Are "whores" and "horse" homophones?

In most varieties of English, these two words are not homophones. But there is an interesting story about why this is so.

In English the alveolar fricatives /s/ and /z/ are contrastive. Notionally, the first is unvoiced and the second voiced. So we can find minimal pairs of words where the difference in voicing results in a change of meaning:

• /su:/
• /zu:/

The first word here is the word sue or Sue, the second is the word zoo.

In a language like Spanish, there is only one alveolar fricative, /s/. This is usually voiceless (although it can become voiced in certain environments). However because there is no phoneme /z/ in Spanish, if a speaker uses a [z] instead of an [s] in a word, it will still be recognised as the same word. So, for example, English speakers of Spanish may well use a [z] in all sorts of words where a Spanish speaker would use an [s], but it will always be recognised as the Spanish phoneme /s/. In other words a Spanish speaker will always map both an [s] and a [z] sound onto the Spanish phoneme /s/.

Because of this lack of contrast in Spanish, it is often difficult for Spanish speakers to differentiate between /s/ and /z/ when listening to English.

Now, many English speakers will tell you that the sound at the end of the word whores is voiced. In actual fact, if it is not followed by a word beginning with a voiced sound, then the /z/ at the end of the word whores will be at least partially, and quite probably completely devoiced. We would show this in a narrow transcription by putting a little circle as a diacritic under the z:

• [hɔ:z̥]

  ^{British English transcription}

A native English speaker will swear blind—I predict at least ten comments under this post saying exactly this—that they can hear the voicing in the sound at the end of the word whores when spoken naturally. However, if you cut that devoiced /z/ off the end of the word and play it back to them without the rest of the word they will confidently tell you that it's an /s/.

Here is what is really happening when a native speaker hears the word whores and why they can differentiate it from the word horse. Unvoiced consonants have an effect on preceding vowels, whereby they cause them to become much shorter. This effect is known as prefortis clipping. If you say the word bead and then the word beat you should be able to hear that the vowel in the word beat is much shorter than the vowel in the word bead. [Try it at home!] It is the unclipped, and therefore relatively long vowel in the word bead that tells your language brain that the phoneme at the end is a 'voiced' /d/ and not an unvoiced /t/. Your brain will trick you into thinking that the sound at the end of the word bead is voiced, but this is just because of the length of the vowel.

So, if a Spanish speaker listening to the word whores and horse puts all their concentration into listening to whether they can hear any voicing in the fricative at the ends of these words, they will not be able to detect any. However, if they concentrate on the contrasting lengths of the vowels, they will notice that the vowel in horse is much shorter than the vowel in whores. This is actually what leads English speakers to hear the difference between the two words.

The story above is true for both standard British and American English (both words will have an /r/ of course in most varieties of American English). However, there may well be varieties of English which use different vowels in each of these two items.

Although the phonetic transcription in a dictionary may be the same, I do hear a difference in most native accents.

• Start: I'd say the H in Horse is a bit stronger than the WH
• Middle: I'd say the O in horse is more of an oe and shorter than in whore.

You would probably hear more difference in some accents than others, though. My2¢

EDIT. This comment by @tchrist explained it far better, in a more technical way:

The voiced /z/ of the WH- word triggers a longer /o/vowel via regressive assimilation, which triggers more pronounced rounding of that /o/, which in turn triggers greater rounding of the initial consonant than occurs in the H- word, again via regressive assimilation

In addition to the excellent points made above, one interesting example of how pronunciation shifts over time is that "whore" and "hour" were homophones in Shakespeare's day, which casts a rather different light on this passage from As You Like It:

And then he drew a dial from his poke,
And, looking on it with lack-lustre eye,
Says very wisely, 'It is ten o'clock:
Thus we may see,' quoth he, 'how the world wags:
'Tis but an hour ago since it was nine,
And after one hour more 'twill be eleven;
And so, from hour to hour, we ripe and ripe,
And then, from hour to hour, we rot and rot;
And thereby hangs a tale.

The forms 'whores' and 'horse' are not homophones in a standard (English/American/Australian) dialect and etymologically speaking the 'w' is a late (C15/16) spurious addition so should not make any difference to pronunciation. However, children learning a rare and archaic word from books/reading may pronounce as 'wars' or transfer stronger rounding from the functional wh* (pronoun/adjective interrogative/relative words - esp. 'who' which originally have a hw* or earlier kw* or qu* type origin). See e.g. etymonline.com under 'whore' and 'wh-'.

The big difference is the longer long 'o' in 'whore' versus the shorter long 'o' in 'horse'. While in both cases the 'o' is long, it is shorter before 's' (or an unvoiced stop) and longer still due to the prevocalized plural /s/ or [zs] (say 'horses' aloud and you should hear your voicing pulse on and off again like [szs]).

Thus the second subtler difference is the (pre)voiced plural morph /s/ = [z] or [zs] in 'whores' (bleeding but diminishing voicing) versus /s's/ = [s's] or [szs] in 'horses'.

As tchrist points out in a comment on another post, the lengthening of the vowel tends to allow stronger rounding and thus licenses the lead 'w' or 'whore' (backward assimilation). Similarly the vowel is more open/lower (bottom lip and jaw drop more). I see these phenomena as deriving from more time to pronounce the word including more time to produce the rounding and height features of the vowel, starting during the preceding consonant.

The plural morpheme takes on the unvoiced form [s] only after an unvoiced consonant, including when doubled as an extra syllable in 'horses' where there is actually no vowel corresponding to the 'e'. The 's' in both 'horse' and 'horses' is unvoiced, so [s] not [z].

Note that quotes as 's' represent what is written ('orthography'), slash bracketing represents conventional (phonemic or morphemic) base representation corresponding to what people think they are saying: /s/ by convention used for the plural morpheme though phonemically arguably /z/ is more appropriate, before any phonetic variation (free variation or coarticulation related to context that changes the phonetic form/morph), while square brackets represent more accurately what is actually pronounced: [z] is the plural morph after vowels or voiced consonants, and [s] is the plural morph with voicing lost after unvoiced consonants similar to the usual pronuncation of the phoneme /s/ in general, including when syllabic ('rose' involves /z/ not /s/ --> [z] with 'roses' as [z'z] versus 'horses' which is /s/ --> [s] and [s's].

Here ['z] and ['s] and ['n] etc. indicate a syllabic form often conventionally replaced by a schwa (rotated e for a minimal vowel representing voicing punctuating but nor interrupting the [szs]) and correctly written with the prime symbol underneath the consonant (not sure I could insert the unicode here).