The parameter
Languages differ in how many meaningful units they pack into a single word, and in how cleanly those units can be separated. The first question is the degree of synthesis: how many morphemes per word. The second is the degree of fusion: when morphemes combine, do they stay recognizable or do they merge into forms that cannot be cut apart.
The two questions are independent, which is why the traditional four-way label set is a simplification rather than a classification. Turkish is highly synthetic and almost perfectly transparent: each suffix carries one function and keeps its shape. Latin is equally synthetic but fusional: a single ending marks case, number and gender at once, with no way to isolate the parts. Mandarin sits at the other end, with roughly one morpheme per word and grammatical relations carried by word order and particles. West Greenlandic incorporates whole arguments into the verb, so a clause can be a word.
No language is a pure type. English is analytic in its verb phrase and fusional in its few remaining inflections; Russian is fusional in its nouns and agglutinative in parts of its verb. The parameter is a tendency, and what matters for a translator is the gap between two tendencies, not either label on its own.
How languages vary
- Isolating: close to one morpheme per word. Mandarin, Vietnamese, Yoruba. Grammatical relations are carried by order and by free particles.
- Agglutinative: many morphemes per word, each with one function and a stable shape. Turkish, Finnish, Hungarian, Swahili, Japanese, Korean.
- Fusional: many morphemes per word, merged so that one form carries several categories at once. Latin, Russian, Polish, Greek, Arabic.
- Polysynthetic: very high synthesis, often with noun incorporation, so that a single verb form corresponds to a full clause elsewhere. West Greenlandic, Mohawk, Chukchi.
The scale is continuous. Most European languages fall between analytic and fusional, which is why translators working only within that range rarely have to think about the parameter, and why it becomes the dominant problem the moment an agglutinative or polysynthetic language enters the pair.
Consequences for translation
The larger the gap in synthesis, the more of the translation consists of repackaging rather than choosing. A word in the source corresponds to a phrase or a clause in the target, and the correspondence is compulsory: no lexical decision will avoid it. That is a unit shift, and where the word class changes with it, a transposition as well.
Three practical consequences follow.
Word counts stop being comparable. Translating out of an agglutinative language routinely produces fifteen to twenty-five per cent more target words than the source contains. Quoting on source words and delivering target words, or the reverse, misprices the work systematically rather than randomly.
Segment alignment degrades. Sentence-aligned memories built across a large synthesis gap contain many one-to-many correspondences at word level, so term extraction and concordance searching both return less than they do within a family.
Compression is not available symmetrically. Going the other way, into the more synthetic language, the translator can often fold a target clause into a single word, which is legitimate, and can also fold in information the source did not carry, which is not. The parameter creates the opportunity for both, and only one of them is a technique.
Examples
TR>EN
- Source: evlerimizdekilere
- Target: to the people in our houses
- Comment: One word, six suffixes, each separable and each with a single function: house, our, in, the one at, plural, dative. English has no comparable packing device, so the correspondence is one word to six. The shift is forced by the system, not chosen by the translator, and nothing in the target should be marked as an addition.
ZH>EN
- Source: 他明天去北京
- Target: He is going to Beijing tomorrow.
- Comment: The Mandarin verb carries no tense marking; future reference comes from the time word. English inflects the verb and cannot leave the time relation unexpressed. What was lexical in the source becomes grammatical in the target, which is the definition of a level shift.
EN>TR
- Source: if we had not been able to come
- Target: gelemeseydik
- Comment: Six English words against one Turkish word: negation, ability, conditional, past and first person plural are all suffixes on a single stem. Translating in this direction, the technique is compression, and the reviser's question is whether anything was quietly lost with the words.
References
- Comrie, B. (1989). Language Universals and Linguistic Typology, 2nd edition. Chicago: University of Chicago Press.
- Croft, W. (2003). Typology and Universals, 2nd edition. Cambridge: Cambridge University Press.
- Whaley, L. J. (1997). Introduction to Typology: The Unity and Diversity of Language. Thousand Oaks: Sage.
- Velupillai, V. (2012). An Introduction to Linguistic Typology. Amsterdam: John Benjamins, Ch. 4.