I know this is 13d old, but language discussions are bait for me.
Disclaimer: I’m discussing this matter from the PoV of conlang designing. I’m aware there are more factors involved in adoption of auxlangs than just the design. I’m not shitting on anyone who wants to learn Esperanto.
Esperanto’s 28 phonemes inventory isn’t “small”; it’s “average”, by world standards. Refer to WALS chapter 1 and chapter 2. Note both English and Polish are poor points of reference because Indo-European languages are strongly biased towards huge phonemic inventories, specially in the European sprachbund.
And average is actually subpar in this case, as you want to stick to consonants most people can tell apart; cue to the /b/ vs. /v/ distinction for Spanish speakers, or /x/ vs. /h/ for English speakers, or /s/ vs. /ʃ/ for Japanese speakers.
I think a 12+5 system like /m n pʰ tʰ kʰ b d g f s h r a e i o u/ would be the sweet spot between not making it too large (so it’s likely to hit contrasts you don’t make) vs. making it too small (so it becomes unflexible). Even then there are some sour points (Arabic speakers for example would have a hard time with /pʰ/).
A few things I’m not including, present in Esperanto:
/w j/ ⟨ŭ j⟩ - it’s better to make syllabification non-phonemic, and let people render /i u/ as [j w] if they want. Much like you mention for ⟨ankaŭ⟩.
/v z ʒ/ - fricative voicing contrast isn’t that common, and this allows the leftover fricatives to adopt voiced allophones if you so desire; specially useful if you’re adapting Romance words with ⟨s⟩ /z/. They can be adapted as /b f u/, /s/ and /i s/ respectively.
/ʃ/ ⟨ĉ⟩ - [ʃ] is excessively common as an allophone of /s/ before front vowels. Specially high ones. Make it /s/.
/t͡s t͡ʃ d͡ʒ/ ⟨c ĉ ĝ⟩ - affricates are better handled as simple sequences of phonemes; render them as /ts ts ds/ (remember what I said about voiced allophones? Nothing wrong if you pronounce /ds/ as [dz], you know…). /d͡ʒ/ could be also rendered as /di/.
/l/ vs. /r/ distinction — I swear, why do people think it’s a good idea? Be glad Esperanto was made by Zamenhof, and not some random Portuguese speaker; otherwise you’d see /ɾ r l ʎ/ instead with /r/ allophonically [h].
A point not-often-mentioned in Esperanto’s design are the phonotactics. Zamenhof never explicit them out, but in practice max onset is 3C (see: ⟨skribi⟩ ) and max coda is CC (see: ⟨ekster⟩ ). I might be biased but I think anything past CCVC is already overboard, there’s Russian/English/Georgian if you want to cram 9001 consonants per syllable.
For the rest, a language do need some diversity in its phonology or the words would have to be very long and least recognizable
“Long” as amount of syllables isn’t that important. When the phonology and phonotactics are simple, people tend to speak more syllables per minute; cue to Spanish and Japanese speakers.
Regarding recognisability, the best thing that can be done is to have predictable rules on how to adapt borrowings.
I know this is 13d old, but language discussions are bait for me.
Disclaimer: I’m discussing this matter from the PoV of conlang designing. I’m aware there are more factors involved in adoption of auxlangs than just the design. I’m not shitting on anyone who wants to learn Esperanto.
Esperanto’s 28 phonemes inventory isn’t “small”; it’s “average”, by world standards. Refer to WALS chapter 1 and chapter 2. Note both English and Polish are poor points of reference because Indo-European languages are strongly biased towards huge phonemic inventories, specially in the European sprachbund.
And average is actually subpar in this case, as you want to stick to consonants most people can tell apart; cue to the /b/ vs. /v/ distinction for Spanish speakers, or /x/ vs. /h/ for English speakers, or /s/ vs. /ʃ/ for Japanese speakers.
I think a 12+5 system like /m n pʰ tʰ kʰ b d g f s h r a e i o u/ would be the sweet spot between not making it too large (so it’s likely to hit contrasts you don’t make) vs. making it too small (so it becomes unflexible). Even then there are some sour points (Arabic speakers for example would have a hard time with /pʰ/).
A few things I’m not including, present in Esperanto:
A point not-often-mentioned in Esperanto’s design are the phonotactics. Zamenhof never explicit them out, but in practice max onset is 3C (see: ⟨skribi⟩ ) and max coda is CC (see: ⟨ekster⟩ ). I might be biased but I think anything past CCVC is already overboard, there’s Russian/English/Georgian if you want to cram 9001 consonants per syllable.
“Long” as amount of syllables isn’t that important. When the phonology and phonotactics are simple, people tend to speak more syllables per minute; cue to Spanish and Japanese speakers.
Regarding recognisability, the best thing that can be done is to have predictable rules on how to adapt borrowings.