Sean M. Burke, 1996-03-25

English -ent and -ant

Abstract: In this paper I will examine generalizations about the meaning and distribution of the English suffixes -ent and -ant, with note to their range of meaning, productivity, and how this relates to when -ent is used as opposed to -ant.

[Note, 1997-7-7: I wrote this paper for a class I once took. I wrote this basically because, even tho' I'd had a year of Latin, I could still never remember what words had -ant and what words had -ent, so I decided to puzzle the issue out. The one word I always trip on is "existence", because I figure it's from stâre and therefore should behave like an -are verb (like constâre > constant, instâre > instant, etc.) However, -sta- shortens (exceptionally?) in forming sistere, which ex(s)istere is based on; so "exist" is an -ere verb, and therefore regularly takes -ent. This's the only real tricky one I know of.]

[Footnote: In developing generalizations about -nt words, I made heavy use of the UNIX wordlist, /usr/dict/words. See the Appendix, attached, for a list of -nt words within the scope of this paper. Forms outside the scope of this paper are -ment words, and words with etymologies which are clearly not of the form V+nt (e.g., elephant).]

Marchand (1960:195)'s account of the distribution of -ent and -ant is basically that it's the conjugational class of the Latin etymon that determines which form (-ant or -ent) appears. In order to evaluate this hypothesis, I will have to say a few words about Latin verbal morphology.

Every Latin verb belongs to one of the several conjugation classes. A given verb's class cannot be predicted based on its semantics, but is a diacritic for each verb in the lexicon. [Footnote: Alternate analyses exist which consider the conjugation class to be determined not by a diacritic, but in an underlying phonological form of the verb. This paper is agnostic as far as which of these models is most correct.] Which class a verb belongs to determines how it is inflected. For our purposes, we can consider there to be three conjugation classes, named after the last three letters of the active present infinitive form:
-are class which produces forms in -ant
-ire class which produces forms in -ient
-ere class which produces forms in -ent (or occasionally -ient)

For example, fevere "to boil" is in the class of -ere verbs and so gives fervent; convenire (to be suitable) is in the class of -ire verbs and so gives convenient, and expectare is in the class of -are verbs and so gives expectant.

Crucially, every Latin verb has exactly one -ant or -ent form, and whether it chooses an -ent or an -ant form is entirely predictable based on the verb class. Consequently, in Latin, no verb ever shows both an -ent and an -ant form. Because Latin -ant and -ent appear in complementary distribution and show a constancy of meaning, we must say that in Latin, there was a morpheme -nt, with allomorphs -ant or -ent, as conditioned by the conjugation class of the verb affixed to.

The meaning of the -nt morpheme in Latin was constant: the adjective produced always meant that "X does V" (habitually or characteristically), where V is the verb -nt is affixed to, and X is whatever the adjective modifies. The theta-roles X can play relative to V seem to be any of the roles subject can play relative to finite verbs. It does not seem that case that, for example, X can only be the agent of V; there are plenty of cases of X being a theme (convenient, X is fitting).

Once the -nt adjective is formed, it can be converted to a noun, as all adjectives in Latin can be. Once converted, an -nt form has the meaning "the person or thing which Xs".

Because every Latin verb can produce an -nt form, it is tempting to call this an inflectional morpheme. However, once formed, -nt forms were free to undergo semantic drift, which leads us to want to call -nt a derivational morpheme instead. For example: serpent and rodent (<serpere, to creep, <rodere, to gnaw) does not mean simply anything or anyone that creeps or gnaws, respectively, but specifically snakes, and a class of small furry mammals.

With this information about Latin morphology, we can formulate and check this generalization about English -ant and -ent:

-ent and -ant in English are allomorphs of one morpheme meaning "X does V", where V is the V affixed to; which of the two forms appears when affixed to a given verb depends on a diacritic on the verb, a diacritic which specifies what conjugation class the Latin etymon belonged to.

This is, in essence, the view Marchand (1960:195) holds. However, it leaves unanswered several questions, which I will now address.

First off, we have to account for what -ant/-ent will do when attached to words which were not Latin. "Not Latin" can mean several things. It can mean a root completely not of any Latin, Latinate, or Romance origin, as in these forms based on English roots of Germanic origin:

coolant, sealant, dopant

"Not Latin" can also mean roots of Romance origin, but not borrowed directly from Latin; the roots would then show signs of phonological change that happened in Vulgar Latin. E.g.:

pleasant, repentant, claimant, pliant, accountant, cognizant, issuant, clairvoyant, combatant, vagrant, entrant, restaurant, lieutenant, puissant, confidant, trenchant, merchant, valiant, sargeant, pursuant

(cf. the form these would have if they were based on the Latin etyma of the Romance roots: *placent, *repenitant, *clamant, *plicant,*accomptant, and *cognoscent, etc.).

We note that these Romance-but-not-directly-Latin forms are all -ant, as are the English-and-not-Latin forms; there is also a case of a Romance form which is not of Latin origin:


In the data so far, if the root is Latin, then the choice of -ant/-ent is determined by the conjugation class of the Latin verb, otherwise it's going to be -ant. To reformalize our latest version of our generalization for the distribution of -ant versus -ent:

Which of the two forms (-ant or -ent) appears when affixed to a given verb depends on the origin of the verb; if the verb is borrowed from Latin, then a conjugation-class diacritic will determine -ant or -ent. If the verb is borrowed from Romance, or is a native (Germanic) English word, then the form will always be -ant.

This approach would neatly explain cognate doublets or pseudo-doublets, like these, where the first is of Romance origin, and the latter is directly from Latin:
repentant, penitent
covenant, convenient
complaisant, complacent
remnant, immanent

With these forms (which are all ultimately from on -ere and -ire class verbs in Latin) we would say that since the roots we see in the left column are from Romance, they are not liable to take an -ent; the "true Latin" forms, by contrast, can take -ent, and in these cases must, since all the roots here are -ere and -ire class. Similarly, with


(which should give *servient) we would say that while there is a Latin verb root serv- (ire class), the root we borrow is the Romance root (identical in form) serv-, and so we have an -ant instead of an -ient.

But consider:


The root react is a true Latin form [Footnote: We have phonological grounds for this: if it had gone thru Romance, the ct cluster would have been simplified, giving something like *rait.] ; and react in Latin is an ere-verb (the infinitive is reagere). So, we would expect *reactent, which we do not find. Possible justification for this is that react- in Latin would not be able to take an -ent since react- is not the infinitive stem [Footnote: In the terms traditionally used, react is the fourth of the four principle parts of a Latin verb, which English generally uses to form -ion forms; whereas it's the first principle part (the present-stem, which for our purposes we'll call the infinitive stem) of the verb that the -ent/-ant gets combined with.]; only the infinitive stem (here, reag-) can combine with the -ent; we would expect *reagent (which seems possible, but is not an occurring form).

We could repair our generalization to accommodate this:

Which of the two forms (-ant or -ent) appears when affixed to a given verb depends on the origin of the verb; if the verb is borrowed from Latin and is the appropriate form of the root to combine with an -ent/-ant, then a conjugation-class diacritic will determine -ant or -ent. Otherwise (i.e., if the verb is Latin but not the right form of the root -- or is native-Germanic English -- or is Romance) then the form will always be -ant.

This generalization covers all of the data I have managed to collect. However, there are serious problems with the implications of this generalization, which I will now elaborate on.

Consider this troubling pair of words, similar to what we saw with servant:
intendant superintendent

To fit this into our generalization, we have to posit that there are two verb roots intend-; both are completely identical in form, and both mean the same thing (i.e., to look after, a sense of intend which has been lost in the free verb morpheme intend); however, our generalization would have us believe that intend- in intendant is marked as being not Latin and therefore always gets the -ant form, whereas intend- in superintendent is marked as being Latin, as being the infinitive stem form, and as being an ere class verb.

All of these diacritics need to be stored in the lexicon to make our generalization work out right; but how could they be acquired? In learning the word intendant, is a child to assume that it is a Romance or English form, an Latin -are class verb, or maybe an -ire or -ere class verb that we don't here have the infinitive stem of? We could collapse these all these diacritics into one, and say that English speakers have one diacritic, which specifies whether or not a given verb is an -ere or -ire class Latin verb which we have the infinitive stem of.

But this is basically no different from saying that we have a diacritic which says either "add -ant" or "add -ent", somewhat of a null hypothesis that I think there is an alternative to, which I will now develop:

I have explained that in Latin, V-nt had the meaning "does V", and have so far implicitly assumed that this is all that can be said about the meanings English -ant/-ent words have.

But consider these words:

coolant, sealant, dopant, depressant, disinfectant, stimulant, refrigerant, surfactant, oxidant, solvent, pollutant, reactant, deodorant, substituent, lubricant, intoxicant, nucleant, fumigant, diluent, propellant

All of these refer to substances (chemical, physical, etc.) which do the action denoted by the verb root (a coolant cools, a dopant dopes, a deodorant deodorizes). They are all also nouns. And none of them can refer to a person who does the action denoted by the verb root, or even a machine which does the action denoted by the verb root; a cooler is not a coolant; a refrigerator is not a refrigerant, etc.

This sub-family of -ant/-ent words can be described (i.e., they refer to substances, not people or machines; and they are all nouns) in a way that other -ant/-ent words can't.

Consider this group of words:

resident, applicant, claimant, litigant, defendant, tenant, inhabitant, accountant, complainant, dependant, informant, consultant, attestant, disputant

All of these are nouns (although resident and dependant can be adjectives); all of them denote people (not substances or machines) which do the action of the verb that -ant/-ent has been attached to. And the actions denoted designate a somewhat cohesive semantic domain, which Marchand (1960:195) characterizes by saying that they are "either outspoken legal terms or carry the stamp of formal procedure". To this group we could probably add:

respondent, contestant

Note also that these forms (which I'll call substance-agent forms and procedure-agent form) have a preference for roots different than -ent/-ant words on the whole.

Of the substance-agent forms and procedure-agent forms, all but three (solvent, surfactant, and tenant) are based on free morphemes words (claim, dope, dilute, etc.); the set of -ant/-ent words as a whole carries a much higher proportion (roughly a 1:2 ratio) of being based on bound (or singleton) to as opposed to free verb roots.

Substance-agent and procedure-agent forms both show a much greater preference for -ant than for -ent; whereas in the group of -ant/-ent words as a whole, there is a nearly equal distribution in -ent and -ant forms.

Moreover, substance-agent forms are more productive than -ant/-ent forms on the whole; coolant, dopant, refrigerant, depressant, lubricant, and many others date only from the past two centuries. (Procedure-agent forms seemed to have had a peak in productivity in the period 1500-1750, but seemed not to have produced much lately.) (Marchand 1960). Almost all other -ant/-ent forms, in contrast, entered the language before 1500.

I also observe that substance-agents and procedure-agents seem generally ineligible for the -nt to -nce/-ncy alternation that Marchand (1960:192) notes:

*coolancy or *coolance, *sealancy or **sealance, refrigerancy,... *applicancy, *litigancy, *[etc]

(exceptions are solvency and residence) whereas most of the larger class of -ent/-ant words can have an -nce or -ncy alternant:

obedience, truancy, exigency, negligence, quiescence, distance, [etc].

Because of these clear differences in productivity, in selection for different types of verbs, in the greater distribution of -ant, and in range of meaning, I believe that the substance-agents and procedure-agents are each different morphemes from the Latin -ant/-ent forms. While this is not to say that a Latin form can't have a substance-agent meaning (e.g., solvent or even possibly birefringent).

My final generalization is this: there are two -nt morphemes:

One has allomorphs -ent and -ant, as determined by a (synchronically) unpredictable diacritic. With few exceptions, this set was borrowed into English from Latin (and some from French) before 1500; about a third of this set is based on bound or singleton forms, as:

fervent, pedant, covenant, merchant, insouciant, vigilant, flagrant, serpent, tangent

Very many non-bound forms have undergone significant semantic drift from the meaning of the verb roots which also survive in English:
-nt formcognate verb
radiant radiate
hydrant hydrate
confident confide
confidant confide
convenient convene

While the original sense of the -ant/-ent suffix was an adjective meaning that X(what the adjective is predicated of) does the action denoted by the verb, this sense is unrecoverable in at least a third of all -ent/-ant words diachronically of the form V+nt. (I count a form as unrecoverable if it is based on a singleton or opaque root (e.g., ambient, transience) or thru having had significant semantic drift take place (e.g., hydrant).)

The other morpheme is another -ant; this group includes the substance-agents and the procedure-agents. This group is much more productive in English than Latin -ant/-ent is, in that it is the basis for recent coinages, and in that it does not restrict for its verb being a "true Latin" form anywhere as tightly as the Latin -ant/-ent class does; it generates almost entirely mostly nouns; its meaning is more predictable, denoting either a person as an agent in a procedural action, or a substance as an agent in a physical/chemical action.


---. ?1987. "/usr/dict/words", SunOS v4.

[various]. ?1989. Oxford English Dictionary, CDROM version.

[various]. ?1993. Hypertext Websters Interface,

Greenough, James Bradstreet and George L. Kittredge. 1895. Latin Grammar. Ginn and Company, Boston.

Marchand, Hans, 1960. The Categories and Types of Present Day English Word-Formation: A Synchronic-Diachronic Apporach. Harrassowitz, Weisbaden.

Skeat, Walter W. 1980. A Concise Etymological Dictionary of the English Language. Perigree, NYC.

Weekley, Ernest. 1967. An Etymological Dictionary of Modern English. Dover, NYC.

Wohlberg, Joseph. 1964. 201 Latin Verbs. Barron, NYC.

Appendix: English -ant/-ent words

derived from /usr/dict/words:

aberrant abeyant abhorrent absent absorbent abstinent abundant acceptant accident accordant accountant adherent adjacent adjutant adolescent advertent afferent affluent agent allegiant ambient ambulant ancient antecedent antiperspirant apparent appellant applicant approximant arctangent ardent arrogant ascendant ascendent aspirant assailant assistant assonant attendant belligerent beneficent birefringent blatant bouffant brilliant buoyant cadent celebrant circulant claimant clairvoyant coalescent coefficient cogent cognizant coherent combatant commandant communicant competent complacent complainant complaisant compliant component concomitant concordant concurrent confidant confident confluent congruent consistent consonant constituent consultant contaminant contestant continent contingent continuant convalescent convenient convergent conversant coolant cotangent covenant credent crescent currant current decadent decedent decent defendant deferent defiant deficient delinquent deliquescent deodorant dependent depressant descendant descendent descent despondent detergent determinant deterrent deviant different diffident diligent diluent discordant discrepant discriminant discussant disputant dissident dissonant distant divergent dominant dopant dormant ebullient efferent efficient efflorescent effluent elegant eloquent emergent emigrant eminent entrant errant evanescent evident excellent excrescent exigent existent exorbitant expectant expectorant expedient exponent extravagant exuberant exultant fervent flagrant flamboyant flippant fluent fluorescent formant fragrant fumigant gallant gallivant gradient habitant hesitant hydrant ignorant immanent immigrant imminent implicant important impudent incandescent incessant incident incipient incumbent independent indicant indigent indignant indolent indulgent infant influent informant ingredient inhabitant inherent innocent insistent insolent insolvent insouciant insurgent intelligent intendant intermittent interpolant intoxicant intransigent irritant issuant itinerant jubilant latent lenient lieutenant litigant lubricant luminescent luxuriant magnificent malevolent malfeasant malignant merchant migrant militant miscreant munificent mutant nascent natant negligent nonchalant nucleant nutrient obedient obeisant observant obsolescent obstruent occident occupant occurrent omniscient opalescent operant opponent orient oxidant parent participant patent patent patient pedant penchant pendant penitent permanent perseverant persistent pertinent pestilent petulant phosphorescent piquant pleasant pliant poignant pollutant potent precedent pregnant preponderant present president proficient prominent propellant proponent protestant protuberant provident prudent prurient pubescent puissant pungent pursuant quiescent radiant rampant reactant recalcitrant recent recipient recombinant recumbent recurrent recusant redundant referent regent registrant relevant reliant reluctant reminiscent remnant repellent repentant repugnant resemblant resident resilient resistant resonant resplendent respondent restaurant resultant resurgent retardant reticent reverent rodent ruminant salient sapient savant sealant segregant sentient sequent sergeant serpent servant sibilant significant silent solvent sonant stagnant stimulant strident stringent student subservient subsistent substituent sufficient superintendent surfactant surveillant tangent tenant tolerant torrent transcendent transient translucent transparent trenchant triumphant truant turbulent urgent vacant vagrant valent valiant variant verdant vibrant vigilant violent, 1997-05-01