This is the second post in a series using Terry Crowley and Claire Bowern's An Introduction to Historical Linguistics as a base to provide a brief introduction to the method to interested parties on the web. The first part is here.
Sounds in human languages change in more-or-less predictable directions. Sound changes are the result of the sum of individual decisions, and so they aren't totally predictable - we could all get together and decide to say 'telephune' instead of 'telephone', for instance, and that would be an isolated and pretty peculiar sound change. But in general they move in similar directions, largely because humans tend to use the same or similar criteria in determining their actions. As I mentioned in my last post, simplicity and relevance are key factors in human communication, and the maximisation of the effort/effect ratio in speech is probably behind much phonological change.
For instance, [p] changing to [w] is a common and well-documented sound change in many unrelated languages. Crowley uses the examples of Uradhi, an indigenous language of Queensland, and Palauan, the national language of Palau, to make the point. An asterisk indicates that the word has been reconstructed. The sound change is from the reconstructed protolanguage to the modern language. We'll come onto how protolanguages are reconstructed later on - these kinds of sound changes are integral to the method, of course, and the process can only be appreciated if you have a real grounding in the types of sound change that there are. (For interest's sake, Uradhi is a Pama-Nyungan language and Palauan is Austronesian. They aren't related, so the sound change can't be something specific to any one language family. Most of Crowley's examples come from Australia and the Pacific.)
*pinta ---- winta 'arm'
*pilu ---- wilu 'hip'
*paqi ---- waɁ 'leg'
*paqit ---- waɁǝð 'bitter'
[p] also changes into other sounds, of course, and if you know Indonesian you may recall that the word for 'bitter' is pahit, preserving the initial [p]. It's quite common for [p] to turn into [f], [b], or [v]. But as Crowley notes, it never seems to turn into [l] or [z] (except by way of other sound changes, perhaps). The fact that there are usually a few alternatives is why sound change is not perfectly predictable even though we can generalise about the direction of change. This is one of many reasons why we can't reconstruct the first language on earth. After a certain point, there are too many alternative sound changes to make accurate reconstruction reliable or verifiable, even with the most advanced statistical methods.
Anyway, Crowley uses this to introduce the first type of sound change: lenition, or 'weakening'.
Some sounds are stronger than others. Most people would say that [p] is a stronger sound than [f] and that [f] is stronger than [h]. Linguists have generalised the properties of phonemes into a hierarchy of relative strength and weakness along these lines. This is known as the sonority hierarchy. The sonority of a sound is down to a combination of 'the loudness of the sound, pitch, and the articulatory effort' (Crowley and Bowern, p. 24). The hierarchy, to copy directly from Crowley, goes like this:
a > e,ɛ > o > i,u > rhotics > laterals > nasals > voiced fricatives > voiceless fricatives > voiced stops > voiceless stops
Wiki gives the following scale:
([a] > [e o] > [i u] > [r] > [l] > [m n] > [z v ð] > [s f θ] > [b d ɡ] > [p t k])
The most sonorous sounds are on the left, the least sonorous on the right. The tendency for change is from less sonorous sounds to more sonorous ones, or from stronger sounds to weaker sounds (ie, the direction on this chart would be from right to left). [p] is more likely to become [f] than [f] [p]. Parts of the hierarchy seem counter-intuitive - [z] has always seemed stronger to me than [s], as it is voiced - but consider articulatory effort and it becomes easier to understand. [s] does seem to take more effort to produce.
The opposite of lenition is fortition. This is sound change on the same scale but in the opposite direction along the sonority hierarchy - from weak sounds to strong sounds. This is rarer, but it is possible, especially in combination without other sound changes (sound changes have effects on other sounds, similar to the manner in which genetic mutations can have repercussions throughout the genome). In the comparative method, which is the normal means of reconstructing past languages and determining relationships, we can look for evidence of lenition or fortition in the pronunciation of the different languages we are comparing, so this isn't a blind process. Principles like sonority hierarchy give historical linguists a guide in looking for relationships between languages. They're not taken on faith.
A specific kind of lenition is called rhotacism. That's where [z] or [s] sounds (voiced or voiceless dental sibilants) change into [r] sounds (rhotics). If you look on the chart, you'll see that it is hard to account for this change wholly in terms of the sonority hierarchy - rhotics and sibilants have a couple of steps in between. Sonority and lenition are more complicated than the simple diagram above. Crowley uses the example of the Latin genitive, but he also provides a much more familiar example:
'Were' is, of course, the plural form of 'to be' in the past. Crowley infers that the 'e' at the end of 'were' once indicated a plural ending, and that the simple past plural of 'to be' may originally have been something like [wase] or [wese]. Lenition acted on the [s] to make it into an [r] as a result of the plural suffix, something that didn't occur in the singular form. (This should give some idea of how complex the inter-relations between phonemes is in configuring sound changes!)
Crowley spoke Australian English, which, like the British English I speak, is arhotic. In dialects like ours, the process has gone even further: 'were' is now pronounced [wɜ:]. We've totally lost the [r]. That will be the subject of the next installment in this series: sound loss.