Are natural languages perverse? Do people in an unco-ordinated way work to their detriment to add unnecessary complications to the languages they speak?
Conventionally Easy and Simple
I’ve achieved basic competency in written toki pona. I can read and write quite a bit. It took about a two months or so. Toki pona is an artificial language with limited vocabulary and grammar. Esperanto is also a constructed language with regularized grammar and a euro-average vocabulary (i.e. lots of words that an educated European might recognize on sight) Both are mostly analytical, i.e. few or no inflections.
Conventional wisdom says that an artificial language should easy, *unlike* languages with lots of inflections, agglutinations, irregular verbs, gendered nouns, etc. Lets see if leaving these features out of a language necessarily makes it easier, or just more analytic (i.e. grammar by word order instead of morphology)
If your brain is processing an irregular verb, it uses a different area than if it is using a grammaticalization (i.e. using grammar to express an idea). So for common words, irregularities may increase speed of recall since you will remember such as verb as half a dozen different words instead of one word that is subject to half a dozen regular transformations.
Agglutination, Inflection and other bound morphemes
Linguists love creating tables of verbs. Seeing every form of a verb is an awesome and scary sight. Reducing the noun or verb’s forms to a single form may or may not simplify things.
Word Generation. Being able to inflect and agglutinate lets people generate words grammatically for when vocabulary fails. I know this from experience using Russian–often I could convey what I wanted to say using an inflected form, even if a native would have used a different word.
Robustness. A language is robust if a listener can understand it despite noise and mistakes. Inflection tends to create redundancy because of agreement and because of encoding of the pronoun in both the subject and the inflection. Removing agreement, and the redundancy means it will be harder to understand under normal conditions of a noisy room and under conditions of language learning (where people are speaking word soup and speaking a new language with the grammar of their mother tongue)
Gender usually means male, female and neuter, but can also be applied to other categories. Gender categories allow your brain to index your vocabulary so that similar words are stored together. One can feel for themselves the near certainty that words are stored logically (if not physically) in different parts of the brain based on their relationships. This is most strongly felt when you are doing code switch, switching from one language to another and for a while when you try to think of a word, the word comes to mind in the other language.
Implications for Language Design
If my wild theories have any merit, then an auxlang (a constructed language that has big hopes for attracting speakers), should not only work on being simple to get started, but have a clear path to complex speech. An analogy would be Basic English. It is easy to get started, but gets progressively harder to move forward (in terms of being expressive) as you gain mastery of the language.
Too Much of a Good Thing
Of course there can be too much of a good thing. Lojban has a reputation for being devilishly hard to speak (D. Harlow’s report of expert Lojbanists only being able to hold a conversation for 15 minutes comes to mind). Japanese has a complicated writing system and I’m not sure how that was a benefit to them (except maybe it meant Japan would need a much higher standard of academics, something that might have unexpected side effects). Similarly, there certainly can be too many words, too many words of limited usage (words for snow that has been thrown over one’s shoulder during early spring), and too many ways to inflect. Finding the happy balance between analytic, agglutinating, inflecting and polysynthetic languages is the task of both the artlang and conlang designer.