Economy of representation in the Esperanto word

Marc van Oostendorp

(Appeared in Linguistics in the Netherlands 1998)


Esperanto words get shortened in certain colloquial styles of speech. It turns out that this shortening is not triggered by prosodic requirements on the word, but rather by a morphological principle of Economy of Representation. Categorial features should not be expressed more than once within a word; because they are always expressed by a categorial suffix, strings of phonological segments get deleted if they get reanalysed as nominalizing, verbalizing and adjectivizing suffixes.


Shortening of words and names is discussed extensively in the recent phonological literature on Prosodic Morphology (e.g. McCarthy and Prince 1986, 1998, Poser 1984, 1990, Kager 1995, Van de Vijver 1997). The picture that emerges from this literature is that shortening is triggered, or at least constrained, by prosodic wellformedness. We can define various phonological templates, such as the `Minimal Word' and the `Loose Minimal Word'. Shortening in this view is inspired by a wish to fit every word or name into a given prosodic template.

In this article I examine another factor that may play a role in shortening: morphological economy of representation. Certain segments within a word may be omitted because they too closely resemble a redundant morpheme.

My data are from what I call `Colloquial Esperanto' (CE), a variant of Esperanto that is used in informal speech by native and near-native speakers. Even if we wish to call Standard Esperanto an artificial language given its history, it does not make sense to talk about CE in the same way when we are interested in its synchronic shape. There is a small group of approximately 1,000 native speakers of CE (in Europe, Lindstedt 1997; see also Versteegh 1993). Furthermore, although the judgements on the data presented below are quite sharp and consistent among speakers of CE, regardless of their native tonge, there is no known descriptive or prescriptive grammar that describes or prescribes them all. The shortenings studied in this article seem to have come into existence quite spontaneously. In addition it may be useful to point out that many of the studies in the Prosodic Morphology literature are based on hypocoristics, secret languages and slangs. It is hard to devise a criterium of `naturallness' which would single out these systems as more natural than CE.

Most of the data given here are excerpted from Philippe (1991); all of them have been checked by the author of this article with four speakers of CE of different linguistic backgrounds (Renato Corsetti, Probal Dasgupta, Jouko Lindstedt, Anna Lowenstein).

Word forms in Colloquial Esperanto

The main characteristic of CE that I am interested in here is that some of its words are shortened in informal style levels:


Standard   Colloquial  
distanca > dista distant
ekspozicio > ekspo expo(sition)
fotografi > foti to photograph
spontanea > spontana > sponta spontaneous
stenografi > steni to stenograph
arkaika > arkaa archaic

In the column under `Standard' I listed the forms such as they occur for instance in the Fundamento de Esperanto (Zamenhof 1991). In the column on the lefthand side I listed the CE forms. In many of the cases under (1) a similar shortening has occurred in the European languages from which the Esperanto lexicon is derived (aùto, ekspo). This is not true for all of these forms however: as far as I know words like dista and foti have no direct correspondents in other languages. It also is important to note that (most of) the forms in (1) were labeled `informal' or `modern' by the informants. In one case, an informant claimed never to use the short forms, but he related this to his own `conservative language use.' I will discuss the influence of style level in section 4.

The CE forms are always shorter than the corresponding forms. In many cases, they are bisyllabic. We might therefore be led to believe that a prosodic analysis is in order. A relatively simple prosodic analysis could be based on two constraints:

(2) PrWd=Ft
A prosodic word contains exactly one foot.

(3)Max(McCarthy and Prince 1994)
All underlying segments should surface.

PrWd=Ft has as an effect that every word consists of at most one foot, two syllables. Since word stress is always on the penultimate syllable both in Standard Esperanto and in CE, it seems straightforward to assume that Esperanto has trochaic feet. All of the bisyllabic forms in (1) comply with this trochaic template. The Max cOnstraint prohibits `unnecessary' deletion of segments. Ranked in the appropriate order, these two constraints have as an effect that CE words surface as exactly bisyllabic:


b. di.sta  
c. dis  

Yet, whatever its merits for other languages, this analysis leaves us with several unsolved problems. For one thing, we may wonder why a more faithfull candidate such as *distan does not come out as the optimal one in the tableau in (4). Yet this problem is not as dramatic as it seems at first sight. The rejection of distan might be analysed as an `emergence of the unmarked' (McCarthy and Prince 1994) effect, due to the interaction between the constraints NoCoda and Max (in that order). The fact that distanca is not shortened to *dita in which also the first syllable is open, can be explained by a constraint Contiguity, which requires a contiguous string of input segments to surface. (We will see below, however, that a morphological account can be given to this example as well.)

More serious is the problem what to do with forms such as steni < stenografi and arkaa (*arka) < arkaika. The first example might lead us to assume that some kind of edge alignment is at work: the last segment in the input word has to be present in the output word. The second form can only be explained under the prosodic analysis if we assume that the template is loosened in certain cases. In order to understand more fully what exactly is going on, it is usefull to have a short excursion into the basic morphology of the CE word.

The morphological shape of the CE word

We first have to note that there is something special about the ending of a lexical (i.e. non-functional) CE word: this ending denotes the categorial value of the word to which it belongs. Nouns end in -o (in the singular, nominative): lingvo (`language'), domo (`house'), adjectives end in -a (in the singular, nominative): eleganta (`elegant'), bela (`beautifull'), verbs end in -i (in the infinitive): labori (`work'), paroli (`speak'). This correlation is bidirectional: all nouns end in -o, all words in -o are nouns.

Probably because of this, CE displays a lot of conversion: once we know the word eleganta, it is straightforward to form a corresponding noun (eleganto `elegance') and verb (eleganti `being elegant'). This does not mean that roots such as elegant- do not have their own inherent categorial specification. This has been pointed out already by Saussure (1910) on the basis of examples such as (5):


bros-i brush-verb 'to brush'
bros-o brush-noun 'a brush'
komb-i comb-verb 'to comb'
komb-o comb-noun 'combing'
bros-ad-o brush-nom-noun 'brushing'
komb-il-o comb-instr-noun 'a comb'

The key observation is that the relation brosi:broso is not the same as kombi:kombo. According to Saussure, this is due to the fact that bros- is inherently nominal, while komb is inherently verbal. The former root means something like `a brush'; if a noun is formed from it, this will mean `to brush'. On the other hand, komb- has an inherent meaning closer to `to comb': if we form a noun on the basis of this, we therefore get a word with the same meaning as the English gerund `combing'. Of course, the difference between the two words might be one of lexical semantics. For the sake of simplicity however, I will assume here that the difference is due to a difference in inherent categorial specification. It is now possible to analyse the two words broso and kombo as follows:


/ \ / \


bros o komb o

It is not possible to use a `redundant' suffix in these cases: brosilo and kombado do not occur, or at least not with the required interpretation. This seems to be due to a form of representational economy in morphology: there can be no superfluous elements in the morphological representation (Chomsky 1995).

(7) Representational Economy: Do not express a syntactic or semantic feature more than once.

A few remarks need to be made about the categorial endings as well. In the first place, it is sometimes asserted that the ending -o may be elided in very formal (`artificial') poetic styles of Esperanto; I have the impression however, that this type of elision is not very common in CE. In the second place, endings are elided in compound forms (maybe subject to some phonological restrictions, see below):


(?)san-stato sano-stato 'state of health'
(?)morti-devulo mort-devulo 'somebody who has to die'
(?*)sama-seksemulo sam-seksemulo 'homosexual' (somebody who likes (persons of) the same sex)

Finally, there are no restrictions on the shape of prepositions and functional elements, except that they may not end in -a, -o or -i: for, antau, ne, post, dum, anstatau, kaj. Taken together, these restrictions seem to mean that every syntactic word -- the lexical elements inserted into syntax and regarded as atomic by that module -- need to have a categorial ending. The first part of a compound does not count as a syntactic word and neither do the closed class items just mentioned. The observation can now be formalised into a constraint:

(9)?License: A feature [+N]/[+V] on a syntactic [+Lex] word must be licensed by a categorial suffix.

The constraint says that all nouns ([+N,-V,+Lex]), verbs ([-N,+V,+Lex]) and adjectives ([+N,+V,+Lex]) should end in a categorial suffix; whereas [-Lex] elements and prepositions ([-V,-N]) are not subject to any such restrictions. This constraint can now explain why the last segment of a shortened CE form corresponds to the last segment of the longer form. This is not due to some input-output alignment, but only to the fact that this latter segment is a licenser for categorial values. We can now set up the following tableau for the word dista:


/distanca/ PrWd=Ft License Max
a. dis.tanc  
b. di.sta    

The question arises to what extent the constraint in (10) is universal and to what extent it is language-specific. On the one hand I don't know any language in which there is such a strong correlation between categorial status and output form; the classical Semitic languages may give us the closest approximation (Aronoff 1995). On the other hand, the type of language in which for instance morphological gender corresponds to some kind of phonological shape is well attested.

An alternative formulation of (10) seems also feasible; we could also try not to analyse -i, -a, -o as suffixes, but as really templatic parts of the word. The License constraint would then be formulated as a correspondence constraint (Jackendoff 1997) between categorial structure and phonological form. A potential problem for this is the productiveness of conversion. Given any word, we can form at will a noun, an adjective or a verb on its basis. E.g. on the basis of the preposition antaù 'before' we can form a noun antaùo `front', an adjective antaùa `previous' or a verb antaùi `precede'. I find it hard to describe these facts without assuming that -a, -o and -i are indeed suffixes.

In any case, the constraints Representational Economy and License are in potential conflict: as far as the former constraint is concerned, we should pronounce *bros rather than broso. The fact that the latter form surfaces can therefore be taken as an indication that License >> Representational Economy.



The constraint ranking would predict that the inflectional ending of the lefthandside component of a compound should not surface:



Although the form without the internal suffix seems to be preferred in most cases, and obligatory in some (*?samaseksemulo), a form like sanastato is not rejected by my informants. It might be that syllable structure reasons plays a variable role in the system; e.g. the ranking of the relevant constraints against complex consonant clusters such as nst may be subject to some variation. Under certain circumstances it may be ranked highly enough to give preference to sanastato over sanstato.

There is a difference between the first components of compounds and the bases for derivation. The latter never get any categorial ending. A word like sanulo (`healthy person', < sana+ulo) pronounced as *sanaulo. This is surprising since the suffix ul --just like almost every other affix of the language -- can behave as an independent word: sana ulo is an acceptable nominal phrase. I will not go into this problem here.

Not all problems have been solved yet. We still do not understand why arkaika turns into arkaa rather than *arka, and we still don't understand why distanca is not shortened to *tanca rather than dista. For the latter case we might assume some position-specific faithfulness constraint such as the one proposed by Beckmann (1997): the first syllable of a word has to stay, regardless of what happens to the others. I believe however that closer analysis of the morphological structure of the CE word will lead us in a different direction.

`Superfluous' suffixes

Next to the categorial suffixes -a, -o, -i, CE has a relatively vast system of productive prefixation and especially suffixation. We have already seen two productive suffixes in () above: the instrumental -il and the nominalizing -ad. Both of them seem inherently nominal. Also the suffix -ul in the `personalizing' words samseksemulo and mortdevulo in (8) is such a suffix. This suffix surfaces whenever a `personal' meaning is intended, even though this causes a violation of Representational Economy. The nominal category is represented twice in samseksemulo: once on the suffix -ul, and once on -o. I assume that the faithfulness constraint Maxis responsible for this.


/samseks/, [+Person]

The majority of Esperanto words is borrowed or derived from other languages, mainly Indo-European. In some of these cases these words have been borrowed with internal morphological structure. Philippe (1991) notes that there are quite a few shortenings where the relevant sequence of segments that get deleted is -ci, sometimes preceded by an a or i.

(14) -ci/-aci/-ici

evolucio > evoluo evolution
polucio > poluo polution
inauguracio > inauguro inauguration
vibracio > vibro vibration
intuicio > intuo intuition

My informants do not agree on all data. One informant reports a meaning difference between poluo `environmental pollution' and polucio `wet dream'; For some informants the word intuo is not acceptable; etc. Most informants seem to think that the long forms inaùguracio and vibracio are completely outdated. If we disregard these idiosyncracies, the overall picture seems to conform the picture in (14).

The deleted sequence -ci is a remnant of the nominalizing suffix -tio(n) in (some of the) the original languages. The preceding a or i are remnants of theme vowels in Latin, from which all of these words eventually derive. The nominal function is however expressed in Esperanto by the -o suffix; -ci is therefore again superfluous.

A morphological factor seems to be involved also in other shortenings. In the examples in (15), for instance, the suffix -ic/-ik is still discernible within the Esperanto word.

(15) -ik

arkaika > arkaa archaic
erotika > erota erotic
klasifiki > klasifi classify

Two observations have to be made about -ik. In the first place, it behaves as an adjectivizing suffix in the European languages. In the second place, it is not a productive suffix of Esperanto; there is a semi-productive suffix -ik, but this is nominal: it is used to form the names of branches of science from the name of a subject field (lingviko `linguistics').

It seems therefore that also in this case the principle of representational economy is decisive. This is so in spite of the fact that the suffixes in question are not part of productive `native' Esperanto morphology. The speakers of CE still seem to have some awareness of them. It is as if some form of `Level I' morphology (Kiparsky 1982) has evolved spontaneously out of the data offered to the language learner. Given this morphological analysis, the observed shortening pattern follows straightforwardly from the constraint ranking already given:


a. evolucio  
b. evoluo  
c. evolo  
d. evoluci

Styles of speech

Especially in the case of -ci/0 alternations, most informants observe that there is a style level diference between the various options: evoluo is supposed to be more informal than evolucio. Van Oostendorp (1997) claims that universally the following generalisation holds for phonological grammars:

(17) The more formal the register, the higher ranked faithfulness.

In this approach, style levels are seen as separate grammars; yet the relation between those grammars is never arbitrary. It is always subject to (17). Van Oostendorp (1997) argues that this generalisation can account for the fact that there is usually less reduction, deletion and epenthesis in formal styles of speech than in informal ones. Here, I wish to claim that the facts under discussion here give additional support to this principle. `Standard' or formal Esperanto could be described by the constraint ranking License, Max >> RE (the relative ranking of License and Max cannot be decided upon, since these constraints are never in a state of conflict):


a. evolucio    
b. evoluo  
c. evolo  
d. evoluci

In the higher style register, it is more important to be faithful to the lexical representation than to be economical.

Remaining problems

In the analysis of evolu(ci)o, arka(ik)a, etc., the prosodic wellformedness constraint PrWd=Ft does not play a role at all. Failed candidate analyses such as *evo and *arka show that the constraint should be ranked below the faithfulness constraint Max in all styles of grammar. There is no evidence that shortening is triggered by prosody, but rather by morphological economy of representation alone. On the other hand, it is of course true that the CE root in all of our examples is only one or two syllables long. It is hard to find relevant examples -- there are not many very long roots in Standard Esperanto -- but the fact that a word like enciklopedio cannot be shortened (*enco, *encio) seems to indicate that prosody is irrelevant indeed.

At this point it is useful to return to the original example dista (< distanca). In this case the sequence -anc which derives from the Latin participial ending -ans (including a theme vowel) seems to have been reanalysed as a suffix. Probably this participial ending is taken as nominal (because distance in French and English, Distanz in German, etc. are nouns). The derivation would then be something like distanco > disto > dista. Interestingly, although most of my informants claim that they use neither distanca nor dista themselves, some of them report a meaning difference, which can be summarized approximately as follows. Distanca seems to mean 'related to distance', whereas dista means `away'. In other words, dista seems to have a typical inherently adjectival meaning, whereas distanca seems to behave as typically nominal.

The explanation of the contrast between dista and *tanca (<distanca) also no longer needs to be prosodic: Beckman's (1997) positional faithfulness is not needed to explain the difference between these forms, because this difference can be described in terms of morphological economy of representation alone.

The last issue is Homonymenflucht. Philippe (1991) claims that shortening does not take place if the resulting noun would be a homonym to some other word:

(19) arbitracio > *arbitro (arbitration)
(musical variation)

Homonymenflucht is not easy to capture in the kind of formal grammar used here: it involves complex comparisons of unrelated words within the lexicon in order to see whether they are homonymous. It should be noted however that the non-shortening words in these examples (and as far as I can see in all the others in Philippe 1991) have a rather specialized and technical meaning. It is very well possible that these technical terms form a special stratum within the Esperanto lexicon that is not sensitive to shortening (e.g. because faithfulness constraints would get a relatively high ranking in this stratum). The Homonymenflucht would in this case be nothing but an artifact.


In this article I have described an ongoing shortening process in the Esperanto lexicon. Many 'Standard' words get shortened in a variant of Esperanto that I call Colloquial. The shortening does not seem to obey any specific prosodic restrictions however. There is no role for prosodic templates in the analysis presented here; shortening is only triggered by a form of morphological economy of representation: the same categorial feature cannot be expressed more than once within the same word.

A point of interest is that the constraint rankings for the two variants of Esperanto do not differ: a speaker of CE has the same grammar as Zamenhof had in all relevant respects. The only real difference is that words such as erotika get assigned a more complex internal structure in CE than in the standard variant. Intuitively, this seems to be connected to the fact that the affixal system of Esperanto is generally much more productive than that of most (Indo-European) languages. The freedom of analysis that a speaker of CE experiences in words such as samseksemulo may be carried over to evolucio as well. It would be interesting to study how this works for speakers who do not speak any other Indo-European languages, but I have not been able to track down such a person.


Aronoff, M. (1995) Morphology by Itself. Cambridge, Mass., The MIT Press.

Beckman, J. (1997) Positional Faithfulness, PhD Dissertation, University of Massachusetts, Amherst.

Chomsky, N. (1995) The Minimalist Program, Cambridge, Mass., The MIT Press.

Kager, R. (1995) `On foot templates and root templates'. In: M. den Dikken and K. Hengeveld (eds.) Linguistics in the Netherlands 1995. Amsterdam, John Benjamins.

Kiparsky, P. (1982) `From Cyclic Phonology to Lexical Phonology', IN: H. van der Hulst and N. Smith (eds.) The Structure of Phonological Representations. Vol I. Foris, Dordrecht. pp. 130-175.

Lindstedt, J. (1997) `Kreoligho kaj spontanea shanghigho de Esperanto', Manuscript, University of Helsinki. Available at (Version in Finnish:

McCarthy, J. and A. Prince (1986) `Prosodic Morphology', Manuscript, University of Massachusetts, Amherst and Brandeis University, Waltham, Mass.

McCarthy, J. and A. Prince (1995) `Faithfulness and Reduplicative Identity', in L. Beckman, et al. (eds.) Papers in Optimality Theory. GLSA UMass, Amherst. p. 252-384.

McCarthy, J. and A. Prince (1998) `Prosodic Morphology', in A. Spencer and A. Zwicky, eds., The Handbook of Morphology, Oxford, Blackwell. p. 283-305.

Oostendorp, M. van (1997) `Style level in conflict resolution'. In: F. Hinskens et al. (eds.) Language Variation and Linguistic Theory.

Philippe, B. (1991) Sprachwandel bei einer Plansprache am Beispiel des Esperanto. Konstanz, Hartung-Gorre Verlag.

Poser, W. (1984) `Hypocoristic formation in Japanese', Proceedings of the 3rd West Coast Conference on Formal Linguistics, 218-229.

Poser, W. (1990) `Evidence for foot structure in Japanese', Language 66:78-105.

Saussure, R. de (1910) La construction logique des mots en espéranto, Paris, UEL.

Versteegh, K. (1993) `Esperanto as a First Language: Acquisition with Restricted Input'. Linguistics 31: 539-555

Vijver, R. van de (1997) `On Dutch clippings'. In: J. Coerts and H. de Hoop (eds.) Linguistics in the Netherlands 1997. Amsterdam, John Benjamins.

Zamenhof, L. (1991) Iom Reviziita Plena Verkaro de Zamenhof [Somewhat revised edition of the complete works of Zamenhof]. Kyoto, Ludovikito.