This is the third part in a series on historical linguistics. Part I - here. Part II - here.
You may have twigged to the fact that much historical linguistic work is applied phonetics. Better understand the nature of sounds in human languages and you'll better understand how they change. Better understanding how sounds change can help you to reconstruct forms used in the past and provide a reasonably reliable image of past times. Sounds almost always change in predictable ways - or, perhaps more importantly for the work of historical linguists, in retrodictable ways. While historical (diachronic) linguists study change in general, most historical linguistic work consists of reconstructing common ancestors of known languages and is therefore retrodictive rather than predictive. Anyway, it's mostly applied phonetics. If you know synchronic phonetics inside-out, applying it to historical cases is quite easy - as we'll see later, this is especially true when looking at fusion and assimilation. In this post, however, we'll be looking at sound loss, which is simpler than fusion if you don't know phonetics so well.
Sound loss is a kind of lenition, because a sound can't really get weaker than non-existence. There are several different kinds of sound loss. Sounds at the ends of words are commonly lost - think of the British/Australian/New York pronunciation of 'car', which has lost the final [r], or French words like cent and chat which have lost the final [t] that was presumably present when French spelling was standardised. Word-final segment loss is known as apocope, and it's probably the most common type. Crowley gives us another example, this time from Vanuatu (Crowley was an acknowledged authority on Vanuatu's languages, including Bislama):
*utu > ut 'lice'
*aŋo > aŋ 'fly'
*asue > asu 'rat'
...and so on. Apocope is easily explicable in terms of the maximisation of effect versus effort; why bother saying the end of the word when context and the prior syllables should have already done the work of communicating the idea for you?
Less explicable is apheresis, which is also much less common. Crowley says that it is especially common in Australian languages, which is a peculiar and interesting phenomenon. It's the inverse of apocope: the loss of an initial segment. Presumably it has to do with stress and rhythm compensating for the loss of the initial part. Crowley's example comes from Angkamuthi, a Pama-Nyungan language of the Cape York Peninsula. Ethnologue tells me that it is extinct.
*maji > aji 'food'
*nani > ani 'ground'
*ŋampu > ampu 'tooth'
...and so on. All the above examples involve the loss of nasals, but this isn't necessarily the case. You could also see the pronunciation of 'specific' and 'specifically' by some British people, in which the word has lost its initial [s] and become a homophone of 'pacifically'. (Mocked here by Mitchell and Webb.)
Syncope is next kind of sound loss. That's the loss of middle segments. 'Specifically' contains an example of syncope - the loss of the [a] in the adverbial suffix (no one says [spə'sɪfɪkæli]). It's fairly common in English, but is found more regularly in other languages. Crowley gives an example provided by Robert Blust, perhaps the greatest of all Austronesian linguists. This comes from proto-North Sarawak, a protolanguage spoken in prehistoric Borneo. The vowel deletion here was the subject of Blust's 1974 PhD dissertation, but Crowley draws the material from a 2002 publication (Kiput historical phonology, Oceanic linguistics 41(2): 384-439).
*baqeRu > *baqRu 'new'
*eRezan > *eRzan 'notched log ladder'
(/R/ represents any liquid - [l] or [r] or anything in between. There seems to have been no distinction between liquids in proto-Malayo-Polynesian - they were allophones.) Crowley also mentions sprite and spirit, both derived from Latin spiritus. 'Sprite' displays syncope (and subsequent diphthongisation).
The next kind of sound loss is cluster reduction. 'Cluster' here refers to consonant clusters, like [spr], the kind of thing Chinese speakers have a problem with (Chinese has no consonant clusters; imagine the nightmare of learning to say the word 'crisps' if you don't deal with consonant clusters regularly). Reduction of consonant clusters refers to the tendency to say 'hambag' instead of 'handbag' or 'goverment' instead of 'government'.
Crowley - again, he worked extensively on south Pacific creoles - notes that 'government' [gʌvənmɛnt] (in American English: [gʌvərnmɛnt], preserving the [r]) typically turns into [gʌvəmɛnt] in English, but syncope sometimes does its work too, such that it is sometimes found as [gʌvmɛnt]. In the English-based creole of Papua New Guinea (often known as tok pisin), this introduced form has turned into [gavman] - cluster reduction, then syncope, apocope, and a vowel change brought about by the structural pressure in south Pacific creoles which lack the distinction between /ɑ/, /æ/, and /ʌ/. (To me [gavman] is still recognisably the same word as 'government'!)
Finally, we have haplology. Haplology is a sporadic, comparatively rare change that occurs when two or more of the same sound occur in a word or phrase. The word 'England' is the product of a change brought about by haplology; originally, it was called Anglaland, the land of the Angles, but the two [l] sounds in one fairly common word proved too much, and were reduced to one. 'Library' pronounced as [laɪbri] is a common sort of haplology. ('February' is a bit different, as the first [r] seems to me to have undergone assimilation instead of loss.) The word 'haplology' is a bit ugly, and I think if it were more common, it would probably also undergo a haplological sound change.
Anyway, that's sound loss, which is a kind of lenition, sort of. It's quite common. I think most of the losses can be explained in terms of simplification, or maximisation of the effort/effect ratio in communication. To sum up, the different kinds are apocope, the loss of word-final segments; apheresis, the loss of word-initial segments; syncope, the loss of middle segments; cluster reduction, the, well, reduction of clusters; and haplology, the loss of one phoneme where it is annoyingly duplicated in a certain context.
Next time: sound addition, metathesis, and an introduction to the idea of assimilation.
If you're interested to see an example of the problems inherent in historical linguistics, take a look at this fantastic series by Piotr Gąsiorowski of Adam Mickiewicz University. The target is a frankly dim-witted study attempting to show that there are some words of Pleistocene heritage still spoken, detectably so, in Eurasian languages. It's excellent stuff.