Some Observations on the Cantonese Lexical Suprafixes

Abstract

Suprafixation (more specifically, tonal affixation) is a word-building strategy that is somewhat frequently employed in Cantonese. This article explores the development of the lexical suprafixes in Standard Cantonese from three perspectives: (a) earlier descriptions of the Cantonese suprafixes and the behaviour of the lexical suprafixes in some older Romanised Cantonese texts; (b) the behaviour of the lexical suprafixes in some non-standard Cantonese varieties; and (c) the behaviour of the suprafixes and diminutives in Yuè and Pínghuà dialects in general. A definite answer cannot be found to the question of what the origin of the Cantonese suprafixes is. Nonetheless, the theory that the Cantonese lexical suprafixes stem from the diminutive suffix *ɲi 兒 (Cantonese ji4, Mandarin ér) remains the most convincing.

1. Introduction and Preliminaries

Affixes do not necessarily involve segments; some affixes consist of only suprasegmental features. An affix of which the form only includes suprasegments is a suprafix. One type of suprasegment is tone, and tonal affixes are not uncommon amongst Sinitic languages (Chappell 2023). Suprafixes (tonal affixes in this case) are reasonably common in Cantonese, especially lexical suprafixes (see Section 1.2 on ‘lexical’ versus ‘non-lexical’ suprafixes). This article looks at the development of the Cantonese lexical suprafixes from different perspectives.

Section 2 looks at what happened within Cantonese itself. Section 2.1 discusses some properties of the lexical suprafixes in modern Standard Cantonese. Section 2.2 looks at the history of the lexical suprafixes in older documents: explicit linguistic descriptions of their phonetics and phonology, and the behaviour of them in older tone-marked Romanised Cantonese texts. Section 2.3 discusses the situation with the lexical suprafixes in some non-standard Cantonese varieties. The data show that the lexical suprafixes on many words do not have a long history. Words can gain or lose suprafixes, although there are more cases of the former. Section 2.2 and Section 2.3 represent original research.

Section 3 looks at what other Yuè dialects and the Pínghuà dialects might tell us about the origin of the Cantonese lexical suprafixes. Section 3.1 discusses briefly the range of diminutive suffixes in Yuè and Pínghuà, Section 3.2 discusses the -ɲi^A and -ɲi^H diminutive suffixes, and Section 3.3 reviews the various opinions on the relationship between the diminutive suffixes and the Cantonese lexical suprafixes. Ultimately, an undisputable answer cannot be found to the question of where the Cantonese lexical suprafixes originated. Nonetheless, the popular theory that the Cantonese lexical suprafixes have the diminutive suffix *ɲi 兒 as their origin (e.g., Whitaker 1955–1956; Kwok Bit-Chee 2016) is still the most-plausible.

Section 3 is mainly a summary of opinions, especially that of Kwok Bit-Chee (2016)’s theory on the development of Yuè lexical suprafixes. It is one of the aims of this article to bring (snapshots of) excellent Chinese articles on this topic—like Kwok Bit-Chee (2016), Kao Wan-yu (2007), and Shào Huìjūn (2005)—to the Anglophone audience. In Section 3.2, new data from Pínghuà dialects and some extreme-western Yuè dialects are presented, and a proposal is made for a small extension to Kwok Bit-Chee (2016)’s theory: before a high-toned -ɲi^H diminutive suffix is developed, there is a preceding step of a -ɲi^A diminutive suffix in its citation tone (see the discussions in Section 3.2).

1.1. ’Cantonese’, and Some Transcription/Presentation Issues

Yuè is a dialect group within the Sinitic language family. Cantonese is the representative of the Yuè dialect group. What is considered ‘Cantonese’ varies greatly; different people use the term ‘Cantonese’ to cover different ranges of Yuè dialects. A somewhat-narrow definition of ‘Cantonese’ is adopted here: Cantonese is the language of Canton (i.e., Guǎngzhōu) and its derivatives (e.g., de Sousa 2022). Since the First Opium War (1839–1842), there have been massive waves of migration in all sorts of directions from the Pearl River Delta region. Migrants from the Canton area were particularly successful in making their version of the Yuè language the dominant speech in many towns and cities in Far Southern China (and also in many Chinatowns overseas). Many Cantonese enclaves can be found throughout Far Southern China. Some examples are Hong Kong, Macau, Zhànjiāng, Běihǎi, and Nánníng. Cantonese pushed the pre-existing Yuè, Hakka, Mǐn, Pínghuà, Southwestern Mandarin and/or Zhuàng varieties in those places away from the town or city centre. The various Cantonese varieties are slightly different from each other, due to, for instance, the difference in the language contact environment, and having preserved different features of earlier Cantonese. Nonetheless, they have remained fairly mutually intelligible, given the short history of divergence, and the (various degrees of) interactions that the different Cantonese enclaves have with each other.

Nowadays, there are two ‘standard’ varieties of Cantonese: Canton and Hong Kong Cantonese. (‘Standard’ is in quotation marks here, as there are only informal standards used for, e.g., broadcast. In the Hong Kong education system, only a small part of Cantonese is codified, namely the pronunciation used for reading Written Chinese. The suprafixes used in colloquial Cantonese are not standardised, although there are strong norms.) Given that Canton and Hong Kong are (traditionally) the two most important commercial centres in the Cantonese world, and that they are geographically not very far away from each other, Canton and Hong Kong have maintained strong contacts with each other, and the difference between these two Cantonese standards is very small (especially in comparison with the other Cantonese varieties). The Cantonese data presented in this article are from Hong Kong Cantonese by default, spoken by speakers under 50 years old or so. For the Hong Kong Cantonese data, consultations have been made with Bauer (2020)’s dictionary and the online dictionary words.hk 粵典.

Broad-phonetic IPA transcription (i.e., phonemes with their ‘main’ allophones) is used for the data provided by the author. When quoting from published sources, non-IPA transcriptions are converted to IPA, except with sources that are more than one hundred years old (see Section 2.2.3). Tonal categories (tonemes, allotones) are notated using full-sized numbers (e.g., nɵy5 女 ‘female’ is in ‘tone 5’), while tone values (phonetic realisations) are notated using superscript numbers, with [⁵] being the highest pitch (in modal register) and [¹] being the lowest pitch (e.g., nɵy¹³ 女 ‘female’ has a low-rising tone, i.e., Standard Cantonese tone 5 is a low-rising tone). Suprafixes in Yuè and Pínghuà dialects are usually substitutive, i.e., the suprafix tone replaces the citation tone (the underlying tone) of the host syllable. They are separated by a hyphen; before the hyphen is the citation tone that is not pronounced, and after the hyphen is the suprafix. For instance, nɵy5-2/nɵy¹³⁻²⁵ 女 ‘daughter’ has a tone 2 suprafix, which is a high-rising tone. (Other than substitutive suprafixes, there are also additive suprafixes, e.g., Cantonese described by Ch’an (1900) (Section 2.2.2) and modern Toishanese (Section 3.2). Additive suprafixes are those where both the citation tone and the suprafix tone are pronounced on a lengthened host syllable.) In older Cantonese, there was a distinction between the suprafix version and the ordinary version of tone 2 and tone 1. When there is a contrast, the suprafix version is notated with a star, e.g., ‘tone 2*’ and ‘tone 1*’, while the ordinary version lacks a star, e.g., ‘tone 2’ and ‘tone 1’.

Modern Standard Cantonese has six tonemes (Section 2.1). Unlike the Sidney Lau Romanisation and Jyutping, which number the tones 1, 2, 3, 4, 5 and 6, this article follows the S. L. Wong Romanisation and Cantonese Pinyin in numbering the tones of obstruent-ending syllables as 7, 8 and 9, while reserving 1, 2, 3, 4, 5 and 6 for sonorant-ending syllables. (When an obstruent-ending syllable takes a tone 2 or 1 suprafix, it is notated as in tip9-2 碟 ‘plate’.) This separation is primarily for philological reasons: the late-nineteenth/early-twentieth century Romanised Cantonese texts transcribed the tones on obstruent-ending versus sonorant-ending syllables separately, following Chinese linguistic traditions. In addition, sometimes they can behave differently. For instance, tones 1 and 7, both high level in pitch, behaved differently in the 1960s Hong Kong Cantonese tone sandhi rules (Section 2.2.1).

Lexical suprafixes are very rarely reflected in the Chinese script. For instance, both nɵy5 ‘female’ and nɵy5-2 ‘daughter’ are written 女. (Nonetheless, nɵy5-1 ‘queen (playing card)’ may be written with a separate character 囡, and nɵy5-4 nɵy5-2 or nɵy5-4 nɵy5-1 ‘darling daughter/girlfriend’ may be written 囡囡 or 女女.) Instead, the study of Cantonese suprafixes primarily relies on Romanised texts that are tone-marked.

The glossing in this article basically follows the Leipzig glossing rules (see also Chén Yùjié et al. (2014) for the application of the Leipzig glossing rules with Sinitic languages). Two ‘non-Leipzig’ abbreviations that are frequently used in this article are supr for suprafix and dim for diminutive.

1.2. Suprafixation versus Tone Sandhi, and Non-Lexical vs. Lexical Suprafixation

The Cantonese suprafixes can cause a change in meaning and/or word class. Cantonese suprafixation is also known in the literature by terms like tone change (e.g., Matthews and Yip 2011, chap. 1.4.2), changed tone (e.g., Jurafsky 1988; Liu Te-hsin 2016; Alderete et al. 2022), morphological tone (e.g., Alan Yu 2007), pin3 jɐm1/binyam/biànyīn 變音 (lit. ‘changed sound’; e.g., Bauer and Benedict 1997, chap. 2.11; Kwok Bit-Chee 2016), and pin3 tiu6/biàndiào 變調 (lit. ‘changed tune’ or ‘changed tone’; e.g., Cheung 2000; Yiu 2010). The last term, pin3 tiu6/biàndiào 變調, is also used to mean tone sandhi. It is important to note the difference between tone sandhi and suprafixation. Tone sandhi is a largely automatic phonological process where one tone transforms into another tone under the conditioning of another tone in the vicinity. (Sometimes there can be morphosyntactic constraints involved.) The Yuè dialects tend to be poor in tone sandhi. Modern Standard Cantonese has no tone sandhi. (However, see Section 2.2.1 for the tone sandhi rules in the 1960s involving tones 1 and 1*.) On the other hand, suprafixation is not triggered by the tones or other phonological features in the vicinity. Suprafixes are just like other affixes; whether one uses a suprafix or not depends on lexical, morphosyntactic, semantic and/or pragmatic considerations. (Although sometimes there can be phonological constraints.)1

The Cantonese suprafixes are here divided into ‘non-lexical’ and ‘lexical’ suprafixes. ‘Non-lexical’ and ‘lexical’ refer to the lexical identity of the base. Non-lexical suprafixes are not constrained by the lexical identity of the base. They often convey grammatical functions. For instance, there is a tone 2 (high-rising) suprafix that signifies the perfective aspect (a contraction of the perfective particle tsɔ2 咗). There are constraints such as a) this suprafix cannot be at the end of an utterance, and b) the host syllable cannot be in tone 1/7 (high level) (e.g., fei1 飛 ‘fly’ → *fei1-2 ‘flew’, tɪk7 剔 ‘tick’ → *tɪk7-2 ‘ticked’). Otherwise, the host can be any verb (or verby adjective), e.g., fɐn3 kau3 瞓覺 ‘sleep’ → fɐn3-2 kau3 ‘slept/fell asleep’, ji4mɐn4 移民 ‘migrate’ → ji4-2mɐn4 ‘migrated’. (In Cantonese, fɐn3 瞓 and ji4 移 are the verb, cf. ji4 tsɔ2 mɐn4 移咗民 ‘migrated’.) Other examples of non-lexical suprafixes include the following: (a) the -tei2 哋 ‘-ish’ construction, which involves the reduplication of a monosyllabic adjective, a tone 2 suprafix on the copy, and a suffix -tei2, e.g., lam4 lam4-2 -tei2 藍藍哋 ‘bluish’; and (b) the contraction of jɐt1 一 ‘one’, e.g., mat8 jɐt1 mat8 kʰɵy5 抹一抹佢 (wipe one wipe 3sg) → mat8-2 mat8 kʰɵy5 ‘wipe it a bit’ (signifying the delimitative aspect in this case).

On the other hand, lexical suprafixes depend on the lexical identity of the base. Whether it is applied to a base or not is mostly unpredictable. For instance, ji5 wan4-2 耳環 ‘ear ring’ has a tone 2 suprafix, but pei6 wan4 鼻環 ‘nose ring’ does not. The use of the lexical suprafix in ji5 wan4-2 耳環 ‘earring’ has nothing to do with the phonological environment. As is sometimes the case with derivational morphology, the application of the lexical suprafix in ji5 wan4-2 耳環 ‘earring’ is not entirely explainable. (The explainable part is that ‘earring’ is a more-familiar concept than ‘nose ring’ and other piercing/accessorial rings. However, being familiar does not guarantee a lexical suprafix.)

The Cantonese lexical suprafixes are quite often considered diminutives (e.g., Chén Zhōngmǐn 2002; Kao Wan-yu 2007; Kwok Bit-Chee 2016). While the lexical suprafixes most probably stem from diminutive suffixes (Section 3), in the majority of cases, the lexical suprafixes no longer have any diminutive-type meaning in Cantonese (Section 2). This article refrains from calling the Cantonese lexical suprafixes diminutives, but it is acknowledged they are at least derivatives of diminutives, regardless of meaning.

2. Lexical Suprafixes in Cantonese

Section 2.1 describes some aspects of the lexical suprafixes in modern Hong Kong Cantonese. Section 2.2 takes a diachronic look at the suprafixes in Standard Cantonese. Section 2.2.1 and Section 2.2.2 look at how there used to be separate suprafix and ordinary versions of tones 1 and 2, respectively, in the twentieth century. Section 2.2.3 looks at how words in modern Cantonese have gained, maintained, lost, or changed suprafixes in comparison with those in Ball (1888, 1894). Section 2.3 looks at the use of lexical suprafixes in some non-standard Cantonese varieties. Section 2.4 is an interim summary.

2.1. Modern Standard Cantonese

Some basic properties of the lexical suprafixes in modern Standard Cantonese are outlined in this section. For fuller accounts, please see the studies quoted in this section.

Table 1. Tones in modern Standard Cantonese in relation to Middle Chinese categories (tone values from Matthews and Yip 2011, p. 28).

	A	B	C	D(L)	D(S)
*voiceless	1 [55]	2 [25~35]	3 [33]	8 [33] (=3)	7 [55] (=1)
*voiced	4 [21~11]	5 [23~13]	6 [22]	9 [22] (=6)

shows the inventory of tones in modern Hong Kong Cantonese, arranged in relation to Middle Chinese categories.2 (The Middle Chinese categories are not essential for the understanding of this article. They are provided here only for the convenience of historical phonologists.)

There are two lexical suprafixes in modern Standard Cantonese: the tone 2 (high-rising) suprafix and the rarer tone 1 (high-level) suprafix.3 They replace the citation tone of the host syllable. Like some other derivational affixes, whether a lexical suprafix can be used or not, which one is used, and what effects it has on the base are not quite predictable. The following are examples of these two lexical suprafixes applied to the noun root mui6 妹 (< Middle Chinese mwoj^C) ‘younger sister’. First of all, in citation tone, the root mui6 ‘younger sister’ is a bound morpheme; most usually it occurs in a compound, e.g.,

(1)	a.	mui6		fu1		妹夫
		y.sis		husband
		‘younger sister’s husband’
	b.	hɪŋ1	tɐi6		tsi2		mui6	兄弟姊妹
		o.bro	y.bro		o.sis		y.sis
		‘siblings’

To say ‘younger sister’ on its own, mui6-2 妹 is used, with a tone 2 suprafix. The tone 2 suprafix is also applied in some other cases unpredictably, e.g., example (2b) below.

(2)	a.	ŋɔ5	kɔ3		mui6-2		我個妹
		1sg	clf		y.sis\supr
		‘my younger sister’
	b.	tsi2		mui6-2		姊妹
		o.sis		y.sis\supr
		‘sisters’ (cf. (1b) above)

With a tone 1 suprafix, -mui6-1 妹 is a suffix meaning ‘-girl’. The tone 1 suprafix is also applied to the root mui6 妹 unpredictably in some nouns, e.g., example (3c) below.

(3)	a.	lʊŋ4 -mui6-1		聾妹
		deaf -y.sis\supr
		‘deaf girl’
	b.	ma5lai1 -mui6-1			馬拉妹
		Malay -y.sis\supr
		‘Malay girl’
	c.	mui6-1 -tsɐi2	妹仔
		y.sis\supr -dim
		‘young female servant’

With the suffix -mui6-1 ‘girl’, suprafixation is applied at the morpheme level (i.e., -1 is applied to -mui6, and then -mui6-1 is suffixed to a base). However, suprafixation is often applied at the word level instead. For instance, with the suffix -pʰɔ4 婆 ‘-woman’, while kʷɐi2 -pʰɔ4 鬼婆 ‘European woman’ and pat3 -pʰɔ4 八婆 ‘bitchy woman’ do not take a lexical suprafix, tin1 -pʰɔ4-2 癲婆 ‘crazy woman’ and jɐu5 tsʰin4-2 -pʰɔ4-2 (have money\supr -woman\supr) 有錢婆 ‘rich woman’ do, for no apparent reason. This arbitrariness can also be demonstrated using place names. The place name ou3mun4-2 澳門 ‘Macau’ is most usually pronounced with a lexical suprafix tone 2 (see also Section 2.3). On the other hand, most other -mun4 門 ‘door’ place names do not carry a suprafix, e.g., tʰyn4 mun4 屯門 ‘Tuen Mun (suburb of Hong Kong)’, tɐu2 mun4 斗門 ‘Dǒumén (district of Zhūhǎi), kɔŋ1 mun4 江門 ‘Kongmoon/Jiāngmén’, and ha6 mun4 廈門 ‘Amoy/Xiàmén’. The place name hɔ4 lan4-1 荷蘭 ‘Holland’ has the rare tone 1 suprafix, but the other -lan4 蘭 place names do not, e.g., fɐn1 lan4 芬蘭 ‘Finland’ and nɐu2 sɐi1 lan4 紐西蘭 ‘New Zealand’.

Only words that depict familiar concepts have a lexical suprafix. For instance, in Canton, hɔ4 nam4-2 河南 (river south), with a lexical suprafix, is the suburbs of Canton south of the river (basically Hǎizhū District), whereas hɔ4 nam4 河南, without a lexical suprafix, is the relatively unfamiliar Hénán province in northern China. In another example, ‘Vancouver’ is wɐn1 kɔ1 wa4 溫哥華 in Cantonese, but a ‘true’ Cantonese Vancouverite calls their city wɐn1 kɔ1 wa4-2, with a tone 2 suprafix.

Historically, the lexical suprafixes probably came from a diminutive suffix (Section 3). In modern Standard Cantonese, lexical suprafixes can sometimes still have diminutive-type meanings. For instance, while lɔ4 籮 is a basket of any size, lɔ4-1 is a basket that is not very big. Some names habitually carry a lexical suprafix (primarily involving tone 4 becoming tone 2), and this clearly has an endearment meaning, e.g., hɔ4-2 pak8 何伯 (Ho\supr old.man) ‘Uncle Ho’ and lɪŋ4 lɪŋ4-2 玲玲 ‘Ling Ling’. Nonetheless, most lexical suprafixes do not have any diminutive-type meaning. For instance, lœŋ4 tʰɪŋ4-2 涼亭 is a gazebo of whatever size, and *lœŋ4 tʰɪŋ4 is not a valid word in colloquial Cantonese (in Hong Kong/Canton). Nouns depicting small things do not necessarily carry a lexical suprafix. For example, looking at two similarly small-sized fruits, while wɔŋ4 pʰei4-2 黃皮 ‘wampi’ carries a lexical suprafix tone 2, lʊŋ4 ŋan5 龍眼 ‘longan’ does not, and cannot, carry a lexical suprafix. Similarly, while pɛ1 lei4-2 啤梨 ‘Western pear’ carries a lexical suprafix, syt8 lei4 雪梨 ‘Chinese pear’ does not.

Correlating with the fact that diminutives are suffixes in Cantonese, the host syllable that obtains a suprafix is often at the end of a word. However, the suprafixed syllable is not necessarily word-final. For instance, a suprafixed base can be followed by a suffix, e.g., tip9-2 -tsɐi2 碟仔 (plate\supr -dim) ‘little plate’. One example that is harder to explain is the name of the suburb of jœŋ4-2 kei1 楊箕 (Yángjī) in Canton; jœŋ4 楊 came from jœŋ4 tʰou4-2 楊桃 ‘star fruit’, i.e., jœŋ4 楊 was not even word-final to begin with.

The citation tone of a suprafixed syllable is usually known by people, as the syllable (morpheme) in citation tone is usually found in some other environments. For instance, ‘egg’ is tan6-2 蛋, and ‘fowl egg’ is kɐi1 tan6-2 雞蛋. The citation-tone morpheme tan6 蛋 is still known by people, as it appears in words like tan6 gou1 蛋糕 ‘cake’ and kɐi1 tan6 -tsɐi2 雞蛋仔 (fowl egg -dim) ‘egg waffle’. (With a lexical suprafix, kɐi1 tan6-2 -tsɐi2 (fowl egg\supr -dim) would be literally ‘little fowl egg’.) However, with some morphemes, the original citation-tone form is lost. One example is tsʰaŋ2 橙 ‘orange’; this morpheme 橙 is now pronounced tsʰaŋ2 in all environments. Very few people know that 橙 is meant to have a citation-tone pronunciation of tsʰaŋ4 (< Middle Chinese ḍɛŋ^A). (The root 橙 ‘orange’ has remained in tone 4 in most other Yuè dialects. One could also figure this citation tone 4 out through cognates in other Sinitic languages, e.g., Mandarin chéng 橙.) In another example, the highest mountain in Hong Kong, tai6 mou6 san1, is sometimes written incorrectly as 大霧山 (big fog mountain), as the pronunciation suggests. Nonetheless, mou6 is actually the citation-tone pronunciation of mou6-2 帽 ‘hat’ (i.e., 大帽山 (big hat mountain))—a fact that is often not consciously known by younger speakers in Hong Kong (unlike, e.g., older speakers in Macau, many of whom still say mou6 for ‘hat’; see Section 2.3.).

With some words, the citation tone and suprafixed versions are in free variation. One example is ap8 ~ ap8-2 鴨 ‘duck’. (However, the diminutive form does not take a suprafix: ap8-tsɐi2 鴨仔 ‘duckling’. There is also the slang ap8-2 鴨 ‘male prostitute’, always with a suprafix.) In some cases, a citation-tone syllable and a suprafixed syllable are both morphologically free words, with different semantics. One example is tʰɔŋ4 糖 ‘sugar’ vs. tʰɔŋ4-2 糖 ‘lolly/candy’. (Cf. one common function of diminutives is denoting a delineated part of a mass (Jurafsky 1996)). With polysyllabic (and polymorphemic) words, not uncommonly the suprafixed version is an established compound with a less-transparent meaning, while the citation-tone version has the literal meaning. For instance, wɔŋ4 pʰei4-2 黃皮 is the fruit ‘wampi’, while wɔŋ4 pʰei4 黃皮 is literally ‘yellow skin’. (The aforementioned kɐi1 tan6 -tsɐi2 雞蛋仔 ‘egg waffle’ versus kɐi1 tan6-2 -tsɐi2 雞蛋仔 ‘little fowl egg’ are counter-examples.)

Occasionally, lexical suprafixes are found with non-noun bases. For instance, 大 tai6 is ‘big’, but tai6-1 is unexpectedly small. (This usage of tai6-1 is now perhaps slightly dated. The distinction is not expressed in the Chinese script.)

(4)	nei5	tsɛk8	kɐu2	kɐm3	dai6-1	tsɛk8	kɛ2?	你隻狗咁大隻嘅？
	2sg	clf	dog	so	big\supr	clf	surprise
	‘Your dog is so small?!’

For non-noun bases, often they become nouns when a lexical suprafix is applied. For instance, verb tsʰat8 刷 ‘to brush’ vs. noun tsʰat8 ~ tsʰat8-2 刷 ‘brush’; classifier tʰʊŋ4 筒 for small cylindrical objects vs. noun tʰʊŋ4-2 筒 ‘tube-shaped object’; and numeral lœŋ5 兩 ‘two’ vs. noun (or unit classifier) lœŋ5-2 兩/両 ‘tael’.

The tone 2 lexical suprafix can be applied to a host syllable in any tone other than tone 1/7 (high level). As for host syllables that are in tone 2 (high rising), there would be no perceptible difference if a tone 2 lexical suprafix is applied to it, and I remain agnostic as to whether a tone 2 host syllable can take a tone 2 suprafix or not in modern Cantonese. (There are no data that suggest that middle-aged and younger speakers produce and perceive a difference between an ordinary tone 2 and a suprafix tone 2 (Section 2.2.2). In the late-nineteenth century when there was such a contrast, on rare occasions, a suprafix tone 2* could be applied to a tone 2 syllable (Section 2.2.3).) In the following (modern) examples, the suprafixed forms can function as independent nouns, while the citation-tone forms cannot, except (5e) ap8 鴨 ‘duck’ and (5d) tɔi6 袋 in the sense of ‘scrotum’. The citation-tone forms can function as bound noun roots, and sometimes as independent words in other word classes.

(5)		citation tone	lexical suprafix tone 2
	a.	片 pʰin3 classifier ‘slice’noun, e.g., 片面 pʰin3 min6 ‘one-sided’	片 pʰin3-2 noun ‘film’, ‘slice’
	b.	房 fɔŋ4 classifier ‘wing of family’noun, e.g., 心房 sɐm1 fɔŋ4 ‘heart atrium’	房 fɔŋ4-2 noun ‘room’
	c.	女 nɵy5 noun, e.g., 女性 nɵy5 sɪŋ3 ‘female sex’	女 nɵy5-2 noun ‘daughter’
	d.	袋 tɔi6 classifier ‘bag’ verb ‘to (put into a) bag’ noun, e.g., (春)袋 (tsʰɵn1) tɔi6 ‘scrotum’	袋 tɔi6-2 noun ‘bag’
	e.	鴨 ap8 noun, e.g., 鴨 ap8 ‘duck’, 鴨屎 ap8 si2 ‘duck faeces’	鴨 ap8-2 noun ‘duck’, ‘male prostitute’
	f.	碟 tip9 classifier ‘plate’ noun, e.g., 碟頭飯 tip9 tʰau4-2 fan6 ‘rice dish’	碟 tip9-2 noun ‘plate’, ‘disc’

The tone 1 suprafix is much rarer. The tone 1 suprafix can be applied to a host syllable in any tone. (Similarly, I remain agnostic as to whether a tone 1 suprafix can be added to a host syllable in tone 1/7 (high level). Examples of tone 8 taking the tone 1 suprafix cannot be found; maybe this is an accidental gap, given the rarity of the tone 1 suprafix, and the relative rarity of tone 8 syllables.) In the following examples, in citation tone, only (6d) mei5 尾 ‘tail’ can function as an independent noun.

(6)		citation tone	lexical suprafix tone 1
	a.	己 kei2 noun, e.g., 私己錢 si1 kei2 tsʰin4-2 ‘(secret) personal money’	自己 tsi6kei2 ~ tsi6kei2-1 pronoun ‘self’
	b.	派 pʰai3 verb ‘distribute’, ‘hand out’ noun ‘faction,’ e.g., 門派 mun4 pʰai3 ‘sect’	派 -pʰai3 ~ -pʰai3-1 ‘-faction,’ e.g., 左派 tsɔ2 pʰai3~1 ‘left faction’
	c.	欄 lan4 noun, e.g., 圍欄 wɐi4 lan4 ‘fence’	欄 lan4-1 noun ‘animal pen’, ‘vegetable market’
	d.	尾 mei5 noun ‘tail’	尾 -mei5-1 e.g., 第尾 tɐi6 -mei5-1 ‘final, last’
	e.	妹 mui6 noun, e.g., 妹夫 mui6 fu1 ‘younger sister’s husband’	妹 -mui6-1 e.g., 泰妹 tʰai3 -mui6-1 ‘Thai girl’
	f.	掠 lœk9 verb, e.g., 搶掠 tsʰœŋ2 lœk9 ‘rob’	掠水 lœk9-1 sɵy2 ‘rob money’

2.2. Earlier Cantonese

In modern Cantonese, the suprafix tone 2 is indistinguishable from the ordinary tone 2 (the tone 2 of syllables that inherently have a tone 2), and for the vast majority of speakers, the suprafix tone 1 is indistinguishable from the ordinary tone 1. On the other hand, in earlier Cantonese, they were different. (When there is a contrast, the suprafix versions are notated as 2*/1*, and the ordinary versions as 2/1. Nonetheless, in the past, there were rare cases where the suprafix was tone 2/1 and not tone 2*/1*; see Section 2.2.3.) Nowadays, a distinction between tones 1 and 1* can still be observed in the speech of some older speakers, but in the last forty years or so, there has been no report of speakers perceiving two different versions of tone 2.

In Section 2.2.1, we shall look at some earlier descriptions of tone 1 and tone 1*, and in Section 2.2.2, some earlier descriptions of tone 2 and tone 2* will be visited. In Section 2.2.3, we shall look at the use of the lexical suprafixes in the late-nineteenth/early-twentieth century.

2.2.1. Tone 1*

The loss of tone 1* as a distinct toneme in Standard Cantonese is rather recent. In fact, impressionistically, there are still older speakers in Canton, Hong Kong, and Macau who maintain a clear contrast between tones 1 and 1*. The following are some descriptions of tone 1 and tone 1* in the literature.

Matthews and Yip (1994, pp. 21–22; 2011, pp. 27–28) describe modern Hong Kong Cantonese tone 1 as a high-level tone. There is only one tone 1. They made a side comment that high falling is an infrequently encountered free variation of tone 1. The side comment was made in the main text in (1994) and in an endnote in (2013). This decrease in prominence impressionistically correlates with an actual decrease in the occurrence of the high-falling realisation. Tang Sze-Wing (2015, p. 359) basically agrees that the high-falling tone has largely disappeared in the speech of younger speakers in Hong Kong, and high falling is not contrastive with high level. Nonetheless, he lists two remaining minimal pairs: noun tsʰɛ⁵⁵ 車 ‘vehicle’ vs. interjection tsʰɛ⁵³ 唓 (expressing disdain), and adverb sin⁵⁵ 先 ‘first, prior’ vs. question particle sin⁵³ 先 ‘actually…?’, e.g., pin⁵⁵kɔ³³ hɵy³³ sin⁵⁵ sin⁵³ 邊個去先先 (who go first actually.q) ‘actually who goes first?’

Zhān Bóhuì et al. (2002, p. 10) mention that in Canton, tone 1 has the free variations of high level and high falling. They also mention that while high level is the norm, in some situations, tone 1 has to be pronounced as high falling. They offer the examples of verb ʃɐi⁵³ 篩 ‘to sieve, to filter’ vs. noun ʃɐi⁵⁵ 篩 ‘sieve’ and verb pau⁵³ 包 ‘to include’ vs. noun pau⁵⁵ 包 ‘bag’.

Bauer and Benedict (1997, p. 117) mention that amongst their participants from Hong Kong, only three had a high-falling tone: one had high falling and high level as free variations, one used the high-falling tone in ‘certain syntactic environments’, and one used the two tones contrastively, similar to their three participants from Canton.

Samuel Hung-nin Cheung ([1972] 2007) claims that high level and high falling are free variations of one toneme (p. 5) but then shows minimal pairs (p. 6). Perhaps speakers who made the distinction, and speakers who did not, were both commonly encountered in Hong Kong in the 1970s.

Cheung Yat-Shing (1969) presents Hong Kong Cantonese as clearly distinguishing tone 1 and tone 1*. In isolation, tone 1 was high falling, and tone 1* was high level. Cheung presents tens of minimal pairs; the following are three examples:

• Hong Kong Cantonese in the 1960s (Cheung Yat-Shing 1969, p. 84).

(7)		Tone 1 [53] (ordinary tone)	Tone 1* [55] (suprafix tone)
	a.	鞭 pin1 ‘whip’ (noun/verb)	辮 pin1* ‘braid’
	b.	忠 tʃʊŋ1 ‘loyal(ty)’	鐘 tʃʊŋ1* ‘clock’
	c.	燒 ʃiu1 ‘burn’	簫 ʃiu1* ‘flute’

Other than being the realisation of suprafix tone 1*, high level could also be a sandhi tone of ordinary tone 1. The rules in Hong Kong were as follows (Cheung Yat-Shing 1969, pp. 94–95):

(8)	a.	53 (tone 1) + 5 (tone 7) → 53 + 5 [no change] e.g., ʃan53 + kʊk5 → ʃan53 kʊk5 山谷 ‘mountain valley’
	b.	53 (tone 1) + 55 (tone 1*) → 55 + 55 e.g., ʃan53 + lʊŋ55 → ʃan55 lʊŋ55 山窿 (mountain hole) ‘cave’
	c.	53 (tone 1) + 53 (tone 1) → 55 + 55 in colloquial context, 53 + 53 in literary context e.g., hœŋ53 + tʃʰyn53 → hœŋ55 tʃʰyn55 ~ hœŋ53 tʃʰyn53 鄉村 ‘rural village’

These 1960s Hong Kong sandhi rules contrasted with those in Canton, where the first syllable became high level in all three cases (rule c: ⁵³ + ⁵³ → ⁵⁵ + ⁵³; Cheung Yat-Shing 1969, p. 94, quoting Zōng Fúbāng 1964).

Tone 1* is here considered a suprafix tone, as it behaves like the modern-day tone 2 suprafix. With a modern tone 2 suprafix (high rising), (a) the resulting base is usually a noun (e.g., verb sou3 掃 ‘brush, sweep’ vs. noun kɐi1 mou4 sou3-2 雞毛掃 (fowl hair brush\supr) ‘feather duster’), and (b) some roots must take a tone 2 suprafix (e.g., tsʰaŋ4-2 橙 ‘orange’), some roots must not take a suprafix tone 2 (e.g., tʰou4 圖 ‘diagram’), and some roots vary, depending on the word (e.g., ap8 鴨 ‘duck’ in siu1 ap8 ~ siu1 ap8-2 燒鴨 ‘roast duck’ vs. tsʰɐu2 siu2 ap8 醜小鴨 (ugly little duck) ‘ugly duckling’). The situation with the 1960s tone 1* (high level) was the same: a) the resulting base is usually a noun (e.g., adjective hɔn4 ʃyn1 寒酸 (cold sour) ‘poverty-stricken’ vs. noun jim4 ʃyn1-1* 鹽酸 (salt sour\supr) ‘hydrochloric acid’, verb ʃan1 閂 ‘close (e.g., door)’ vs. noun mun4 ʃan1-1* 門閂 (door close\supr) ‘door lock’), and b) some roots must take a suprafix tone 1* (e.g., mau4-1* 貓 ‘cat’), some roots (in a non-sandhi environment) must not have a high-level tone (e.g., tʰin1 天 ‘sky’), and some roots vary, depending on the word (e.g., fʊŋ1 峯 ‘peak’ in kou1 + fʊŋ1 → kou⁵⁵ fʊŋ⁵⁵ ~ kou⁵³ fʊŋ⁵³ 高峯 ‘tall peak’ (sandhi rule 8c) vs. ʃan1 + fʊŋ1-1* → ʃan⁵³⁻⁵⁵ fʊŋ⁵⁵ 山峯 ‘mountain peak’ (sandhi rule 8b)).

Nonetheless, Cheung Yat-Shing (1969) analyses tones 1 and 1* as two ordinary tones in Hong Kong Cantonese. Lǐ Xīnkuí et al. (1995) have a similar stance for Canton Cantonese. While they acknowledge that tone 1* started off as a diminutive suprafix, the use of tone 1* had become so ubiquitous that people no longer had a sense that tone 1* was a derivative of tone 1, so it is better to treat them as two ordinary tones.

Shī Qíshēng (2004) looks at the situation with tones 1 and 1* in Ball (1883) and Ball (1912). The situation was similar. Ball (1888) (a newer edition of Ball 1883) offers minimal pairs like shí1 尸 ‘corpse’ vs. shí1* 詩 ‘hymn’ and tsz1 貲 ‘wealth’ vs. tsz1* 資 ‘postage’ (p. xxxiii). One difference with later descriptions is that, in Ball (1888), suprafix tone 1* syllables could only have citation tone 1, and occasionally tone 4 (see Section 2.2.3), whereas in Cheung Yat-Shing (1969), a wider range of citation tones were possible (although the citation tone was still predominantly tone 1).

There are also descriptions that describe tone 1* as an extra-high-level tone. Gāo Huánián (1980, p. 7) mentions that in Canton, tone 1 could be high falling or high level, while the high-level suprafix (tone 1*) was even higher than the normal high-level tone. Wong Shik-Ling ([1941] 1997) also describes the suprafix tone 1* as extra-high level. (Wong was from Canton and moved to Hong Kong in the early 1950s.)

2.2.2. Tone 2*

There has been no report of speakers perceiving two different types of tone 2 in the last forty years or so. However, there have been some rather-recent reports of speakers still producing such a distinction.

Liu Te-hsin (2016) measures the speech of a 77-year-old native speaker ‘born in the Province of Guangdong’, and the length of suprafixed tone 2 syllables (i.e., tone 2 syllables with a different citation tone) were on average double that of ordinary tone 2 syllables. (In contrast, with the ten subjects aged 20 to 25 years old, the difference was minimal.) Also quoted by Liu Te-hsin (2016) is Yu (2007)’s study: although the subjects produced the two types of high-rising tone (tone 2) with subtle differences, they were not able to perceive the difference between the two types of high-rising tone.

Both Matthews and Yip (1994, pp. 21–22; 2011, pp. 27–28) and Bauer and Benedict (1997, p. 116) mention that there is only one high-rising tone (tone 2) in Hong Kong.

Gāo Huánián (1980, p. 7) mentions that in Canton, the high-rising suprafix tone (tone 2*) rose even higher than the ordinary high-rising tone (tone 2).

Chao Yuen Ren (1947, pp. 34–35) describes tone 2* as having a larger rise [²⁵] than tone 2 [³⁵]. Bauer and Benedict (1997, p. 170) mention that Benedict (1942) came to a similar conclusion. They also mention Søren Egerod’s personal communication with Robert Bauer, where Egerod observed this difference while learning Cantonese in Canton in 1949/1950.

Kwok Bit-Chee (2009), in his article on the Cantonese grammatical suprafixes (a type of non-lexical suprafixes), quotes the phonetic descriptions in Ch’an Chan Sene (1900) (also quoted by Ball 1907).4 In Ch’an (1900)’s description, except for the suprafix version (⁵⁵) of tone 1 (⁵³), the high-tone suprafix was an additive suprafix: the citation tone was pronounced, and then the pitch rose. The syllable as a whole was lengthened. (This is similar to the modern day Toishanese high-rising suprafixes; see Section 3.2.) For Cantonese, this represents an intermediate stage in which the tone merger had not yet completed.

Ball (1888, p. xxxi) describes the ‘third rising tone’ as ‘a longer tone than the other two Rising Tones [tones 2 and 5], banning lower and ascending higher than either of them’. In other words, tone 2* was a ‘dramatic’ dip, something like [³¹⁵]. Ball (1888) provides minimal pairs like fong2 訪 ‘inquire’ vs. fong2* 房 ‘room’ and yöng2 抰 ‘shake (e.g., cloth)’ vs. yöng2* 樣 ‘pattern’.

2.2.3. Suprafixes in Ball’s Cantonese Textbooks

James Dyer Ball (1847–1919) was born in Canton. He grew up in Canton and spoke Canton Cantonese natively. He worked as a civil servant in Hong Kong between the 1870s and the 1900s. Ball was considered the most-knowledgable Westerner on Cantonese at the time. Amongst his many publications were descriptions of Cantonese, some other Yuè dialects (e.g., pre-Cantonesised Macau), and Hakka. (He has also published on a wide range of other topics.) The language publications include Chinese characters, Romanisations that are fully tone-marked, glossing, and translations. In this section, we will discuss some similarities and differences in the usage of suprafixes in some of Ball’s Cantonese publications versus usage in modern-day Standard Cantonese.

Probably the best known amongst Ball’s language publications are his Cantonese Made Easy textbooks; the first to third editions were published in 1883, 1888, and 1907, and a fourth edition was published posthumously (Ball 1924). In this section, we will primarily look at the second edition (Ball 1888). Also discussed in this section is Readings in Cantonese Colloquial (Ball 1894, fourth edition). In the data presented below, his Romanisation is used for the segments, while the tones are notated using numerals like the rest of this article.5

Looking at Ball (1888), host syllables with suprafix tone 1* most usually had tone 1 as the citation tone, and rarely tone 4.6 (Kwok Bit-Chee (2016, p. 291) makes the observation that, amongst the cases of suprafix tone 1* syllables with a non-tone-1 citation tone listed by Mài Yún (1995, p. 267), the onset is most usually sonorant. In other words, most usually, only non-tone-1 syllables with a sonorant onset can take either suprafix tone 1* or tone 2*. Otherwise, there is usually only one choice, with tone 1/7 taking tone 1* and tones 3/8, 4, 5, and 6/9 taking tone 2*.) In Ball (1888), there were some cases where the suprafix was tone 1 (high falling) and not tone 1* (high level), e.g., sz1 nái5-1 師奶 ‘lady’ and man4-1 蚊 ‘mosquito’.

With the suprafix tone 2*, the citation tone was one of the low tones—usually tone 4 or 6, sometimes tone 5, and there was one case of tone 9: ts’ak9-2 賊 ‘thief’. (This last case was so rare that Ball (1888) had a footnote about it (p. 31).) Importantly, the suprafix tone 2* did not have tone 2 as its citation tone. (With one exception: the verb ts’ui2-2* 娶 ‘get wife’. This was a lexical suprafix, as ts’ui2-2* 娶 had a suprafix in isolation and in whatever aspect-mood. That this is a verb perhaps also explains why this tone 2 could take a suprafix tone 2* unexpectedly; perhaps this was originally a non-lexical suprafix, and non-lexical suprafixes did not have this constraint.) Other than suprafix tone 2*, there were cases where the suprafix was tone 2 (see

Table A4. Words that were pronounced in suprafix tone 2 in Ball (1888) and are now pronounced in suprafix tone 2 or another tone (‘suprafix maintained’ or ‘suprafix lost’).

	Ball (1888)		Modern Cantonese		Ball (1888)		Modern Cantonese
tael	両	löng5-2	lœŋ5-2	cess-pool	氹	t’am5-2	tʰɐm5
antithetical sentence	對	tui3-2	tɵy3-2	coiffure	髻	kai3-2	kɐi3

in Appendix A). While there are discussions in the literature on tones 1* and 2* as suprafixes, there do not seem to be discussions on tones 1 and 2 being lexical suprafixes during times when tones 1/2 and 1*/2* were distinguished.

In comparison with Ball (1888), Ball (1907) sees more cases of tone 1*. The following are some examples.

(9)			Ball (1888)	Ball (1907)
	a.	師奶 ‘lady’	sz1 nái4-1	sz1 nái4-1~1*
	b.	蚊 ‘mosquito’	man4-1	man4-1*
	c.	啲 (mass classifier, comparative)	ti1 ~ tí1	ti1* ~ tí1*
	d.	都 (‘all’)	tò1	tò1~1*
	e.	資 ‘postage’	tsz1~1*	tsz1*
	f.	艙 ‘cabin/hold’	ch’ong1 ~ ts’ong1*	ts’ong1*

(See also Shī Qíshēng (2004)’s study comparing tones 1 and 1* in Ball (1883) and another textbook (Ball 1912), with the same conclusion that there was an increase in tone 1*.) Yiu (2010) (which, in a sense, is a follow up of Cheung (2000)) looks at Romanised Cantonese texts in different time periods, and shows that cases of tone 1* (high level) have continuously increased, at the expense of tone 1 (high falling). In other words, tone 1 [⁵³] has merged into tone 1* [⁵⁵]. The high-falling tone [⁵³] has now basically disappeared; at most, it is a rarely used free variation of a tone that we now call tone 1. With differences,7 tone 2 [³⁵] has also merged into tone 2* [²⁵] (using Chao Yuen Ren (1947)’s tone values), becoming what we now call tone 2 [²⁵].

Comparing the lexical items in Ball (1888, 1894) with their modern equivalents, one can see that many words that had a lexical suprafix back in the late-nineteenth century still have a lexical suprafix nowadays. Some words have since gained a suprafix. Importantly, the opposite has also occurred: some words have lost their suprafixes, usually reverting back to their citation-tone pronunciation. The

Table 2. Number of unique lexical items in Ball (1888) that have gained, maintained, lost, or changed a lexical suprafix (see Table A1, Table A2, Table A3, Table A4, Table A5 and Table A6 for the lexical items involved).

Items Shown in Table:	Tone in Ball (1888)	Tone in Modern Cantonese	Suprafix:	Number of Items
A1	citation tone	suprafix tone 2	gained	33
A2	suprafix tone 2*	suprafix tone 2	maintained	34
A3	suprafix tone 2*	citation tone	lost	19
A3	suprafix tone 2*	suprafix tone 1	changed	2
A4	suprafix tone 2	suprafix tone 2	maintained	2
A4	suprafix tone 2	citation tone	lost	2
A5	suprafix tone 1*	suprafix tone 1	maintained	28
A5	suprafix tone 1	suprafix tone 1	maintained	1
A6	suprafix	(word is now obsolete)	?	11

and

Table 3. Number of unique lexical items in Ball (1894) that have gained, maintained, or lost a lexical suprafix (see Table A7, Table A8, Table A9, Table A10 and Table A11 for the lexical items involved).

Items Shown in Table:	Tone in Ball (1894)	Tone in Modern Cantonese	Suprafix:	Number of Items
A7	citation tone	suprafix tone 2	gained	14
A8	suprafix tone 2*	suprafix tone 2	maintained	27
A9	suprafix tone 2*	citation tone	lost	6
A10	suprafix tone 2	suprafix tone 2	maintained	1
A10	suprafix tone 2	citation tone	lost	1
A11	citation tone	suprafix tone 1	gained	1
A11	suprafix tone 1*	suprafix tone 1	maintained	7
A11	suprafix tone 1*	citation tone	lost	1
A11	suprafix tone 1	suprafix tone 1	maintained	1

show the number of unique lexical items in Ball (1888, 1894) that have gained, maintained, lost, or changed lexical suprafixes in comparison with modern Cantonese. The lexical items themselves are shown in

Table A1. Words that were pronounced in citation tone in Ball (1888) but are now usually pronounced in suprafix tone 2 (‘suprafix gain’).

	Ball (1888)		Modern Cantonese		Ball (1888)		Modern Cantonese
bangle	鈪	ák8	(ŋ)ak8-2	probably	約嗼	yök8 mok8	jœk8 mɔk8-2
duck	鴨	áp8	ap8/2	bear	熊人	hung4 yan4 A23	huŋ4 jɒn4-2
tongs	鉗	k’im4	kʰim4-2	share (n.)	股份	kwú2 fan6 A2	ku2 fɐn6-2
case	案件	òn3 kín6	ɔn3 kin6-2	jest	笑話	síú3 wá6 A2	siu3 wa6-2
daughter	女	nui5	nɵy5-2	sister	姊妹	tsz2 múi6 A2	tsi2 mui(6/)2
old woman	伯爺婆	pák8 ye4-1* p’o4	pak8 jɛ4-1 pʰɔ4-2	domino	骨牌	kwat7 p’ái4/2*A2	kʷɐt1 pʰai4-2
building site	盤	p’un4	pʰun4-2	(playing) cards	紙牌	chí p’ái4/2*	tsi2 pʰai4-2
string	繩	shing4	sɪŋ4-2	Yamen	衙門	ngá4 mún4/2*	ŋa4 mun4-2
sail	行船	háng4 shún4	haŋ4 syn4-2	thief	賊	ts’ák9/2*	tsʰak9-2
plate	碟	tip9	tip9-2	home country	鄉吓	höng1 há5/2*	hœŋ1 ha5-2
peach	桃	t’ò4	tʰou(4/)2	chess	碁	k’éi4/2*	kʰei4-2
sheep	綿羊	min4 yöng4	min4 jœŋ4-2	temple	廟	míú6/2*	miu6-2
mattress	牀褥	ch’ong4 yuk9	tsʰɔŋ4 jʊk9-2	pill	丸	yün4/2*	jyn4-2
Macao	澳門	Ò3 mún4	ou3 mun(4/)2	table	檯/臺	t’oi4/2*	tʰɔi4-2
B.A.	秀才	sau3 ts’oi4	sau3 tsʰɔi(4/)2	city	城	sheng4/2*	sɛŋ4-2
general assistant	打雜	tá2 tsáp9	ta2 tsap9-2	dollars	銀	ngan4/2*	ŋɐn4-2
monastery	祠堂	ts’z4 t’ong4	tsʰi4 tʰɔŋ4-2

^A23 This root is also featured in another word in Table A2 and Table A3. ^A2 This root is also featured in another word in Table A2. yök8 mok8 約嗼 ‘probably’: one would expect yök8 mok9 from usual sound-change rules from Middle Chinese; perhaps the tone in mok9 was assimilated by the tone of yök8.

,

Table A2. Words that were pronounced in suprafix tone 2* in Ball (1888) but are now usually pronounced in suprafix tone 2 (‘suprafix maintained’).

	Ball (1888)		Modern Cantonese		Ball (1888)		Modern Cantonese
man	人	yan4-2* A13	jɒn4/2	compradore	買辦	mái5 pán6-2*	mai5 pan6-2
share (n.)	份	fan6-2* A1	fɐn6-2	priest	和尚	wo4 shöng6-2*	wɔ4 sœŋ6-2
(-language)	話	-wá6-2* A1	-wa6-2	who	乜誰	mi1 shui4-2*	mɐt1 sɵy4-2
younger sister	妹	múi6-2* A1	mui6-2	egg	蛋	tán6-2*	tan6-2
card	牌	p’ái4-2* A1	pʰai4-2	master	事頭	sz6 t’au4-2*	si6 tʰɐu4-2
Yamen	衙門	ngá4 mún4/2*	ŋa4 mun4-2	sun	熱頭	yit9 t’au4-2*	jit6 tʰɐu4-2
thief	賊	ts’ák9/2*	tsʰak9-2	peach	桃	t’ò4-2*	tʰou4-2
home country	鄉下	höng1 há5/2*	hœŋ1 ha5-2	persimmon	柿	ts’z5-2*	tsʰi5-2
chess	碁	k’éi4/2*	kʰei4-2	vase, pot	壺	ú4-2*	wu4-2
temple	廟	míú6/2*	miu6-2	picture	畫	wá6-2*	wa6-2
pill	丸	yün4/2*	jyn4-2	kite	鳶	yíú6-2*	-jiu6-2
table	檯/臺	t’oi4/2*	tʰɔi4-2	pattern	樣	yöng6-2*	jœŋ6-2
city	城	sheng4/2*	sɛŋ4-2	fish	魚	yü4-2*	jy4-2
orange	橙	ch’ang4-2*	tsʰaŋ4-2	garden	園	yün4-2*	jyn4-2
prisoner	犯	fán6-2*	fan6-2	college	院	yün6-2*	jyn6-2
room	房	fong4-2*	fɔŋ4-2	commission	用	yung6-2*	jʊŋ6-2
hong	行	hong4-2*	hɔŋ4-2	ferry	渡	tò6-2*	-tou6/2
plan	計	kaí3-2*	kɒi3-2	(be a) cook		tsò6 ch’ü4-2*	tsou6 tsʰy4-2
sedan	轎	kíú6-2*	k(ʰ)iu6-2	(that time)	嗰陣時	ko3 chan6 shí4-2*	kɔ2 tsɐn6 si4/2
accountant	掌櫃	chöng2 kwai6-2*	tsœŋ2 kʷɐi6-2	(clf)	位	wai6-2*	wɐi6-2
basket	籃	lám4-2*	lam4-2	about	度	tò?-2*	tou2
fox	狸	léi4-2*	-lei4-2	get wife	娶	ts’ui2-2*	tsʰɵy2
blinds	㡘	lím4-2*	lim4-2	allow	由	yau4-2*	jɐu4-2
name	名	meng4-2*	mɛŋ4-2	certain	定	ting6-2*	tɪŋ6-2
face, top	面	min6-2*	min6-2	cook (n.)	火頭	fo2 t’au4-2* A3	fɔ2 tʰɐu4-2
hat	帽	mò6-2*	mou6-2	house	樓	lau4-2* A3	lɐu4-2
pass book	部	pò6-2*	pou6-2	money	銀	ngan4-2* A3	ŋɐn4-2

^A13 This root is also featured in another word in Table A1 and Table A3. ^A1 This root is also featured in another word in Table A1. ^A3 This root is also featured in another word in Table A3. tɪŋ6-2 定 ‘certain’ sounds slightly old, but it is still used.

,

Table A3. Words that were pronounced in suprafix tone 2* in Ball (1888) but are now usually pronounced in another tone (‘suprafix loss’ or ‘suprafix changed’).

	Ball (1888)		Modern Cantonese		Ball (1888)		Modern Cantonese
who	乜人	mi1 yan4-2* A12	mɐt1 jɐn4	together	一齊	yat7 ts’ai4-2*	jɐt7 tsʰɐi4
beginning	起頭	héi2 t’au4-2* A2	hei2 tʰɐu4	rope	纜	lám6-2*	lam6
upstairs	樓上	lau4-2* shöng6 A2	lɐu4 sœŋ6	shanty	寮	líú4-2*	liu4
money	銀錢	ngan4/2* ts’ín4-2* A2	ŋɐn4 tsʰin4-2	pineapple	波羅	po1 lo4-2*	pɔ1 lɔ4
ferry	渡	tò6-2*	-tou6/2	furnace	爐	lò4-2*	lou4
(that time)	嗰陣時	ko3 chan6 shí4-2*	kɔ2 tsɐn6 si4/2	bridge of nose	鼻梁	pei6 löng4-2*	pei6 lœŋ4
now	呢陣	ni1 chan6-2*	ni1 tsɐn6/2	place	定	teng6-2*	tɛŋ6
how long (time)	幾耐	kéi2 noi6-2*	kei2 nɔi6	rhyme	韻	wan5-2*	wɐn5
(verb clf)	吓	há5/2*	ha5	tonight	今晚	kam1 mán5-2*	kɐm1 man5/1
sit	坐	ts’o5-2*	tsʰɔ5
heavy	重	ch’ung5-2*	tsʰʊŋ5	sister-in-law	姨	yí4-2*	ji4-1

^A12 This root is also featured in another word in Table A1 and Table A2. ^A2 This root is also featured in another word in Table A2.

, Table A4,

Table A5. Words that were pronounced in suprafix tone 1* (or 1) in Ball (1888) but are now usually pronounced in suprafix tone 1 (‘suprafix maintained’).

	Ball (1888)		Modern Cantonese		Ball (1888)		Modern Cantonese
tetter	瘡	chong1/1*	tsʰɔŋ1	cloves	丁香	ting1-1* höng1-1*	tɪŋ1 hœŋ1
plaster	膏	kò1-1*	kou1	colour of nature	青	ts’ing1-1*	tsʰɪŋ1
market	欄	lán4-1*	lan4-1	salad	英	yíng1-1*	jɪŋ1
dollar	文	man4-1*	mɐn4-1	hold/cabin	艙	ts’ong1-1*/ch’ong1	tsʰɔŋ1
cat	貓	máu4-1/1*	mau4-1	go-down	貨倉	fo3 ts’ong1-1*	fɔ3 tsʰɔŋ1
thirty	卅	sá1/1*	sa1	note paper	箋	tsín1-1*	tsin1
star	星	seng1/1*	sɪŋ1	old man	伯爺公	pák8 ye4-1* kung1	pak8 jɛ4-1 kʊŋ1
hymn	詩	shí1/1*	si1	old woman	伯爺婆	pák8 ye4-1* p’o4	pak8 jɛ4-1 pʰɔ4-2
court	廳	t’eng1/1*	tʰɛŋ1	beggar	乞兒	hat7 yí4-1*	hɐt1 ji4-1
gun	鎗	ts’öng1/1*	tsʰœŋ1	hawk	鷹	ying1-1*	jɪŋ1
postage	(信)資	(sun3) tsz1/1*	sɵn3 tsi1	fly	烏蠅	wú1 ying4-1*	wu1 jɪŋ4-1
jingling	玎	ting1/1*	tɪŋ1	slight rain	雨微	yü5 méi4-1*	jy5 mei4-1
throughout	通	t’ung1/1*	tʰʊŋ1	scorch	燶	nung?-1*	nʊŋ1
cent	仙	sin1* (loanword)	sin1
hand gong	玎璫	ting1-1* tong1-1*	tɪŋ1 tɔŋ1	mosquito	蚊	man4-1	mɐn4-1

ts’ing1* 青 ‘colour of nature’ was different from ts’eng1 青 ‘(dark) blue’ in Ball (1888). 英 ‘salad’ can perhaps be considered obsolete; in words.hk, 英 ‘salad’ can be found in the entry kʷa1 jɪŋ1 瓜英 ‘papaya slices preserved by sugaring’.

and

Table A6. Words with lexical suprafixes in Ball (1888) that are now too dated.

	Ball (1888)			Ball (1888)
tetter	火疔瘡	fo2 teng1-1* chong1-1*	cash	X個錢	X ko3 ts’ín4-2*
posts of a certain frame	文𥲤	man4 kui6-2*	dollar	X個銀	X ko3 ngan4-2*
capital	本銀	pún2 ngan4-2*	pò-tsz	寶字	pò2 tsz6-2*
ingot	錠	ting3-2*	six-barrelled revolver	六口連	lúk9 hau2 lín~m4-2*
shop-coolie		kún2-tím3/2*	sailing ship	桅棒船	wai4 p’ang5-2 shün4
mistress	東家婆	tung1 ká1 p’o4-2*

for Ball (1888) and

Table A7. Words that were pronounced in citation tone in Ball (1894) but are now usually pronounced in suprafix tone 2 (‘suprafix gain’).

	Ball (1894)		Modern Cantonese		Ball (1894)		Modern Cantonese
mother	老母	lò5 mò5	lou5 mou5-2	mother-in-law	家婆	ká1 p’o4	ka1 pʰɔ(4/)2
pear	沙梨	shá1 léi4	sa1 lei4/2	granny	婆婆	p’o4 p’o4	pʰɔ4 pʰɔ4-2~1
jasmine	茉莉花	mút9 léi4 fá1	mut9 lei4-2 fa1	oppress	難為	nán4 waí6	nan4 wɐi4/2
daughter	女	nuí5	nɵy5-2	bear	熊人	hung4 yan4 A89	huŋ4 jɒn4-2
city	城	sheng4	sɛŋ4-2	money	錢銀	ts’ín4ngan4-2* A89	tsʰin4-2ŋɐn4-2
church	教會	káu3 wúi6	kau3 wui(6/)2	such as	之類	chí1 luí6/2*	tsi1 lɵy6/2
holy church	聖公會	shing3 kung1 wúi6	sɪŋ3 kʊŋ1 wui6-2	spice	香料	höng1 líú6/2	hœŋ1 liu6-2

^A89 This root is also featured in another word in Table A8 and Table A9.

,

Table A8. Words that were pronounced in suprafix tone 2* in Ball (1894) but are now usually pronounced in suprafix tone 2 (‘suprafix maintained’).

	Ball (1894)		Modern Cantonese		Ball (1894)		Modern Cantonese
man	人	yan4-2* A79	jan4/2	room	房	fong4-2*	fɔŋ4-2
money	錢	ts’ín4-2* A79	tsʰin4-2	money	錢銀	ts’ín4 ngan4-2*	tsʰin4-2 ŋɐn4-2
spice	香料	höng1 líú6/2*	hœŋ1 liu6-2	and be done with	罷	pá6-2*	pa6-2
peach	桃	t’ò4-2*	tʰou4-2	court	衙門	ngá4 mún4/2*	ŋa4 mun4-2
bean	豆	tau6-2*	tɐu6-2	disorderly	亂	lün6-2*	lyn6-2
flower garden	花園	fá1 yün4-2*	fa1 jyn4-2	a little	略略	lök9 lök9-2*	lœk9 lœk9-2
not (very) long	冇(幾)耐	mò5 (kéi2) noí6-2*	mou5 (kei2) nɔi6-2~1	pattern	樣	yöng6-2* A9	jœŋ6-2
sun	熱頭	yit9 t’au4-2*	jit9 tʰɐu4-2	name	名	meng4-2* A9	mɛŋ4-2
who	乜誰	mat7 shui4-2*	mɐt7 sɵy4-2	potato	薯	shü4-2* A9	-sy4/2
widow	寡母婆	kwá2 mò5 p’o4-2*	kʷa2 mou5 pʰɔ4-2	such as	之類	chí1 luí6/2* A9	tsi1 lɵy6/2
thing	物件	mat9 kín6-2*	mɐt6 kin6-2	probably	大概	tái6 k’oi3-2* A9	tai6 kʰɔi3-2
old (man)	老大	lò5 tái6-2*	lou5 tai6-2	once	一排	yat7 p’ái4-2* A9	jɐt1 pʰai4/2
(classifier)	位	waí6-2*	wɐi6-2	gradually	漸漸	tsím6 tsím6-2* A9	tsim6 tsim6/2
				that (clf) time	個陣時	ko3 chan6 shí4-2* A9	kɔ2 tsɐn6 si4/2

^A79 This root is also featured in another word in Table A7 and Table A9. ^A9 This root is also featured in another word in Table A9.

,

Table A9. Words that were pronounced in suprafix tone 2* in Ball (1894) but are now usually pronounced in another tone (‘suprafix loss’).

	Ball (1894)		Modern Cantonese		Ball (1894)		Modern Cantonese
who	乜人	mi1 yan4-2* A78	mɐt1 jɐn4	probably	大概	tái6 k’oi3-2* A8	tai6 khoi3/2
wages	工錢	kung1 ts’ín4-2*A78	kʊŋ1 tsʰin4	once	一排	yat7 p’ái4-2* A8	jɐt1 pʰai4/2
appearance	樣子	yöng6-2* tsz2 A8	jœŋ6 tsi2	gradually	漸漸	tsím6 tsím6-2* A8	tsim6 tsim6/2
Lord’s holy name	主之聖名	chü chí1 shing3 meng4-2*A8	tsy2 tsi1 sɪŋ3 mɪŋ4	that (clf) time	個陣時	ko3 chan6 shí4-2* A8	kɔ2 tsɐn6 si4/2
potato	薯	shü4-2* A8	-sy4/2	constantly	常常	shöng6 shöng4-2*	sœŋ4 sœŋ4
such as	之類	chí1 luí6/2* A8	tsi1 lɵy6/2	sit	坐	ts’o5-2*	tsʰɔ5

^A78 This root is also featured in another word in Table A7 and Table A8. ^A8 This root is also featured in another word in Table A8.

,

Table A10. Words that were pronounced in suprafix tone 2 in Ball (1894) but are now pronounced in suprafix tone 2 or another tone (‘suprafix maintained’ or ‘suprafix lost’).

	Ball (1894)		Modern Cantonese		Ball (1894)		Modern Cantonese
cross	十字架	shap9 tsz6 ká3/2	sɐp6 tsi6 ka3/2	stomach	肚	t’ò5-2	tʰou5

and

Table A11. Words that were pronounced in suprafix tone 1* (or 1) in Ball (1894) and are now pronounced in suprafix tone 1 or citation tone (‘suprafix maintained’, ‘suprafix lost’), or words that have gained a suprafix tone 1 (‘suprafix gained’).

	Ball (1894)		Modern Cantonese		Ball (1894)		Modern Cantonese
eh?	呢	ni1*	nɛ1	bend down head	垂低頭	shuí4 tai1-1* t’au4	sɵy4 tɐi1 tʰɐu4
descendent	子/仔孫	tsz/tsaí 2 sün1-1*	tsi2 syn1	finally	收尾	shau1 méi5-1*	sɐu1 mei(5/)1
side	側邊	chak7 pín1-1*	tsɐk7 pin1	mosquito net	蚊帳	man4-1 chöng3	mɐn4-1 tsœŋ3
which-	邊	pín1-1*-	pin1-
spear	鎗	ts’öng1/1*	tsʰœŋ1	evening	挨晚	ái1 mán5	ai1 man5-1

for Ball (1894) in the appendix. (Some lexical items are shown in more than one table, e.g., free variations. Cases of non-lexical suprafixes are ignored. Some verbs and adjectives always carried a lexical suprafix, regardless of the grammatical environment.)

Obviously, textbooks only demonstrate a small portion of the vocabulary used during a particular time period. One cannot conclude from simply comparing the suprafixed words in old textbooks versus modern language that the number of suprafixed words overall has increased or decreased. Nonetheless, that there are more gains than losses within each textbook (here in Section 2.2.3), and the lower prevalence of lexical suprafixation in other Cantonese varieties (Section 2.3 below), strongly suggests that, on average, there has been an increase in the number of suprafixed words in Standard Cantonese.

(As discussed in Yiu (2010, pp. 24–25), prior to Ball (1883), the earliest publications with Romanised texts of Cantonese which displayed suprafixation were Dennys (1874) and Eitel (1877). Earlier publications with Romanised texts were either not tone-marked or documented only literary language, where suprafixation either did not occur or was not recorded.)

2.3. Other Cantonese Varieties

(See Section 1.1 above for a definition of ‘Cantonese’.) There is another perspective that enables us to look into the development of the lexical suprafixes in Standard Cantonese, albeit somewhat indirectly: the behaviour of the lexical suprafixes in other Cantonese varieties. There are many enclaves of Cantonese speakers outside the Canton area (e.g., de Sousa 2022). Except the Wúzhōu and Hèzhōu Cantonese enclaves, the Cantonese enclaves started developing after the First Opium War (1839–1842). The city centre of Hong Kong became Cantonese-dominant relatively early, in the last decades of the 19th century (e.g., Ball 1883). Most other cities became Cantonese-dominant later. For instance, Cantonese migrants started arriving in Tsamkong 湛江 (Zhànjiāng/Fort-Bayard) after the start of French colonisation in 1898.8 Macau (Zhān Bóhuì et al. 2002, pp. 201–2) and the city centre of Nánníng (Lì Mínghuì 2008) did not become Cantonese-dominant until the 1940s. Canton and Hong Kong Cantonese are at the forefront of the development of lexical suprafixes; there are no reports of Cantonese varieties that have more suprafixed words than Canton and Hong Kong Cantonese. The following are quotes commenting on the lower number of suprafixed words in the Cantonese of Macau, Nánníng, and Běihǎi.

[In Macau Cantonese, in contrast to Hong Kong Cantonese:]

人们戏称，说 [ɔu³³ mun¹¹ jan¹¹] 者是真正的澳门人，说 [ɔu³³ mun³⁵ jan¹¹] 者是假冒的澳门人。不过这种差别正在缩小。

‘People joke that, one who says ou3 mun4 jɐn4 is a real Macau person, one who says ou3 mun4-2 jɐn4 is a fake Macau person. But this type of difference is narrowing.’

(Wong Yee 2007, p. 91)

[In Nánníng Cantonese, the high-rising (tone 2) suprafix:]

使用远无广州、香港白话那么普遍，基本可以穷尽。

‘The usage is far less common than in Canton and Hong Kong Cantonese, [examples of suprafixed words] can basically be exhaustively listed.’

(Lín Yì and Qín Fèngyú 2008, p. 90)

[In Běihǎi Cantonese, the high-level (tone 1) and high-rising (tone 2) suprafixes:]

例子很少，在我们的词汇表里，我们一共记录了七个有变调的词。

‘There are very few examples, in our vocabulary list [of 2179 items], we have recorded seven items with a suprafix in total.’

(Chén Xiǎojǐn and Chén Tāo 2005, p. 26)

The following is a small sample of words in three varieties of Cantonese that the author has some familiarity with: (a) Nánníng Cantonese (exposure in tandem with fieldwork on Nánníng Pínghuà; data here are from Lín Yì and Qín Fèngyú 2008); (b) ‘Older Macau Cantonese’ (the speech of the author’s parents, born around the wars in the 1940s); and (c) Hong Kong Cantonese (exposure since birth). Traditionally, Nánníng and Macau Cantonese have far fewer suprafixed words than Hong Kong Cantonese. (Although Macau Cantonese is quickly ‘catching up’ due to the strong influence of nearby Hong Kong.) Most words in the ‘Older Macau Cantonese’ list below are still commonly used amongst older speakers in Macau. In comparison, the Macau-type of suprafix-less-ness is very rare in Hong Kong and Canton. The data below demonstrate some suprafixed words (nouns) in Hong Kong Cantonese, as well as their equivalents in Nánníng and Older Macau Cantonese. Nánníng and/or Macau have the same nouns in citation tone, except (10j) hɐt7-ji4-1 乞兒 ‘beggar’.

(10)	nouns:	Nánníng Canto.	Older Macau Canto.	Hong Kong Canto.
a.	繩 ‘string’	ʃeŋ21 (4)	sɪŋ21 (4)	sɪŋ25 (4-2)
b.	鉗 ‘pliers’	kʰɛm21 (4)	kʰim21 (4)	kʰim25 (4-2)
c.	碟 ‘plate’	tip2 (9)	tip2 (9)	tip25 (9-2)
d.	澳門 ‘Macau’	u33 mun21 (4)	ou33 mun21~25 (4 ~ 2)	ou33 mun25 (4-2)
e.	名 ‘name’	mɛŋ21 (4)	mɛŋ21~25 (4 ~ 2)	mɛŋ25 (4-2)
f.	帽 ‘hat’	mu22 (6)	mou22 (6)	mou25 (6-2)
g.	枱 ‘table’	tʰɔi21 (4)	tʰɔi21~25 (4 ~ 2)	tʰɔi25 (4-2)
h.	枱布 ‘table cloth’	tʰɔi21 (4) pu33	tʰɔi21 (4) pou33	tʰɔi25 (4-2) pou33
i.	-話 ‘-speech’	-wa22 (6)	-wa25 (6-2)	-wa25 (6-2)
j.	乞兒 ‘beggar’	hɐt5-ji55 (4-1)	hat5-ji55 (4-1)	hɐt5-ji55 (4-1)
k.	乞兒仔 ‘little beggar’	?	hat5-ji21 (4) -tsɐi25 (beg-dim-dim)	hɐt5-ji55 (4-1) -tsɐi25
l.	眼鏡 ‘eye glasses’	ŋɛn24 kɛŋ33 (3)	ŋan13 kɛŋ33 (3)	ŋan13 kɛŋ25 (3-2)
m.	袋 ‘bag’	tɔi22 (6)	tɔi22 (6)	tɔi25 (6-2)
n.	-文 ‘-script/-language’	-mɐn21 (4)	-mɐn21 (~25) (4 (~2))	-mɐn(21~) 25 ((4~) 2)

There are two probable reasons why non-standard Cantonese varieties have fewer suprafixed words than Standard Cantonese. Firstly, when comparing with the late-nineteenth century data (Section 2.2.3), in many cases, the non-standard Cantonese varieties have preserved the older citation-tone version of the words used in Canton (or the less-urban parts of the Canton area where many migrants came from). Secondly, the Sinitic languages originally spoken in these new locations, with far fewer cases of lexical suprafixes, would have helped with the preservation of the citation-tone version of the Cantonese words to some degree. Perhaps there were also cases of superfixed words reverting back to their citation-tone pronunciations in these newer Cantonese varieties, due to many people applying sound correspondence rules from their first language to Cantonese. Comparing the words in (10) with those in Ball (1888) (see Section 2.2.3 above), (a) to (d)—繩 ‘string’, 鉗 ‘pliers’, 碟 ‘plate’, and 澳門 ‘Macau’—did not have a suprafix in Ball (1888), while (e) to (j)—名 ‘name’, 帽 ‘hat’, 枱 ‘table’, 枱布 ‘table cloth’, -話 ‘-language’, and 乞兒 ‘beggar’—did. (Examples (k) to (n)—乞兒仔 ‘little beggar’, 眼鏡 ‘eye glasses’, 袋 ‘bag’, and -文 ‘-script’—are not found in Ball (1888).) Given that Nánníng and Macau did not become Cantonese-dominant until the 1940s (see above), this suggests that many words that have a suprafix nowadays in Canton (or the less-urban parts of the Canton area, where many of the migrants came from) did not have a suprafix as late as the 1940s or slightly prior. (On the other hand, Hong Kong’s city centre became Cantonesised more than 50 years prior and always had stronger commercial contacts with Canton. This caused Canton and Hong Kong Cantonese to remain linguistically very close to each other, including having a very similar range of suprafixed words.)

That the lexical suprafixes in Hong Kong Cantonese are older than those in Older Macau Cantonese is also reflected in the morphological level where they are applied. In Hong Kong Cantonese, the lexical suprafixes are often applied at the morpheme level. For example (‘>2’ here indicates the step where the suprafixation of tone 2 is applied),

• Hong Kong Cantonese

(11)	i.	tʰɔi4>2	table 枱
		‘table’

(12)	i.	tʰɔi4>2	table 枱
	ii.	mat8 tʰɔi2	wipe table 抺枱
	iii.	mat8 tʰɔi2 pou3	wipe table cloth 抺枱布
		‘table-cleaning cloth’

(The Standard Cantonese in Ball (1888) was similar: tʰɔi4-2* ‘table’ (p. xlii), tʰɔi4-2* pou3 ‘table cloth’ (p. 17). But ‘on the table’ could be tʰɔi4 min2* or tʰɔi4-2* min2* (p. 11).) On the other hand, in Older Macau Cantonese, the lexical suprafix is often applied at a later step, typically at the word level. In example (13), the tone 2 suprafix can be applied to the word tʰɔi4 ‘table’ optionally. (The verb phrase ‘wipe table’ would be mat8 tʰɔi4 or mat8 tʰɔi4-2.) In example (14), in the noun compound mat8 tʰɔi4 pou3 ‘table-cleaning cloth’, the lexical suprafix did not have a chance of being applied to tʰɔi4 ‘table’ at the morpheme level, and at the word level, the head pou3 ‘cloth’ is not a morpheme that can take a lexical suprafix.

• Older Macau Cantonese

(13)	i.	tʰɔi4 ~ tʰɔi4>2	table 枱
		‘table’

(14)	i.	tʰɔi4	table 枱
	ii.	mat8 tʰɔi4	wipe table 抺枱
	iii.	mat8 tʰɔi4 pou3 (*>2)	wipe table cloth 抺枱布
		‘table-cleaning cloth’

Similarly, in Hong Kong Cantonese, ‘Macau’ is usually ou3 mun4-2 澳門, and noun compounds containing ou3 mun4-2 ‘Macau’ also have a lexical suprafix on the morpheme mun4-2 (unless they are imitating a Macau accent), e.g., ou3 mun4-2 tsɪŋ3 fu2 澳門政府 ‘Macau government’, ou3 mun4-2 jɐt9 bou3 澳門日報 ‘Macao Daily News’, ou3 mun4-2 tai6 hɔk9 澳門大學 ‘Universidade de Macau’, and pʰou4-sʊk9 ou3 mun4-2 葡屬澳門 ‘Portuguese Macau’. On the other hand, in Older Macau Cantonese, ‘Macau’ is ou3 mun4 or ou3 mun4-2 澳門, but the lexical suprafix cannot apply to ou3 mun4 ‘Macau’ in compounds, unless it is at the end of a compound: ou3 mun4 tsɪŋ3 fu2 澳門政府 ‘Macau government’, ou3 mun4 jɐt9 bou3 澳門日報 ‘Macao Daily News’, ou3 mun4 tai6 hɔk9 澳門大學 ‘Universidade de Macau’, and pʰou4-sʊk9 ou3 mun4(-2) 葡屬澳門 ‘Portuguese Macau’. Examples (10j) and (10k) above demonstrate the same point, but with a tone 1 suprafix: in Hong Kong Cantonese—hɐt1 -ji4-1 乞兒 (beg -dim\supr) ‘beggar’ and hɐt1 -ji4-1 -tsɐi2 乞兒仔 (beg -dim\supr -dim) ‘little beggar’; in Older Macau Cantonese—hat1 -ji4-1 乞兒 (beg -dim\supr) ‘beggar’ and hat1 -ji4 -tsɐi2 乞兒仔 (beg -dim -dim) ‘little beggar’.

Looking back at Wong Yee (2007)’s quote above, which says that one who says ou3mun4 jɐn4 (Macau person) is a real Macau person and one who says ou3mun4-2 jɐn4 (Macau\supr person) is a ‘fake’ Macau person, this indeed used to be the case. Nevertheless, the Hong Kong-type suprafix patterns are becoming more and more common in Macau these days. For instance, my impressionistic observation as a native speaker is that ou3mun4-2 jɐn4 is more common than ou3mun4 jɐn4 in Macau these days.

Lastly, one interesting fact is that there are a few words that are suprafixed in Nánníng Cantonese but not in Standard Cantonese. In other words, the non-standard Cantonese varieties do not always follow the lead of Standard Cantonese in terms of suprafixation. Examples of this are Nánníng Cantonese pu³⁵ ly²⁴⁻⁵⁵ 堡壘 ‘fortress’ and ʃaŋ³³ ŋaŋ²²⁻³⁵ 生硬 ‘stiff, unnatural’ (Lín Yì and Qín Fèngyú (2008, p. 90); the tones in 壘 ly⁵⁵, 生 ʃaŋ³³ and 硬 ŋaŋ³⁵ are probably influenced by the cognates in Old Nánníng Mandarin, a type of Southwestern Mandarin), versus Standard Cantonese pou²⁵ lɵy¹³ 堡壘 ‘fortress’ and saŋ⁵⁵ ŋaŋ²² 生硬 ‘still, unnatural’. These examples further exemplify the fact that the suprafixes in some words do not have a very long history (Section 2.2.3, Section 2.3 and Section 2.4), given the short time of divergence between Canton Cantonese and Nánníng Cantonese.

2.4. Interim Summary

In Section 2.1, some properties of the lexical suprafixes in modern Standard Cantonese—the tone 2 suprafix [²⁵] and the rarer tone 1 suprafix [⁵⁵]—are discussed. There is no difference between the suprafix tones 2/1 and the ordinary tones 2/1, respectively. In Section 2.2.1, the distinction between ordinary tone 1 [⁵³] and suprafix tone 1* [⁵⁵] in earlier Cantonese is discussed. In Section 2.2.2, the earlier distinction between ordinary tone 2 [³⁵] and suprafix tone 2* ([²⁵] in the 1940s and [³¹⁵] in the 1900s) is discussed. In Section 2.2.3, we look at the use of lexical suprafixes in some late-nineteenth century Cantonese textbooks. Many suprafixed words back then have remained suprafixed in modern Standard Cantonese. Some non-suprafixed words have since gained a suprafix, and some suprafixed words reverted back to their citation-tone pronunciation in modern Cantonese. Section 2.3 discusses how non-standard Cantonese varieties like Nánníng Cantonese and Older Macau Cantonese have far fewer suprafixed words than the Standard Cantonese in Canton and Hong Kong. The facts demonstrated in Section 2.2 and Section 2.3 strongly suggest that Canton and Hong Kong Cantonese had considerably fewer suprafixed words one century ago.

The Cantonese lexical suprafixes probably came from diminutives (Section 3). Maybe the Cantonese lexical suprafixes developed independently within Cantonese, or maybe they have the same origin as the diminutives in some other Sinitic varieties—for instance, the rhotacisation (ér-huà 兒化) in Northern Mandarin. It is beyond the scope of this article to address this question properly. Obviously, this is an important research topic. If one were to pursue this line of inquiry (e.g., Simmons, forthcoming), one needs to compare the range of suprafixed words in Cantonese with the range of suprafixed and/or diminutive words in other Sinitic varieties. However, as we have seen above, the suprafixes in many words in modern Cantonese do not have a long history, and words can gain or lose suprafixes rather quickly. If suitable historical material is available, it would be more useful to look at the range of suprafixed/diminutive words in earlier periods of a speech variety. It would also be useful to look into the range of words in which the suprafixation/diminutivisation is more stable and compare these across Sinitic varieties. (This would be particularly difficult for suprafixation, as suprafixation is rarely reflected in Chinese script, and not many Sinitic varieties have tone-marked written records that are ‘pre-modern’.)

3. Lexical Suprafixes and Diminutives in Yuè and Pínghuà Dialects

3.1. Yuè and Pínghuà Diminutives

We note in Section 2.2.2 that tone 2* in Cantonese was a long dipping tone in the late-nineteenth century/early-twentieth century. The citation tone of a lexical suprafix tone 2* syllable was most usually a low tone (tone 4 [²¹], 5 [¹³], 6 [²²], or 9 [²]). With tone 1*, we know that at least the suprafixed version of tone 7 [⁵] had to be long (see endnote 6). The extra length and high pitch at the end of a suprafixed syllable suggest that such syllables were originally two syllables, made up of a host syllable in citation tone followed by a suffix syllable in high tone. The segments of the suffix disappeared, leaving a high tone at the end of a compensatory-lengthened host syllable. Given the diminutive-type meaning that Cantonese lexical suprafixes sometimes have, most theories hypothesise that the high-toned suffix was a diminutive suffix (e.g., Chen Zhōngmǐn 1999; Chén 2002), with most hypothesising that this was a high-tone version of Middle Chinese ɲe^A 兒 (Cantonese ji4, Mandarin ér, e.g., Whitaker 1955–1956; Chao Yuen Ren 1959; Mài Yún 1995; Wang Pen-Ying 1995; Zhōu Liètíng 2002, pp. 154–71; Kao Wan-yu 2007; Kwok Bit-Chee 2016). The morpheme 兒 (Old Chinese *ŋe, Middle Chinese ɲe^A)9 originally meant ‘male child’ and then later broadened to ‘child’. Looking at the reflexes in Yuè and Pínghuà dialects, the lexical word 兒 can be reconstructed as *ɲi^A. In the discussions below, when the tone is immaterial, this morpheme is referred to as ‘*ɲi 兒’. When 兒 in citation tone (tone A) is referred to, this is rendered ‘*ɲi^A 兒’. When it has a high-tone suprafix (usually high level or high falling), this is rendered ‘*ɲi^H 兒’.

The various versions of the *ɲi 兒 theory hinge on the behaviour of the diminutives in the western Yuè dialects. Standard Cantonese is spoken close to the eastern edge of the Yuè-speaking territory. In Standard Cantonese, there are many cases of lexical suprafixation, and the default diminutive suffix is *tsVj 仔 (see below). The old *ɲi^H 兒 suffix is only found in some fossilised expressions. As one heads towards the west, starting from approximately Guǎngxī and southwestern Guǎngdōng (Màomíng and Zhànjiāng Prefectures), *ɲi 兒 becomes the dominant diminutive suffix. Below, we will base our discussions on Kwok Bit-Chee (2016)’s cline of grammaticalisation from *ɲi^H 兒 to the Cantonese lexical suprafixes. (Obviously I do not cover everything mentioned in Kwok Bit-Chee (2016); the reader is encouraged to read Kwok Bit-Chee (2016). Kwok Bit-Chee (2016) surveys nineteen varieties of Yuè, and their *ɲi^H 兒 suffixes always occur with a high-tone suprafix. I expanded the search westward and found some Yuè dialects further to the west, and many Pínghuà dialects, with an unsuprafixed *ɲi^A 兒 diminutive suffix (in their citation tone, Lower tone A). The Pínghuà dialects, as defined by the Language Atlas of China (Wurm and Li 1987; Zhāng Zhènxīng et al. 2012), are basically western continuations of the Pearl-River Yuè dialect chain. Ignoring the Cantonese enclaves in Guǎngxī and western Guǎngdōng (which are relatively recent transplants from the Canton area), as one travels west from the heart of the Pearl River Delta, Cantonese/Guǎngfǔ Yuè gradually becomes Gōulòu Yuè in Guǎngdōng and Guǎngxī, and Gōulòu Yuè in Guǎngxī gradually becomes Southern Pínghuà. (On the other hand, ‘Northern Pínghuà’ presents other complications, but this is not something we need to elaborate here. See, e.g., (de Sousa, forthcoming). In this article, only data from the core ‘undisputed’ Northern Pínghuà dialects are presented.) My proposal of a small extension to Kwok Bit-Chee (2016)’s theory is that these unsuprafixed *ɲi^A 兒 diminutives represent one step prior to Kwok Bit-Chee (2016)’s grammaticalisation cline, which begins with the suprafixed *ɲi^H 兒 diminutives.

There are three diminutive suffixes that are commonly found amongst Yuè dialects (e.g., Wang Pen-Ying 1995; Kao Wan-yu 2007) and Pínghuà dialects. They all grammaticalised from words that meant something like ‘child’. ‘Child’ is a very common source of diminutives (e.g., Jurafsky 1996; Heine and Kuteva 2002, pp. 65–67). The first suffix is the *ɲi^A/H 兒 ‘(male) child’ suffix introduced above. This diminutive suffix is only found in very few fossilised expressions in Standard Cantonese in the east, but *ɲi^A/H 兒 diminutives are still very commonly used amongst the western Yuè and Pínghuà dialects in the west. The second suffix is 仔 ‘child’ (maybe etymologically the same as 崽 ‘young animal’, Middle Chinese ṣɛ(y)^A). Looking at the reflexes in Yuè dialects, the proto-form of 仔 in Yuè would be something like *tsaj^B (e.g., Cantonese tsɐi²⁵) or *tsoj^B (e.g., Toishanese tᵘɔi⁵⁵). The *tsVj^B 仔 suffix is a newer eastern innovation, and it is now the dominant diminutive suffix in the east (e.g., Zhān Bóhuì et al. 2002, pp. 183–84). Nonetheless, due to the strong influence of Cantonese, the *tsVj^B 仔 suffix is also borrowed into a great number of Yuè and Pínghuà dialects in the west (coexisting with the older *ɲi^A/H 兒 suffix). The third suffix is 子 ‘child’ (Old Chinese *tsəʔ, Middle Chinese tsi^B). As a suffix, *tsi^B 子 is rarely used in Standard Cantonese. There are Yuè and Pínghuà dialects where the *tsi^B 子 suffix is commonly used, e.g., Yángjiāng Yuè (Kao Wan-yu 2007, p. 238; Zhān Bóhuì et al. 2002, p. 210), Běihǎi Cantonese (Chén Xiǎojǐn and Chén Tāo 2005, pp. 390–91), and Nánníng Pínghuà (de Sousa, forthcoming). However, there is no obvious geographical pattern for where the *tsi^B 子 suffix is more prevalent amongst Yuè and Pínghuà dialects. (Except in some cases where the *tsi^B 子 suffix is an obvious Hakka influence, e.g., the Yuè dialects around Wúchuān and Huàzhōu (Kwok Bit-Chee 2016, p. 285).) The discussions below mostly revolve around *ɲi 兒, while *tsVj^B 仔 is also discussed to a small degree. The *tsi^B 子 suffix is only discussed briefly towards the end of Section 3.3.

3.2. The *ɲi^A and *ɲi^H Diminutives

In the west, many Pínghuà dialects have a diminutive suffix *ɲi^A 兒 in its citation tone, Lower tone A (yáng píng, i.e., Cantonese tone 4, normal cases of Mandarin tone 2). This also occurs in some extreme-western Yuè dialects. (Middle Chinese tones A, B, C and D have usually developed an ‘Upper’ (yīn) and a ‘Lower’ (yáng) variant, which may or may not become phonemicised in the modern Sinitic languages. Usually the Upper tones occur with syllables with a voiceless onset in Middle Chinese, and Lower tones occur with a voiced onset in Middle Chinese). The norm amongst Pínghuà and Yuè dialects is that all their Lower tones are lower in pitch than their Upper counterparts, i.e., the tone in the *ɲi^A 兒 suffix is relatively low in pitch.

Table 4. Examples of Pínghuà and Yuè dialects with a diminutive suffix *ɲi^A 兒.

Dialect Group	Locality/Name of Dialect	Form of Suffix	Tone Category
Pínghuà, Northern	臨桂五通 Línguì Wǔtōng	兒 -ɲi31	Lower tone A
Pínghuà, Northern	永福桃城 Yǒngfú Táochéng	兒 -ɲi23	Lower tone A
Yuè, Gōulòu	平樂本地話 Pínglè ‘Běndìhuà’	兒 -ɲi31	Lower tone A
Yuè, Gōulòu	鍾山本地話 Zhōngshān ‘Běndìhuà’	兒 -ɲi23	Lower tone A
Pínghuà, Southern	三江六甲話Sānjiāng ‘Liùjiǎhuà’	兒 -ɲi31	Lower tone A
Pínghuà, Southern	宜州得勝 Yízhōu Déshèng	兒 -ŋi31	Lower tone A
Pínghuà, Southern	百色那畢 Bǎisè Nàbì	兒 -ɲi31	Lower tone A
Pínghuà, Southern	崇左江州 Chóngzuǒ Jiāngzhōu	兒 -ɲi32	Lower tone A
Pínghuà, Southern	崇左新和 Chóngzuǒ Xīnhé	兒 -ɲi21	Lower tone A
Yuè, Qīnlián	合浦(廉州) Hépǔ (Liánzhōu)	兒 -ŋi44	Lower tone A

shows some Pínghuà and Yuè dialects with a citation-tone *ɲi^A 兒 diminutive suffix.

Some comments are needed for the Yuè varieties. As argued convincingly by Qín Yuǎnxióng (2019), the Běndìhuà 本地話 ‘local language’ varieties in places like Pínglè and Zhōngshān in Guǎngxī are Gōulòu Yuè varieties that are recently Mandarinised, and not Northern Pínghuà as per the Language Atlas of China. In Hépǔ (Liánzhōu), the Lower tone A [⁴⁴] sounds rather high, but it is already lower than the Upper tone A [⁴⁵].

The localities listed in Table 4 above are relatively peripheral (north, west, and south) in Guǎngxī. In mid-southern Guǎngxī, there are speech varieties like Nánníng Pínghuà (spoken in the suburbs of Nánníng, the capital of Guǎngxī) and Héngzhōu Cantonese (spoken in the city centre of Héngzhōu) where one finds a *ɲi^A 兒 suffix in its citation tone, Lower tone A, and also a *ɲi^H 兒 suffix in a high-tone suprafix. These are shown in

Table 5. Examples of Pínghuà and Yuè dialects with diminutive suffixes *ɲi^A 兒 and *ɲi^H 兒.

Dialect Group	Locality/Lect Name	Form of Suffix	Tone Category
Pínghuà, Southern	南寧位子碌 Nánníng Wèizǐlù	兒 -ɲi21 兒 -ɲi21⁻53	Lower tone A Upper tone A
Yuè, Yōngxún	橫洲白話 Héngzhōu Cantonese	兒 -ɲi232 兒 -ɲi232⁻55	Lower tone A Upper tone A

.

Why would a diminutive suffix develop a high-tone pronunciation? One observation that is relevant is that diminutive affixes are especially susceptible to developing a high-tone pronunciation (e.g., Nichols 1971; Zhū Xiǎonóng 2004). This is perhaps due to the iconicity link between high pitch and smallness (e.g., Ultan 1978; Perlman and Cain 2014). One related point of discussion is that suprafixes are normally understood as having developed from the tone of an affix that has its segments deleted (e.g., Chappell 2023). On the other hand, at least in a speech variety with both -ɲi^A and -ɲi^H diminutives, the -ɲi^H suffix can be analysed as -ɲi^A plus a suprafix H. There is no evidence that this suprafix H came from another suffix. This would be a non-typical case of a suprafix not having developed from a segmental affix.

In Nánníng Wèizǐlù Pínghuà, whether a diminutive suffix can be used or not, and which diminutive suffix is used with what noun, are not quite predictable. Some nouns are compatible with more than one diminutive suffix (free variation). The citation tone -ɲi²¹ 兒 is more commonly used with animal nouns, e.g., kɐi⁵³ -ɲi²¹ 雞兒 (fowl -dim) ‘chick’ and jɐŋ²¹ -ɲi²¹ 羊兒 (goat -dim) ‘kid’, but some inanimate nouns can also take -ɲi²¹ 兒, e.g., tʃʰɛ⁵³ -ɲi²¹ 車兒 (car -dim) ‘little car’ and tau⁵³ -ɲi²¹ 刀兒 (knife -dim) ‘little knife’. The suprafixed -ɲi²¹⁻⁵³ 兒 is primarily used with inanimate nouns, e.g., tʃʰɛ⁵³ -ɲi²¹⁻⁵³ 車兒 (car -dim\supr) ‘little car’, ʊk³ -ɲi²¹⁻⁵³ 屋兒 (house -dim\supr) ‘little house’, and tʃʰɛŋ⁵³ -ɲi⁵³ 鐺兒 (wok -dim\supr) ‘little wok’, but -ɲi²¹⁻⁵³ 兒 can also be used with some animal nouns, e.g., kɐu³³ -ɲi²¹⁻⁵³ 狗兒 (dog -dim\supr) ‘puppy’ (although kɐu³³ -ɲi²¹ 狗兒 (dog -dim) ‘puppy’ is more common). There is also a newer diminutive suffix, -tʃai³³ 仔, inspired by Nánníng Cantonese -tʃɐi³⁵ 仔, e.g., Nánníng Wèizǐlù Pínghuà kɐi⁵³ -tʃai³³ 雞仔 (fowl -dim) ‘chick’, tʃəi⁵³ -tʃai³³ 豬仔 (pig -dim) ‘piglet’, tau⁵³ -tʃai³³ 刀仔 (knife -dim) ‘little knife’, and un³³ -tʃai³³ 碗仔 (bowl -dim) ‘little bowl’.

Nánníng Wèizǐlù Pínghuà also has some cases of suprafixes being applied directly on roots. The suprafix is usually a high-falling tone (Upper tone A, like that in the -ɲi²¹⁻⁵³ 兒 suffix), e.g., mɐn²¹⁻⁵³ 文 ‘Yuán’ (cf. Cantonese mɐn²¹⁻⁵⁵ 文 ‘Yuán/dollar’) and lau¹³ nɛŋ²¹⁻⁵³ 老娘 (old lady\supr) ‘mother’. There are also rare cases of a high-level suprafix (unaspirated Upper tone C), e.g., həi²¹ nɛŋ²¹⁻⁵⁵ 姨娘 (aunt lady\supr) ‘mother’s younger sister’ and nai¹³⁻⁵⁵ nai¹³⁻⁵⁵ 奶奶 ‘paternal grandmother’ (inspired by Old Nánníng Mandarin nai³⁵ nai³⁵ 奶奶 ‘paternal grandmother’), and a high-rising suprafix (aspirated Upper tone C) in Cantonese loanwords (phonetically the same as the Cantonese Upper tone B/tone 2 suprafix), e.g., jɛ²¹ jɛ²¹⁻³⁵ 爺爺 ‘paternal grandfather’. (See de Sousa, forthcoming, for further discussions on the suprafixes and diminutive suffixes in Nánníng Wèizǐlù Pínghuà.)

East of Nánníng are Pínghuà and Yuè dialects that have a suprafixed *ɲi^H 兒 suffix but no citation tone *ɲi^A 兒 suffix. (The citation-tone pronunciation of *ɲi^A 兒 is still found in literary lexical words in these dialects, e.g., Cantonese ji²¹ tʰʊŋ²¹ 兒童 ‘juvenile’.) Also belonging to this type is Hèzhōu Guìlǐng Yuè in the northeastern corner of Guǎngxī (near the tripoint with Guǎngdōng and Húnán). Examples of these are shown in

Table 6. Examples of Pínghuà and Yuè dialects with a diminutive suffix *ɲi^H 兒.

Dialect Group	Locality/Lect Name	Form of Suffix	Tone Category
Yuè, Gōulòu	賀州桂嶺 Hèzhōu Guìlǐng	兒 -ŋi52	Upper tone A
Pínghuà, Southern	賓陽新橋 Bīnyáng Xīnqiáo	兒 -ŋi55	Upper tone C
Pínghuà, Southern	橫洲嶺鷯 Héngzhōu Lǐngliáo	兒 -ji24	Upper tone A
Yuè, Gōulòu	貴港 Guìgǎng	兒 -ŋi55 ~ -ɲi55	Upper tone A
Yuè, Yōngxún	桂平白話 Guìpíng Cantonese	兒 -ŋi55	Upper tone A

. This step is where Bit-Chee (2016)’s cline of grammaticalisation begins.

According to the *ɲi 兒 theory, the idea of having a high-tone suprafix has to develop in the *ɲi^H 兒 diminutive suffix first. After the step of having a *ɲi^H 兒 diminutive suffix, there are two main types of development: ‘nasal rise’ and applying a suprafix directly on the root. Nasal rise involves cases where the segments of *ɲi^H 兒 are eroded to only a nasal segment or a nasal feature, which is then attached to the host syllable, and the host syllable has some sort of rising tone. Examples of nasal rise can be found in southwestern Guǎngdōng and neighbouring southeastern Guǎngxī. For instance, in Màomíng 茂名 Yuè (Guǎngdōng), there are some nouns that freely alternate between a *ɲi^H 兒 suffix and a nasal rise, e.g., ʃiak² -ɲi⁵⁵ 勺兒 ~ ʃiaŋ↗︎ ‘little spoon’ (Shào Huìjūn and Gān Yū’ēn 2007, p. 21). (↗ signifies an extra-high rise; it rises to above the pitch of a normal high-level tone, often falsetto.) The case in nearby Xìnyí 信宜 Yuè is oft-quoted: a. zero-coda syllables add -n and an ↗ suprafix; b. plosive-coda syllables add [+nasal] to the coda and an ↗ suprafix; and c. nasal-coda syllables and glide-coda syllables (-j -w, here rendered -i -u as per the norm in Chinese linguistics) add only an ↗ suprafix (c. is not nasal rising).

• Xìnyí Yuè (Shào Huìjūn and Gān Yū’ēn 2007, p. 21)

(15)		citation form	diminutive
	ai.	tʃʰœ23	tʃʰœn↗	坐 ‘sit’
	aii.	tʃy53	tʃyn↗	豬 ‘pig’
	aiii.	ku53	kun↗	姑 ‘aunt’
	bi.	ap3	am↗	鴨 ‘duck’
	bii.	fut3	fun↗	闊 ‘wide’
	biii.	kiak3	kiaŋ↗	脚 ‘foot/leg’
	ci.	pʰiɛn33	pʰiɛn↗	片 ‘slice’
	cii.	pui53	pui↗	杯 ‘cup’
	ciii.	tʰɐu23	tʰɐu↗	頭 ‘head’

Xìnyí Yuè also has other diminutives—for instance, a -ɲi⁵³ 兒 suffix for animal newborns, e.g., tʃy⁵³ -ɲi⁵³ 豬兒 ‘newborn piglet’. (There is also tʃy⁵³ -ɲin↗ (pig -dim\dim) ‘tiny weeny newborn piglet’.)

A preceding step can be shown in some words in nearby Huàzhōu Liángguāng 化州良光 Yuè. There, the *ɲi^H 兒 suffix is usually still a separate syllable, but in the form of a syllabic nasal -ŋ²³⁻⁵³. The coda of the preceding syllable is then nasal-assimilated, and the tone of that syllable becomes high rising or high level, e.g., dit² -ŋ²³⁻⁵³ → diŋ³⁵ -ŋ²³⁻⁵³ 碟兒 ‘little plate’ (Kwok Bit-Chee 2016, p. 300; see also Lǐ Jiàn 2014). Presumably, Xìnyí Yuè experienced something similar, and the suffix was later deleted.

The rules of nasal rising are slightly different in the various Yuè varieties in this southwestern Guǎngdōng/southeastern Guǎngxī region. For instance, the rules in nearby Róng County 容縣 (Róngxiàn; Guǎngxī) Yuè is similar to those in Xìnyí Yuè, except that a nasal segment is not added to vowel-ending host syllables (Zhōu Zǔyáo 1987). In Yùlín 玉林 (Guǎngxī) Yuè, the situation with their nasal rising and suprafixes (and also tone sandhi) are rather complicated (see Zhōu Liètíng 2002, pp. 148–54).

It seems that the Yuè dialects that have nasal rising also have cases where they apply suprafixes directly on the root (e.g., (15c) above for Xìnyí Yuè). However, having suprafixes on roots does not infer having nasal rising. To the east of the nasal-rising area, the *ɲi^H 兒 diminutive is only found in some fossilised expressions,10 and the *tsVj^B 仔 diminutive dominates.

Suprafixes can be applied on lexical roots in the vast majority of Yuè dialects, although in many of these dialects, suprafixes are only used sparingly. For instance, to the east of the nasal-rising area, Zhān Bóhuì et al. (2002, pp. 179–80, 98) mention that lexical suprafixes are not used very often in the following Yuè dialects in mid-western and central Guǎngdōng: Luódìng 羅定, Yúnfú 雲浮, Xīnxīng 新興, Guǎngníng 廣寧, Sìhuì 四會, and Zhōngshān 中山. Standard Cantonese in the east peaks in the prevalence of lexical suprafixation.

There are some scholars, e.g., Wáng Fútáng (2005, p. 177) and Chén Xiǎomíng (2007), who argue that the grammaticalisation pathway is *ɲi^H 兒 → nasal rising → suprafix directly on the root. On the other hand, Kwok Bit-Chee (2016, pp. 305–6) argues that the two latter steps are independent of each other, i.e., X-ɲi^H can develop into Xn↗, and X-ɲi^H can develop directly into X↗ or X^H. Nasal rising cannot be a prerequisite of suprafixing directly on the root in Yuè dialects because the regular deletion of a nasal coda, or denasalisation of another sort, is exceedingly rare amongst Yuè dialects. In other words, if Xn↗ has developed, in the vast majority of Yuè dialects, there are no general sound change rules that delete the nasal coda in a Xn↗ syllable to become X↗ or denasalise a Xn↗ syllable to become Xt↗.

Amongst the nineteen Yuè varieties surveyed by Kwok Bit-Chee (2016), lexical suprafixation is found in nearly all of them (albeit lexical suprafixation is rare in some varieties). From a phonetic point of view, the lexical suprafixes can be divided into two types. (Some dialects have one, and some, like Cantonese, have both.) One type is the high-level or high-falling suprafixes. It is the same high-level or high-falling tone that the suprafixed *ɲi^H 兒 diminutive suffix has, and no Yuè dialect is found to have both a high-level and a high-falling suprafix (see also Section 3.3). The segments of the diminutive suffix were deleted, and its tone replaces that of the preceding host syllable. Another type is the high-rising suprafix. The situation in Toishanese (Táishān Yuè) probably shows how the substitutive high-rising suprafixes in the other Yuè dialects were developed. Toishanese has additive suprafixes: the citation tone of the host syllable is pronounced, and then the tone rises to a high end point. The high pitch at the end is the tone of what used to be a diminutive suffix. Toishanese has the rise and dip-rise suprafixes of [³⁵], [²⁵], [¹¹⁵] and [³¹⁵]; these are the suprafixed versions of the four tonemes [³³], [²²], [¹¹] and [³¹] (the remaining fifth toneme, [⁵⁵], cannot take a high suprafix; Kwok Bit-Chee 2016, p. 289, quoting Gān Yū’ēn 2010). Ch’an (1900) also describes similar additive suprafixes in earlier Cantonese. Presumably, other Yuè dialects with high-rising suprafixes have gone through the same path, and they have later simplified the rising suprafixes into one or two rising contours.

3.3. Possible Origins of the Cantonese Lexical Suprafixes

Amongst the theories on the origin of the Cantonese lexical suprafixes,11 the most-common theory is that they developed from a *ɲi 兒 diminutive suffix (e.g., Whitaker 1955–1956; Chao Yuen Ren 1959; Mài Yún 1995; Wang Pen-Ying 1995; Zhōu Liètíng 2002, pp. 154–71; Kao Wan-yu 2007; Kwok Bit-Chee 2016). It seems that none of the studies on the Yuè diminutives and suprafixes extended their investigation to include the Pínghuà dialects. Here, the Pínghuà dialects are included, as Pínghuà, Gōulòu Yuè, and Guǎngfǔ Yuè (including Standard Cantonese) lie on the same dialect continuum. In fact, the majority opinion amongst Chinese dialectologists these days is that at least Southern Pínghuà is part of the Yuè dialect group (e.g., Qín Yuǎnxióng 2000; Xiè Jiànyóu 2007; Hóu Xīngquán 2015; Liú Lěi 2015; see also de Sousa, forthcoming). With additional data from Pínghuà and some western Yuè dialects, Bit-Chee (2016)’s cline of grammaticalisation, and other characterisations of the diminutives and suprafixes in Yuè, are still mostly valid. One small difference with Kwok Bit-Chee (2016), as shown in Section 3.2, is that a preceding step of having an unsuprafixed *ɲi^A 兒 diminutive suffix is proposed here (Bit-Chee (2016)’s grammaticalisation cline starts with a suprafixed *ɲi^H 兒). The following two of Kwok Bit-Chee (2016)’s characterisations for Yuè dialects also largely hold: (a) a speech variety can have a high-level or a high-falling suprafix, but not both, and (b) the high-level or high-falling suprafix that can be used on roots is the same as the high-level or high-falling suprafix that is used on the *ɲi^H 兒 diminutive suffix. As discussed in Section 3.2, Nánníng Wèizǐlù Pínghuà has both a high-falling and a high-level suprafix. Nonetheless, to be fair, the high-level suprafix is rare, and the more commonly used high-falling suprafix is indeed the same as that in the -ɲi²¹⁻⁵³ 兒 diminutive suffix. Given that the vast majority of Pínghuà speakers live in places where they are the minority, and the much higher linguistic diversity in Guǎngxī (than e.g., Guǎngdōng), the language contact situation that Pínghuà dialects face are on average much more complex than that faced by Yuè dialects. That Nánníng Wèizǐlù Pínghuà has a high-level suprafix (in addition to the expected high-falling suprafix) could be due to the influences from the other Nánníng languages, especially Nánníng Cantonese (see de Sousa 2015 on the language contact situation in Nánníng).

The theory here that the unsuprafixed *ɲi^A 兒 suffix develops into the suprafixed *ɲi^H 兒 suffix faces one slight problem. This theory predicts that speech varieties that have not developed a suprafixed *ɲi^H 兒 suffix would not have developed lexical suprafixes that can be used directly on the root (i.e., the *ɲi^H 兒 step would be skipped). There are indeed counterexamples. For instance, in the Southern Pínghuà of Chóngzuǒ Jiāngzhōu, there are a few cases of human nouns having a high-level suprafix that expresses affection (Lǐ Liánjìn and Zhū Yàn’é 2009, p. 16). In nearby Chóngzuǒ Xīnhé (Liáng Wěihuá and Lín Yì 2009, pp. 152–53), one finds examples of a suprafixed tʃɛk³⁻³⁵ or tʃʰɛk³⁻³⁵ 雀 ‘bird’ in, e.g., tʃʰɛk³⁻³⁵ -ɲi²¹ tɐu³⁵ 雀兒竇 (bird\supr -dim nest) ‘bird nest’, and ma³¹ tʃɛk³⁻³⁵ 麻雀 ‘sparrow’ (cf. the citation-tone pronunciation in kʰoŋ³³ tʃʰɛk³ 孔雀 ‘peacock’). Nonetheless, it is true that these speech varieties, which have not developed a suprafix in the *ɲi^A 兒 suffix, are very poor in lexical suprafixation in general (based on data seen so far). Speakers of Chóngzuǒ Pínghuà would have come across the suprafixes in the Cantonese and Zhuàng varieties spoken in Chóngzuǒ. The (limited amount of) lexical suprafixes in Chóngzuǒ Pínghuà could easily be a contact-induced phenomenon. In addition, the Chóngzuǒ Pínghuà varieties are in contact with other Pínghuà varieties spoken along the Left River, all the way downriver to at least the western suburbs of Nánníng, where their ancestors came from. (In Nánníng Wèizǐlù Pínghuà, similarly, there are the suprafixed tʃɛk³⁻⁵ -ɲi²¹ 雀兒 (bird -dim) ‘little bird’ and ma²¹ tʃɛk³⁻⁵ 麻雀 ‘sparrow’, and the unsuprafixed kʰʊŋ³³ tʃʰɛk³ 孔雀 ‘peacock’.) Kwok Bit-Chee (2016)’s arguments for the Yuè dialects are still largely applicable to the Pínghuà dialects.

Countering the common theory that the Cantonese lexical suprafixes came from an earlier ɲi^H 兒 diminutive suffix, Shào Huìjūn (e.g., 2005; Shào Huìjūn and Gān Yū’ēn 2007) argues that the Cantonese lexical suprafixes are not related to the *ɲi^H 兒 diminutives and nasal rising found in the western Yuè dialects. Her first main argument is that, if the Cantonese lexical suprafixes came from a *ɲi^H 兒 suffix, one would see a gradual decrease in the use of *ɲi^H 兒 diminutives as one heads east from Guǎngxī and southwestern Guǎngdōng. However, east of the *ɲi^H 兒/nasal-rising area in Guǎngxī and southwestern Guǎngdōng, there is a sudden disappearance of a productive ɲi^H 兒 suffix. Her second main argument is that the range of suprafixed words in Standard Cantonese does not match well with those in the Yuè dialects in southwestern Guǎngdōng (in which Shào is an expert).

Both of these are facts that need to be addressed. Shào’s first point perhaps simply reflects the fact that the influence of Cantonese is extraordinarily strong. The eastern Yuè dialects have developed a newer diminutive suffix, *tsVj^B 仔, e.g., Cantonese tsɐi²⁵. Theoretically, the influence from the heart of the Pearl River Delta is so strong that *tsVj^B 仔 has become the dominant diminutive suffix in most Guǎngdōng Yuè dialects, causing the obsolescence of the earlier ɲi^H 兒 suffix in these dialects. Hence, one sees a large area without a productive ɲi^H 兒 suffix. As for Shào’s second point, while having highly similar ranges of suprafixed words can indicate a high level of relatedness (e.g., Canton and Hong Kong Cantonese), not having similar ranges of suprafixed words cannot be used to demonstrate non-relatedness, cf. the case of Hong Kong and Macau Cantonese demonstrated in Section 2.3. Canton, Hong Kong, and Macau Cantonese are highly similar to each other. One cannot draw meaningful conclusions on the level of (un)relatedness from just the differences in the modern range of suprafixed words, as both the gaining and losing of suprafix can happen rather quickly.

Nonetheless, there are reasons to be skeptical like Shào. There is no a priori reason why the eastern Yuè dialects must have gone through the same grammaticalisation pathway as the western Yuè dialects. Also, it is indeed hard to prove that the eastern Yuè dialects had an earlier stage when the ɲi^H 兒 suffix was prevalent. Looking at older Romanised Cantonese texts like Ball (1888) (Section 2.2.3), with non-lexical suprafixes, we know which grammatical morpheme a high-rising suprafix alternated with. For instance, it was clearly stated (p. 15) that the non-lexical suprafix in ká2* ‘married’ (woman getting a husband) was in free variation with the perfective marker cho2 in ká3 cho2 嫁唨 ‘married’. On the other hand, with lexical suprafixes, while we know that suprafix tone 2* was a long tone, and the extra length must have come from a suffix, there is no evidence of what the identity of this suffix was.

There are two other diminutive suffixes that are commonly found amongst Yuè dialects: *tsVj^B 仔 and *tsi^B 子. The *tsVj^B 仔 suffix (e.g., Cantonese tsɐi²⁵) is now the dominant diminutive suffix in the eastern Yuè dialects. While Kwok Bit-Chee (2016) primarily argues for the *ɲi^H 兒 theory, he has nonetheless also presented some arguments for and against *tsVj^B 仔 being the origin of the Cantonese lexical suprafixes. In a footnote (p. 307), some alternating pairs were presented, e.g., ma²¹⁻³⁵ 麻 and ma²¹-tʃɐi³⁵ 麻仔 ‘measles’. These suggest that -tʃɐi³⁵ 仔 could be the source of the Cantonese high-rising lexical suprafix, in at least some words. One argument against the *tsVj^B 仔 theory is that, while the high-rising tone in *tsVj^B 仔 explains the high-rising suprafixes well, it does not explain the high-level and high-falling suprafixes. Also, looking at the Yuè varieties listed in Kwok Bit-Chee (2016)’s Table 4 (p. 288), they all have high-level or high-falling suprafixes but not necessarily high-rising suprafixes. This further lowers the explanatory power of the *tsVj^B 仔 theory. (Nonetheless, the suprafixes amongst Yuè dialects do not need to all stem from one single source; the *ɲi^H 兒 theory and the *tsVj^B 仔 theory need not be mutually exclusive.)

There is also Maurice Wong (1982; quoted by Alan Yu 2007, p. 202), who argues that the Cantonese lexical suprafixes came from the *tsi^B 子 suffix. He offers alternating pairs like min²²⁻³⁵ 面 and min²²-tsi³⁵ 面子 ‘face, honour’ as examples. While there are Yuè (e.g., Kao Wan-yu 2007) and Pínghuà dialects (e.g., Nánníng Pínghuà; de Sousa, forthcoming) where the *tsi^B 子 suffix is commonly used, the *tsi^B 子 suffix is infrequently used in Standard Cantonese and other Yuè dialects in the Pearl River Delta. In the Pearl River Delta Yuè dialects, the *tsi^B 子 suffix is largely restricted to literary expressions, and there are very few alternating pairs between a high-rising suprafix and a *tsi^B 子 suffix. In addition, the *tsi^B 子 theory suffers the same problem as the *tsVj^B 仔 theory, given its high-rising tone.

Looking at the Cantonese data in Ball (1888), only one word had the *ɲi^H 兒 suffix: hat7 -yi4-1* 乞兒 (beg -dim\supr) ‘beggar’. The *ɲi^H 兒 theory argues that the old -ɲi^H suffix had already transformed into the lexical suprafixes in Cantonese; the -yi4-1* 兒 suffix in hat7 -yi4-1* 乞兒 (beg -dim\supr) ‘beggar’ is a very rare remnant of the old -ɲi^H suffix. At the time of Ball (1888), the lexical suprafixes had already lost most of their diminutive meanings, and the primary diminutive suffix was already -tsai2 仔. There were no cases of lexical suprafixes freely alternating with -tsai2 仔, and -tsai2 仔 could be suffixed to suprafixed roots, e.g., t’oi4-2* tsai2 檯仔 ‘small table’ and pò6-2* tsai2 部仔 ‘small manuscript book’ (p. 41). This testifies that -tsai2 仔 was not the source of the lexical suprafixes in Cantonese at the time of Ball (1888). (However, this does not preclude the possibility that -tsɐi²⁵ 仔 caused more words to acquire a high-rising suprafix in later Cantonese.)

It is notable that -tsai2 仔 did not have a suprafix in Ball (1888). Nonetheless, this is not a problem for the *tsVj^B 仔 theory, as the perfective marker cho2 唨 also did not have a suprafix, and it alternated with the (non-lexical) high-rising suprafix.

The -tsz2 子 suffix was rarely used in Ball (1888). There were no examples of -tsz2子 alternating with a suprafix. The -tsz2 子 suffix was most probably not the source of the lexical suprafixes in Ball (1888).

Based on all the points presented above, despite the flaws of the *ɲi 兒 theory, the *ɲi 兒 suffix remains the most likely source of the Cantonese lexical suprafixes.

4. Conclusions

This article discussed some synchronic and diachronic aspects of the Cantonese lexical suprafixes. We began by looking at the definitions of ‘Cantonese’ (Section 1.1) and the difference between non-lexical and lexical suprafixes. Lexical suprafixes are derivational tonal affixes that are lexically determined (Section 1.2).

The lexical suprafixes most likely originated from diminutive suffixes. Sometimes, the lexical suprafixes still have diminutive-type meanings, but usually there is no diminutive-type meaning left. They are ‘just there’ and may change the meaning and/or word class. In modern Cantonese, there is a tone 2 (high-rising) suprafix and a rarer tone 1 (high-level) suprafix (Section 2.1). Looking at descriptions in the twentieth century and late-nineteenth century, there used to be a distinction between suprafix tone 1* (high level) and ordinary tone 1 (high falling) (Section 2.2.1) and a distinction between suprafix tone 2* ([²⁵] in the 1940s and [³¹⁵] in the late-nineteenth century) and ordinary tone 2 [³⁵] (Section 2.2.2). We looked at the lexical items in some late-nineteenth century Cantonese textbooks. In comparison with modern Cantonese, some words that had a lexical suprafix back then are still suprafixed nowadays, some words have gained a suprafix, and some words have lost a suprafix (usually reverting back to their citation-tone pronunciation). Looking at just the words found in these texts, there has been an increase in the number of suprafixed words (Section 2.2.3). In Section 2.3, we looked at the use of suprafixes in some non-standard Cantonese varieties like Nánníng Cantonese and Older Macau Cantonese. They developed later than Canton and Hong Kong Cantonese (the standard varieties of Cantonese), and they have far fewer suprafixed words than Standard Cantonese. The facts discussed in Section 2.2 and Section 2.3 strongly suggest that in Standard Cantonese, overall, there has been an increase in the number of suprafixed words, and that in many words, their lexical suprafixes do not have a long history.

In Section 3.1, we looked at the three most-commonly used diminutive suffixes amongst Pínghuà and Yuè dialects: *ɲi^A/H 兒, *tsVj^B 仔, and *tsi^B 子. (Pínghuà and Yuè basically lie on a dialect continuum, with Pínghuà in the west and Standard Cantonese close to the eastern edge of the dialect continuum.) The majority opinion is that the lexical suprafixes in Standard Cantonese came from a *ɲi^H 兒 diminutive suffix. In Section 3.2, based on Kwok Bit-Chee (2016)’s grammaticalisation cline from *ɲi^H 兒 to the lexical suprafixes in Yuè dialects, this study expands westward and includes data from some Pínghuà and some extreme-western Yuè dialects. Many Pínghuà dialects and some extreme-western Yuè dialects only have an unsuprafixed *ɲi^A 兒 diminutive suffix in its citation tone—Lower tone A. Then, in mid-southern Guǎngxī, one finds speech varieties like Nánníng Wèizǐlù Pínghuà and Héngzhōu Cantonese; other than the unsuprafixed *ɲi^A 兒 suffix, they have also developed a suprafixed *ɲi^H 兒 diminutive suffix. Some Pínghuà and Yuè dialects further east only have a suprafixed *ɲi^H 兒 diminutive suffix. From *ɲi^H 兒, there are two independent development pathways: ‘nasal rising’ (a nasal coda or nasal feature attached to a root, plus a rising tone) and suprafixing directly on a root. It seems that speech varieties with the former also have cases of the latter, but having the latter does not infer the former. Geographically, nasal rising is only found in southwestern Guǎngdōng and neighbouring southeastern Guǎngxī. On the other hand, suprafixing directly on a root is found in the vast majority of Yuè dialects, and also in some Pínghuà dialects. Nonetheless, in many Yuè and Pínghuà dialects, the use of lexical suprafixes is rather marginal. Lexical suprafixation is the most developed in Standard Cantonese.

Ultimately, an undisputable answer cannot be found to the question of the origin of the Cantonese lexical suprafixes. With non-lexical suprafixes, we have textual evidence of what grammatical particles they alternated with. On the other hand, with lexical suprafixes, there is no such textual evidence. Section 3.3 discusses the merits and flaws of the *ɲi^A 兒, *tsVj^B 仔, and *tsi^B 子 origin theories. Out of these three (types of) theories, the *ɲi^A 兒 theory remains the most promising.12