Here I list words with three of the same consonant in a row, like mamamahayag and pirarara. Many languages have geminate consonants, which in many ways behave like single consonants. I will consider these to be single instances, and only consider consonants to be separate instances if they are separated by the physical tongue and lip movement of an intervening vowel. Therefore Finnish totta counts as two ts, while Finnish totota counts as three.
I expect languages with small consonant inventories to dominate, with Finnish taking a clear lead over everything else. Finnish also has particles -ko and -pa, which could potentially make its listed words even longer. However, Rotokas might be the greatest performer of all, since it has quite a few entries on this list despite having a much smaller corpus than the other languages.
Strict rules prevail, so for example reduplication is not allowed, nor are contrived compounds with no practical use. For example, popup is certainly a word in English even though it's a compound, but *poppy plant is not commonly used and does not count as an instance of 3 /p/'s.
The word bibimbap comes close to qualifying for this list, and without its /m/ it would arguably have four /b/'s in a row since there are no final voiced stops in Korean and the final voiceless stops most closely resemble the voiced (tenuis) stops in pronunciation. (This sound is sometimes spelled p in phonetic transcriptions because the voicing is allophonic.) It is possible that 밥 "rice" is actually a direct parallel of English pap meaning baby food, but even if this is discounted the initial consonant would still count as a /b/. The stem of the verb appears to just be bibi-, but I dont think there's any natural Korean expression that would get all the consonants together.
This would have more entries were I not so reluctant to include triplets based on morphology. But here are some examples:
The Loup language, possibly Nipmuc, provides us the name Chargoggagoggmanchoggagogg (and sometimes even longer forms are seen). The language is extinct, and it is a matter of debate whether these /g/'s should actually be analyzed as /k/'s. There is no /r/ in the Wikipedia writeup for the language, just as most Algonquian languages don't have it, so that is probably a nonrhotic representation of /a/. Debate remains about the authenticity of the name, but the spelling above was in use even in the late 1700s before tourism provided an incentive to further exaggerate the length of the name.
A translative affix -ksi can add a fourth consecutive /k/ to some of these words.
An oft-repeated Finnish sentence is Kokoa koko kokko kokoon. It translates to "Put together the whole bonfire" and is usually seen as part of a longer tongue-twister. At first blush this sentence seems far superior to everything on the list. Nonetheless, it may still need to share its prize, as it appears that three of the four words in the sentence are derived from the same root koko and that the strictest possible rules would dictate that only koko kokko is valid as a set phrase. To this, I could add the question particle -ko, and arrive with five /k/'s in a row with no other intervening consonants: Koko kokkoko? This sentence would be "A full bonfire?" There are most likely other nouns with two /k/'s, but the tongue twister focused on also getting the vowels to line up.
Possibly kokoke unless there is reduplication
Many Japanese words with /k/, especially those with /k/ in consecutive syllables, are derived from Chinese where at one point various Chinese dorsal sounds were all represented as /k/ in Japanese. Thus one hears jiang ge in modern Mandarin but the word is pronounced /kōku/ in modern Japanese.
国語学 has been used in text to mean Japanese language studies; it is pronounced /kokugogaku/, with five velars in a row, but there is most likely no stage of the Japanese language in which they would have all been /k/ without also having other consonants. Japanese linguists Kyōsuke and Haruhiko Kindaichi (father and son) used this word as an example how they disliked the sound of their native language here, and the word kokugogaku (in Roman letters) appears in Haruhiko's Wikipedia article today.
The above words show the Finnish words are not just dependent on a single root. /lello/ might be a word for "game, toy" in Votic, which would then take case markers beginning with /l/. However the transcription seems to mix Cyrillic and Roman letters so there may be two phonemes
ʻiole liʻiliʻi though it deserves two cautions, one for having reduplication and one for having glottal stops. The glottal stop can contrast with hiatus.
I have read that ממון can take a prefix mi-, which would mean three /m/'s in a row
Many terms in this section are likely sound symbolism (consider the bouba/kiki effect), but I still place sound symbolism on a higher tier than reduplication so long as they represent single sounds and not repeated sounds. Maori has pēpepe as one word for butterfly, and I dont know if it uses reduplication or not.
The name of the World Health Organization in Khmer is អង្គការសុខភាពពិភពលោក, which is pronounced /ɑŋ kaː so.kʰaʔ.pʰiəp pi.pʰup loːk/. Arguably this contains 4 or even 5 consecutive /p/'s, but Khmer /ph/ is usually analyzed as a consonant cluster, and if this analysis is rejected, it usually means that /p/ and /pʰ/ are being analyzed as different phonemes.
Likewise, 표범피부 (pyobeompibu) does not contain more than 2 of any one consonant, but it does have 4 syllables in a row beginning with bilabial stops.
The phrase kapaipiyapai ipitsi (4 /p/ in a row if treating <y> as part of a diphthong) appears on the meyeixapai page.
this document claims pepapa ~ pepapo ~ pepape along with some longer forms like pampapombe are all verbal forms in some language of SE Asia, possibly w:Pamona language.
This section is hard to search for because of frequent use of /r/ in morphology.
This section is difficult to search for. Japanese may have compounds involving /tatsu/ or /tachi/, while English words are dominated by morphology.
Examples abound because of morphological affixes, for example substitute is just one of many such words, and statutory is another. Content-only words seem hard to come by. teetotal is reduplication.
See also #Hiatus below.
In theory, the null consonant Ø could be considered here, as it counts for alliteration purposes in English poetry. Just as geminate consonants are common, so too are long vowels. It is difficult sometimes to draw the line between a vowel sequence, a long vowel, and a diphthong. For example, Finnish long vowels are almost never considered to be sequences of two short vowels, yet Finnish diphthongs often are, even though many of them arose from historical long vowels. Thus riiuuyöaie could be counted as having five, seven, or nine vowels. Moreover, some languages have triple length vowels. I will leave this section ungraded as to number.
Note that many Polynesian words that appear to have long vowel sequences may actually have glottal stops. For example, the town of Kaaawa, Hawaii is actually Kaʻaʻawa and Kalaeoio is Kalaeʻōʻio. It is possible that Rotokas' vowel sequences are real, both in the sense that they have no glottal stops and that the /i/ and /u/ do not reduce to glides.
Some people might analyze /j/ and /w/ as consonants.
honorable mention for wiijiiw ... a palindrome with five dotted letters in a row, though it is a /dʒ/ in the middle, so this is not hiatus by even the loosest definition of the term (some people would accept /i:ji:/ as a sequence of two /i:/ for example).
This is just a loose list without strict criteria. Some languages have vowel harmony, for example.
Check to see if there exist shells with higher numbers due to allowing diphthongs, e.g. town. No words end in /auk/ or /aup/ so I consider the /p/ and /k/ vowel slots filled.
Note, none of these words distinguish between COT and CAUGHT, because that is not a robust distinction and is merged in many dialects as well. t_n:
k_n:
p_t:
m_s:
One reason for the dominance of /p_p/ and /k_k/ may be sound symbolism. In fact it is possible that every single one of the /p/ words derives from either onomatopoeia or expressive language (e.g. baby talk) rather than having a conventional PIE etymology. Similarly, w_p has almost as full of a set as these others, but more than half are onomatopoeia.
However, toit and tout are words, so maybe /t/ is the winner: