zompist bboard

http://www.tofugu.com/japanese/japanese-punctuation/

If you see this smaller version of the hiragana つ, it is not pronounced "tsu" (ever!). […] If you see it at the end of a word (before the particle と in many onomatopoeia) then it's a glottal stop. That means it's kind of like a constricted sound in your throat (that's your glottis in there, thus the name). The katakana version looks like this ッ.

True or not?

I seem to remember reading that it makes the preceding vowel shorter, but yeah, that's probably true.

True at the end of an utterance. Utterance-final short vowels are generally followed a glottal stop anyway, but this makes it more "emphatic" like maybe a shorter vowel, exaggerated intonation, audible breathy release of the glottal stop, etc.

When followed by another consonant it's just a geminate though.

I just happened to see a perfect example so I took a picture. You can see they Romanized ぐるっと as grutto ("guru" to "gru" is just to make ir Englishy though). Anyway, just a geminate.

It's weird when they do it with a voiced consonant, though, like レッド or キッズ because then it sounds like a glottal stop again [reʔdo] or [kiʔdzu] or maybe an unreleased voiceless consonant, or the voicing just disappears. Like it's definitely not a geminate [red:o] in this case.

Wouldn't [kiʔdzu] be キッヅ?

ヅ and ズ are pronounced identically as /zu/. The distinction between them is lost. Standard transliterations from English into katakana rarely use ヅ, and on the rare occasion they do (and I can't think of any), it's a stylistic choice. The standard way to render "kids" is キッズ, but this is clearly not pronounced [kiz:u]; it's much more like what finlay indicated.

See The Phonetics of sokuon, or geminate obstruents, by Shigeto Kawahara, and specifically sections 5.3 & 5.4

The summary is that a some have argued in the past that geminates in Japanese involve some form of glottal or laryngeal constriction. But a few others recently found no evidence for any such constriction.

So the answer for now is that no, obstruent geminates in Japanese don't involve full glottal stops.

the paragraph ends with "still to be explored" though, and the first section starts with "it would be interesting to investigate ..." – these are unsubtle code words that mean "we don't know"