How do conlangs/natlangs have prefixes suffixes and not get them jumbled up?

Question

When thinking how to build a conlang system of prefixes/suffixes, I get stuck right at the beginning: how do you prevent clashes within the prefix/suffix/base system?

By that I mean the following. Say you have these prefixes:

• a
• an
• ani
• amo
• am
• ami
• amo
• ach
• acho

Then say you have these word "bases":

• ma
• man
• mother
• noise
• nose
• norm
• match
• cheek
• chim
• chin
• aman
• amish
• achy
• norma
• macha

And you have these suffixes:

• chum
• ma
• ama
• amo
• ana
• vana
• chana

I just picked random ones for demonstrative purposes. These will come into conflict in the following scenarios:

• a + amish = aamish?
• an + noise = annoise?
• am + macha + ama = ammachaama?

The point is, some of them won't be able to combine:

• If one ends with a vowel and the other starts with the same vowel (or even a vowel at all), and you don't have long vowels, then what do you do?
• If one ends with a consonant and the other starts with the same consonant (and you don't have long/geminate consonants), then what do you do?

I am vaguely starting to imagine a system where you have two forms to each prefix/base/suffix, so like:

1. If ending in a consonant, it can end in a sibilant/fricative like s or f, or a nasal like m or n, so you have pairs like "mas" and "mam", and you pick one depending on the following word chunk you are working with.
2. If ending in a vowel, and the next begins with a vowel, you put a y or w between. So a + ama would become "awama".

Otherwise, if you just let them merge together (and remove the head/tail of the joining pair), then you lose some information.

How do natural languages (or conlangs for that matter) deal with this problem? Like in Sanskrit, or Latin, I see these sorts of prefixes/suffixes like my examples, and yet somehow the system seems to work (I haven't studied enough examples in each language though to know how/why it seems to work). What constraints do they place on the system? Etc. How do they create a prefix/suffix system that just works?

I ended up making all prefixes/suffixes/bases have to start/end with a consonant, and to combine you join with a vowel. That means for extra syllables, which I would like to get rid of. So instead of "kan" and "van" combining to form "kanavan", for example, I would like to just do "kanvan". Or even better, just have things that don't start/end with consonants, like "ka" and "an", and have them be, I don't know, "kan" or "kawan". How do you accomplish such a system? What are the range of possibilities which you can use (taking inspiration from natlangs or conlangs the like)?

Answer 1

If ending in a consonant, it can end in a sibilant/fricative like s or f, or a nasal like m or n, so you have pairs like "mas" and "mam", and you pick one depending on the following word chunk you are working with.

This is called "allomorphy" and it's quite common. For example, in Ancient Greek, the negative prefix is /a/ before a consonant (asymmetric, atheism) but /an/ before a vowel (anemia, anorexic). Inuktitut, a famously agglutinating language, has a lot of allomorphy rules.

Usually this sort of thing results from some sort of phonological pattern. The negative prefix in Greek came from PIE *ṇ-, which historically became /a/ in certain environments and /an/ in others. But by the time of Plato, it no longer seems to be a part of the language's phonological system; it's just a quirk of this one prefix.

If ending in a vowel, and the next begins with a vowel, you put a y or w between. So a + ama would become "awama".

This is called "epenthesis" and it's one type of phonological process, which is also quite common. The difference between this and allomorphy is that these phonological processes seem to happen at a different level; it doesn't matter that it's specifically the prefix a- being used here, it just matters that two /a/ phonemes ended up next to each other, and that's not allowed.

Otherwise, if you just let them merge together (and remove the head/tail of the joining pair), then you lose some information.

Sure, but natural languages have a high degree of redundancy, so losing a couple bits of information here isn't a problem.

How do natural languages (or conlangs for that matter) deal with this problem? Like in Sanskrit, or Latin, I see these sorts of prefixes/suffixes like my examples, and yet somehow the system seems to work (I haven't studied enough examples in each language though to know how/why it seems to work). What constraints do they place on the system? Etc. How do they create a prefix/suffix system that just works?

In Latin, at least, only the last phoneme of a prefix ever gets changed when it attaches, and there usually isn't much functional load on this last phoneme. This makes sense, because if the assimilation had ever created an unacceptable ambiguity, speakers would have found some way to fix that (by, say, ceasing to use one of the ambiguous prefixes).

So when we do occasionally see a high functional load on the last phoneme of a prefix (ad-esse "to be present" vs ab-esse "to be absent"), the system must have some way of keeping them distinct; in this case, ad-esse is able to assimilate in forms like assum "I am present" and affuī "I was present", while ab-esse is not (absum, abfuī). If it didn't, speakers would have found, created, and used workarounds over the centuries.

Answer 2

Otherwise, if you just let them merge together (and remove the head/tail of the joining pair), then you lose some information.

This happens in natlangs all the time. Consider French: the overabundance of silent letters at the ends of words means, for example, chien (dog) and chiens (dogs) are pronounced exactly the same. The language gets around it through the articles; you know that "le chien" is "the (male) dog" and "les chiens" is "the (male) dogs", thus in this case the difference is indicated by the difference between [lə] and [le].

And in some cases, it can get ridiculous. Consider the conjugations of the present tense "to run":

• je cours (I run)
• tu cours (you run)
• il court (he runs)
• elle court (she runs)
• ils courent (the males run)
• elles courent (the females run)

Every single one of those verbs are pronounced exactly the same, [kuʁ]. The only way you can differentiate them in speech is by the pronouns. Except, well, il/ils is pronounced the same, as is elle/elles. So there is, just from isolation, no way for someone to differentiate between "he is running" and "they are running".

So, what can you do? Well, context. Sometimes it's clear what you're talking about. Sometimes you have to add words to make the context clear: "Ils courent" (They are running) vs "Louis, il court" (Louis, he is running). Sometimes you can use other adjective which may have different words for male/female or singular/plural. All the normal tricks languages use.

ADDENDUM

I just realized, with my example of the conjugations of "courir", that it provides a good example of how someone trying to understand the linguistics of another language could get it really wrong from a limited sample size. We know French is written that way because of its historical development, and that it conjugates the verbs for different persons, and has gender and number coded in the pronouns, which the written language maintains. But if French wasn't written down (or it was written in a script that was strictly phonetic), and you only had a small sample size, you might come to the conclusion that the verb "to run", koor, wasn't conjugated in the present tense, and that French didn't differentiate between singular and plural in the third person.

Similarly, if you only had the written text, you might think that my examples above were pronounced je koors, too koors, ill koort, ell koort, ills koor-ent and ells koor-ent (or maybe koo-rent).