Conlinguistic diachronics is a fickle thing - many of those conlangers who do diachronic conlangs do it in a simple step by step process: first, they make their proto-language, often unattested and not really elaborated upon; second, they make descendants; third, they finalise their creations, then rinse and repeat from two.
It's very common to see this - I know I've been "guilty" of it - but I don't really think it should go that way. In my opinion, some prior planning is needed to make the results stand out in the end.
The planning I talk about isn't planning of the proto-language (although, you need to plan that one out just as well), but the idea which paths the descendant languages will take. You can do it at a whim, of course, but the method rarely produces any truly amazing results.
Now, I'm not saying you should meticulously plan out every single sound change to the minutest of details, but that you should have an idea where it's all going.
This ties in to the title of this post, namely what I like to call diachronic "traits". These traits are something language families possess and towards which the sound changes move.
For an example, let's take the various branches of Indo-European: the Baltic, Slavic, Germanic and Romantic (Romance) languages.
When compared side by side, all four of these are related clades, although some are more closely related than others. The Baltic and Slavic languages fit in a great Balto-Slavic family while the Germanic and Romantic families are relative isolates. Looking at these language families and their sub-families and descendants, we can find some common traits and the absence of other common traits.
First, we can draw the line between the Baltic and Slavic languages on one side and the Germanic and Romantic languages on the other, as they stand at opposite sides of the Centum-Satem isogloss (possibly a Satem-Nonsatem isogloss, but won't talk about that peculiarity here) - this defines their structures rather well - but we also see the Balto-Slavic group sharing some fundamental verb morphology with the Indo-Aryan languages (Satem), something we don't really find amongst the Centum languages.
In the Germanic group we find the North, West and (extinct) East Germanic languages, and nowadays the first two divide into even smaller categories, but later on that. I'll focus on the North Germanic languages a bit more than on either of the East/West duo, as their traits are more prominent.
When looked at as a whole, the North Germanic group of languages is the most coherent today, primarily because it split up around 1000AD; the prominence of their traits is primarily due to this rather small temporal depth. Modern Nordic languages all share the same realisation of the basic Germanic feature of definiteness - they use a suffix.
This small snippet of information might not be much, even more so as they already have preposited definiteness particles, but this one suffix is the primary thing setting them apart from the other Germanic languages. Syntax of the Nordic languages is generally closer to the syntax of English and West Frisian than it is to non-Anglo-Frisian languages, having a peculiar, in Germanic terms, placement of the verbal infinitive directly after the governing verb.
Now, let's take a look at Slavic languages.
The Slavic languages are divided into three groups just as well, except mirrored: there's a Western, an Eastern and a Southern group. The main split here is between the Southern and non-Southern group (which I will call Northern for convenience). The Northern Slavic languages have been in contact with each other for centuries, but have been isolated from the Southern languages for almost a millenium.
This has lead to the wildly divergent Southern Slavic languages, with feature continuums ranging from seven cases and three numbers (Slovene) to one, maybe two cases, two numbers and a distance-definiteness conflation. The Northern languages, on the other hand, have a continuum between six and seven cases.
Modern Southern Slavic languages do not distinguish between palatalised obstruents, and some have dropped palatalisation completely, while the Northern Slavic languages have either retained the Old Slavic system (post-Havlík), or expanded it (final vowel elision of Russian).
The Baltic languages (Lithuanian and Latvian) belong to the same branch of the Baltic languages and share many features with each other. There had been a dialect continuum between the two up to at least the early 1600s.
The Romantic languages are divided into multiple subclades which I won't explain here.
Now, most of these traits mentioned are synchronic, not diachronic. Synchronic traits are real-time, not historic as diachronic are - as diachronic traits represent shared features over a course of time, synchronic traits represent shared features in a cut-out.
The actual diachronic traits of these language families are best exemplified by the Germanic and Slavic language families because of their relatively clear divisions.
Slavic languages in general are less unified than the Nordic languages, while they share a lot more features. While that may be counter-intuitive, it's certainly true - the Slavic clade is on a higher tier than the Nordic subclade.
From the Proto-Indo-European period, there were the Centum and the Satem languages, which continued to develop independently after the split. I'm a subscriber of a certain theory about the Satem languages, the one that says that they developed together for a longer period of time after they then split up into the various subclades.
I've noticed that the Satem languages have throughout history strived to simplify their vowels and expand consonant inventories - the Indo-Aryan levelling, the Slavic vowel shifts and the Baltic nasal levelling - while the non-Anatolian Centum languages never really levelled their vowel systems to such an extent, instead opting for consonant reduction, as exemplified by Germanic umlaut and a-influence.The non-Anatolian Centum languages (let's call them neo-Centum for convenience) generally have extensive systems of monophthongs and diphthongs, with some members having a number above fifteen monophthongs.
This might have roots in the earliest of sound changes, including the Balto-Slavic syllabic laryngeal levelling and the Indo-Iranian o/e levelling.
This gives us our first trait division - the vocalic Centum and the consonantal Satem. This division runs through all of the subclades of Indo-European, giving rise to some rather unusual traits as consequences.
By me, the weirdest of these traits is the constant need of the Slavic languages to shed vowel features and press them into nearby consonants, followed by the uncomfortable Germanic vowel system expansions.
Fun fact: today's Slavic languages have a vowel count generally ranging from 5 to 6 vowels - there are no attested Germanic languages with 6 vowels or less, and it was likely that Proto-Germanic had six or more.
Looking at the Germanic family, we can see some deviation from this norm, such as the stress-based consonant voicing and early Proto-Germanic consonant gradation.
Italic languages might also have had such minor deviations, but from what we can see from Latin, we don't see much of it.
The traits I mentioned in the paragraph about the Centum and Satem languages are very prevalent through the history of these languages, and they somewhat guide the sound changes of these languages; you won't see Russian instantly dropping all the suprasegmentals to increase its vowel inventory, will you? Not really, no - Russian today is undergoing a stress-based monophthong reduction which is currently increasing the saliency of palatalisation of its consonants.
As we go further forward in time, we get to some more modern traits. For example, let's take the Germanic languages.
All of the current Germanic languages have umlaut, whether it is in traces or a full-blown system. The umlaut system is dying out nowadays, being developed into an ablaut (say, Faroese, Icelandic, certain varieties of English) or just becoming fossilised and more and more obsolete (certain varieties of German).
That's because this umlaut phenomenon comes from ancient vowel reductions which started around 150 AD (therefore excluding Gothic and Vandalic) - these reductions have either become minor (say, Swedish) or very prevalent (English) as they developed over the course of time.
This progress of older sound changes triggers an entire set of new sound changes that hinge on something that over time becomes obsolete, therefore prompting the new set to gradually become more and more obsolete.
Umlaut is a classic example of this - it had once been regular (German, Norse) but is now becoming a diachronic thing as analogy levels it. Umlaut shows a progression of a synchronic sound law, also known as a morphophonological law, into a diachronic sound change.
At the moment it was morphophonological law, it had absolutely no effect upon the phonemic system of the language nor on the language's morphology, but as it becomes a diachronic change (when its conditioning factor disappears), it begins to have a larger effect on the morphology and phonemics of the language.
Diachronic traits can be seen as direct continuations of any diachronic change, whether it is morphological or sound-based, as when a major change occurs and spreads to become near-universal, it becomes very salient and new changes often attach to it.
These new changes don't have to be as salient as the change that triggered them, but they can grow to it over time.
An example of this is the Nordic postpositional definite.
In Swedish and Norwegian, there's a definite suffix, but also a prepositioned article - both of these are obligatory - while Danish has either a suffix or a preposition; Faroese still has special rules as to the choice between the preposition and the suffix, while Icelandic is currently dropping the dual system in favour of a system that is exclusively suffixing.
This has sources in word order changes in early Old Norse - there, the definite article could be either prepositioned or postpositioned. Over a rather short period of time, by way of phonological bleaching, the postpositioned definite article became analysed as a suffix and that lead to a series of changes by analogy that together made a complete system of suffixing definiteness.
This was triggered by an irregular change, and lead to a flurry of reanalysis of certain features and analogical levelling and extension which continues in modern-day Icelandic but has fallen short in continental Nordic languages.
Old Norse, as a result of this new addition to its morphology, developed a strange variety of double marking - double definiteness; this way, the definite article could come before a definitely-suffixed noun. It's from this system that the definiteness varieties of Nordic languages come from.
This trait of suffixing definiteness is something that stands out in the Germanic family. At the moment, it is both a synchronic and a diachronic trait, but I'm more interested in the second kind.
Icelandic today has developed a system of possessives based on this suffixing definiteness, and has some rather strange rules regarding the genitive and definiteness.
So far, introducing this concept of a synchronic and diachronic trait, without actually explaining how they actually relate to conlanging. My approach to a posteriori conlanging is almost always to work backwards. That becomes really hard when you have two languages which are supposed to be related, but it's manageable with traits.
My first conlinguistic reconstruction involved my starship language, Kti, and another one named Dni. The two are rather far apart, but both came to share the same traits, conveniently enough.
It required some convoluted explanations and distant etymologies (not to mention coining new words and lots of trial and error), but the process had me restructure Kti a bit and really flesh out Dni. The concept of certain shared features helped smooth things along, as I never really came about to completely redesign both languages thus leaving them with some very basic but deeply-rooted deviations (including a good part of morphosyntax and the entire verbal agreement system), but those were synchronic traits. I took the things they had in common, reworked some, and introduced some very durable diachronic traits as explanations (linking the ergative in one with the nominative/instrumental in the other, for example) and got good material for reconstruction.
The traits are, in this case, just some recurring patterns on top of which other things in language built themselves. Some traits, when introduced, can go away really quickly, such as consonantal umlauts in Icelandic, but others can stay for very long periods of time and become much more than what they initially were.
(More to come)
This was a great read - can't wait to see the rest of it! I have to admit I use phonetic changes to some extent, but mostly in developing dialects rather than descendant languages, though over the long term they may develop as such. It's hard to create a truly descendant conlang without understanding how it works in natlangs, and many conlangers probably don't study that side of things as much as how to create working phonology and morphology rules, which are much easier to grasp and regulate.
Phonetic changes as-is aren't bad - they're incredibly useful. But using only them at the expense of syntactic and morphological changes, that's bad.
Making purely phonetic changes in conlang decendants does seem to be a prevelant bad habit that many conlangers have.