Making up around 1-2% of Taiwan’s indigenous population, the Saisiyat people numbered 6743 in 2020. Among them, around 2000 were native speakers of the Saisiyat language, according to the Council of Indigenous Peoples Taiwan in 2015. The UNESCO Atlas of the World’s Languages in Danger has classified this language as “severely endangered”, with many of its youth preferring to use Hakka or Atayal instead, with few children speaking Saisiyat. Despite this, according to the Endangered Languages Project, it was noted that the Saisiyat are rather involved in preserving their language with support from the Taiwanese government, although a challenge was bridging phonological gaps between the over-80s and under-80s.

Today, the areas where Saisiyat is spoken are quite small, located in northwest Taiwan, in close proximity with Hakka and Atayal-speaking regions. Namely, these counties are the Miaoli (苗栗縣) and Hsinchu (新竹縣) counties. Saisiyat is split into two dialects, the Ta’ai or North Saisiyata (Saykilapaa), and the Tungho or South Saisiyat (Saymahahebon) dialects. Ta’ai is spoken in the townships of Nanchuang (南庄鄉) and Shihtan (獅潭鄉) in Miaoli county, while Tungho is spoken in Wufeng township (五峰鄉) in Hsinchu county.

Saisiyat has a total of 16 consonants and six vowels, with some variations in the “s” and “z” sounds, which can be pronounced as [θ] and [ð] respectively. Like all other Formosan languages, these sounds are represented using the Latin alphabet, from which around 18 letters are used. One of the more remarkable orthographical features is the use of both the uppercase and lowercase “s”, to distinguish the “sh” and “s/θ” sounds respectively. However, the letter “l” is used to represent both the “l” and retroflex “l” sounds. Diagraphs, or pairs of letters that make one sound, include “ae”, “oe”, and “ng”, as well as long vowels like “aa”, “ee”, and “ii”.

There are no adverbs in Saisiyat, consistent with many Formosan languages. Instead, verbs and other expressions are used to present the concept of the adverb. There are case endings in Saisiyat, although inflections only apply to pronouns. For other kinds of nouns, however, case markers are used instead. These are:

Noun form	Nominative	Accusative	Genitive	Dative	Possessive	Locative
Personal pronoun	Ø, hi	hi	ni	‘an-a	‘ini’	kan, kala
Common noun	Ø, ka	ka	noka	‘an noka-a	no	ray

Saisiyat allows for verb-initial constructions, like several other Formosan languages, but linguists have noticed a marked shift towards an accusative language, while retaining some features of a construction known as “split ergativity”. This is a term used to denote languages with features of ergativity in some sentences, and nominative-accusative in other kinds of sentences. Saisiyat is a subject-verb-object language, with the only other known Formosan languages with this feature being Thao, and the now-extinct Pazeh languages.

As many Saisiyat speakers are able to speak Hakka, Atayal, Mandarin Chinese, and Taiwanese Minnan in addition, one might expect to see influences of these languages on Saisiyat itself. However, I have yet to come across linguistic studies covering this topic, and so I am unable to ascertain if there is indeed some sort of linguistic influence, or there is a definitive lack of studies dedicated to studying this aspect. Should you come across articles or journals discussing this, please share or recommend some to me, either in the comments below, or through the Contact Us page. Thank you so much.

If you want to learn the Saisiyat language, I have to tell you that many resources are in traditional Chinese. However, there is one central site that documents the words, dialogues, audio and pronunciation of Saisiyat, along with most of the other Formosan languages. With multimedia materials, translations of books and stories into respective indigenous languages, this serves as a platform for people to learn about the cultures and languages of Taiwan, as well as preserving the Formosan languages in the digital world. Access it here at: http://web.klokah.tw/. This is so detailed, that even known individual dialects are compiled, and users are prompted to choose a dialect to learn that falls under the target language.

Featured image: Shihtan Township roadsign, seen along one of the connecting roads between Nanchuang and Shihtan townships in Miaoli county, Taiwan, seen in Google street view.