Most conlangs give some thought to euphony, but what if that was the only thing you cared about?? That's the core thought of this post.

Sindarin is a famous conlang designed largely for euphony, but there were also etymological and other constraints to balance. What if euphony is the ONLY goal??

I just had this idea yesterday so I haven't gotten much further than that.

First directive: Be non-tonal

https://pmc.ncbi.nlm.nih.gov/articles/PMC10151606/ is the most thorough research on what languages are nice. They report a negative finding! They couldn't find any clues "apart from a possible slight preference for nontonal languages".

https://pmc.ncbi.nlm.nih.gov/articles/PMC7940689/ is a less sophisticated study, taking less pains to control for rater-bias, but also finds, perhaps counter-intuïtively, that melody is bad, "We Like It Flat and Fast, but Not Melodious"

Implosives probably good, ejectives bad

Fig. 3. in the first-linked paper above indicates non-significant findings that implosives and clicks are good, but non-pulmonic consonants are bad.

[Non-pulmonic consonants] means [implosives + ejectives + clicks].

So the stats show that people gave a low rating to ejectives (understandable: they intuïtively sound ugly) and high ratings to clicks and implosives.

Vowels

https://web.archive.org/web/20200530122205if_/http://www.davidcrystal.com/?fileid=-4009 was hunting for nice words rather than nice languages – an easier job.

It ranks the vowels in order that occur in pretty words —

• the schwa from the French word 'je' (central, mid) ə
• the vowel in 'lid' (near-front, near-high) ɪ
• Fonzie's vowel (front, high-mid) e
• the diphthong in 'lime' (front-low then front-high) aɪ
• the vowel in 'gut' (back, low-mid) ʌ
• the diphthong in 'make' (front,high-mid then front-high) eɪ
• the vowel in 'free' (front, high) i

It says pretty words tend to start with front vowels, and move to back vowels.

It says pretty words tend to start with mid or low vowels, and move to high vowels

Consonants

The same paper ranks in order:

• l
• m
• s
• n
• r
• k
• t
• d

But the paper makes a weird and massive oversight on glides (w and y) which I'm going to have to fix myself.

Using common words to bias the sound of the language

Look at https://en.wiktionary.org/wiki/Wiktionary:Frequency_lists/PG/2006/04/1-10000 — maybe a quarter of words in a text are:

• definite and indefinite article
• pronouns
• 'and'
• a few prepositions like 'of', 'at', 'for', 'in'
• the verbs 'to be' and 'to have'

(That's for English. Languages with things like tense markers would have other verycommon words.)

So these should be words like —

• 'le' like the French masculine definite article, and rhyming monosyllables beginning with m,n,r,k,t, and d. These can be our pronouns and 'to be'
• maymuh (vowel sequence goes backwards and upwards) Paste meɪmə here
• lilluh (vowel sequence goes backwards and upwards). Paste lɪlə here

Other ideas

I'm think alternating geminated and ungeminated consonants might be a good idea.