs. This isn't particularly easy to pronounce (try it), but it's by no means impossible. However, it breaks a basic rule of syllable structure in English, and indeed of almost every spoken language in the world:
- Once the voicing is turned off in a syllable coda, it cannot go back on again.
In other words, English will not permit a sequence of voiceless and then voiced sounds within a syllable coda. One way of thinking of a syllable is to think of it as a peak in sonority. In other words, the syllable gets more sonorant as it approaches the vowel and less sonorant as it moves away from it. Having a voiceless consonant followed by a voiced one breaks this basic organizational principle of syllables, as the syllable would be starting to get more sonorous again before it ended.
For this reason, we have a rule in the phonology (the rules of the sounds in the language) which replaces the ending /d/ with its non-voiced counterpart /t/ after a voiceless consonant. Notice that /d/ and /t/ are identical apart from the fact that /d/ is voiced and involves the vibration of the vocal cords, whereas /t/ is unvoiced and does not. If we tack a /t/ onto the end of a voiceless consonant at the end of the syllable, this does not break the rule described above. So, basically, to make a past tense form of a verb ending in a voiceless consonant, we tack /t/ onto the end instead of /d/ so as not to break the rules of English syllables (the phonotactic constraints of English).
There is still one problematic case, however. English does not allow ɢᴇᴍɪɴᴀᴛᴇ consonants within syllables. A geminate consonant is just a double length or 'twin' consonant. Geminate consonants are permissible in many different languages, but not in English. Now if we apply the rules described above, we should append a /d/ onto words that end in voiced consonants such as mend and a /t/ onto words ending with voiceless consonants such as the verb rest. However, this would result in the following:
- mend: /mend/--> */mendd/ (badly formed)
- rest: /rest/--> * /restt/ (badly formed)
As we can see this would result in a geminate /dd/ in the first example and a geminate /tt/ in the second. When I'm teaching English to language students, I like to point out that this would make the final /d/ or /t/ ending difficult to distinguish (and if we wanted to release the first /t/ or /d/ before the second, difficult to pronounce). However, this is not a good argument for why this doesn't happen in English. Firstly, it does not seem to be very important to hear or pronounce past tense endings in English. If the base form of a verb ends in a consonant, then we are always allowed to omit regular /t/ or /d/ endings in spoken English so long as the following word begins with a consonant. Additionally, many irregular verbs in English have a past tense which is identical to the base form or present tense, for example the verbs cut or put—we just tell which is which from the context. The real reason might seem to be just that English never allows geminate consonants in the first place.
The strategy used by English with regard to regular verbs to get round this is to insert a KIT vowel, /⁠ɪ⁠/, between the base and the ending. For this reason we get the following:
- /mend/ --> /mendɪd/
- /rest/ --> /restɪd/
Notice that now there is a voiced /ɪ/ before the final consonant we can use /d/ again. We don't need to use its voiceless counterpart.
Conclusion
In addition to /d/ we need two extra endings to accommodate the phonotactic constraints of English. We cannot turn the voicing back on once it has been turned off in a syllable coda. We cannot have geminate consonants in English either.
Transcription note
I've used Brtish English transcription of the type used by Wells in The Longman Pronunciation Dictionary. Nothing much hinges on this.
Yes. Look up sandhi, or just start at alternation.
This is nothing specific to the final -ed. It's the same reason why you say "houzez" rather than actually "houses", and a ton of other things. (Like, right there: "thingz", for that matter, and not actually "things". And if you do try saying "things" with an "s", you'll end up saying "thinks".)
If you have a consonant, it's just natural to make it voiced if it follows a vowel or a voiced consonant. Conversely, it's easier to make it unvoiced if it's preceded by another unvoiced consonant. That's just how your mouth works, really. Again, try it yourself right now. Say /gz/, then /ks/. And now try saying /gs/, then /kz/. Good luck.
Likewise for -ed. Try saying "workd" and "beggt" vs. "workt" and "beggd".
Happens in English, happens in French, happens in Russian, all over the place, everywhere. Nothing to write home about. That Wikipedia article has quite a few more examples, if you're interested.
It arose from a natural process of reduction of endings. In Early Middle English, the suffix -ed was pronounced /ɛd/, no matter what sounds it came after. But people are lazy, and we started leaving out the vowel when we could. However, after /t/ or /d/, it's very hard to say /stɑrtd/ (for example) with no vowel between the /t/ and the /d/, so we kept the vowel /-ɪd/ after /t/ and /d/. And it's easier to say /maskt/ than /maskd/, so /d/ turned into /t/ after voiceless consonants.
This happens in other languages, too. For example, the French spell the third person plural ending for verbs (ils passent) as -ent , but don't pronounce any of these three letters. That's because all of these letters used to be pronounced in Old French, and the spelling was never changed.
In Early Modern English, it appears that there were two pronunciations in use for many words: in his sonnets, Shakespeare pronounced masked as /maskɪd/ or /maskt/, depending on which one fit the meter (and distinguished them by spelling them masked and mask'd).