Morphology is the part of grammar dealing with the ways individual words are constructed out of components technically referred to as “morphemes”. Morphemes may be things that are capable of standing as independently viable words in their own right (“free morphemes”, like word), or they may not (“bound morphemes”, like the plural ending ‑s); and it's useful to recognise two distinct kinds of process for combining them:
- DERIVATIONAL morphology deals with the ways fresh words are coined (e.g. by compounding), or existing ones converted from one word class to another. (The “words” I'm talking about here aren't the things you hear spoken in sentences; they're more abstract linguistic units known technically as “lexemes”, though we might as well just call them “dictionary entries”.) Derivational processes have cumulative, reorderable effects (unrepacked and reunpacked have two different meanings), and as long as they're “productive” – that is, still in use building new words – they tend to be regular in form; however, in natural languages they can also become “non‐productive”, leaving the examples that are already in the lexicon to erode away into irregularity over the centuries. A newly invented language has no need to carry that sort of historical baggage, but derivational processes also routinely tend to be slightly unpredictable in application. They can never be entirely governed by predefined rules, since derivation has the task of labelling all the disorderly and erratic features of the real world.
- INFLECTIONAL morphology deals with the ways words are modified to fit into a particular sentence by adding case endings, person agreement, and the like; picking the appropriate inflection to apply to a word in context is usually grammatically compulsory. Inflectional changes tend to be applicable to all members of a word class, with a completely predictable meaning but often (in natural languages) a mildly or extravagantly irregular form. They apply as prefixes or (especially) suffixes to a lexeme as a whole, not to its component morphemes; and when a word carries more than one inflectional marker they either have a fixed order or sometimes fuse into a single affix.
One of Esperanto's selling points is that it allegedly has completely tidy, automatic, productive morphology. Mind you, it's more work to invent a fully functional language with eccentrically organic morphology than it is to make it meticulously regular, so this was the lazy approach! Unfortunately, Zamenhof's scheme was sabotaged from the start by the fact that he was unaware of the distinction between inflectional and derivational morphology. Inflection in a constructed language can be perfectly systematic; if you assemble a word like lern‐ant‑a‑j‑n = “learning (obj. pl.)”, its meaning follows routinely from the grammar rules. Derivation isn't necessarily like that – instead much of the time meanings are a matter of precedent. If the lexeme lern‐ej‑o = (literally) “learning‐place” wasn't already in use, you could coin it as a new word to refer to any old location where somebody once did some learning (just as an ek‐ir‐ej‑o becomes a “starting‐place” as soon as anybody or anything sets off from there); but as it happens the word has the established idiomatic meaning of “school”. And that's despite the fact that most people learning the word aren't doing it in a school!
The ‐ej suffix in lernejo is on the official list of specialised affixes, along with just a couple of dozen others. Given the support of a word‐class marker vowel, these can stand as words in their own right (ej‑o = “a place”, ul‑o = “a person”), which means they're as close to being free morphemes as are ordinary Esperanto roots (like lern‑), if not quite as free as the words that can appear as bare roots (like the article la = “the”). In many cases the affixes are shadowed by non‐affixes with the same senses (lok‑o = “a place”, hom‑o = “a person”), which can themselves be used regularly and productively to form compounds, making it unclear why we'd need both; the official affixes are a rather pointless intermediate category in between the necessarily open class of ordinary dictionary entries and the necessarily closed class of special inflectional endings. Yes, the affixes have a characteristic form intended to make them easier to pronounce in compounds… but this is just an indictment of the poor design of Esperanto roots. The Fundamento confuses things further with a proclamation that all morphemes are equal (“Grammatical terminations are considered as independent words”), but this is nonsense: the “grammatical terminations” on a word like lernantajn are bound morphemes.
Constructing new words with ejo and ulo and so on may be a process of derivational compounding, but there are a couple of morphemes on the list that people sometimes argue are inflectional: ‑ig = “render” and ‑iĝ = “become”. There are languages with causatives/inchoatives that constitute an integral part of the inflectional system, and it has been suggested that one of them – Hebrew – may have been in the back of Zamenhof's mind when he added ‑ig to Esperanto. But if it was like the Hebrew “hiphʿil” form (which for a start isn't a suffix) then it would automatically turn transitive verbs into ditransitives (with two simultaneous direct objects, illegal in Esperanto). Instead it behaves more like the derivational causative suffixes in languages such as Lithuanian: kalt‑as = kulp‑a = “guilty”; kalt‐in‑ti = kulp‐ig‑i = “to accuse”. It's not an inflected form of the adjective, it's a separate lexeme derived slightly unpredictably from that adjective.
And then there are those word‐class marker vowels. The way the noun lok‑o = “a place” relates to the adjective lok‑a = “local” and the verb lok‑i = “to (set in) place” is about as diagnostically derivational as it's possible to get, but Zamenhof couldn't see this. He thought of being a noun or adjective or verb as a transient grammatical attribute added to an underlyingly categoryless root, so he handled it the same way as he handled verb conjugations: with a set of inflectional endings, which are expected to work uniformly and predictably, replacing one another rather than stacking to mark the word's history (the way deriv‐ation‐al‐ly carries the traces of being an adverb formed from an adjective formed from a noun formed from a verb). In reality, as even the Academy of Esperanto now admits, roots start with an inherent lexical category (loko is a noun‐root), and this entire mechanism of systematic zero derivation was broken by design.