One of the signature features that distinguishes
a naturalistic artlang from other types of conlang is the demand for irregularity. A word is considered irregular if it doesn’t
fit the pattern of inflection one would expect it to. In English, the standard past tense is formed
by adding the ‘-ed’ suffix, and the vast majority of verbs follow this pattern. However, there are a number of verbs that
exhibit completely different, sometimes totally unique patterns for the past tense, and are
therefore considered irregular. All natural languages are irregular to some
extent, and therefore, so should a naturalistic conlang. One easy way this can be done is by completely
arbitrarily choosing a word and making up an irregular pattern for it. While this does work, at least on a superficial
level, without an understanding of how the irregularity came about, it can look very
forced and shallow under scrutiny. After all, irregularity doesn’t just appear
out of the blue, it’s a natural and unavoidable consequence of linguistic evolution. One of the great advantages of using a proto-language
is that, as you evolve your conlang, irregularity should arise naturally without you necessarily
having to do anything. Let’s say we’ve got a proto-language in
which the only consonant that can serve as a coda is /n/, just like Japanese. We’ll coin a couple of words and a create
a basic tense suffix of the form ‘-lu’. If we implement a sound change wherein /l/
following /n/ becomes /d/, then any word that ends in /n/ will have a past tense suffix
of the form ‘-du’ rather than the ‘-lu’ suffix one would expect. Although it might look like it, this, technically,
isn’t irregularity in the truest sense, because the irregularity can be predicted
from the phonetic environment without exception, and can thus be described with a simple allophonic
rule. This is what George Corley of the Conlangery
Podcast terms “Regular irregularity”; it looks irregular at first, but there’s
actually a regular and predictable underlying pattern to it. To create genuine irregularity, we need to
remove the element of predictability. If we implement word-final vowel loss, we
see that the vowels these words used to end in now only surface in the suffixed forms,
and the only way of knowing which pattern a given verb follows, at least without knowing
its etymology, is just through sheer memorization. This gives us three different conjugations
based on which vowel appears in the suffixed form. This is one very easy way to create multiple
paradigms, in fact, this is very similar to the process that produced the four conjugations
of Latin. However, this still doesn’t count as true
irregularity, because none of these patterns are the “default”, a verb is just as likely
to belong in one conjugation as any another, so they’re all treated as regular patterns. What makes a word irregular is the fact that
it doesn’t follow any of the expected patterns. However, while in the proto-lang we knew for
certain that any word that ended with a /n/ would take the /-du/ suffix, now that word-final
vowels have been lost, we have some words that end in /n/ that don’t take /-du/, but
instead fit the regular patterns. So at this point in the language’s history,
most words ending in /n/ will be regular, but a small number of them will fit a pattern
that seems totally arbitrary and unrelated. With just two sound changes, we’ve managed
to get three conjugations and a class of irregular verbs. This kind of phonologically-driven irregularity
happens when a sound change affects only a small subset of words, and then the feature
of the phonetic environment that makes the change in pattern predictable is lost. As I say, this may happen without you even
intending it as you apply sound changes. All you have to is keep an eye out for when
it occurs. But this is only one method of generating
irregularity, and it’s far from the full story. If we look at English, we see that just about
all of our most commonly used verbs are irregular, and it’s most often the same in other languages
too, much to the chagrin of language learners everywhere. This isn’t due to phonology or sound changes,
it’s the byproduct of another process. Many of English’s irregular verbs are derived
from Germanic Strong Verbs, in which the regular past tense was marked by umlaut. However, at some point a different strategy
for marking the past tense came about, one by adding a final ‘-ed’, and it began
to steadily replace the pattern of vowel alternation. Those verbs that retain the vowel alternation
today comprise the verbs that were used often enough to resist the change, and even though
they were once considered regular, the relative rarity of their occurrence in the modern language
means they have now become irregular. This is an extremely common way for irregularity
to manifest. When a new grammatical construction arises,
it may spread and replace the old system, becoming the new regular pattern. However, in especially common words, the speakers
may be so used to using the old forms, that the new forms never become associated with
them, and thus they consequently become viewed as irregular. For your conlang, keep track of any words
that survive from your proto-language into the modern language, and have them keep the
inflected forms they had in the proto-language. These words that once represented the regular
pattern will now form a small class of extremely common irregulars. But there’s another factor that contributes
to frequently used words becoming irregular. Common words often simplify independently
of sound changes due to the fact that they’re used so often. In Middle English, the verb “make” was
completely regular, having a past tense formed with the ‘-ed’ suffix, but people ended
up using the word so frequently it reduced from “maked” to “made”. However, this didn’t affect less common
verbs, so the past tense of “bake” is still “baked”, not “bade”. The key takeaway here is that there is a very
strong correlation between a word’s frequency of use and likelihood of being irregular. Beyond the typical collection of common words,
the most likely candidates for irregularity are auxiliary verbs, and, perhaps the most
cross-linguistically irregular verb on earth, “to be”. If we look at the conjugation of “to be”
in English, we can see that it’s by far the most irregular verb, and it’s normally
the same way in other languages. In the case of English, “to be” actually
has the most forms of any verb, as a result of a long history that stretches all the way
back to proto-Indo-European, with numerous fossilized forms along the way. But some of the forms that survive to the
modern day owe their divergence to another, particularly striking form of irregularity;
In English, while our regular past tense is formed by adding ‘-ed’, many of our irregulars
are formed through vowel alternation, like see and saw, but then we have ‘go’ and…
‘went’. How did we get from “go” to “went”? Well it turns out, this isn’t the product
of some fiendish sequence of sound changes, nor is it some fossilized conjugation, but
rather, originally, ‘go’ and ‘went’ were entirely separate words with their own
independent conjugations, which later ended up merging, with ‘went’ being conflated
as a past-tense form of ‘go’. This phenomenon is called suppletion, wherein
two unrelated, usually at least partly synonymous words are reanalyzed as being part of the
same paradigm. Suppletion doesn’t occur very often, but
when it does, the split may occur between singular and plural forms of nouns, different
tenses, or in number agreement for subjects or objects. This might seem pretty inexplicable, but we
can get some insight into how this happens thanks to the ridiculously awesome way that
Navajo derives verb meanings. To convey the concept of one person running,
Navajo uses a verb root that actually means “to bend” plus a reflexive, creating the
meaning of “to bend or flex oneself”, referring to the bending of one’s legs. Whereas for two people running, it uses a
root that means “to flee” with a reciprocal, which here means “to chase each other.” These two entirely unrelated verb roots are
being used to communicate the same concept, the only difference being the number of subjects. You can very easily see how this sort of metaphorical
extension could ultimately lead to suppletion. Suppletion tends to occur most frequently
in non-specific or otherwise semantically weak words, especially words for which many
synonyms exist, with the prime candidate once again being “to be”, in which case, past
tense forms may come from independent words that mean “to stay”, “remain”, or
“to live”, while future forms can evolve from “to become” or “to grow”. This suppletion, combined with simplification
from overuse, as well as preservation of defunct conjugations, on top of any phonological shenanigans
results in a really, really irregular pattern. If any word in your conlang is going to be
irregular, it’s going to be ‘to be’, so don’t be afraid to use all of these processes
to make it as irregular as you please. So in summary, to generate irregularity in
a naturalistic conlang, allow any old words that survive from the proto-language to maintain
inflections that have since been lost in other words, simplify particularly common words
independent of any sound changes, especially those that occur as part of grammatical constructions,
combine semantically weak and synonymous words through suppletion, and keep track of any
phonological irregularities that arise along the way as you implement sound changes. If you stick to these principles, you’ll
be well on your way to making a conlang that replicates the same kind of irregularity that
we see in natural languages, and that will make your language feel all the more authentic
and real.
