Since its beginning in the 19th century, morphological typology has postulated a universal scale of less vs. more tightly packed word forms. The scale ranges from isolating to agglutinative to fusional to introflexive, and is canonically exemplified by Chinese (isolating), Turkish (agglutinative), Latin (fusional) and Modern Standard Arabic (introflexive). Recent research has shown that such a scale conflates many different typological variables and incorrectly assumes that these parameters covary universally (see Plank 1999, Bickel and Nichols 2005, among others). Three prominent variables involved in this are phonological fusion, formative exponence, and flexivity (i.e. allomorphy, inflectional classes). In this chapter we concentrate on fusion. For exponence, see Chapter 21, and for one aspect of flexivity, see Chapter 59 (on possessive classification).

1. Defining fusion types

Fusion refers to the degree to which grammatical markers (called formatives in the following) are phonologically connected to a host word or stem. There are three basic values: isolating, concatenative, and nonlinear.

Isolating formatives are full-fledged phonological words of their own. In Fijian, all formatives with more than one mora are isolating. An example is the past tense formative aa:

(1) Boumaa Fijian (Dixon 1988: 53)

Au	aa	soli-a	a=niu	vei	ira.
1sg	pst	give-tr	art=coconut	to	3pl
‘I gave the coconut to them.’

Concatenative formatives are phonologically bound. They need some other host word for their pronunciation and form one single phonological word together with that host. The usual effects of this are that concatenative formatives cannot be individually stressed, and that the combination of formative and host undergoes various phonological adjustments. The past tense marker of Turkish, for example, undergoes vowel harmony and assimilates in consonant voicing to the host stem. Thus, the past tense formative is -ti after a stem with unrounded front vowels and a voiceless final consonant (e.g. git-ti ‘go-past’), -tı after a stem with unrounded back vowels and a voiceless final consonant (e.g. yap-tı ‘do-past’), -di after a stem with unrounded front vowels and a voiced final consonant (e.g. gel-di ‘come-past’), and so on. A subset of concatenative markers is constituted by cliticized words. The Spanish object marker a, for example, is syntactically a word (preposition) but phonologically it is a clitic and thus concatenative.

Once the phonological alternations are properly analyzed, strings of concatenative formatives can be segmented into clear-cut morphemes. Nonlinear formatives are not amenable to this because they are realized not in linear sequence but by direct modification of their host. In our sample, we found two subtypes of nonlinear formatives: ablaut and tonal. Modern Hebrew illustrates the ablaut type. The past (“perfect”) vs. future (“imperfect”) opposition, for example, is expressed by (i) the choice of a stem template (e.g. CaCVC in the past, CCVC in the future) and (ii) the choice of agreement affixes (entirely suffixes in the past, mostly prefixes in the future).

(2) Modern Hebrew (Orin Gensler, p.c.)

a.	šamar-ti	b.	ʔe-šmor
	guard.pst-1sg.pst		1sg.fut-guard.fut
	‘I guarded’		‘I will guard’

Neither affix nor stem choice appears to be basic. Tense is not marked in this language by an extractable morpheme but by the complete affix-plus-stem pattern as a whole. For purposes of this survey we call this ablaut morphology.

Suprasegmental nonlinear formatives chiefly involve tonal modification. In Kisi (Atlantic; Guinea), most tense-aspect opposition are expressed by tone, and tone alone: 

(3) Kisi (Childs 1995: 220-3)

a.	Ò	cìmbù.
	3sg	leave.pres.habitual
	‘She (usually) leaves.’
b.	Ò	cìmbú.
	3sg	leave.pst.pfv
	‘She left.’

Here, present habitual is expressed by low tone on the last syllable (3a); past perfective is expressed by high tone (3b).

2. Sampling procedure and feature values

Languages were surveyed for the case and tense-aspect-mood exemplars as defined in the chapter on “Sampling case and tense formatives”. Nearly 90% of the languages in our sample have the same values for case and tense-aspect-mood, and therefore we combined the two values into one overall value: 

Values of Map 20A. Fusion of Selected Inflectional Formatives

Go to map
	Value	Representation
	Exclusively concatenative	125
	Exclusively isolating	16
	Exclusively tonal	3
	Tonal/isolating	1
	Tonal/concatenative	2
	Ablaut/concatenative	5
	Isolating/concatenative	13
Total:		165

Mixed types mean that case and tense-aspect-mood differ from each other, and the type list exhausts what combinations are attested in our sample. In a few instances, however, mixed types refer to languages where there is conflicting evidence for the fusion type of at least one of the formatives. Conflicting evidence is found, for example, in Lakhota (Siouan; North and South Dakota), where the future tense marker -kta is part of the same phonological word as the verb stem with regard to morphophonological rules, but not apparently with regard to syllabification (see Russell 1999 for careful discussion). Other instances of conflicting evidence that we found include Beja, Thai, Chamorro, and Gooniyandi.

The color scheme on the map is set up so as to highlight the presence of some isolating and the presence of some nonlinear formatives. 

3. Results and discussion

Most of the languages in our sample (75%) rely exclusively on concatenative morphology for case and tense-aspect-mood. Languages with some isolating or nonlinear formatives are much rarer and have limited areal distribution. 

Languages with isolating formatives, or traces of isolating structure in mixed types, are mostly confined to the Sahel Belt of West Africa and to Southeast Asia and the Pacific. Apart from this, there are outliers in southern Africa (Khoekhoe), Australia (Gooniyandi), and the Americas (several instances).

In our sample, ablaut morphology is always mixed with concatenative morphology and appears as an African singularity limited to representatives of Afro-Asiatic (Hebrew, Egyptian Arabic, Middle Berber Atlas, Beja) and the Central Sudanic branch of Nilo-Saharan (Lugbara). Similarly, tonal morphology is also by and large restricted in our sample to African languages (Niger-Congo and Nilotic). An outlier with tonal formatives is Iau (Lakes Plain; Papua, Indonesia). 

4. Acknowledgements

This research was supported by U.S. National Science Foundation Grant No. 96-16448 (Nichols, P.I.), Swiss National Science Foundation Grants No. 08210-053455 and 610-062717 (Bickel, P.I.), and the Institute for Slavic, Eurasian, and East European Studies, UC Berkeley. We thank Aimee Lahaussois Bartosik, Dave Peterson, and Suzanne Wilhite for help with data collection.