1. Identifying verbal person syncretism

This map records instances of syncretism in the inflectional marking of subject person in verbs. For example, in German, the same form is regularly used for both 1st and 3rd person subjects in the plural, e.g. machen  ‘(we/they) make’ or machten  ‘(we/they) made’. The map makes a three-way distinction:

Values of Map 29A. Syncretism in Verbal person Marking

Go to map
	Value	Representation
	No subject person/number marking	57
	Subject person/number marking is syncretic	60
	Subject person/number marking is never syncretic	81
Total:		198

In order to identify a form as syncretic one needs some evidence that there are multiple person values underlying a single form. This is straightforward when the syncretic pattern is somehow restricted in scope. For example, it may be found only in certain tense/aspect/mood paradigms, as in Aymara, where 1sg=2sg in the future perfect but not the present (see 1); or it may be found only in certain inflection classes, as in Kunama, where 2pl=3pl in verbs of conjugation class I but not those of class IIb (see 2); or it may be limited to a particular gender, as in Iraqw, where 2sg=3sg for feminine subjects but not for masculine ones (see 3).

(1) Aymara (Bolivia and Peru; Deza Galindo 1992: 103-105)

‘love’	future perfect	present
1sg	munchiyäta	muntua
2sg	munchiyäta	muntahua
3sg	munchïna	munihua

(2) Kunama (Nilo-Saharan; Eritrea; Bender 1996: 32-33)

aorist	class I ‘sat’	class IIb‘had’
1pl	gomake	maináke
2pl	goŋke	meináke
3pl	goŋke	oináke

(3) Iraqw (Southern Cushitic; Tanzania; Mous 1993: 162)

‘get up’	feminine	masculine
1sg	tláw	tláw
2sg	tléer	tléer
3sg	tléer	tláy

However, in some cases a given syncretic pattern is systematic throughout a language, so there is no direct evidence for the distinction. For example, in Kobon, 2du and 3du are always identical (see 4). Nevertheless, there is indirect evidence, in as much as 2nd and 3rd person are distinct in the singular and plural; the same is true of the 1pl/3pl German form machen (see above), which corresponds to two forms in the singular, namely mache  ‘I make’ and macht  ‘s/he makes’.

(4) Kobon (Trans-New Guinea; Davies 1981b: 166)

‘went’	du	sg	pl
1	arlo	arnö	arno
2	arlö	arna	arbe
3	arlö	ara	arla

In the remainder of this chapter, we will discuss a number of further generalizations that are not directly reflected on the map. 

2. Common patterns of syncretism

Although it would appear that any combination of person values is a possible one, there are certain patterns which are more commonly attested than others. One relevant parameter is number. In the singular, by far the most common pattern is for 2nd and 3rd person to be identical, as in Iraqw (see 3), while in the non-singular, the most common patterns are for 1st or 2nd person to be identical with 3rd, as in German, Kobon (see 4) or Kunama (see 2). Another relevant parameter is the scope a given syncretism has within the language, i.e. whether the syncretism is limited in scope (as in 1-3) or systematic throughout the language, as in Kobon (see 4). Where syncretism is limited in scope, the singular is the most likely to be affected, accounting for more than half of the examples in the sample. On the other hand, where syncretism is systematic, it is rare for it to affect singular persons, accounting for only three examples in the sample (Hindi, Nivkh and Dongolese Nubian), all of them involving 2sg=3sg.

The rarest type of syncretism is that where a single form combines different person values with conflicting values for another feature, typically number or gender. For example, in the "full form" paradigm in Diola-Fogny (used in certain syntactic contexts), the prefix nu-  is used for both 2sg and 1pl (see 5). In the prefix conjugation of the Semitic languages, 2sg masculine may be identical to the 3sg feminine, as in Egyptian Arabic tiktib  ‘write(s)’ (Mitchell 1962: 73). Such examples are not numerous enough to reveal any striking cross-linguistic tendencies.

(5) Diola-Fogny (Northern Atlantic; Senegal; Sapir 1965: 90)

‘broke’	sg	pl
1	nitigɛr	nutigɛr
2	nutigɛr	jitigɛr
3	natigɛr	kutigɛr

3. Geographical and genealogical tendencies

person syncretism is widespread: of the 140 languages which mark person inflectionally (out of 197 total in the sample), roughly two-fifths (60 languages) display it. In most of these (54 languages) the syncretic pattern is restricted in scope, as shown in (1-3). The most syncretism-prone areas are Africa, New Guinea, western and northern Europe and, to a lesser extent, South and Central America, and the Indian subcontinent. Least syncretism-prone are North America, Eurasia (with the exception of the above-mentioned parts of Europe) and the Pacific (with the exception of New Guinea).

In genealogical terms, person syncretism is widely found in nearly all of the sample languages from the Trans-New Guinea phylum, and is almost as widespread in the Nilo-Saharan, Niger-Congo and Indo-European families. Some patterns can be said to be typical of certain families, at least as represented in the sample. Afro-Asiatic has 1sg or 2sg=3sg, restricted by gender. In Bantu, one of the gender/class markers used for 3rd person is often identical to the marker for 2nd or 1st person. In Carib languages, 1pl exclusive and 3rd person are identical, opposed to a distinct 1st inclusive. In the Trans-New Guinea phylum, 2nd and 3rd person are identical in some or all non-singular numbers.

4. Theoretical implications

A central question concerning syncretism is whether or not syncretic forms are indicative of some underlying semantic or morphological relationship. The data allow us to consider four possible interpretations. 

First, one could assume that identity of form indicates identity of function, so that the syncretic form reflects a natural semantic or morphosyntactic class. For example, one could characterize the person values in terms of component features, so that the identity of 1st and 2nd person in the singular in Aymara (see 1) would be attributed to a shared feature such as [+discourse participant], while the identity of 2nd and 3rd person in the dual in Kobon (see 4) could be attributed to a feature such as [–speaker].

Second, the syncretic form could be construed as unspecified for person. An obvious example inviting this interpretation is English, where the ending -s signals a 3sg subject in the present, while the bare stem is used for all other person values. Similar examples occur with 1st or 2nd person: an overt 2sg marker contrasting with no marker for the other singular persons is found in the non-past in Ika (n˄-; Frank 1990: 51), in the preterite in German (-st), and in various tense/aspect/mood paradigms in Spanish (-s); while an overt 1sg contrasting with no marker for other singular persons occurs in negative verbs in Karok (ná-; Macaulay 1992a: 183).

Third, one could say that the syncretic form actually “belongs” to one of its component values. Such an analysis is suggested by examples where a form which is non-syncretic in one paradigm is used syncretically in another. For example, in Iraqw (see 3), the syncretic 2sg/3sg form used with feminines is identical to the distinct 2sg form used with masculines. There are other similar examples in the sample, which likewise involve 2nd and 3rd person, both in the singular (Igbo) and the plural (Carib, Lower Grand Valley Dani and Suena).

Fourth, syncretism could be the result of purely fortuitous homophony. Such an interpretation is most obvious where the syncretism is attributable to a morphophonological rule. For example, in Daga, a form for 1sg/3pl is opposed to a form for 3sg in the past tense of class A verbs by the alternation -a- ~ -e-; this is neutralized in class E verbs, which have non-alternating -e-  throughout (Murane 1974: 52-53).

The data from the sample do not fully support any one single analysis. Each of the first two, by themselves, would be overly restrictive, given the range of patterns discussed in §2. The third possible analysis finds direct support only in a small number of cases. The fourth fails to take account of the existence of patterns which reoccur cross-linguistically. We suspect that all the analyses are valid in principle, being appropriate for different sorts of data.